1 @c Copyright (C) 1988-2016 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Extensions to the C Language Family
8 @cindex extensions, C language
9 @cindex C language extensions
12 GNU C provides several language features not found in ISO standard C@.
13 (The @option{-pedantic} option directs GCC to print a warning message if
14 any of these features is used.) To test for the availability of these
15 features in conditional compilation, check for a predefined macro
16 @code{__GNUC__}, which is always defined under GCC@.
18 These extensions are available in C and Objective-C@. Most of them are
19 also available in C++. @xref{C++ Extensions,,Extensions to the
20 C++ Language}, for extensions that apply @emph{only} to C++.
22 Some features that are in ISO C99 but not C90 or C++ are also, as
23 extensions, accepted by GCC in C90 mode and in C++.
26 * Statement Exprs:: Putting statements and declarations inside expressions.
27 * Local Labels:: Labels local to a block.
28 * Labels as Values:: Getting pointers to labels, and computed gotos.
29 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
30 * Constructing Calls:: Dispatching a call to another function.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
33 * __int128:: 128-bit integers---@code{__int128}.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Floating Types:: Additional Floating Types.
37 * Half-Precision:: Half-Precision Floating Point.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Named Address Spaces::Named address spaces.
42 * Zero Length:: Zero-length arrays.
43 * Empty Structures:: Structures with no members.
44 * Variable Length:: Arrays whose length is computed at run time.
45 * Variadic Macros:: Macros with a variable number of arguments.
46 * Escaped Newlines:: Slightly looser rules for escaped newlines.
47 * Subscripting:: Any array can be subscripted, even if not an lvalue.
48 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
49 * Pointers to Arrays:: Pointers to arrays with qualifiers work as expected.
50 * Initializers:: Non-constant initializers.
51 * Compound Literals:: Compound literals give structures, unions
53 * Designated Inits:: Labeling elements of initializers.
54 * Case Ranges:: `case 1 ... 9' and such.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Mixed Declarations:: Mixing declarations and code.
57 * Function Attributes:: Declaring that functions have no side effects,
58 or that they can never return.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Label Attributes:: Specifying attributes on labels.
62 * Enumerator Attributes:: Specifying attributes on enumerators.
63 * Attribute Syntax:: Formal syntax for attributes.
64 * Function Prototypes:: Prototype declarations and old-style definitions.
65 * C++ Comments:: C++ comments are recognized.
66 * Dollar Signs:: Dollar sign is allowed in identifiers.
67 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
68 * Alignment:: Inquiring about the alignment of a type or variable.
69 * Inline:: Defining inline functions (as fast as macros).
70 * Volatiles:: What constitutes an access to a volatile object.
71 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
72 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
73 * Incomplete Enums:: @code{enum foo;}, with details to follow.
74 * Function Names:: Printable strings which are the name of the current
76 * Return Address:: Getting the return or frame address of a function.
77 * Vector Extensions:: Using vector instructions through built-in functions.
78 * Offsetof:: Special syntax for implementing @code{offsetof}.
79 * __sync Builtins:: Legacy built-in functions for atomic memory access.
80 * __atomic Builtins:: Atomic built-in functions with memory model.
81 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
82 arithmetic overflow checking.
83 * x86 specific memory model extensions for transactional memory:: x86 memory models.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
87 * Cilk Plus Builtins:: Built-in functions for the Cilk Plus language extension.
88 * Other Builtins:: Other built-in functions.
89 * Target Builtins:: Built-in functions specific to particular targets.
90 * Target Format Checks:: Format checks specific to particular targets.
91 * Pragmas:: Pragmas accepted by GCC.
92 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
93 * Thread-Local:: Per-thread variables.
94 * Binary constants:: Binary constants using the @samp{0b} prefix.
98 @section Statements and Declarations in Expressions
99 @cindex statements inside expressions
100 @cindex declarations inside expressions
101 @cindex expressions containing statements
102 @cindex macros, statements in expressions
104 @c the above section title wrapped and causes an underfull hbox.. i
105 @c changed it from "within" to "in". --mew 4feb93
106 A compound statement enclosed in parentheses may appear as an expression
107 in GNU C@. This allows you to use loops, switches, and local variables
108 within an expression.
110 Recall that a compound statement is a sequence of statements surrounded
111 by braces; in this construct, parentheses go around the braces. For
115 (@{ int y = foo (); int z;
122 is a valid (though slightly more complex than necessary) expression
123 for the absolute value of @code{foo ()}.
125 The last thing in the compound statement should be an expression
126 followed by a semicolon; the value of this subexpression serves as the
127 value of the entire construct. (If you use some other kind of statement
128 last within the braces, the construct has type @code{void}, and thus
129 effectively no value.)
131 This feature is especially useful in making macro definitions ``safe'' (so
132 that they evaluate each operand exactly once). For example, the
133 ``maximum'' function is commonly defined as a macro in standard C as
137 #define max(a,b) ((a) > (b) ? (a) : (b))
141 @cindex side effects, macro argument
142 But this definition computes either @var{a} or @var{b} twice, with bad
143 results if the operand has side effects. In GNU C, if you know the
144 type of the operands (here taken as @code{int}), you can define
145 the macro safely as follows:
148 #define maxint(a,b) \
149 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
152 Embedded statements are not allowed in constant expressions, such as
153 the value of an enumeration constant, the width of a bit-field, or
154 the initial value of a static variable.
156 If you don't know the type of the operand, you can still do this, but you
157 must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).
159 In G++, the result value of a statement expression undergoes array and
160 function pointer decay, and is returned by value to the enclosing
161 expression. For instance, if @code{A} is a class, then
170 constructs a temporary @code{A} object to hold the result of the
171 statement expression, and that is used to invoke @code{Foo}.
172 Therefore the @code{this} pointer observed by @code{Foo} is not the
175 In a statement expression, any temporaries created within a statement
176 are destroyed at that statement's end. This makes statement
177 expressions inside macros slightly different from function calls. In
178 the latter case temporaries introduced during argument evaluation are
179 destroyed at the end of the statement that includes the function
180 call. In the statement expression case they are destroyed during
181 the statement expression. For instance,
184 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
185 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
195 has different places where temporaries are destroyed. For the
196 @code{macro} case, the temporary @code{X} is destroyed just after
197 the initialization of @code{b}. In the @code{function} case that
198 temporary is destroyed when the function returns.
200 These considerations mean that it is probably a bad idea to use
201 statement expressions of this form in header files that are designed to
202 work with C++. (Note that some versions of the GNU C Library contained
203 header files using statement expressions that lead to precisely this
206 Jumping into a statement expression with @code{goto} or using a
207 @code{switch} statement outside the statement expression with a
208 @code{case} or @code{default} label inside the statement expression is
209 not permitted. Jumping into a statement expression with a computed
210 @code{goto} (@pxref{Labels as Values}) has undefined behavior.
211 Jumping out of a statement expression is permitted, but if the
212 statement expression is part of a larger expression then it is
213 unspecified which other subexpressions of that expression have been
214 evaluated except where the language definition requires certain
215 subexpressions to be evaluated before or after the statement
216 expression. In any case, as with a function call, the evaluation of a
217 statement expression is not interleaved with the evaluation of other
218 parts of the containing expression. For example,
221 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
225 calls @code{foo} and @code{bar1} and does not call @code{baz} but
226 may or may not call @code{bar2}. If @code{bar2} is called, it is
227 called after @code{foo} and before @code{bar1}.
230 @section Locally Declared Labels
232 @cindex macros, local labels
234 GCC allows you to declare @dfn{local labels} in any nested block
235 scope. A local label is just like an ordinary label, but you can
236 only reference it (with a @code{goto} statement, or by taking its
237 address) within the block in which it is declared.
239 A local label declaration looks like this:
242 __label__ @var{label};
249 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
252 Local label declarations must come at the beginning of the block,
253 before any ordinary declarations or statements.
255 The label declaration defines the label @emph{name}, but does not define
256 the label itself. You must do this in the usual way, with
257 @code{@var{label}:}, within the statements of the statement expression.
259 The local label feature is useful for complex macros. If a macro
260 contains nested loops, a @code{goto} can be useful for breaking out of
261 them. However, an ordinary label whose scope is the whole function
262 cannot be used: if the macro can be expanded several times in one
263 function, the label is multiply defined in that function. A
264 local label avoids this problem. For example:
267 #define SEARCH(value, array, target) \
270 typeof (target) _SEARCH_target = (target); \
271 typeof (*(array)) *_SEARCH_array = (array); \
274 for (i = 0; i < max; i++) \
275 for (j = 0; j < max; j++) \
276 if (_SEARCH_array[i][j] == _SEARCH_target) \
277 @{ (value) = i; goto found; @} \
283 This could also be written using a statement expression:
286 #define SEARCH(array, target) \
289 typeof (target) _SEARCH_target = (target); \
290 typeof (*(array)) *_SEARCH_array = (array); \
293 for (i = 0; i < max; i++) \
294 for (j = 0; j < max; j++) \
295 if (_SEARCH_array[i][j] == _SEARCH_target) \
296 @{ value = i; goto found; @} \
303 Local label declarations also make the labels they declare visible to
304 nested functions, if there are any. @xref{Nested Functions}, for details.
306 @node Labels as Values
307 @section Labels as Values
308 @cindex labels as values
309 @cindex computed gotos
310 @cindex goto with computed label
311 @cindex address of a label
313 You can get the address of a label defined in the current function
314 (or a containing function) with the unary operator @samp{&&}. The
315 value has type @code{void *}. This value is a constant and can be used
316 wherever a constant of that type is valid. For example:
324 To use these values, you need to be able to jump to one. This is done
325 with the computed goto statement@footnote{The analogous feature in
326 Fortran is called an assigned goto, but that name seems inappropriate in
327 C, where one can do more than simply store label addresses in label
328 variables.}, @code{goto *@var{exp};}. For example,
335 Any expression of type @code{void *} is allowed.
337 One way of using these constants is in initializing a static array that
338 serves as a jump table:
341 static void *array[] = @{ &&foo, &&bar, &&hack @};
345 Then you can select a label with indexing, like this:
352 Note that this does not check whether the subscript is in bounds---array
353 indexing in C never does that.
355 Such an array of label values serves a purpose much like that of the
356 @code{switch} statement. The @code{switch} statement is cleaner, so
357 use that rather than an array unless the problem does not fit a
358 @code{switch} statement very well.
360 Another use of label values is in an interpreter for threaded code.
361 The labels within the interpreter function can be stored in the
362 threaded code for super-fast dispatching.
364 You may not use this mechanism to jump to code in a different function.
365 If you do that, totally unpredictable things happen. The best way to
366 avoid this is to store the label address only in automatic variables and
367 never pass it as an argument.
369 An alternate way to write the above example is
372 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
374 goto *(&&foo + array[i]);
378 This is more friendly to code living in shared libraries, as it reduces
379 the number of dynamic relocations that are needed, and by consequence,
380 allows the data to be read-only.
381 This alternative with label differences is not supported for the AVR target,
382 please use the first approach for AVR programs.
384 The @code{&&foo} expressions for the same label might have different
385 values if the containing function is inlined or cloned. If a program
386 relies on them being always the same,
387 @code{__attribute__((__noinline__,__noclone__))} should be used to
388 prevent inlining and cloning. If @code{&&foo} is used in a static
389 variable initializer, inlining and cloning is forbidden.
391 @node Nested Functions
392 @section Nested Functions
393 @cindex nested functions
394 @cindex downward funargs
397 A @dfn{nested function} is a function defined inside another function.
398 Nested functions are supported as an extension in GNU C, but are not
399 supported by GNU C++.
401 The nested function's name is local to the block where it is defined.
402 For example, here we define a nested function named @code{square}, and
407 foo (double a, double b)
409 double square (double z) @{ return z * z; @}
411 return square (a) + square (b);
416 The nested function can access all the variables of the containing
417 function that are visible at the point of its definition. This is
418 called @dfn{lexical scoping}. For example, here we show a nested
419 function which uses an inherited variable named @code{offset}:
423 bar (int *array, int offset, int size)
425 int access (int *array, int index)
426 @{ return array[index + offset]; @}
429 for (i = 0; i < size; i++)
430 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
435 Nested function definitions are permitted within functions in the places
436 where variable definitions are allowed; that is, in any block, mixed
437 with the other declarations and statements in the block.
439 It is possible to call the nested function from outside the scope of its
440 name by storing its address or passing the address to another function:
443 hack (int *array, int size)
445 void store (int index, int value)
446 @{ array[index] = value; @}
448 intermediate (store, size);
452 Here, the function @code{intermediate} receives the address of
453 @code{store} as an argument. If @code{intermediate} calls @code{store},
454 the arguments given to @code{store} are used to store into @code{array}.
455 But this technique works only so long as the containing function
456 (@code{hack}, in this example) does not exit.
458 If you try to call the nested function through its address after the
459 containing function exits, all hell breaks loose. If you try
460 to call it after a containing scope level exits, and if it refers
461 to some of the variables that are no longer in scope, you may be lucky,
462 but it's not wise to take the risk. If, however, the nested function
463 does not refer to anything that has gone out of scope, you should be
466 GCC implements taking the address of a nested function using a technique
467 called @dfn{trampolines}. This technique was described in
468 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
469 C++ Conference Proceedings, October 17-21, 1988).
471 A nested function can jump to a label inherited from a containing
472 function, provided the label is explicitly declared in the containing
473 function (@pxref{Local Labels}). Such a jump returns instantly to the
474 containing function, exiting the nested function that did the
475 @code{goto} and any intermediate functions as well. Here is an example:
479 bar (int *array, int offset, int size)
482 int access (int *array, int index)
486 return array[index + offset];
490 for (i = 0; i < size; i++)
491 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
495 /* @r{Control comes here from @code{access}
496 if it detects an error.} */
503 A nested function always has no linkage. Declaring one with
504 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
505 before its definition, use @code{auto} (which is otherwise meaningless
506 for function declarations).
509 bar (int *array, int offset, int size)
512 auto int access (int *, int);
514 int access (int *array, int index)
518 return array[index + offset];
524 @node Constructing Calls
525 @section Constructing Function Calls
526 @cindex constructing calls
527 @cindex forwarding calls
529 Using the built-in functions described below, you can record
530 the arguments a function received, and call another function
531 with the same arguments, without knowing the number or types
534 You can also record the return value of that function call,
535 and later return that value, without knowing what data type
536 the function tried to return (as long as your caller expects
539 However, these built-in functions may interact badly with some
540 sophisticated features or other extensions of the language. It
541 is, therefore, not recommended to use them outside very simple
542 functions acting as mere forwarders for their arguments.
544 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
545 This built-in function returns a pointer to data
546 describing how to perform a call with the same arguments as are passed
547 to the current function.
549 The function saves the arg pointer register, structure value address,
550 and all registers that might be used to pass arguments to a function
551 into a block of memory allocated on the stack. Then it returns the
552 address of that block.
555 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
556 This built-in function invokes @var{function}
557 with a copy of the parameters described by @var{arguments}
560 The value of @var{arguments} should be the value returned by
561 @code{__builtin_apply_args}. The argument @var{size} specifies the size
562 of the stack argument data, in bytes.
564 This function returns a pointer to data describing
565 how to return whatever value is returned by @var{function}. The data
566 is saved in a block of memory allocated on the stack.
568 It is not always simple to compute the proper value for @var{size}. The
569 value is used by @code{__builtin_apply} to compute the amount of data
570 that should be pushed on the stack and copied from the incoming argument
574 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
575 This built-in function returns the value described by @var{result} from
576 the containing function. You should specify, for @var{result}, a value
577 returned by @code{__builtin_apply}.
580 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
581 This built-in function represents all anonymous arguments of an inline
582 function. It can be used only in inline functions that are always
583 inlined, never compiled as a separate function, such as those using
584 @code{__attribute__ ((__always_inline__))} or
585 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
586 It must be only passed as last argument to some other function
587 with variable arguments. This is useful for writing small wrapper
588 inlines for variable argument functions, when using preprocessor
589 macros is undesirable. For example:
591 extern int myprintf (FILE *f, const char *format, ...);
592 extern inline __attribute__ ((__gnu_inline__)) int
593 myprintf (FILE *f, const char *format, ...)
595 int r = fprintf (f, "myprintf: ");
598 int s = fprintf (f, format, __builtin_va_arg_pack ());
606 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
607 This built-in function returns the number of anonymous arguments of
608 an inline function. It can be used only in inline functions that
609 are always inlined, never compiled as a separate function, such
610 as those using @code{__attribute__ ((__always_inline__))} or
611 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
612 For example following does link- or run-time checking of open
613 arguments for optimized code:
616 extern inline __attribute__((__gnu_inline__)) int
617 myopen (const char *path, int oflag, ...)
619 if (__builtin_va_arg_pack_len () > 1)
620 warn_open_too_many_arguments ();
622 if (__builtin_constant_p (oflag))
624 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
626 warn_open_missing_mode ();
627 return __open_2 (path, oflag);
629 return open (path, oflag, __builtin_va_arg_pack ());
632 if (__builtin_va_arg_pack_len () < 1)
633 return __open_2 (path, oflag);
635 return open (path, oflag, __builtin_va_arg_pack ());
642 @section Referring to a Type with @code{typeof}
645 @cindex macros, types of arguments
647 Another way to refer to the type of an expression is with @code{typeof}.
648 The syntax of using of this keyword looks like @code{sizeof}, but the
649 construct acts semantically like a type name defined with @code{typedef}.
651 There are two ways of writing the argument to @code{typeof}: with an
652 expression or with a type. Here is an example with an expression:
659 This assumes that @code{x} is an array of pointers to functions;
660 the type described is that of the values of the functions.
662 Here is an example with a typename as the argument:
669 Here the type described is that of pointers to @code{int}.
671 If you are writing a header file that must work when included in ISO C
672 programs, write @code{__typeof__} instead of @code{typeof}.
673 @xref{Alternate Keywords}.
675 A @code{typeof} construct can be used anywhere a typedef name can be
676 used. For example, you can use it in a declaration, in a cast, or inside
677 of @code{sizeof} or @code{typeof}.
679 The operand of @code{typeof} is evaluated for its side effects if and
680 only if it is an expression of variably modified type or the name of
683 @code{typeof} is often useful in conjunction with
684 statement expressions (@pxref{Statement Exprs}).
685 Here is how the two together can
686 be used to define a safe ``maximum'' macro which operates on any
687 arithmetic type and evaluates each of its arguments exactly once:
691 (@{ typeof (a) _a = (a); \
692 typeof (b) _b = (b); \
693 _a > _b ? _a : _b; @})
696 @cindex underscores in variables in macros
697 @cindex @samp{_} in variables in macros
698 @cindex local variables in macros
699 @cindex variables, local, in macros
700 @cindex macros, local variables in
702 The reason for using names that start with underscores for the local
703 variables is to avoid conflicts with variable names that occur within the
704 expressions that are substituted for @code{a} and @code{b}. Eventually we
705 hope to design a new form of declaration syntax that allows you to declare
706 variables whose scopes start only after their initializers; this will be a
707 more reliable way to prevent such conflicts.
710 Some more examples of the use of @code{typeof}:
714 This declares @code{y} with the type of what @code{x} points to.
721 This declares @code{y} as an array of such values.
728 This declares @code{y} as an array of pointers to characters:
731 typeof (typeof (char *)[4]) y;
735 It is equivalent to the following traditional C declaration:
741 To see the meaning of the declaration using @code{typeof}, and why it
742 might be a useful way to write, rewrite it with these macros:
745 #define pointer(T) typeof(T *)
746 #define array(T, N) typeof(T [N])
750 Now the declaration can be rewritten this way:
753 array (pointer (char), 4) y;
757 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
758 pointers to @code{char}.
761 In GNU C, but not GNU C++, you may also declare the type of a variable
762 as @code{__auto_type}. In that case, the declaration must declare
763 only one variable, whose declarator must just be an identifier, the
764 declaration must be initialized, and the type of the variable is
765 determined by the initializer; the name of the variable is not in
766 scope until after the initializer. (In C++, you should use C++11
767 @code{auto} for this purpose.) Using @code{__auto_type}, the
768 ``maximum'' macro above could be written as:
772 (@{ __auto_type _a = (a); \
773 __auto_type _b = (b); \
774 _a > _b ? _a : _b; @})
777 Using @code{__auto_type} instead of @code{typeof} has two advantages:
780 @item Each argument to the macro appears only once in the expansion of
781 the macro. This prevents the size of the macro expansion growing
782 exponentially when calls to such macros are nested inside arguments of
785 @item If the argument to the macro has variably modified type, it is
786 evaluated only once when using @code{__auto_type}, but twice if
787 @code{typeof} is used.
791 @section Conditionals with Omitted Operands
792 @cindex conditional expressions, extensions
793 @cindex omitted middle-operands
794 @cindex middle-operands, omitted
795 @cindex extensions, @code{?:}
796 @cindex @code{?:} extensions
798 The middle operand in a conditional expression may be omitted. Then
799 if the first operand is nonzero, its value is the value of the conditional
802 Therefore, the expression
809 has the value of @code{x} if that is nonzero; otherwise, the value of
812 This example is perfectly equivalent to
818 @cindex side effect in @code{?:}
819 @cindex @code{?:} side effect
821 In this simple case, the ability to omit the middle operand is not
822 especially useful. When it becomes useful is when the first operand does,
823 or may (if it is a macro argument), contain a side effect. Then repeating
824 the operand in the middle would perform the side effect twice. Omitting
825 the middle operand uses the value already computed without the undesirable
826 effects of recomputing it.
829 @section 128-bit Integers
830 @cindex @code{__int128} data types
832 As an extension the integer scalar type @code{__int128} is supported for
833 targets which have an integer mode wide enough to hold 128 bits.
834 Simply write @code{__int128} for a signed 128-bit integer, or
835 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
836 support in GCC for expressing an integer constant of type @code{__int128}
837 for targets with @code{long long} integer less than 128 bits wide.
840 @section Double-Word Integers
841 @cindex @code{long long} data types
842 @cindex double-word arithmetic
843 @cindex multiprecision arithmetic
844 @cindex @code{LL} integer suffix
845 @cindex @code{ULL} integer suffix
847 ISO C99 supports data types for integers that are at least 64 bits wide,
848 and as an extension GCC supports them in C90 mode and in C++.
849 Simply write @code{long long int} for a signed integer, or
850 @code{unsigned long long int} for an unsigned integer. To make an
851 integer constant of type @code{long long int}, add the suffix @samp{LL}
852 to the integer. To make an integer constant of type @code{unsigned long
853 long int}, add the suffix @samp{ULL} to the integer.
855 You can use these types in arithmetic like any other integer types.
856 Addition, subtraction, and bitwise boolean operations on these types
857 are open-coded on all types of machines. Multiplication is open-coded
858 if the machine supports a fullword-to-doubleword widening multiply
859 instruction. Division and shifts are open-coded only on machines that
860 provide special support. The operations that are not open-coded use
861 special library routines that come with GCC@.
863 There may be pitfalls when you use @code{long long} types for function
864 arguments without function prototypes. If a function
865 expects type @code{int} for its argument, and you pass a value of type
866 @code{long long int}, confusion results because the caller and the
867 subroutine disagree about the number of bytes for the argument.
868 Likewise, if the function expects @code{long long int} and you pass
869 @code{int}. The best way to avoid such problems is to use prototypes.
872 @section Complex Numbers
873 @cindex complex numbers
874 @cindex @code{_Complex} keyword
875 @cindex @code{__complex__} keyword
877 ISO C99 supports complex floating data types, and as an extension GCC
878 supports them in C90 mode and in C++. GCC also supports complex integer data
879 types which are not part of ISO C99. You can declare complex types
880 using the keyword @code{_Complex}. As an extension, the older GNU
881 keyword @code{__complex__} is also supported.
883 For example, @samp{_Complex double x;} declares @code{x} as a
884 variable whose real part and imaginary part are both of type
885 @code{double}. @samp{_Complex short int y;} declares @code{y} to
886 have real and imaginary parts of type @code{short int}; this is not
887 likely to be useful, but it shows that the set of complex types is
890 To write a constant with a complex data type, use the suffix @samp{i} or
891 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
892 has type @code{_Complex float} and @code{3i} has type
893 @code{_Complex int}. Such a constant always has a pure imaginary
894 value, but you can form any complex value you like by adding one to a
895 real constant. This is a GNU extension; if you have an ISO C99
896 conforming C library (such as the GNU C Library), and want to construct complex
897 constants of floating type, you should include @code{<complex.h>} and
898 use the macros @code{I} or @code{_Complex_I} instead.
900 @cindex @code{__real__} keyword
901 @cindex @code{__imag__} keyword
902 To extract the real part of a complex-valued expression @var{exp}, write
903 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
904 extract the imaginary part. This is a GNU extension; for values of
905 floating type, you should use the ISO C99 functions @code{crealf},
906 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
907 @code{cimagl}, declared in @code{<complex.h>} and also provided as
908 built-in functions by GCC@.
910 @cindex complex conjugation
911 The operator @samp{~} performs complex conjugation when used on a value
912 with a complex type. This is a GNU extension; for values of
913 floating type, you should use the ISO C99 functions @code{conjf},
914 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
915 provided as built-in functions by GCC@.
917 GCC can allocate complex automatic variables in a noncontiguous
918 fashion; it's even possible for the real part to be in a register while
919 the imaginary part is on the stack (or vice versa). Only the DWARF
920 debug info format can represent this, so use of DWARF is recommended.
921 If you are using the stabs debug info format, GCC describes a noncontiguous
922 complex variable as if it were two separate variables of noncomplex type.
923 If the variable's actual name is @code{foo}, the two fictitious
924 variables are named @code{foo$real} and @code{foo$imag}. You can
925 examine and set these two fictitious variables with your debugger.
928 @section Additional Floating Types
929 @cindex additional floating types
930 @cindex @code{__float80} data type
931 @cindex @code{__float128} data type
932 @cindex @code{__ibm128} data type
933 @cindex @code{w} floating point suffix
934 @cindex @code{q} floating point suffix
935 @cindex @code{W} floating point suffix
936 @cindex @code{Q} floating point suffix
938 As an extension, GNU C supports additional floating
939 types, @code{__float80} and @code{__float128} to support 80-bit
940 (@code{XFmode}) and 128-bit (@code{TFmode}) floating types.
941 Support for additional types includes the arithmetic operators:
942 add, subtract, multiply, divide; unary arithmetic operators;
943 relational operators; equality operators; and conversions to and from
944 integer and other floating types. Use a suffix @samp{w} or @samp{W}
945 in a literal constant of type @code{__float80} or type
946 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
948 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
949 types using the corresponding internal complex type, @code{XCmode} for
950 @code{__float80} type and @code{TCmode} for @code{__float128} type:
953 typedef _Complex float __attribute__((mode(TC))) _Complex128;
954 typedef _Complex float __attribute__((mode(XC))) _Complex80;
957 In order to use @code{__float128} and @code{__ibm128} on PowerPC Linux
958 systems, you must use the @option{-mfloat128}. It is expected in
959 future versions of GCC that @code{__float128} will be enabled
960 automatically. In addition, there are currently problems in using the
961 complex @code{__float128} type. When these problems are fixed, you
962 would use the following syntax to declare @code{_Complex128} to be a
963 complex @code{__float128} type:
965 On the PowerPC Linux VSX targets, you can declare complex types using
966 the corresponding internal complex type, @code{KCmode} for
967 @code{__float128} type and @code{ICmode} for @code{__ibm128} type:
970 typedef _Complex float __attribute__((mode(KC))) _Complex_float128;
971 typedef _Complex float __attribute__((mode(IC))) _Complex_ibm128;
974 Not all targets support additional floating-point types.
975 @code{__float80} and @code{__float128} types are supported on x86 and
976 IA-64 targets. The @code{__float128} type is supported on hppa HP-UX.
977 The @code{__float128} type is supported on PowerPC 64-bit Linux
978 systems by default if the vector scalar instruction set (VSX) is
981 On the PowerPC, @code{__ibm128} provides access to the IBM extended
982 double format, and it is intended to be used by the library functions
983 that handle conversions if/when long double is changed to be IEEE
984 128-bit floating point.
987 @section Half-Precision Floating Point
988 @cindex half-precision floating point
989 @cindex @code{__fp16} data type
991 On ARM targets, GCC supports half-precision (16-bit) floating point via
992 the @code{__fp16} type. You must enable this type explicitly
993 with the @option{-mfp16-format} command-line option in order to use it.
995 ARM supports two incompatible representations for half-precision
996 floating-point values. You must choose one of the representations and
997 use it consistently in your program.
999 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
1000 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
1001 There are 11 bits of significand precision, approximately 3
1004 Specifying @option{-mfp16-format=alternative} selects the ARM
1005 alternative format. This representation is similar to the IEEE
1006 format, but does not support infinities or NaNs. Instead, the range
1007 of exponents is extended, so that this format can represent normalized
1008 values in the range of @math{2^{-14}} to 131008.
1010 The @code{__fp16} type is a storage format only. For purposes
1011 of arithmetic and other operations, @code{__fp16} values in C or C++
1012 expressions are automatically promoted to @code{float}. In addition,
1013 you cannot declare a function with a return value or parameters
1014 of type @code{__fp16}.
1016 Note that conversions from @code{double} to @code{__fp16}
1017 involve an intermediate conversion to @code{float}. Because
1018 of rounding, this can sometimes produce a different result than a
1021 ARM provides hardware support for conversions between
1022 @code{__fp16} and @code{float} values
1023 as an extension to VFP and NEON (Advanced SIMD). GCC generates
1024 code using these hardware instructions if you compile with
1025 options to select an FPU that provides them;
1026 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1027 in addition to the @option{-mfp16-format} option to select
1028 a half-precision format.
1030 Language-level support for the @code{__fp16} data type is
1031 independent of whether GCC generates code using hardware floating-point
1032 instructions. In cases where hardware support is not specified, GCC
1033 implements conversions between @code{__fp16} and @code{float} values
1037 @section Decimal Floating Types
1038 @cindex decimal floating types
1039 @cindex @code{_Decimal32} data type
1040 @cindex @code{_Decimal64} data type
1041 @cindex @code{_Decimal128} data type
1042 @cindex @code{df} integer suffix
1043 @cindex @code{dd} integer suffix
1044 @cindex @code{dl} integer suffix
1045 @cindex @code{DF} integer suffix
1046 @cindex @code{DD} integer suffix
1047 @cindex @code{DL} integer suffix
1049 As an extension, GNU C supports decimal floating types as
1050 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1051 floating types in GCC will evolve as the draft technical report changes.
1052 Calling conventions for any target might also change. Not all targets
1053 support decimal floating types.
1055 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1056 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1057 @code{float}, @code{double}, and @code{long double} whose radix is not
1058 specified by the C standard but is usually two.
1060 Support for decimal floating types includes the arithmetic operators
1061 add, subtract, multiply, divide; unary arithmetic operators;
1062 relational operators; equality operators; and conversions to and from
1063 integer and other floating types. Use a suffix @samp{df} or
1064 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1065 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1068 GCC support of decimal float as specified by the draft technical report
1073 When the value of a decimal floating type cannot be represented in the
1074 integer type to which it is being converted, the result is undefined
1075 rather than the result value specified by the draft technical report.
1078 GCC does not provide the C library functionality associated with
1079 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1080 @file{wchar.h}, which must come from a separate C library implementation.
1081 Because of this the GNU C compiler does not define macro
1082 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1083 the technical report.
1086 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1087 are supported by the DWARF debug information format.
1093 ISO C99 supports floating-point numbers written not only in the usual
1094 decimal notation, such as @code{1.55e1}, but also numbers such as
1095 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1096 supports this in C90 mode (except in some cases when strictly
1097 conforming) and in C++. In that format the
1098 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1099 mandatory. The exponent is a decimal number that indicates the power of
1100 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1107 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1108 is the same as @code{1.55e1}.
1110 Unlike for floating-point numbers in the decimal notation the exponent
1111 is always required in the hexadecimal notation. Otherwise the compiler
1112 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1113 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1114 extension for floating-point constants of type @code{float}.
1117 @section Fixed-Point Types
1118 @cindex fixed-point types
1119 @cindex @code{_Fract} data type
1120 @cindex @code{_Accum} data type
1121 @cindex @code{_Sat} data type
1122 @cindex @code{hr} fixed-suffix
1123 @cindex @code{r} fixed-suffix
1124 @cindex @code{lr} fixed-suffix
1125 @cindex @code{llr} fixed-suffix
1126 @cindex @code{uhr} fixed-suffix
1127 @cindex @code{ur} fixed-suffix
1128 @cindex @code{ulr} fixed-suffix
1129 @cindex @code{ullr} fixed-suffix
1130 @cindex @code{hk} fixed-suffix
1131 @cindex @code{k} fixed-suffix
1132 @cindex @code{lk} fixed-suffix
1133 @cindex @code{llk} fixed-suffix
1134 @cindex @code{uhk} fixed-suffix
1135 @cindex @code{uk} fixed-suffix
1136 @cindex @code{ulk} fixed-suffix
1137 @cindex @code{ullk} fixed-suffix
1138 @cindex @code{HR} fixed-suffix
1139 @cindex @code{R} fixed-suffix
1140 @cindex @code{LR} fixed-suffix
1141 @cindex @code{LLR} fixed-suffix
1142 @cindex @code{UHR} fixed-suffix
1143 @cindex @code{UR} fixed-suffix
1144 @cindex @code{ULR} fixed-suffix
1145 @cindex @code{ULLR} fixed-suffix
1146 @cindex @code{HK} fixed-suffix
1147 @cindex @code{K} fixed-suffix
1148 @cindex @code{LK} fixed-suffix
1149 @cindex @code{LLK} fixed-suffix
1150 @cindex @code{UHK} fixed-suffix
1151 @cindex @code{UK} fixed-suffix
1152 @cindex @code{ULK} fixed-suffix
1153 @cindex @code{ULLK} fixed-suffix
1155 As an extension, GNU C supports fixed-point types as
1156 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1157 types in GCC will evolve as the draft technical report changes.
1158 Calling conventions for any target might also change. Not all targets
1159 support fixed-point types.
1161 The fixed-point types are
1162 @code{short _Fract},
1165 @code{long long _Fract},
1166 @code{unsigned short _Fract},
1167 @code{unsigned _Fract},
1168 @code{unsigned long _Fract},
1169 @code{unsigned long long _Fract},
1170 @code{_Sat short _Fract},
1172 @code{_Sat long _Fract},
1173 @code{_Sat long long _Fract},
1174 @code{_Sat unsigned short _Fract},
1175 @code{_Sat unsigned _Fract},
1176 @code{_Sat unsigned long _Fract},
1177 @code{_Sat unsigned long long _Fract},
1178 @code{short _Accum},
1181 @code{long long _Accum},
1182 @code{unsigned short _Accum},
1183 @code{unsigned _Accum},
1184 @code{unsigned long _Accum},
1185 @code{unsigned long long _Accum},
1186 @code{_Sat short _Accum},
1188 @code{_Sat long _Accum},
1189 @code{_Sat long long _Accum},
1190 @code{_Sat unsigned short _Accum},
1191 @code{_Sat unsigned _Accum},
1192 @code{_Sat unsigned long _Accum},
1193 @code{_Sat unsigned long long _Accum}.
1195 Fixed-point data values contain fractional and optional integral parts.
1196 The format of fixed-point data varies and depends on the target machine.
1198 Support for fixed-point types includes:
1201 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1203 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1205 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1207 binary shift operators (@code{<<}, @code{>>})
1209 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1211 equality operators (@code{==}, @code{!=})
1213 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1214 @code{<<=}, @code{>>=})
1216 conversions to and from integer, floating-point, or fixed-point types
1219 Use a suffix in a fixed-point literal constant:
1221 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1222 @code{_Sat short _Fract}
1223 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1224 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1225 @code{_Sat long _Fract}
1226 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1227 @code{_Sat long long _Fract}
1228 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1229 @code{_Sat unsigned short _Fract}
1230 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1231 @code{_Sat unsigned _Fract}
1232 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1233 @code{_Sat unsigned long _Fract}
1234 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1235 and @code{_Sat unsigned long long _Fract}
1236 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1237 @code{_Sat short _Accum}
1238 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1239 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1240 @code{_Sat long _Accum}
1241 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1242 @code{_Sat long long _Accum}
1243 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1244 @code{_Sat unsigned short _Accum}
1245 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1246 @code{_Sat unsigned _Accum}
1247 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1248 @code{_Sat unsigned long _Accum}
1249 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1250 and @code{_Sat unsigned long long _Accum}
1253 GCC support of fixed-point types as specified by the draft technical report
1258 Pragmas to control overflow and rounding behaviors are not implemented.
1261 Fixed-point types are supported by the DWARF debug information format.
1263 @node Named Address Spaces
1264 @section Named Address Spaces
1265 @cindex Named Address Spaces
1267 As an extension, GNU C supports named address spaces as
1268 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1269 address spaces in GCC will evolve as the draft technical report
1270 changes. Calling conventions for any target might also change. At
1271 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1272 address spaces other than the generic address space.
1274 Address space identifiers may be used exactly like any other C type
1275 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1276 document for more details.
1278 @anchor{AVR Named Address Spaces}
1279 @subsection AVR Named Address Spaces
1281 On the AVR target, there are several address spaces that can be used
1282 in order to put read-only data into the flash memory and access that
1283 data by means of the special instructions @code{LPM} or @code{ELPM}
1284 needed to read from flash.
1286 Per default, any data including read-only data is located in RAM
1287 (the generic address space) so that non-generic address spaces are
1288 needed to locate read-only data in flash memory
1289 @emph{and} to generate the right instructions to access this data
1290 without using (inline) assembler code.
1294 @cindex @code{__flash} AVR Named Address Spaces
1295 The @code{__flash} qualifier locates data in the
1296 @code{.progmem.data} section. Data is read using the @code{LPM}
1297 instruction. Pointers to this address space are 16 bits wide.
1304 @cindex @code{__flash1} AVR Named Address Spaces
1305 @cindex @code{__flash2} AVR Named Address Spaces
1306 @cindex @code{__flash3} AVR Named Address Spaces
1307 @cindex @code{__flash4} AVR Named Address Spaces
1308 @cindex @code{__flash5} AVR Named Address Spaces
1309 These are 16-bit address spaces locating data in section
1310 @code{.progmem@var{N}.data} where @var{N} refers to
1311 address space @code{__flash@var{N}}.
1312 The compiler sets the @code{RAMPZ} segment register appropriately
1313 before reading data by means of the @code{ELPM} instruction.
1316 @cindex @code{__memx} AVR Named Address Spaces
1317 This is a 24-bit address space that linearizes flash and RAM:
1318 If the high bit of the address is set, data is read from
1319 RAM using the lower two bytes as RAM address.
1320 If the high bit of the address is clear, data is read from flash
1321 with @code{RAMPZ} set according to the high byte of the address.
1322 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1324 Objects in this address space are located in @code{.progmemx.data}.
1330 char my_read (const __flash char ** p)
1332 /* p is a pointer to RAM that points to a pointer to flash.
1333 The first indirection of p reads that flash pointer
1334 from RAM and the second indirection reads a char from this
1340 /* Locate array[] in flash memory */
1341 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1347 /* Return 17 by reading from flash memory */
1348 return array[array[i]];
1353 For each named address space supported by avr-gcc there is an equally
1354 named but uppercase built-in macro defined.
1355 The purpose is to facilitate testing if respective address space
1356 support is available or not:
1360 const __flash int var = 1;
1367 #include <avr/pgmspace.h> /* From AVR-LibC */
1369 const int var PROGMEM = 1;
1373 return (int) pgm_read_word (&var);
1375 #endif /* __FLASH */
1379 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1380 locates data in flash but
1381 accesses to these data read from generic address space, i.e.@:
1383 so that you need special accessors like @code{pgm_read_byte}
1384 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1385 together with attribute @code{progmem}.
1388 @b{Limitations and caveats}
1392 Reading across the 64@tie{}KiB section boundary of
1393 the @code{__flash} or @code{__flash@var{N}} address spaces
1394 shows undefined behavior. The only address space that
1395 supports reading across the 64@tie{}KiB flash segment boundaries is
1399 If you use one of the @code{__flash@var{N}} address spaces
1400 you must arrange your linker script to locate the
1401 @code{.progmem@var{N}.data} sections according to your needs.
1404 Any data or pointers to the non-generic address spaces must
1405 be qualified as @code{const}, i.e.@: as read-only data.
1406 This still applies if the data in one of these address
1407 spaces like software version number or calibration lookup table are intended to
1408 be changed after load time by, say, a boot loader. In this case
1409 the right qualification is @code{const} @code{volatile} so that the compiler
1410 must not optimize away known values or insert them
1411 as immediates into operands of instructions.
1414 The following code initializes a variable @code{pfoo}
1415 located in static storage with a 24-bit address:
1417 extern const __memx char foo;
1418 const __memx void *pfoo = &foo;
1422 Such code requires at least binutils 2.23, see
1423 @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1426 On the reduced Tiny devices like ATtiny40, no address spaces are supported.
1427 Data can be put into and read from flash memory by means of
1428 attribute @code{progmem}, see @ref{AVR Variable Attributes}.
1432 @subsection M32C Named Address Spaces
1433 @cindex @code{__far} M32C Named Address Spaces
1435 On the M32C target, with the R8C and M16C CPU variants, variables
1436 qualified with @code{__far} are accessed using 32-bit addresses in
1437 order to access memory beyond the first 64@tie{}Ki bytes. If
1438 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1441 @subsection RL78 Named Address Spaces
1442 @cindex @code{__far} RL78 Named Address Spaces
1444 On the RL78 target, variables qualified with @code{__far} are accessed
1445 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1446 addresses. Non-far variables are assumed to appear in the topmost
1447 64@tie{}KiB of the address space.
1449 @subsection SPU Named Address Spaces
1450 @cindex @code{__ea} SPU Named Address Spaces
1452 On the SPU target variables may be declared as
1453 belonging to another address space by qualifying the type with the
1454 @code{__ea} address space identifier:
1461 The compiler generates special code to access the variable @code{i}.
1462 It may use runtime library
1463 support, or generate special machine instructions to access that address
1466 @subsection x86 Named Address Spaces
1467 @cindex x86 named address spaces
1469 On the x86 target, variables may be declared as being relative
1470 to the @code{%fs} or @code{%gs} segments.
1475 @cindex @code{__seg_fs} x86 named address space
1476 @cindex @code{__seg_gs} x86 named address space
1477 The object is accessed with the respective segment override prefix.
1479 The respective segment base must be set via some method specific to
1480 the operating system. Rather than require an expensive system call
1481 to retrieve the segment base, these address spaces are not considered
1482 to be subspaces of the generic (flat) address space. This means that
1483 explicit casts are required to convert pointers between these address
1484 spaces and the generic address space. In practice the application
1485 should cast to @code{uintptr_t} and apply the segment base offset
1486 that it installed previously.
1488 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1489 defined when these address spaces are supported.
1493 @section Arrays of Length Zero
1494 @cindex arrays of length zero
1495 @cindex zero-length arrays
1496 @cindex length-zero arrays
1497 @cindex flexible array members
1499 Zero-length arrays are allowed in GNU C@. They are very useful as the
1500 last element of a structure that is really a header for a variable-length
1509 struct line *thisline = (struct line *)
1510 malloc (sizeof (struct line) + this_length);
1511 thisline->length = this_length;
1514 In ISO C90, you would have to give @code{contents} a length of 1, which
1515 means either you waste space or complicate the argument to @code{malloc}.
1517 In ISO C99, you would use a @dfn{flexible array member}, which is
1518 slightly different in syntax and semantics:
1522 Flexible array members are written as @code{contents[]} without
1526 Flexible array members have incomplete type, and so the @code{sizeof}
1527 operator may not be applied. As a quirk of the original implementation
1528 of zero-length arrays, @code{sizeof} evaluates to zero.
1531 Flexible array members may only appear as the last member of a
1532 @code{struct} that is otherwise non-empty.
1535 A structure containing a flexible array member, or a union containing
1536 such a structure (possibly recursively), may not be a member of a
1537 structure or an element of an array. (However, these uses are
1538 permitted by GCC as extensions.)
1541 Non-empty initialization of zero-length
1542 arrays is treated like any case where there are more initializer
1543 elements than the array holds, in that a suitable warning about ``excess
1544 elements in array'' is given, and the excess elements (all of them, in
1545 this case) are ignored.
1547 GCC allows static initialization of flexible array members.
1548 This is equivalent to defining a new structure containing the original
1549 structure followed by an array of sufficient size to contain the data.
1550 E.g.@: in the following, @code{f1} is constructed as if it were declared
1556 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1559 struct f1 f1; int data[3];
1560 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1564 The convenience of this extension is that @code{f1} has the desired
1565 type, eliminating the need to consistently refer to @code{f2.f1}.
1567 This has symmetry with normal static arrays, in that an array of
1568 unknown size is also written with @code{[]}.
1570 Of course, this extension only makes sense if the extra data comes at
1571 the end of a top-level object, as otherwise we would be overwriting
1572 data at subsequent offsets. To avoid undue complication and confusion
1573 with initialization of deeply nested arrays, we simply disallow any
1574 non-empty initialization except when the structure is the top-level
1575 object. For example:
1578 struct foo @{ int x; int y[]; @};
1579 struct bar @{ struct foo z; @};
1581 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1582 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1583 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1584 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1587 @node Empty Structures
1588 @section Structures with No Members
1589 @cindex empty structures
1590 @cindex zero-size structures
1592 GCC permits a C structure to have no members:
1599 The structure has size zero. In C++, empty structures are part
1600 of the language. G++ treats empty structures as if they had a single
1601 member of type @code{char}.
1603 @node Variable Length
1604 @section Arrays of Variable Length
1605 @cindex variable-length arrays
1606 @cindex arrays of variable length
1609 Variable-length automatic arrays are allowed in ISO C99, and as an
1610 extension GCC accepts them in C90 mode and in C++. These arrays are
1611 declared like any other automatic arrays, but with a length that is not
1612 a constant expression. The storage is allocated at the point of
1613 declaration and deallocated when the block scope containing the declaration
1619 concat_fopen (char *s1, char *s2, char *mode)
1621 char str[strlen (s1) + strlen (s2) + 1];
1624 return fopen (str, mode);
1628 @cindex scope of a variable length array
1629 @cindex variable-length array scope
1630 @cindex deallocating variable length arrays
1631 Jumping or breaking out of the scope of the array name deallocates the
1632 storage. Jumping into the scope is not allowed; you get an error
1635 @cindex variable-length array in a structure
1636 As an extension, GCC accepts variable-length arrays as a member of
1637 a structure or a union. For example:
1643 struct S @{ int x[n]; @};
1647 @cindex @code{alloca} vs variable-length arrays
1648 You can use the function @code{alloca} to get an effect much like
1649 variable-length arrays. The function @code{alloca} is available in
1650 many other C implementations (but not in all). On the other hand,
1651 variable-length arrays are more elegant.
1653 There are other differences between these two methods. Space allocated
1654 with @code{alloca} exists until the containing @emph{function} returns.
1655 The space for a variable-length array is deallocated as soon as the array
1656 name's scope ends, unless you also use @code{alloca} in this scope.
1658 You can also use variable-length arrays as arguments to functions:
1662 tester (int len, char data[len][len])
1668 The length of an array is computed once when the storage is allocated
1669 and is remembered for the scope of the array in case you access it with
1672 If you want to pass the array first and the length afterward, you can
1673 use a forward declaration in the parameter list---another GNU extension.
1677 tester (int len; char data[len][len], int len)
1683 @cindex parameter forward declaration
1684 The @samp{int len} before the semicolon is a @dfn{parameter forward
1685 declaration}, and it serves the purpose of making the name @code{len}
1686 known when the declaration of @code{data} is parsed.
1688 You can write any number of such parameter forward declarations in the
1689 parameter list. They can be separated by commas or semicolons, but the
1690 last one must end with a semicolon, which is followed by the ``real''
1691 parameter declarations. Each forward declaration must match a ``real''
1692 declaration in parameter name and data type. ISO C99 does not support
1693 parameter forward declarations.
1695 @node Variadic Macros
1696 @section Macros with a Variable Number of Arguments.
1697 @cindex variable number of arguments
1698 @cindex macro with variable arguments
1699 @cindex rest argument (in macro)
1700 @cindex variadic macros
1702 In the ISO C standard of 1999, a macro can be declared to accept a
1703 variable number of arguments much as a function can. The syntax for
1704 defining the macro is similar to that of a function. Here is an
1708 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1712 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1713 such a macro, it represents the zero or more tokens until the closing
1714 parenthesis that ends the invocation, including any commas. This set of
1715 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1716 wherever it appears. See the CPP manual for more information.
1718 GCC has long supported variadic macros, and used a different syntax that
1719 allowed you to give a name to the variable arguments just like any other
1720 argument. Here is an example:
1723 #define debug(format, args...) fprintf (stderr, format, args)
1727 This is in all ways equivalent to the ISO C example above, but arguably
1728 more readable and descriptive.
1730 GNU CPP has two further variadic macro extensions, and permits them to
1731 be used with either of the above forms of macro definition.
1733 In standard C, you are not allowed to leave the variable argument out
1734 entirely; but you are allowed to pass an empty argument. For example,
1735 this invocation is invalid in ISO C, because there is no comma after
1742 GNU CPP permits you to completely omit the variable arguments in this
1743 way. In the above examples, the compiler would complain, though since
1744 the expansion of the macro still has the extra comma after the format
1747 To help solve this problem, CPP behaves specially for variable arguments
1748 used with the token paste operator, @samp{##}. If instead you write
1751 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1755 and if the variable arguments are omitted or empty, the @samp{##}
1756 operator causes the preprocessor to remove the comma before it. If you
1757 do provide some variable arguments in your macro invocation, GNU CPP
1758 does not complain about the paste operation and instead places the
1759 variable arguments after the comma. Just like any other pasted macro
1760 argument, these arguments are not macro expanded.
1762 @node Escaped Newlines
1763 @section Slightly Looser Rules for Escaped Newlines
1764 @cindex escaped newlines
1765 @cindex newlines (escaped)
1767 The preprocessor treatment of escaped newlines is more relaxed
1768 than that specified by the C90 standard, which requires the newline
1769 to immediately follow a backslash.
1770 GCC's implementation allows whitespace in the form
1771 of spaces, horizontal and vertical tabs, and form feeds between the
1772 backslash and the subsequent newline. The preprocessor issues a
1773 warning, but treats it as a valid escaped newline and combines the two
1774 lines to form a single logical line. This works within comments and
1775 tokens, as well as between tokens. Comments are @emph{not} treated as
1776 whitespace for the purposes of this relaxation, since they have not
1777 yet been replaced with spaces.
1780 @section Non-Lvalue Arrays May Have Subscripts
1781 @cindex subscripting
1782 @cindex arrays, non-lvalue
1784 @cindex subscripting and function values
1785 In ISO C99, arrays that are not lvalues still decay to pointers, and
1786 may be subscripted, although they may not be modified or used after
1787 the next sequence point and the unary @samp{&} operator may not be
1788 applied to them. As an extension, GNU C allows such arrays to be
1789 subscripted in C90 mode, though otherwise they do not decay to
1790 pointers outside C99 mode. For example,
1791 this is valid in GNU C though not valid in C90:
1795 struct foo @{int a[4];@};
1801 return f().a[index];
1807 @section Arithmetic on @code{void}- and Function-Pointers
1808 @cindex void pointers, arithmetic
1809 @cindex void, size of pointer to
1810 @cindex function pointers, arithmetic
1811 @cindex function, size of pointer to
1813 In GNU C, addition and subtraction operations are supported on pointers to
1814 @code{void} and on pointers to functions. This is done by treating the
1815 size of a @code{void} or of a function as 1.
1817 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1818 and on function types, and returns 1.
1820 @opindex Wpointer-arith
1821 The option @option{-Wpointer-arith} requests a warning if these extensions
1824 @node Pointers to Arrays
1825 @section Pointers to Arrays with Qualifiers Work as Expected
1826 @cindex pointers to arrays
1827 @cindex const qualifier
1829 In GNU C, pointers to arrays with qualifiers work similar to pointers
1830 to other qualified types. For example, a value of type @code{int (*)[5]}
1831 can be used to initialize a variable of type @code{const int (*)[5]}.
1832 These types are incompatible in ISO C because the @code{const} qualifier
1833 is formally attached to the element type of the array and not the
1838 transpose (int N, int M, double out[M][N], const double in[N][M]);
1842 transpose(3, 2, y, x);
1846 @section Non-Constant Initializers
1847 @cindex initializers, non-constant
1848 @cindex non-constant initializers
1850 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1851 automatic variable are not required to be constant expressions in GNU C@.
1852 Here is an example of an initializer with run-time varying elements:
1855 foo (float f, float g)
1857 float beat_freqs[2] = @{ f-g, f+g @};
1862 @node Compound Literals
1863 @section Compound Literals
1864 @cindex constructor expressions
1865 @cindex initializations in expressions
1866 @cindex structures, constructor expression
1867 @cindex expressions, constructor
1868 @cindex compound literals
1869 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1871 A compound literal looks like a cast of a brace-enclosed aggregate
1872 initializer list. Its value is an object of the type specified in
1873 the cast, containing the elements specified in the initializer.
1874 Unlike the result of a cast, a compound literal is an lvalue. ISO
1875 C99 and later support compound literals. As an extension, GCC
1876 supports compound literals also in C90 mode and in C++, although
1877 as explained below, the C++ semantics are somewhat different.
1879 Usually, the specified type of a compound literal is a structure. Assume
1880 that @code{struct foo} and @code{structure} are declared as shown:
1883 struct foo @{int a; char b[2];@} structure;
1887 Here is an example of constructing a @code{struct foo} with a compound literal:
1890 structure = ((struct foo) @{x + y, 'a', 0@});
1894 This is equivalent to writing the following:
1898 struct foo temp = @{x + y, 'a', 0@};
1903 You can also construct an array, though this is dangerous in C++, as
1904 explained below. If all the elements of the compound literal are
1905 (made up of) simple constant expressions suitable for use in
1906 initializers of objects of static storage duration, then the compound
1907 literal can be coerced to a pointer to its first element and used in
1908 such an initializer, as shown here:
1911 char **foo = (char *[]) @{ "x", "y", "z" @};
1914 Compound literals for scalar types and union types are also allowed. In
1915 the following example the variable @code{i} is initialized to the value
1916 @code{2}, the result of incrementing the unnamed object created by
1917 the compound literal.
1920 int i = ++(int) @{ 1 @};
1923 As a GNU extension, GCC allows initialization of objects with static storage
1924 duration by compound literals (which is not possible in ISO C99 because
1925 the initializer is not a constant).
1926 It is handled as if the object were initialized only with the brace-enclosed
1927 list if the types of the compound literal and the object match.
1928 The elements of the compound literal must be constant.
1929 If the object being initialized has array type of unknown size, the size is
1930 determined by the size of the compound literal.
1933 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1934 static int y[] = (int []) @{1, 2, 3@};
1935 static int z[] = (int [3]) @{1@};
1939 The above lines are equivalent to the following:
1941 static struct foo x = @{1, 'a', 'b'@};
1942 static int y[] = @{1, 2, 3@};
1943 static int z[] = @{1, 0, 0@};
1946 In C, a compound literal designates an unnamed object with static or
1947 automatic storage duration. In C++, a compound literal designates a
1948 temporary object that only lives until the end of its full-expression.
1949 As a result, well-defined C code that takes the address of a subobject
1950 of a compound literal can be undefined in C++, so G++ rejects
1951 the conversion of a temporary array to a pointer. For instance, if
1952 the array compound literal example above appeared inside a function,
1953 any subsequent use of @code{foo} in C++ would have undefined behavior
1954 because the lifetime of the array ends after the declaration of @code{foo}.
1956 As an optimization, G++ sometimes gives array compound literals longer
1957 lifetimes: when the array either appears outside a function or has
1958 a @code{const}-qualified type. If @code{foo} and its initializer had
1959 elements of type @code{char *const} rather than @code{char *}, or if
1960 @code{foo} were a global variable, the array would have static storage
1961 duration. But it is probably safest just to avoid the use of array
1962 compound literals in C++ code.
1964 @node Designated Inits
1965 @section Designated Initializers
1966 @cindex initializers with labeled elements
1967 @cindex labeled elements in initializers
1968 @cindex case labels in initializers
1969 @cindex designated initializers
1971 Standard C90 requires the elements of an initializer to appear in a fixed
1972 order, the same as the order of the elements in the array or structure
1975 In ISO C99 you can give the elements in any order, specifying the array
1976 indices or structure field names they apply to, and GNU C allows this as
1977 an extension in C90 mode as well. This extension is not
1978 implemented in GNU C++.
1980 To specify an array index, write
1981 @samp{[@var{index}] =} before the element value. For example,
1984 int a[6] = @{ [4] = 29, [2] = 15 @};
1991 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1995 The index values must be constant expressions, even if the array being
1996 initialized is automatic.
1998 An alternative syntax for this that has been obsolete since GCC 2.5 but
1999 GCC still accepts is to write @samp{[@var{index}]} before the element
2000 value, with no @samp{=}.
2002 To initialize a range of elements to the same value, write
2003 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2004 extension. For example,
2007 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2011 If the value in it has side-effects, the side-effects happen only once,
2012 not for each initialized field by the range initializer.
2015 Note that the length of the array is the highest value specified
2018 In a structure initializer, specify the name of a field to initialize
2019 with @samp{.@var{fieldname} =} before the element value. For example,
2020 given the following structure,
2023 struct point @{ int x, y; @};
2027 the following initialization
2030 struct point p = @{ .y = yvalue, .x = xvalue @};
2037 struct point p = @{ xvalue, yvalue @};
2040 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2041 @samp{@var{fieldname}:}, as shown here:
2044 struct point p = @{ y: yvalue, x: xvalue @};
2047 Omitted field members are implicitly initialized the same as objects
2048 that have static storage duration.
2051 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2052 @dfn{designator}. You can also use a designator (or the obsolete colon
2053 syntax) when initializing a union, to specify which element of the union
2054 should be used. For example,
2057 union foo @{ int i; double d; @};
2059 union foo f = @{ .d = 4 @};
2063 converts 4 to a @code{double} to store it in the union using
2064 the second element. By contrast, casting 4 to type @code{union foo}
2065 stores it into the union as the integer @code{i}, since it is
2066 an integer. (@xref{Cast to Union}.)
2068 You can combine this technique of naming elements with ordinary C
2069 initialization of successive elements. Each initializer element that
2070 does not have a designator applies to the next consecutive element of the
2071 array or structure. For example,
2074 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2081 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2084 Labeling the elements of an array initializer is especially useful
2085 when the indices are characters or belong to an @code{enum} type.
2090 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2091 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2094 @cindex designator lists
2095 You can also write a series of @samp{.@var{fieldname}} and
2096 @samp{[@var{index}]} designators before an @samp{=} to specify a
2097 nested subobject to initialize; the list is taken relative to the
2098 subobject corresponding to the closest surrounding brace pair. For
2099 example, with the @samp{struct point} declaration above:
2102 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2106 If the same field is initialized multiple times, it has the value from
2107 the last initialization. If any such overridden initialization has
2108 side-effect, it is unspecified whether the side-effect happens or not.
2109 Currently, GCC discards them and issues a warning.
2112 @section Case Ranges
2114 @cindex ranges in case statements
2116 You can specify a range of consecutive values in a single @code{case} label,
2120 case @var{low} ... @var{high}:
2124 This has the same effect as the proper number of individual @code{case}
2125 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2127 This feature is especially useful for ranges of ASCII character codes:
2133 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2134 it may be parsed wrong when you use it with integer values. For example,
2149 @section Cast to a Union Type
2150 @cindex cast to a union
2151 @cindex union, casting to a
2153 A cast to union type looks similar to other casts, except that the type
2154 specified is a union type. You can specify the type either with the
2155 @code{union} keyword or with a @code{typedef} name that refers to
2156 a union. A cast to a union actually creates a compound literal and
2157 yields an lvalue, not an rvalue like true casts do.
2158 (@xref{Compound Literals}.)
2160 The types that may be cast to the union type are those of the members
2161 of the union. Thus, given the following union and variables:
2164 union foo @{ int i; double d; @};
2170 both @code{x} and @code{y} can be cast to type @code{union foo}.
2172 Using the cast as the right-hand side of an assignment to a variable of
2173 union type is equivalent to storing in a member of the union:
2178 u = (union foo) x @equiv{} u.i = x
2179 u = (union foo) y @equiv{} u.d = y
2182 You can also use the union cast as a function argument:
2185 void hack (union foo);
2187 hack ((union foo) x);
2190 @node Mixed Declarations
2191 @section Mixed Declarations and Code
2192 @cindex mixed declarations and code
2193 @cindex declarations, mixed with code
2194 @cindex code, mixed with declarations
2196 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2197 within compound statements. As an extension, GNU C also allows this in
2198 C90 mode. For example, you could do:
2207 Each identifier is visible from where it is declared until the end of
2208 the enclosing block.
2210 @node Function Attributes
2211 @section Declaring Attributes of Functions
2212 @cindex function attributes
2213 @cindex declaring attributes of functions
2214 @cindex @code{volatile} applied to function
2215 @cindex @code{const} applied to function
2217 In GNU C, you can use function attributes to declare certain things
2218 about functions called in your program which help the compiler
2219 optimize calls and check your code more carefully. For example, you
2220 can use attributes to declare that a function never returns
2221 (@code{noreturn}), returns a value depending only on its arguments
2222 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2224 You can also use attributes to control memory placement, code
2225 generation options or call/return conventions within the function
2226 being annotated. Many of these attributes are target-specific. For
2227 example, many targets support attributes for defining interrupt
2228 handler functions, which typically must follow special register usage
2229 and return conventions.
2231 Function attributes are introduced by the @code{__attribute__} keyword
2232 on a declaration, followed by an attribute specification inside double
2233 parentheses. You can specify multiple attributes in a declaration by
2234 separating them by commas within the double parentheses or by
2235 immediately following an attribute declaration with another attribute
2236 declaration. @xref{Attribute Syntax}, for the exact rules on
2237 attribute syntax and placement.
2239 GCC also supports attributes on
2240 variable declarations (@pxref{Variable Attributes}),
2241 labels (@pxref{Label Attributes}),
2242 enumerators (@pxref{Enumerator Attributes}),
2243 and types (@pxref{Type Attributes}).
2245 There is some overlap between the purposes of attributes and pragmas
2246 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2247 found convenient to use @code{__attribute__} to achieve a natural
2248 attachment of attributes to their corresponding declarations, whereas
2249 @code{#pragma} is of use for compatibility with other compilers
2250 or constructs that do not naturally form part of the grammar.
2252 In addition to the attributes documented here,
2253 GCC plugins may provide their own attributes.
2256 * Common Function Attributes::
2257 * AArch64 Function Attributes::
2258 * ARC Function Attributes::
2259 * ARM Function Attributes::
2260 * AVR Function Attributes::
2261 * Blackfin Function Attributes::
2262 * CR16 Function Attributes::
2263 * Epiphany Function Attributes::
2264 * H8/300 Function Attributes::
2265 * IA-64 Function Attributes::
2266 * M32C Function Attributes::
2267 * M32R/D Function Attributes::
2268 * m68k Function Attributes::
2269 * MCORE Function Attributes::
2270 * MeP Function Attributes::
2271 * MicroBlaze Function Attributes::
2272 * Microsoft Windows Function Attributes::
2273 * MIPS Function Attributes::
2274 * MSP430 Function Attributes::
2275 * NDS32 Function Attributes::
2276 * Nios II Function Attributes::
2277 * Nvidia PTX Function Attributes::
2278 * PowerPC Function Attributes::
2279 * RL78 Function Attributes::
2280 * RX Function Attributes::
2281 * S/390 Function Attributes::
2282 * SH Function Attributes::
2283 * SPU Function Attributes::
2284 * Symbian OS Function Attributes::
2285 * V850 Function Attributes::
2286 * Visium Function Attributes::
2287 * x86 Function Attributes::
2288 * Xstormy16 Function Attributes::
2291 @node Common Function Attributes
2292 @subsection Common Function Attributes
2294 The following attributes are supported on most targets.
2297 @c Keep this table alphabetized by attribute name. Treat _ as space.
2299 @item alias ("@var{target}")
2300 @cindex @code{alias} function attribute
2301 The @code{alias} attribute causes the declaration to be emitted as an
2302 alias for another symbol, which must be specified. For instance,
2305 void __f () @{ /* @r{Do something.} */; @}
2306 void f () __attribute__ ((weak, alias ("__f")));
2310 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2311 mangled name for the target must be used. It is an error if @samp{__f}
2312 is not defined in the same translation unit.
2314 This attribute requires assembler and object file support,
2315 and may not be available on all targets.
2317 @item aligned (@var{alignment})
2318 @cindex @code{aligned} function attribute
2319 This attribute specifies a minimum alignment for the function,
2322 You cannot use this attribute to decrease the alignment of a function,
2323 only to increase it. However, when you explicitly specify a function
2324 alignment this overrides the effect of the
2325 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2328 Note that the effectiveness of @code{aligned} attributes may be
2329 limited by inherent limitations in your linker. On many systems, the
2330 linker is only able to arrange for functions to be aligned up to a
2331 certain maximum alignment. (For some linkers, the maximum supported
2332 alignment may be very very small.) See your linker documentation for
2333 further information.
2335 The @code{aligned} attribute can also be used for variables and fields
2336 (@pxref{Variable Attributes}.)
2339 @cindex @code{alloc_align} function attribute
2340 The @code{alloc_align} attribute is used to tell the compiler that the
2341 function return value points to memory, where the returned pointer minimum
2342 alignment is given by one of the functions parameters. GCC uses this
2343 information to improve pointer alignment analysis.
2345 The function parameter denoting the allocated alignment is specified by
2346 one integer argument, whose number is the argument of the attribute.
2347 Argument numbering starts at one.
2352 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2356 declares that @code{my_memalign} returns memory with minimum alignment
2357 given by parameter 1.
2360 @cindex @code{alloc_size} function attribute
2361 The @code{alloc_size} attribute is used to tell the compiler that the
2362 function return value points to memory, where the size is given by
2363 one or two of the functions parameters. GCC uses this
2364 information to improve the correctness of @code{__builtin_object_size}.
2366 The function parameter(s) denoting the allocated size are specified by
2367 one or two integer arguments supplied to the attribute. The allocated size
2368 is either the value of the single function argument specified or the product
2369 of the two function arguments specified. Argument numbering starts at
2375 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2376 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2380 declares that @code{my_calloc} returns memory of the size given by
2381 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2382 of the size given by parameter 2.
2385 @cindex @code{always_inline} function attribute
2386 Generally, functions are not inlined unless optimization is specified.
2387 For functions declared inline, this attribute inlines the function
2388 independent of any restrictions that otherwise apply to inlining.
2389 Failure to inline such a function is diagnosed as an error.
2390 Note that if such a function is called indirectly the compiler may
2391 or may not inline it depending on optimization level and a failure
2392 to inline an indirect call may or may not be diagnosed.
2395 @cindex @code{artificial} function attribute
2396 This attribute is useful for small inline wrappers that if possible
2397 should appear during debugging as a unit. Depending on the debug
2398 info format it either means marking the function as artificial
2399 or using the caller location for all instructions within the inlined
2402 @item assume_aligned
2403 @cindex @code{assume_aligned} function attribute
2404 The @code{assume_aligned} attribute is used to tell the compiler that the
2405 function return value points to memory, where the returned pointer minimum
2406 alignment is given by the first argument.
2407 If the attribute has two arguments, the second argument is misalignment offset.
2412 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2413 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2417 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2418 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2421 @item bnd_instrument
2422 @cindex @code{bnd_instrument} function attribute
2423 The @code{bnd_instrument} attribute on functions is used to inform the
2424 compiler that the function should be instrumented when compiled
2425 with the @option{-fchkp-instrument-marked-only} option.
2428 @cindex @code{bnd_legacy} function attribute
2429 @cindex Pointer Bounds Checker attributes
2430 The @code{bnd_legacy} attribute on functions is used to inform the
2431 compiler that the function should not be instrumented when compiled
2432 with the @option{-fcheck-pointer-bounds} option.
2435 @cindex @code{cold} function attribute
2436 The @code{cold} attribute on functions is used to inform the compiler that
2437 the function is unlikely to be executed. The function is optimized for
2438 size rather than speed and on many targets it is placed into a special
2439 subsection of the text section so all cold functions appear close together,
2440 improving code locality of non-cold parts of program. The paths leading
2441 to calls of cold functions within code are marked as unlikely by the branch
2442 prediction mechanism. It is thus useful to mark functions used to handle
2443 unlikely conditions, such as @code{perror}, as cold to improve optimization
2444 of hot functions that do call marked functions in rare occasions.
2446 When profile feedback is available, via @option{-fprofile-use}, cold functions
2447 are automatically detected and this attribute is ignored.
2450 @cindex @code{const} function attribute
2451 @cindex functions that have no side effects
2452 Many functions do not examine any values except their arguments, and
2453 have no effects except the return value. Basically this is just slightly
2454 more strict class than the @code{pure} attribute below, since function is not
2455 allowed to read global memory.
2457 @cindex pointer arguments
2458 Note that a function that has pointer arguments and examines the data
2459 pointed to must @emph{not} be declared @code{const}. Likewise, a
2460 function that calls a non-@code{const} function usually must not be
2461 @code{const}. It does not make sense for a @code{const} function to
2466 @itemx constructor (@var{priority})
2467 @itemx destructor (@var{priority})
2468 @cindex @code{constructor} function attribute
2469 @cindex @code{destructor} function attribute
2470 The @code{constructor} attribute causes the function to be called
2471 automatically before execution enters @code{main ()}. Similarly, the
2472 @code{destructor} attribute causes the function to be called
2473 automatically after @code{main ()} completes or @code{exit ()} is
2474 called. Functions with these attributes are useful for
2475 initializing data that is used implicitly during the execution of
2478 You may provide an optional integer priority to control the order in
2479 which constructor and destructor functions are run. A constructor
2480 with a smaller priority number runs before a constructor with a larger
2481 priority number; the opposite relationship holds for destructors. So,
2482 if you have a constructor that allocates a resource and a destructor
2483 that deallocates the same resource, both functions typically have the
2484 same priority. The priorities for constructor and destructor
2485 functions are the same as those specified for namespace-scope C++
2486 objects (@pxref{C++ Attributes}).
2488 These attributes are not currently implemented for Objective-C@.
2491 @itemx deprecated (@var{msg})
2492 @cindex @code{deprecated} function attribute
2493 The @code{deprecated} attribute results in a warning if the function
2494 is used anywhere in the source file. This is useful when identifying
2495 functions that are expected to be removed in a future version of a
2496 program. The warning also includes the location of the declaration
2497 of the deprecated function, to enable users to easily find further
2498 information about why the function is deprecated, or what they should
2499 do instead. Note that the warnings only occurs for uses:
2502 int old_fn () __attribute__ ((deprecated));
2504 int (*fn_ptr)() = old_fn;
2508 results in a warning on line 3 but not line 2. The optional @var{msg}
2509 argument, which must be a string, is printed in the warning if
2512 The @code{deprecated} attribute can also be used for variables and
2513 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2515 @item error ("@var{message}")
2516 @itemx warning ("@var{message}")
2517 @cindex @code{error} function attribute
2518 @cindex @code{warning} function attribute
2519 If the @code{error} or @code{warning} attribute
2520 is used on a function declaration and a call to such a function
2521 is not eliminated through dead code elimination or other optimizations,
2522 an error or warning (respectively) that includes @var{message} is diagnosed.
2524 for compile-time checking, especially together with @code{__builtin_constant_p}
2525 and inline functions where checking the inline function arguments is not
2526 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2528 While it is possible to leave the function undefined and thus invoke
2529 a link failure (to define the function with
2530 a message in @code{.gnu.warning*} section),
2531 when using these attributes the problem is diagnosed
2532 earlier and with exact location of the call even in presence of inline
2533 functions or when not emitting debugging information.
2535 @item externally_visible
2536 @cindex @code{externally_visible} function attribute
2537 This attribute, attached to a global variable or function, nullifies
2538 the effect of the @option{-fwhole-program} command-line option, so the
2539 object remains visible outside the current compilation unit.
2541 If @option{-fwhole-program} is used together with @option{-flto} and
2542 @command{gold} is used as the linker plugin,
2543 @code{externally_visible} attributes are automatically added to functions
2544 (not variable yet due to a current @command{gold} issue)
2545 that are accessed outside of LTO objects according to resolution file
2546 produced by @command{gold}.
2547 For other linkers that cannot generate resolution file,
2548 explicit @code{externally_visible} attributes are still necessary.
2551 @cindex @code{flatten} function attribute
2552 Generally, inlining into a function is limited. For a function marked with
2553 this attribute, every call inside this function is inlined, if possible.
2554 Whether the function itself is considered for inlining depends on its size and
2555 the current inlining parameters.
2557 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2558 @cindex @code{format} function attribute
2559 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2561 The @code{format} attribute specifies that a function takes @code{printf},
2562 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2563 should be type-checked against a format string. For example, the
2568 my_printf (void *my_object, const char *my_format, ...)
2569 __attribute__ ((format (printf, 2, 3)));
2573 causes the compiler to check the arguments in calls to @code{my_printf}
2574 for consistency with the @code{printf} style format string argument
2577 The parameter @var{archetype} determines how the format string is
2578 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2579 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2580 @code{strfmon}. (You can also use @code{__printf__},
2581 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2582 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2583 @code{ms_strftime} are also present.
2584 @var{archetype} values such as @code{printf} refer to the formats accepted
2585 by the system's C runtime library,
2586 while values prefixed with @samp{gnu_} always refer
2587 to the formats accepted by the GNU C Library. On Microsoft Windows
2588 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2589 @file{msvcrt.dll} library.
2590 The parameter @var{string-index}
2591 specifies which argument is the format string argument (starting
2592 from 1), while @var{first-to-check} is the number of the first
2593 argument to check against the format string. For functions
2594 where the arguments are not available to be checked (such as
2595 @code{vprintf}), specify the third parameter as zero. In this case the
2596 compiler only checks the format string for consistency. For
2597 @code{strftime} formats, the third parameter is required to be zero.
2598 Since non-static C++ methods have an implicit @code{this} argument, the
2599 arguments of such methods should be counted from two, not one, when
2600 giving values for @var{string-index} and @var{first-to-check}.
2602 In the example above, the format string (@code{my_format}) is the second
2603 argument of the function @code{my_print}, and the arguments to check
2604 start with the third argument, so the correct parameters for the format
2605 attribute are 2 and 3.
2607 @opindex ffreestanding
2608 @opindex fno-builtin
2609 The @code{format} attribute allows you to identify your own functions
2610 that take format strings as arguments, so that GCC can check the
2611 calls to these functions for errors. The compiler always (unless
2612 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2613 for the standard library functions @code{printf}, @code{fprintf},
2614 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2615 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2616 warnings are requested (using @option{-Wformat}), so there is no need to
2617 modify the header file @file{stdio.h}. In C99 mode, the functions
2618 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2619 @code{vsscanf} are also checked. Except in strictly conforming C
2620 standard modes, the X/Open function @code{strfmon} is also checked as
2621 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2622 @xref{C Dialect Options,,Options Controlling C Dialect}.
2624 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2625 recognized in the same context. Declarations including these format attributes
2626 are parsed for correct syntax, however the result of checking of such format
2627 strings is not yet defined, and is not carried out by this version of the
2630 The target may also provide additional types of format checks.
2631 @xref{Target Format Checks,,Format Checks Specific to Particular
2634 @item format_arg (@var{string-index})
2635 @cindex @code{format_arg} function attribute
2636 @opindex Wformat-nonliteral
2637 The @code{format_arg} attribute specifies that a function takes a format
2638 string for a @code{printf}, @code{scanf}, @code{strftime} or
2639 @code{strfmon} style function and modifies it (for example, to translate
2640 it into another language), so the result can be passed to a
2641 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2642 function (with the remaining arguments to the format function the same
2643 as they would have been for the unmodified string). For example, the
2648 my_dgettext (char *my_domain, const char *my_format)
2649 __attribute__ ((format_arg (2)));
2653 causes the compiler to check the arguments in calls to a @code{printf},
2654 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2655 format string argument is a call to the @code{my_dgettext} function, for
2656 consistency with the format string argument @code{my_format}. If the
2657 @code{format_arg} attribute had not been specified, all the compiler
2658 could tell in such calls to format functions would be that the format
2659 string argument is not constant; this would generate a warning when
2660 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2661 without the attribute.
2663 The parameter @var{string-index} specifies which argument is the format
2664 string argument (starting from one). Since non-static C++ methods have
2665 an implicit @code{this} argument, the arguments of such methods should
2666 be counted from two.
2668 The @code{format_arg} attribute allows you to identify your own
2669 functions that modify format strings, so that GCC can check the
2670 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2671 type function whose operands are a call to one of your own function.
2672 The compiler always treats @code{gettext}, @code{dgettext}, and
2673 @code{dcgettext} in this manner except when strict ISO C support is
2674 requested by @option{-ansi} or an appropriate @option{-std} option, or
2675 @option{-ffreestanding} or @option{-fno-builtin}
2676 is used. @xref{C Dialect Options,,Options
2677 Controlling C Dialect}.
2679 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2680 @code{NSString} reference for compatibility with the @code{format} attribute
2683 The target may also allow additional types in @code{format-arg} attributes.
2684 @xref{Target Format Checks,,Format Checks Specific to Particular
2688 @cindex @code{gnu_inline} function attribute
2689 This attribute should be used with a function that is also declared
2690 with the @code{inline} keyword. It directs GCC to treat the function
2691 as if it were defined in gnu90 mode even when compiling in C99 or
2694 If the function is declared @code{extern}, then this definition of the
2695 function is used only for inlining. In no case is the function
2696 compiled as a standalone function, not even if you take its address
2697 explicitly. Such an address becomes an external reference, as if you
2698 had only declared the function, and had not defined it. This has
2699 almost the effect of a macro. The way to use this is to put a
2700 function definition in a header file with this attribute, and put
2701 another copy of the function, without @code{extern}, in a library
2702 file. The definition in the header file causes most calls to the
2703 function to be inlined. If any uses of the function remain, they
2704 refer to the single copy in the library. Note that the two
2705 definitions of the functions need not be precisely the same, although
2706 if they do not have the same effect your program may behave oddly.
2708 In C, if the function is neither @code{extern} nor @code{static}, then
2709 the function is compiled as a standalone function, as well as being
2710 inlined where possible.
2712 This is how GCC traditionally handled functions declared
2713 @code{inline}. Since ISO C99 specifies a different semantics for
2714 @code{inline}, this function attribute is provided as a transition
2715 measure and as a useful feature in its own right. This attribute is
2716 available in GCC 4.1.3 and later. It is available if either of the
2717 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2718 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2719 Function is As Fast As a Macro}.
2721 In C++, this attribute does not depend on @code{extern} in any way,
2722 but it still requires the @code{inline} keyword to enable its special
2726 @cindex @code{hot} function attribute
2727 The @code{hot} attribute on a function is used to inform the compiler that
2728 the function is a hot spot of the compiled program. The function is
2729 optimized more aggressively and on many targets it is placed into a special
2730 subsection of the text section so all hot functions appear close together,
2733 When profile feedback is available, via @option{-fprofile-use}, hot functions
2734 are automatically detected and this attribute is ignored.
2736 @item ifunc ("@var{resolver}")
2737 @cindex @code{ifunc} function attribute
2738 @cindex indirect functions
2739 @cindex functions that are dynamically resolved
2740 The @code{ifunc} attribute is used to mark a function as an indirect
2741 function using the STT_GNU_IFUNC symbol type extension to the ELF
2742 standard. This allows the resolution of the symbol value to be
2743 determined dynamically at load time, and an optimized version of the
2744 routine can be selected for the particular processor or other system
2745 characteristics determined then. To use this attribute, first define
2746 the implementation functions available, and a resolver function that
2747 returns a pointer to the selected implementation function. The
2748 implementation functions' declarations must match the API of the
2749 function being implemented, the resolver's declaration is be a
2750 function returning pointer to void function returning void:
2753 void *my_memcpy (void *dst, const void *src, size_t len)
2758 static void (*resolve_memcpy (void)) (void)
2760 return my_memcpy; // we'll just always select this routine
2765 The exported header file declaring the function the user calls would
2769 extern void *memcpy (void *, const void *, size_t);
2773 allowing the user to call this as a regular function, unaware of the
2774 implementation. Finally, the indirect function needs to be defined in
2775 the same translation unit as the resolver function:
2778 void *memcpy (void *, const void *, size_t)
2779 __attribute__ ((ifunc ("resolve_memcpy")));
2782 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2783 and GNU C Library version 2.11.1 are required to use this feature.
2786 @itemx interrupt_handler
2787 Many GCC back ends support attributes to indicate that a function is
2788 an interrupt handler, which tells the compiler to generate function
2789 entry and exit sequences that differ from those from regular
2790 functions. The exact syntax and behavior are target-specific;
2791 refer to the following subsections for details.
2794 @cindex @code{leaf} function attribute
2795 Calls to external functions with this attribute must return to the
2796 current compilation unit only by return or by exception handling. In
2797 particular, a leaf function is not allowed to invoke callback functions
2798 passed to it from the current compilation unit, directly call functions
2799 exported by the unit, or @code{longjmp} into the unit. Leaf functions
2800 might still call functions from other compilation units and thus they
2801 are not necessarily leaf in the sense that they contain no function
2804 The attribute is intended for library functions to improve dataflow
2805 analysis. The compiler takes the hint that any data not escaping the
2806 current compilation unit cannot be used or modified by the leaf
2807 function. For example, the @code{sin} function is a leaf function, but
2808 @code{qsort} is not.
2810 Note that leaf functions might indirectly run a signal handler defined
2811 in the current compilation unit that uses static variables. Similarly,
2812 when lazy symbol resolution is in effect, leaf functions might invoke
2813 indirect functions whose resolver function or implementation function is
2814 defined in the current compilation unit and uses static variables. There
2815 is no standard-compliant way to write such a signal handler, resolver
2816 function, or implementation function, and the best that you can do is to
2817 remove the @code{leaf} attribute or mark all such static variables
2818 @code{volatile}. Lastly, for ELF-based systems that support symbol
2819 interposition, care should be taken that functions defined in the
2820 current compilation unit do not unexpectedly interpose other symbols
2821 based on the defined standards mode and defined feature test macros;
2822 otherwise an inadvertent callback would be added.
2824 The attribute has no effect on functions defined within the current
2825 compilation unit. This is to allow easy merging of multiple compilation
2826 units into one, for example, by using the link-time optimization. For
2827 this reason the attribute is not allowed on types to annotate indirect
2831 @cindex @code{malloc} function attribute
2832 @cindex functions that behave like malloc
2833 This tells the compiler that a function is @code{malloc}-like, i.e.,
2834 that the pointer @var{P} returned by the function cannot alias any
2835 other pointer valid when the function returns, and moreover no
2836 pointers to valid objects occur in any storage addressed by @var{P}.
2838 Using this attribute can improve optimization. Functions like
2839 @code{malloc} and @code{calloc} have this property because they return
2840 a pointer to uninitialized or zeroed-out storage. However, functions
2841 like @code{realloc} do not have this property, as they can return a
2842 pointer to storage containing pointers.
2845 @cindex @code{no_icf} function attribute
2846 This function attribute prevents a functions from being merged with another
2847 semantically equivalent function.
2849 @item no_instrument_function
2850 @cindex @code{no_instrument_function} function attribute
2851 @opindex finstrument-functions
2852 If @option{-finstrument-functions} is given, profiling function calls are
2853 generated at entry and exit of most user-compiled functions.
2854 Functions with this attribute are not so instrumented.
2856 @item no_profile_instrument_function
2857 @cindex @code{no_profile_instrument_function} function attribute
2858 The @code{no_profile_instrument_function} attribute on functions is used
2859 to inform the compiler that it should not process any profile feedback based
2860 optimization code instrumentation.
2863 @cindex @code{no_reorder} function attribute
2864 Do not reorder functions or variables marked @code{no_reorder}
2865 against each other or top level assembler statements the executable.
2866 The actual order in the program will depend on the linker command
2867 line. Static variables marked like this are also not removed.
2868 This has a similar effect
2869 as the @option{-fno-toplevel-reorder} option, but only applies to the
2872 @item no_sanitize_address
2873 @itemx no_address_safety_analysis
2874 @cindex @code{no_sanitize_address} function attribute
2875 The @code{no_sanitize_address} attribute on functions is used
2876 to inform the compiler that it should not instrument memory accesses
2877 in the function when compiling with the @option{-fsanitize=address} option.
2878 The @code{no_address_safety_analysis} is a deprecated alias of the
2879 @code{no_sanitize_address} attribute, new code should use
2880 @code{no_sanitize_address}.
2882 @item no_sanitize_thread
2883 @cindex @code{no_sanitize_thread} function attribute
2884 The @code{no_sanitize_thread} attribute on functions is used
2885 to inform the compiler that it should not instrument memory accesses
2886 in the function when compiling with the @option{-fsanitize=thread} option.
2888 @item no_sanitize_undefined
2889 @cindex @code{no_sanitize_undefined} function attribute
2890 The @code{no_sanitize_undefined} attribute on functions is used
2891 to inform the compiler that it should not check for undefined behavior
2892 in the function when compiling with the @option{-fsanitize=undefined} option.
2894 @item no_split_stack
2895 @cindex @code{no_split_stack} function attribute
2896 @opindex fsplit-stack
2897 If @option{-fsplit-stack} is given, functions have a small
2898 prologue which decides whether to split the stack. Functions with the
2899 @code{no_split_stack} attribute do not have that prologue, and thus
2900 may run with only a small amount of stack space available.
2902 @item no_stack_limit
2903 @cindex @code{no_stack_limit} function attribute
2904 This attribute locally overrides the @option{-fstack-limit-register}
2905 and @option{-fstack-limit-symbol} command-line options; it has the effect
2906 of disabling stack limit checking in the function it applies to.
2909 @cindex @code{noclone} function attribute
2910 This function attribute prevents a function from being considered for
2911 cloning---a mechanism that produces specialized copies of functions
2912 and which is (currently) performed by interprocedural constant
2916 @cindex @code{noinline} function attribute
2917 This function attribute prevents a function from being considered for
2919 @c Don't enumerate the optimizations by name here; we try to be
2920 @c future-compatible with this mechanism.
2921 If the function does not have side-effects, there are optimizations
2922 other than inlining that cause function calls to be optimized away,
2923 although the function call is live. To keep such calls from being
2930 (@pxref{Extended Asm}) in the called function, to serve as a special
2933 @item nonnull (@var{arg-index}, @dots{})
2934 @cindex @code{nonnull} function attribute
2935 @cindex functions with non-null pointer arguments
2936 The @code{nonnull} attribute specifies that some function parameters should
2937 be non-null pointers. For instance, the declaration:
2941 my_memcpy (void *dest, const void *src, size_t len)
2942 __attribute__((nonnull (1, 2)));
2946 causes the compiler to check that, in calls to @code{my_memcpy},
2947 arguments @var{dest} and @var{src} are non-null. If the compiler
2948 determines that a null pointer is passed in an argument slot marked
2949 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2950 is issued. The compiler may also choose to make optimizations based
2951 on the knowledge that certain function arguments will never be null.
2953 If no argument index list is given to the @code{nonnull} attribute,
2954 all pointer arguments are marked as non-null. To illustrate, the
2955 following declaration is equivalent to the previous example:
2959 my_memcpy (void *dest, const void *src, size_t len)
2960 __attribute__((nonnull));
2964 @cindex @code{noplt} function attribute
2965 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
2966 Calls to functions marked with this attribute in position-independent code
2971 /* Externally defined function foo. */
2972 int foo () __attribute__ ((noplt));
2975 main (/* @r{@dots{}} */)
2984 The @code{noplt} attribute on function @code{foo}
2985 tells the compiler to assume that
2986 the function @code{foo} is externally defined and that the call to
2987 @code{foo} must avoid the PLT
2988 in position-independent code.
2990 In position-dependent code, a few targets also convert calls to
2991 functions that are marked to not use the PLT to use the GOT instead.
2994 @cindex @code{noreturn} function attribute
2995 @cindex functions that never return
2996 A few standard library functions, such as @code{abort} and @code{exit},
2997 cannot return. GCC knows this automatically. Some programs define
2998 their own functions that never return. You can declare them
2999 @code{noreturn} to tell the compiler this fact. For example,
3003 void fatal () __attribute__ ((noreturn));
3006 fatal (/* @r{@dots{}} */)
3008 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3014 The @code{noreturn} keyword tells the compiler to assume that
3015 @code{fatal} cannot return. It can then optimize without regard to what
3016 would happen if @code{fatal} ever did return. This makes slightly
3017 better code. More importantly, it helps avoid spurious warnings of
3018 uninitialized variables.
3020 The @code{noreturn} keyword does not affect the exceptional path when that
3021 applies: a @code{noreturn}-marked function may still return to the caller
3022 by throwing an exception or calling @code{longjmp}.
3024 Do not assume that registers saved by the calling function are
3025 restored before calling the @code{noreturn} function.
3027 It does not make sense for a @code{noreturn} function to have a return
3028 type other than @code{void}.
3031 @cindex @code{nothrow} function attribute
3032 The @code{nothrow} attribute is used to inform the compiler that a
3033 function cannot throw an exception. For example, most functions in
3034 the standard C library can be guaranteed not to throw an exception
3035 with the notable exceptions of @code{qsort} and @code{bsearch} that
3036 take function pointer arguments.
3039 @cindex @code{optimize} function attribute
3040 The @code{optimize} attribute is used to specify that a function is to
3041 be compiled with different optimization options than specified on the
3042 command line. Arguments can either be numbers or strings. Numbers
3043 are assumed to be an optimization level. Strings that begin with
3044 @code{O} are assumed to be an optimization option, while other options
3045 are assumed to be used with a @code{-f} prefix. You can also use the
3046 @samp{#pragma GCC optimize} pragma to set the optimization options
3047 that affect more than one function.
3048 @xref{Function Specific Option Pragmas}, for details about the
3049 @samp{#pragma GCC optimize} pragma.
3051 This attribute should be used for debugging purposes only. It is not
3052 suitable in production code.
3055 @cindex @code{pure} function attribute
3056 @cindex functions that have no side effects
3057 Many functions have no effects except the return value and their
3058 return value depends only on the parameters and/or global variables.
3059 Such a function can be subject
3060 to common subexpression elimination and loop optimization just as an
3061 arithmetic operator would be. These functions should be declared
3062 with the attribute @code{pure}. For example,
3065 int square (int) __attribute__ ((pure));
3069 says that the hypothetical function @code{square} is safe to call
3070 fewer times than the program says.
3072 Some common examples of pure functions are @code{strlen} or @code{memcmp}.
3073 Interesting non-pure functions are functions with infinite loops or those
3074 depending on volatile memory or other system resource, that may change between
3075 two consecutive calls (such as @code{feof} in a multithreading environment).
3077 @item returns_nonnull
3078 @cindex @code{returns_nonnull} function attribute
3079 The @code{returns_nonnull} attribute specifies that the function
3080 return value should be a non-null pointer. For instance, the declaration:
3084 mymalloc (size_t len) __attribute__((returns_nonnull));
3088 lets the compiler optimize callers based on the knowledge
3089 that the return value will never be null.
3092 @cindex @code{returns_twice} function attribute
3093 @cindex functions that return more than once
3094 The @code{returns_twice} attribute tells the compiler that a function may
3095 return more than one time. The compiler ensures that all registers
3096 are dead before calling such a function and emits a warning about
3097 the variables that may be clobbered after the second return from the
3098 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3099 The @code{longjmp}-like counterpart of such function, if any, might need
3100 to be marked with the @code{noreturn} attribute.
3102 @item section ("@var{section-name}")
3103 @cindex @code{section} function attribute
3104 @cindex functions in arbitrary sections
3105 Normally, the compiler places the code it generates in the @code{text} section.
3106 Sometimes, however, you need additional sections, or you need certain
3107 particular functions to appear in special sections. The @code{section}
3108 attribute specifies that a function lives in a particular section.
3109 For example, the declaration:
3112 extern void foobar (void) __attribute__ ((section ("bar")));
3116 puts the function @code{foobar} in the @code{bar} section.
3118 Some file formats do not support arbitrary sections so the @code{section}
3119 attribute is not available on all platforms.
3120 If you need to map the entire contents of a module to a particular
3121 section, consider using the facilities of the linker instead.
3124 @cindex @code{sentinel} function attribute
3125 This function attribute ensures that a parameter in a function call is
3126 an explicit @code{NULL}. The attribute is only valid on variadic
3127 functions. By default, the sentinel is located at position zero, the
3128 last parameter of the function call. If an optional integer position
3129 argument P is supplied to the attribute, the sentinel must be located at
3130 position P counting backwards from the end of the argument list.
3133 __attribute__ ((sentinel))
3135 __attribute__ ((sentinel(0)))
3138 The attribute is automatically set with a position of 0 for the built-in
3139 functions @code{execl} and @code{execlp}. The built-in function
3140 @code{execle} has the attribute set with a position of 1.
3142 A valid @code{NULL} in this context is defined as zero with any pointer
3143 type. If your system defines the @code{NULL} macro with an integer type
3144 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3145 with a copy that redefines NULL appropriately.
3147 The warnings for missing or incorrect sentinels are enabled with
3151 @itemx simd("@var{mask}")
3152 @cindex @code{simd} function attribute
3153 This attribute enables creation of one or more function versions that
3154 can process multiple arguments using SIMD instructions from a
3155 single invocation. Specifying this attribute allows compiler to
3156 assume that such versions are available at link time (provided
3157 in the same or another translation unit). Generated versions are
3158 target-dependent and described in the corresponding Vector ABI document. For
3159 x86_64 target this document can be found
3160 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3162 The optional argument @var{mask} may have the value
3163 @code{notinbranch} or @code{inbranch},
3164 and instructs the compiler to generate non-masked or masked
3165 clones correspondingly. By default, all clones are generated.
3167 The attribute should not be used together with Cilk Plus @code{vector}
3168 attribute on the same function.
3170 If the attribute is specified and @code{#pragma omp declare simd} is
3171 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3172 switch is specified, then the attribute is ignored.
3175 @cindex @code{stack_protect} function attribute
3176 This attribute adds stack protection code to the function if
3177 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3178 or @option{-fstack-protector-explicit} are set.
3180 @item target (@var{options})
3181 @cindex @code{target} function attribute
3182 Multiple target back ends implement the @code{target} attribute
3183 to specify that a function is to
3184 be compiled with different target options than specified on the
3185 command line. This can be used for instance to have functions
3186 compiled with a different ISA (instruction set architecture) than the
3187 default. You can also use the @samp{#pragma GCC target} pragma to set
3188 more than one function to be compiled with specific target options.
3189 @xref{Function Specific Option Pragmas}, for details about the
3190 @samp{#pragma GCC target} pragma.
3192 For instance, on an x86, you could declare one function with the
3193 @code{target("sse4.1,arch=core2")} attribute and another with
3194 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3195 compiling the first function with @option{-msse4.1} and
3196 @option{-march=core2} options, and the second function with
3197 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3198 to make sure that a function is only invoked on a machine that
3199 supports the particular ISA it is compiled for (for example by using
3200 @code{cpuid} on x86 to determine what feature bits and architecture
3204 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3205 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3208 You can either use multiple
3209 strings separated by commas to specify multiple options,
3210 or separate the options with a comma (@samp{,}) within a single string.
3212 The options supported are specific to each target; refer to @ref{x86
3213 Function Attributes}, @ref{PowerPC Function Attributes},
3214 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3217 @item target_clones (@var{options})
3218 @cindex @code{target_clones} function attribute
3219 The @code{target_clones} attribute is used to specify that a function
3220 be cloned into multiple versions compiled with different target options
3221 than specified on the command line. The supported options and restrictions
3222 are the same as for @code{target} attribute.
3224 For instance, on an x86, you could compile a function with
3225 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3226 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3227 It also creates a resolver function (see the @code{ifunc} attribute
3228 above) that dynamically selects a clone suitable for current architecture.
3231 @cindex @code{unused} function attribute
3232 This attribute, attached to a function, means that the function is meant
3233 to be possibly unused. GCC does not produce a warning for this
3237 @cindex @code{used} function attribute
3238 This attribute, attached to a function, means that code must be emitted
3239 for the function even if it appears that the function is not referenced.
3240 This is useful, for example, when the function is referenced only in
3243 When applied to a member function of a C++ class template, the
3244 attribute also means that the function is instantiated if the
3245 class itself is instantiated.
3247 @item visibility ("@var{visibility_type}")
3248 @cindex @code{visibility} function attribute
3249 This attribute affects the linkage of the declaration to which it is attached.
3250 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3251 (@pxref{Common Type Attributes}) as well as functions.
3253 There are four supported @var{visibility_type} values: default,
3254 hidden, protected or internal visibility.
3257 void __attribute__ ((visibility ("protected")))
3258 f () @{ /* @r{Do something.} */; @}
3259 int i __attribute__ ((visibility ("hidden")));
3262 The possible values of @var{visibility_type} correspond to the
3263 visibility settings in the ELF gABI.
3266 @c keep this list of visibilities in alphabetical order.
3269 Default visibility is the normal case for the object file format.
3270 This value is available for the visibility attribute to override other
3271 options that may change the assumed visibility of entities.
3273 On ELF, default visibility means that the declaration is visible to other
3274 modules and, in shared libraries, means that the declared entity may be
3277 On Darwin, default visibility means that the declaration is visible to
3280 Default visibility corresponds to ``external linkage'' in the language.
3283 Hidden visibility indicates that the entity declared has a new
3284 form of linkage, which we call ``hidden linkage''. Two
3285 declarations of an object with hidden linkage refer to the same object
3286 if they are in the same shared object.
3289 Internal visibility is like hidden visibility, but with additional
3290 processor specific semantics. Unless otherwise specified by the
3291 psABI, GCC defines internal visibility to mean that a function is
3292 @emph{never} called from another module. Compare this with hidden
3293 functions which, while they cannot be referenced directly by other
3294 modules, can be referenced indirectly via function pointers. By
3295 indicating that a function cannot be called from outside the module,
3296 GCC may for instance omit the load of a PIC register since it is known
3297 that the calling function loaded the correct value.
3300 Protected visibility is like default visibility except that it
3301 indicates that references within the defining module bind to the
3302 definition in that module. That is, the declared entity cannot be
3303 overridden by another module.
3307 All visibilities are supported on many, but not all, ELF targets
3308 (supported when the assembler supports the @samp{.visibility}
3309 pseudo-op). Default visibility is supported everywhere. Hidden
3310 visibility is supported on Darwin targets.
3312 The visibility attribute should be applied only to declarations that
3313 would otherwise have external linkage. The attribute should be applied
3314 consistently, so that the same entity should not be declared with
3315 different settings of the attribute.
3317 In C++, the visibility attribute applies to types as well as functions
3318 and objects, because in C++ types have linkage. A class must not have
3319 greater visibility than its non-static data member types and bases,
3320 and class members default to the visibility of their class. Also, a
3321 declaration without explicit visibility is limited to the visibility
3324 In C++, you can mark member functions and static member variables of a
3325 class with the visibility attribute. This is useful if you know a
3326 particular method or static member variable should only be used from
3327 one shared object; then you can mark it hidden while the rest of the
3328 class has default visibility. Care must be taken to avoid breaking
3329 the One Definition Rule; for example, it is usually not useful to mark
3330 an inline method as hidden without marking the whole class as hidden.
3332 A C++ namespace declaration can also have the visibility attribute.
3335 namespace nspace1 __attribute__ ((visibility ("protected")))
3336 @{ /* @r{Do something.} */; @}
3339 This attribute applies only to the particular namespace body, not to
3340 other definitions of the same namespace; it is equivalent to using
3341 @samp{#pragma GCC visibility} before and after the namespace
3342 definition (@pxref{Visibility Pragmas}).
3344 In C++, if a template argument has limited visibility, this
3345 restriction is implicitly propagated to the template instantiation.
3346 Otherwise, template instantiations and specializations default to the
3347 visibility of their template.
3349 If both the template and enclosing class have explicit visibility, the
3350 visibility from the template is used.
3352 @item warn_unused_result
3353 @cindex @code{warn_unused_result} function attribute
3354 The @code{warn_unused_result} attribute causes a warning to be emitted
3355 if a caller of the function with this attribute does not use its
3356 return value. This is useful for functions where not checking
3357 the result is either a security problem or always a bug, such as
3361 int fn () __attribute__ ((warn_unused_result));
3364 if (fn () < 0) return -1;
3371 results in warning on line 5.
3374 @cindex @code{weak} function attribute
3375 The @code{weak} attribute causes the declaration to be emitted as a weak
3376 symbol rather than a global. This is primarily useful in defining
3377 library functions that can be overridden in user code, though it can
3378 also be used with non-function declarations. Weak symbols are supported
3379 for ELF targets, and also for a.out targets when using the GNU assembler
3383 @itemx weakref ("@var{target}")
3384 @cindex @code{weakref} function attribute
3385 The @code{weakref} attribute marks a declaration as a weak reference.
3386 Without arguments, it should be accompanied by an @code{alias} attribute
3387 naming the target symbol. Optionally, the @var{target} may be given as
3388 an argument to @code{weakref} itself. In either case, @code{weakref}
3389 implicitly marks the declaration as @code{weak}. Without a
3390 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3391 @code{weakref} is equivalent to @code{weak}.
3394 static int x() __attribute__ ((weakref ("y")));
3395 /* is equivalent to... */
3396 static int x() __attribute__ ((weak, weakref, alias ("y")));
3398 static int x() __attribute__ ((weakref));
3399 static int x() __attribute__ ((alias ("y")));
3402 A weak reference is an alias that does not by itself require a
3403 definition to be given for the target symbol. If the target symbol is
3404 only referenced through weak references, then it becomes a @code{weak}
3405 undefined symbol. If it is directly referenced, however, then such
3406 strong references prevail, and a definition is required for the
3407 symbol, not necessarily in the same translation unit.
3409 The effect is equivalent to moving all references to the alias to a
3410 separate translation unit, renaming the alias to the aliased symbol,
3411 declaring it as weak, compiling the two separate translation units and
3412 performing a reloadable link on them.
3414 At present, a declaration to which @code{weakref} is attached can
3415 only be @code{static}.
3420 @c This is the end of the target-independent attribute table
3422 @node AArch64 Function Attributes
3423 @subsection AArch64 Function Attributes
3425 The following target-specific function attributes are available for the
3426 AArch64 target. For the most part, these options mirror the behavior of
3427 similar command-line options (@pxref{AArch64 Options}), but on a
3431 @item general-regs-only
3432 @cindex @code{general-regs-only} function attribute, AArch64
3433 Indicates that no floating-point or Advanced SIMD registers should be
3434 used when generating code for this function. If the function explicitly
3435 uses floating-point code, then the compiler gives an error. This is
3436 the same behavior as that of the command-line option
3437 @option{-mgeneral-regs-only}.
3439 @item fix-cortex-a53-835769
3440 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3441 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3442 applied to this function. To explicitly disable the workaround for this
3443 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3444 This corresponds to the behavior of the command line options
3445 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3448 @cindex @code{cmodel=} function attribute, AArch64
3449 Indicates that code should be generated for a particular code model for
3450 this function. The behavior and permissible arguments are the same as
3451 for the command line option @option{-mcmodel=}.
3454 @cindex @code{strict-align} function attribute, AArch64
3455 Indicates that the compiler should not assume that unaligned memory references
3456 are handled by the system. The behavior is the same as for the command-line
3457 option @option{-mstrict-align}.
3459 @item omit-leaf-frame-pointer
3460 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3461 Indicates that the frame pointer should be omitted for a leaf function call.
3462 To keep the frame pointer, the inverse attribute
3463 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3464 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3465 and @option{-mno-omit-leaf-frame-pointer}.
3468 @cindex @code{tls-dialect=} function attribute, AArch64
3469 Specifies the TLS dialect to use for this function. The behavior and
3470 permissible arguments are the same as for the command-line option
3471 @option{-mtls-dialect=}.
3474 @cindex @code{arch=} function attribute, AArch64
3475 Specifies the architecture version and architectural extensions to use
3476 for this function. The behavior and permissible arguments are the same as
3477 for the @option{-march=} command-line option.
3480 @cindex @code{tune=} function attribute, AArch64
3481 Specifies the core for which to tune the performance of this function.
3482 The behavior and permissible arguments are the same as for the @option{-mtune=}
3483 command-line option.
3486 @cindex @code{cpu=} function attribute, AArch64
3487 Specifies the core for which to tune the performance of this function and also
3488 whose architectural features to use. The behavior and valid arguments are the
3489 same as for the @option{-mcpu=} command-line option.
3493 The above target attributes can be specified as follows:
3496 __attribute__((target("@var{attr-string}")))
3504 where @code{@var{attr-string}} is one of the attribute strings specified above.
3506 Additionally, the architectural extension string may be specified on its
3507 own. This can be used to turn on and off particular architectural extensions
3508 without having to specify a particular architecture version or core. Example:
3511 __attribute__((target("+crc+nocrypto")))
3519 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3520 extension and disables the @code{crypto} extension for the function @code{foo}
3521 without modifying an existing @option{-march=} or @option{-mcpu} option.
3523 Multiple target function attributes can be specified by separating them with
3524 a comma. For example:
3526 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3534 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3535 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3537 @subsubsection Inlining rules
3538 Specifying target attributes on individual functions or performing link-time
3539 optimization across translation units compiled with different target options
3540 can affect function inlining rules:
3542 In particular, a caller function can inline a callee function only if the
3543 architectural features available to the callee are a subset of the features
3544 available to the caller.
3545 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3546 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3547 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3548 because the all the architectural features that function @code{bar} requires
3549 are available to function @code{foo}. Conversely, function @code{bar} cannot
3550 inline function @code{foo}.
3552 Additionally inlining a function compiled with @option{-mstrict-align} into a
3553 function compiled without @code{-mstrict-align} is not allowed.
3554 However, inlining a function compiled without @option{-mstrict-align} into a
3555 function compiled with @option{-mstrict-align} is allowed.
3557 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3558 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3559 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3560 architectural feature rules specified above.
3562 @node ARC Function Attributes
3563 @subsection ARC Function Attributes
3565 These function attributes are supported by the ARC back end:
3569 @cindex @code{interrupt} function attribute, ARC
3570 Use this attribute to indicate
3571 that the specified function is an interrupt handler. The compiler generates
3572 function entry and exit sequences suitable for use in an interrupt handler
3573 when this attribute is present.
3575 On the ARC, you must specify the kind of interrupt to be handled
3576 in a parameter to the interrupt attribute like this:
3579 void f () __attribute__ ((interrupt ("ilink1")));
3582 Permissible values for this parameter are: @w{@code{ilink1}} and
3588 @cindex @code{long_call} function attribute, ARC
3589 @cindex @code{medium_call} function attribute, ARC
3590 @cindex @code{short_call} function attribute, ARC
3591 @cindex indirect calls, ARC
3592 These attributes specify how a particular function is called.
3593 These attributes override the
3594 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3595 command-line switches and @code{#pragma long_calls} settings.
3597 For ARC, a function marked with the @code{long_call} attribute is
3598 always called using register-indirect jump-and-link instructions,
3599 thereby enabling the called function to be placed anywhere within the
3600 32-bit address space. A function marked with the @code{medium_call}
3601 attribute will always be close enough to be called with an unconditional
3602 branch-and-link instruction, which has a 25-bit offset from
3603 the call site. A function marked with the @code{short_call}
3604 attribute will always be close enough to be called with a conditional
3605 branch-and-link instruction, which has a 21-bit offset from
3609 @node ARM Function Attributes
3610 @subsection ARM Function Attributes
3612 These function attributes are supported for ARM targets:
3616 @cindex @code{interrupt} function attribute, ARM
3617 Use this attribute to indicate
3618 that the specified function is an interrupt handler. The compiler generates
3619 function entry and exit sequences suitable for use in an interrupt handler
3620 when this attribute is present.
3622 You can specify the kind of interrupt to be handled by
3623 adding an optional parameter to the interrupt attribute like this:
3626 void f () __attribute__ ((interrupt ("IRQ")));
3630 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3631 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3633 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3634 may be called with a word-aligned stack pointer.
3637 @cindex @code{isr} function attribute, ARM
3638 Use this attribute on ARM to write Interrupt Service Routines. This is an
3639 alias to the @code{interrupt} attribute above.
3643 @cindex @code{long_call} function attribute, ARM
3644 @cindex @code{short_call} function attribute, ARM
3645 @cindex indirect calls, ARM
3646 These attributes specify how a particular function is called.
3647 These attributes override the
3648 @option{-mlong-calls} (@pxref{ARM Options})
3649 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3650 @code{long_call} attribute indicates that the function might be far
3651 away from the call site and require a different (more expensive)
3652 calling sequence. The @code{short_call} attribute always places
3653 the offset to the function from the call site into the @samp{BL}
3654 instruction directly.
3657 @cindex @code{naked} function attribute, ARM
3658 This attribute allows the compiler to construct the
3659 requisite function declaration, while allowing the body of the
3660 function to be assembly code. The specified function will not have
3661 prologue/epilogue sequences generated by the compiler. Only basic
3662 @code{asm} statements can safely be included in naked functions
3663 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3664 basic @code{asm} and C code may appear to work, they cannot be
3665 depended upon to work reliably and are not supported.
3668 @cindex @code{pcs} function attribute, ARM
3670 The @code{pcs} attribute can be used to control the calling convention
3671 used for a function on ARM. The attribute takes an argument that specifies
3672 the calling convention to use.
3674 When compiling using the AAPCS ABI (or a variant of it) then valid
3675 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3676 order to use a variant other than @code{"aapcs"} then the compiler must
3677 be permitted to use the appropriate co-processor registers (i.e., the
3678 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3682 /* Argument passed in r0, and result returned in r0+r1. */
3683 double f2d (float) __attribute__((pcs("aapcs")));
3686 Variadic functions always use the @code{"aapcs"} calling convention and
3687 the compiler rejects attempts to specify an alternative.
3689 @item target (@var{options})
3690 @cindex @code{target} function attribute
3691 As discussed in @ref{Common Function Attributes}, this attribute
3692 allows specification of target-specific compilation options.
3694 On ARM, the following options are allowed:
3698 @cindex @code{target("thumb")} function attribute, ARM
3699 Force code generation in the Thumb (T16/T32) ISA, depending on the
3703 @cindex @code{target("arm")} function attribute, ARM
3704 Force code generation in the ARM (A32) ISA.
3706 Functions from different modes can be inlined in the caller's mode.
3709 @cindex @code{target("fpu=")} function attribute, ARM
3710 Specifies the fpu for which to tune the performance of this function.
3711 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3712 command-line option.
3718 @node AVR Function Attributes
3719 @subsection AVR Function Attributes
3721 These function attributes are supported by the AVR back end:
3725 @cindex @code{interrupt} function attribute, AVR
3726 Use this attribute to indicate
3727 that the specified function is an interrupt handler. The compiler generates
3728 function entry and exit sequences suitable for use in an interrupt handler
3729 when this attribute is present.
3731 On the AVR, the hardware globally disables interrupts when an
3732 interrupt is executed. The first instruction of an interrupt handler
3733 declared with this attribute is a @code{SEI} instruction to
3734 re-enable interrupts. See also the @code{signal} function attribute
3735 that does not insert a @code{SEI} instruction. If both @code{signal} and
3736 @code{interrupt} are specified for the same function, @code{signal}
3737 is silently ignored.
3740 @cindex @code{naked} function attribute, AVR
3741 This attribute allows the compiler to construct the
3742 requisite function declaration, while allowing the body of the
3743 function to be assembly code. The specified function will not have
3744 prologue/epilogue sequences generated by the compiler. Only basic
3745 @code{asm} statements can safely be included in naked functions
3746 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3747 basic @code{asm} and C code may appear to work, they cannot be
3748 depended upon to work reliably and are not supported.
3752 @cindex @code{OS_main} function attribute, AVR
3753 @cindex @code{OS_task} function attribute, AVR
3754 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3755 do not save/restore any call-saved register in their prologue/epilogue.
3757 The @code{OS_main} attribute can be used when there @emph{is
3758 guarantee} that interrupts are disabled at the time when the function
3759 is entered. This saves resources when the stack pointer has to be
3760 changed to set up a frame for local variables.
3762 The @code{OS_task} attribute can be used when there is @emph{no
3763 guarantee} that interrupts are disabled at that time when the function
3764 is entered like for, e@.g@. task functions in a multi-threading operating
3765 system. In that case, changing the stack pointer register is
3766 guarded by save/clear/restore of the global interrupt enable flag.
3768 The differences to the @code{naked} function attribute are:
3770 @item @code{naked} functions do not have a return instruction whereas
3771 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3772 @code{RETI} return instruction.
3773 @item @code{naked} functions do not set up a frame for local variables
3774 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3779 @cindex @code{signal} function attribute, AVR
3780 Use this attribute on the AVR to indicate that the specified
3781 function is an interrupt handler. The compiler generates function
3782 entry and exit sequences suitable for use in an interrupt handler when this
3783 attribute is present.
3785 See also the @code{interrupt} function attribute.
3787 The AVR hardware globally disables interrupts when an interrupt is executed.
3788 Interrupt handler functions defined with the @code{signal} attribute
3789 do not re-enable interrupts. It is save to enable interrupts in a
3790 @code{signal} handler. This ``save'' only applies to the code
3791 generated by the compiler and not to the IRQ layout of the
3792 application which is responsibility of the application.
3794 If both @code{signal} and @code{interrupt} are specified for the same
3795 function, @code{signal} is silently ignored.
3798 @node Blackfin Function Attributes
3799 @subsection Blackfin Function Attributes
3801 These function attributes are supported by the Blackfin back end:
3805 @item exception_handler
3806 @cindex @code{exception_handler} function attribute
3807 @cindex exception handler functions, Blackfin
3808 Use this attribute on the Blackfin to indicate that the specified function
3809 is an exception handler. The compiler generates function entry and
3810 exit sequences suitable for use in an exception handler when this
3811 attribute is present.
3813 @item interrupt_handler
3814 @cindex @code{interrupt_handler} function attribute, Blackfin
3815 Use this attribute to
3816 indicate that the specified function is an interrupt handler. The compiler
3817 generates function entry and exit sequences suitable for use in an
3818 interrupt handler when this attribute is present.
3821 @cindex @code{kspisusp} function attribute, Blackfin
3822 @cindex User stack pointer in interrupts on the Blackfin
3823 When used together with @code{interrupt_handler}, @code{exception_handler}
3824 or @code{nmi_handler}, code is generated to load the stack pointer
3825 from the USP register in the function prologue.
3828 @cindex @code{l1_text} function attribute, Blackfin
3829 This attribute specifies a function to be placed into L1 Instruction
3830 SRAM@. The function is put into a specific section named @code{.l1.text}.
3831 With @option{-mfdpic}, function calls with a such function as the callee
3832 or caller uses inlined PLT.
3835 @cindex @code{l2} function attribute, Blackfin
3836 This attribute specifies a function to be placed into L2
3837 SRAM. The function is put into a specific section named
3838 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3843 @cindex indirect calls, Blackfin
3844 @cindex @code{longcall} function attribute, Blackfin
3845 @cindex @code{shortcall} function attribute, Blackfin
3846 The @code{longcall} attribute
3847 indicates that the function might be far away from the call site and
3848 require a different (more expensive) calling sequence. The
3849 @code{shortcall} attribute indicates that the function is always close
3850 enough for the shorter calling sequence to be used. These attributes
3851 override the @option{-mlongcall} switch.
3854 @cindex @code{nesting} function attribute, Blackfin
3855 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3856 Use this attribute together with @code{interrupt_handler},
3857 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3858 entry code should enable nested interrupts or exceptions.
3861 @cindex @code{nmi_handler} function attribute, Blackfin
3862 @cindex NMI handler functions on the Blackfin processor
3863 Use this attribute on the Blackfin to indicate that the specified function
3864 is an NMI handler. The compiler generates function entry and
3865 exit sequences suitable for use in an NMI handler when this
3866 attribute is present.
3869 @cindex @code{saveall} function attribute, Blackfin
3870 @cindex save all registers on the Blackfin
3871 Use this attribute to indicate that
3872 all registers except the stack pointer should be saved in the prologue
3873 regardless of whether they are used or not.
3876 @node CR16 Function Attributes
3877 @subsection CR16 Function Attributes
3879 These function attributes are supported by the CR16 back end:
3883 @cindex @code{interrupt} function attribute, CR16
3884 Use this attribute to indicate
3885 that the specified function is an interrupt handler. The compiler generates
3886 function entry and exit sequences suitable for use in an interrupt handler
3887 when this attribute is present.
3890 @node Epiphany Function Attributes
3891 @subsection Epiphany Function Attributes
3893 These function attributes are supported by the Epiphany back end:
3897 @cindex @code{disinterrupt} function attribute, Epiphany
3898 This attribute causes the compiler to emit
3899 instructions to disable interrupts for the duration of the given
3902 @item forwarder_section
3903 @cindex @code{forwarder_section} function attribute, Epiphany
3904 This attribute modifies the behavior of an interrupt handler.
3905 The interrupt handler may be in external memory which cannot be
3906 reached by a branch instruction, so generate a local memory trampoline
3907 to transfer control. The single parameter identifies the section where
3908 the trampoline is placed.
3911 @cindex @code{interrupt} function attribute, Epiphany
3912 Use this attribute to indicate
3913 that the specified function is an interrupt handler. The compiler generates
3914 function entry and exit sequences suitable for use in an interrupt handler
3915 when this attribute is present. It may also generate
3916 a special section with code to initialize the interrupt vector table.
3918 On Epiphany targets one or more optional parameters can be added like this:
3921 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3924 Permissible values for these parameters are: @w{@code{reset}},
3925 @w{@code{software_exception}}, @w{@code{page_miss}},
3926 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3927 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3928 Multiple parameters indicate that multiple entries in the interrupt
3929 vector table should be initialized for this function, i.e.@: for each
3930 parameter @w{@var{name}}, a jump to the function is emitted in
3931 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3932 entirely, in which case no interrupt vector table entry is provided.
3934 Note that interrupts are enabled inside the function
3935 unless the @code{disinterrupt} attribute is also specified.
3937 The following examples are all valid uses of these attributes on
3940 void __attribute__ ((interrupt)) universal_handler ();
3941 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3942 void __attribute__ ((interrupt ("dma0, dma1")))
3943 universal_dma_handler ();
3944 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3945 fast_timer_handler ();
3946 void __attribute__ ((interrupt ("dma0, dma1"),
3947 forwarder_section ("tramp")))
3948 external_dma_handler ();
3953 @cindex @code{long_call} function attribute, Epiphany
3954 @cindex @code{short_call} function attribute, Epiphany
3955 @cindex indirect calls, Epiphany
3956 These attributes specify how a particular function is called.
3957 These attributes override the
3958 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3959 command-line switch and @code{#pragma long_calls} settings.
3963 @node H8/300 Function Attributes
3964 @subsection H8/300 Function Attributes
3966 These function attributes are available for H8/300 targets:
3969 @item function_vector
3970 @cindex @code{function_vector} function attribute, H8/300
3971 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3972 that the specified function should be called through the function vector.
3973 Calling a function through the function vector reduces code size; however,
3974 the function vector has a limited size (maximum 128 entries on the H8/300
3975 and 64 entries on the H8/300H and H8S)
3976 and shares space with the interrupt vector.
3978 @item interrupt_handler
3979 @cindex @code{interrupt_handler} function attribute, H8/300
3980 Use this attribute on the H8/300, H8/300H, and H8S to
3981 indicate that the specified function is an interrupt handler. The compiler
3982 generates function entry and exit sequences suitable for use in an
3983 interrupt handler when this attribute is present.
3986 @cindex @code{saveall} function attribute, H8/300
3987 @cindex save all registers on the H8/300, H8/300H, and H8S
3988 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3989 all registers except the stack pointer should be saved in the prologue
3990 regardless of whether they are used or not.
3993 @node IA-64 Function Attributes
3994 @subsection IA-64 Function Attributes
3996 These function attributes are supported on IA-64 targets:
3999 @item syscall_linkage
4000 @cindex @code{syscall_linkage} function attribute, IA-64
4001 This attribute is used to modify the IA-64 calling convention by marking
4002 all input registers as live at all function exits. This makes it possible
4003 to restart a system call after an interrupt without having to save/restore
4004 the input registers. This also prevents kernel data from leaking into
4008 @cindex @code{version_id} function attribute, IA-64
4009 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4010 symbol to contain a version string, thus allowing for function level
4011 versioning. HP-UX system header files may use function level versioning
4012 for some system calls.
4015 extern int foo () __attribute__((version_id ("20040821")));
4019 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4022 @node M32C Function Attributes
4023 @subsection M32C Function Attributes
4025 These function attributes are supported by the M32C back end:
4029 @cindex @code{bank_switch} function attribute, M32C
4030 When added to an interrupt handler with the M32C port, causes the
4031 prologue and epilogue to use bank switching to preserve the registers
4032 rather than saving them on the stack.
4034 @item fast_interrupt
4035 @cindex @code{fast_interrupt} function attribute, M32C
4036 Use this attribute on the M32C port to indicate that the specified
4037 function is a fast interrupt handler. This is just like the
4038 @code{interrupt} attribute, except that @code{freit} is used to return
4039 instead of @code{reit}.
4041 @item function_vector
4042 @cindex @code{function_vector} function attribute, M16C/M32C
4043 On M16C/M32C targets, the @code{function_vector} attribute declares a
4044 special page subroutine call function. Use of this attribute reduces
4045 the code size by 2 bytes for each call generated to the
4046 subroutine. The argument to the attribute is the vector number entry
4047 from the special page vector table which contains the 16 low-order
4048 bits of the subroutine's entry address. Each vector table has special
4049 page number (18 to 255) that is used in @code{jsrs} instructions.
4050 Jump addresses of the routines are generated by adding 0x0F0000 (in
4051 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4052 2-byte addresses set in the vector table. Therefore you need to ensure
4053 that all the special page vector routines should get mapped within the
4054 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4057 In the following example 2 bytes are saved for each call to
4058 function @code{foo}.
4061 void foo (void) __attribute__((function_vector(0x18)));
4072 If functions are defined in one file and are called in another file,
4073 then be sure to write this declaration in both files.
4075 This attribute is ignored for R8C target.
4078 @cindex @code{interrupt} function attribute, M32C
4079 Use this attribute to indicate
4080 that the specified function is an interrupt handler. The compiler generates
4081 function entry and exit sequences suitable for use in an interrupt handler
4082 when this attribute is present.
4085 @node M32R/D Function Attributes
4086 @subsection M32R/D Function Attributes
4088 These function attributes are supported by the M32R/D back end:
4092 @cindex @code{interrupt} function attribute, M32R/D
4093 Use this attribute to indicate
4094 that the specified function is an interrupt handler. The compiler generates
4095 function entry and exit sequences suitable for use in an interrupt handler
4096 when this attribute is present.
4098 @item model (@var{model-name})
4099 @cindex @code{model} function attribute, M32R/D
4100 @cindex function addressability on the M32R/D
4102 On the M32R/D, use this attribute to set the addressability of an
4103 object, and of the code generated for a function. The identifier
4104 @var{model-name} is one of @code{small}, @code{medium}, or
4105 @code{large}, representing each of the code models.
4107 Small model objects live in the lower 16MB of memory (so that their
4108 addresses can be loaded with the @code{ld24} instruction), and are
4109 callable with the @code{bl} instruction.
4111 Medium model objects may live anywhere in the 32-bit address space (the
4112 compiler generates @code{seth/add3} instructions to load their addresses),
4113 and are callable with the @code{bl} instruction.
4115 Large model objects may live anywhere in the 32-bit address space (the
4116 compiler generates @code{seth/add3} instructions to load their addresses),
4117 and may not be reachable with the @code{bl} instruction (the compiler
4118 generates the much slower @code{seth/add3/jl} instruction sequence).
4121 @node m68k Function Attributes
4122 @subsection m68k Function Attributes
4124 These function attributes are supported by the m68k back end:
4128 @itemx interrupt_handler
4129 @cindex @code{interrupt} function attribute, m68k
4130 @cindex @code{interrupt_handler} function attribute, m68k
4131 Use this attribute to
4132 indicate that the specified function is an interrupt handler. The compiler
4133 generates function entry and exit sequences suitable for use in an
4134 interrupt handler when this attribute is present. Either name may be used.
4136 @item interrupt_thread
4137 @cindex @code{interrupt_thread} function attribute, fido
4138 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4139 that the specified function is an interrupt handler that is designed
4140 to run as a thread. The compiler omits generate prologue/epilogue
4141 sequences and replaces the return instruction with a @code{sleep}
4142 instruction. This attribute is available only on fido.
4145 @node MCORE Function Attributes
4146 @subsection MCORE Function Attributes
4148 These function attributes are supported by the MCORE back end:
4152 @cindex @code{naked} function attribute, MCORE
4153 This attribute allows the compiler to construct the
4154 requisite function declaration, while allowing the body of the
4155 function to be assembly code. The specified function will not have
4156 prologue/epilogue sequences generated by the compiler. Only basic
4157 @code{asm} statements can safely be included in naked functions
4158 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4159 basic @code{asm} and C code may appear to work, they cannot be
4160 depended upon to work reliably and are not supported.
4163 @node MeP Function Attributes
4164 @subsection MeP Function Attributes
4166 These function attributes are supported by the MeP back end:
4170 @cindex @code{disinterrupt} function attribute, MeP
4171 On MeP targets, this attribute causes the compiler to emit
4172 instructions to disable interrupts for the duration of the given
4176 @cindex @code{interrupt} function attribute, MeP
4177 Use this attribute to indicate
4178 that the specified function is an interrupt handler. The compiler generates
4179 function entry and exit sequences suitable for use in an interrupt handler
4180 when this attribute is present.
4183 @cindex @code{near} function attribute, MeP
4184 This attribute causes the compiler to assume the called
4185 function is close enough to use the normal calling convention,
4186 overriding the @option{-mtf} command-line option.
4189 @cindex @code{far} function attribute, MeP
4190 On MeP targets this causes the compiler to use a calling convention
4191 that assumes the called function is too far away for the built-in
4195 @cindex @code{vliw} function attribute, MeP
4196 The @code{vliw} attribute tells the compiler to emit
4197 instructions in VLIW mode instead of core mode. Note that this
4198 attribute is not allowed unless a VLIW coprocessor has been configured
4199 and enabled through command-line options.
4202 @node MicroBlaze Function Attributes
4203 @subsection MicroBlaze Function Attributes
4205 These function attributes are supported on MicroBlaze targets:
4208 @item save_volatiles
4209 @cindex @code{save_volatiles} function attribute, MicroBlaze
4210 Use this attribute to indicate that the function is
4211 an interrupt handler. All volatile registers (in addition to non-volatile
4212 registers) are saved in the function prologue. If the function is a leaf
4213 function, only volatiles used by the function are saved. A normal function
4214 return is generated instead of a return from interrupt.
4217 @cindex @code{break_handler} function attribute, MicroBlaze
4218 @cindex break handler functions
4219 Use this attribute to indicate that
4220 the specified function is a break handler. The compiler generates function
4221 entry and exit sequences suitable for use in an break handler when this
4222 attribute is present. The return from @code{break_handler} is done through
4223 the @code{rtbd} instead of @code{rtsd}.
4226 void f () __attribute__ ((break_handler));
4229 @item interrupt_handler
4230 @itemx fast_interrupt
4231 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4232 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4233 These attributes indicate that the specified function is an interrupt
4234 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4235 used in low-latency interrupt mode, and @code{interrupt_handler} for
4236 interrupts that do not use low-latency handlers. In both cases, GCC
4237 emits appropriate prologue code and generates a return from the handler
4238 using @code{rtid} instead of @code{rtsd}.
4241 @node Microsoft Windows Function Attributes
4242 @subsection Microsoft Windows Function Attributes
4244 The following attributes are available on Microsoft Windows and Symbian OS
4249 @cindex @code{dllexport} function attribute
4250 @cindex @code{__declspec(dllexport)}
4251 On Microsoft Windows targets and Symbian OS targets the
4252 @code{dllexport} attribute causes the compiler to provide a global
4253 pointer to a pointer in a DLL, so that it can be referenced with the
4254 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4255 name is formed by combining @code{_imp__} and the function or variable
4258 You can use @code{__declspec(dllexport)} as a synonym for
4259 @code{__attribute__ ((dllexport))} for compatibility with other
4262 On systems that support the @code{visibility} attribute, this
4263 attribute also implies ``default'' visibility. It is an error to
4264 explicitly specify any other visibility.
4266 GCC's default behavior is to emit all inline functions with the
4267 @code{dllexport} attribute. Since this can cause object file-size bloat,
4268 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4269 ignore the attribute for inlined functions unless the
4270 @option{-fkeep-inline-functions} flag is used instead.
4272 The attribute is ignored for undefined symbols.
4274 When applied to C++ classes, the attribute marks defined non-inlined
4275 member functions and static data members as exports. Static consts
4276 initialized in-class are not marked unless they are also defined
4279 For Microsoft Windows targets there are alternative methods for
4280 including the symbol in the DLL's export table such as using a
4281 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4282 the @option{--export-all} linker flag.
4285 @cindex @code{dllimport} function attribute
4286 @cindex @code{__declspec(dllimport)}
4287 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4288 attribute causes the compiler to reference a function or variable via
4289 a global pointer to a pointer that is set up by the DLL exporting the
4290 symbol. The attribute implies @code{extern}. On Microsoft Windows
4291 targets, the pointer name is formed by combining @code{_imp__} and the
4292 function or variable name.
4294 You can use @code{__declspec(dllimport)} as a synonym for
4295 @code{__attribute__ ((dllimport))} for compatibility with other
4298 On systems that support the @code{visibility} attribute, this
4299 attribute also implies ``default'' visibility. It is an error to
4300 explicitly specify any other visibility.
4302 Currently, the attribute is ignored for inlined functions. If the
4303 attribute is applied to a symbol @emph{definition}, an error is reported.
4304 If a symbol previously declared @code{dllimport} is later defined, the
4305 attribute is ignored in subsequent references, and a warning is emitted.
4306 The attribute is also overridden by a subsequent declaration as
4309 When applied to C++ classes, the attribute marks non-inlined
4310 member functions and static data members as imports. However, the
4311 attribute is ignored for virtual methods to allow creation of vtables
4314 On the SH Symbian OS target the @code{dllimport} attribute also has
4315 another affect---it can cause the vtable and run-time type information
4316 for a class to be exported. This happens when the class has a
4317 dllimported constructor or a non-inline, non-pure virtual function
4318 and, for either of those two conditions, the class also has an inline
4319 constructor or destructor and has a key function that is defined in
4320 the current translation unit.
4322 For Microsoft Windows targets the use of the @code{dllimport}
4323 attribute on functions is not necessary, but provides a small
4324 performance benefit by eliminating a thunk in the DLL@. The use of the
4325 @code{dllimport} attribute on imported variables can be avoided by passing the
4326 @option{--enable-auto-import} switch to the GNU linker. As with
4327 functions, using the attribute for a variable eliminates a thunk in
4330 One drawback to using this attribute is that a pointer to a
4331 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4332 address. However, a pointer to a @emph{function} with the
4333 @code{dllimport} attribute can be used as a constant initializer; in
4334 this case, the address of a stub function in the import lib is
4335 referenced. On Microsoft Windows targets, the attribute can be disabled
4336 for functions by setting the @option{-mnop-fun-dllimport} flag.
4339 @node MIPS Function Attributes
4340 @subsection MIPS Function Attributes
4342 These function attributes are supported by the MIPS back end:
4346 @cindex @code{interrupt} function attribute, MIPS
4347 Use this attribute to indicate that the specified function is an interrupt
4348 handler. The compiler generates function entry and exit sequences suitable
4349 for use in an interrupt handler when this attribute is present.
4350 An optional argument is supported for the interrupt attribute which allows
4351 the interrupt mode to be described. By default GCC assumes the external
4352 interrupt controller (EIC) mode is in use, this can be explicitly set using
4353 @code{eic}. When interrupts are non-masked then the requested Interrupt
4354 Priority Level (IPL) is copied to the current IPL which has the effect of only
4355 enabling higher priority interrupts. To use vectored interrupt mode use
4356 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4357 the behavior of the non-masked interrupt support and GCC will arrange to mask
4358 all interrupts from sw0 up to and including the specified interrupt vector.
4360 You can use the following attributes to modify the behavior
4361 of an interrupt handler:
4363 @item use_shadow_register_set
4364 @cindex @code{use_shadow_register_set} function attribute, MIPS
4365 Assume that the handler uses a shadow register set, instead of
4366 the main general-purpose registers. An optional argument @code{intstack} is
4367 supported to indicate that the shadow register set contains a valid stack
4370 @item keep_interrupts_masked
4371 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4372 Keep interrupts masked for the whole function. Without this attribute,
4373 GCC tries to reenable interrupts for as much of the function as it can.
4375 @item use_debug_exception_return
4376 @cindex @code{use_debug_exception_return} function attribute, MIPS
4377 Return using the @code{deret} instruction. Interrupt handlers that don't
4378 have this attribute return using @code{eret} instead.
4381 You can use any combination of these attributes, as shown below:
4383 void __attribute__ ((interrupt)) v0 ();
4384 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4385 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4386 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4387 void __attribute__ ((interrupt, use_shadow_register_set,
4388 keep_interrupts_masked)) v4 ();
4389 void __attribute__ ((interrupt, use_shadow_register_set,
4390 use_debug_exception_return)) v5 ();
4391 void __attribute__ ((interrupt, keep_interrupts_masked,
4392 use_debug_exception_return)) v6 ();
4393 void __attribute__ ((interrupt, use_shadow_register_set,
4394 keep_interrupts_masked,
4395 use_debug_exception_return)) v7 ();
4396 void __attribute__ ((interrupt("eic"))) v8 ();
4397 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4403 @cindex indirect calls, MIPS
4404 @cindex @code{long_call} function attribute, MIPS
4405 @cindex @code{near} function attribute, MIPS
4406 @cindex @code{far} function attribute, MIPS
4407 These attributes specify how a particular function is called on MIPS@.
4408 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4409 command-line switch. The @code{long_call} and @code{far} attributes are
4410 synonyms, and cause the compiler to always call
4411 the function by first loading its address into a register, and then using
4412 the contents of that register. The @code{near} attribute has the opposite
4413 effect; it specifies that non-PIC calls should be made using the more
4414 efficient @code{jal} instruction.
4418 @cindex @code{mips16} function attribute, MIPS
4419 @cindex @code{nomips16} function attribute, MIPS
4421 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4422 function attributes to locally select or turn off MIPS16 code generation.
4423 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4424 while MIPS16 code generation is disabled for functions with the
4425 @code{nomips16} attribute. These attributes override the
4426 @option{-mips16} and @option{-mno-mips16} options on the command line
4427 (@pxref{MIPS Options}).
4429 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4430 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4431 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4432 may interact badly with some GCC extensions such as @code{__builtin_apply}
4433 (@pxref{Constructing Calls}).
4435 @item micromips, MIPS
4436 @itemx nomicromips, MIPS
4437 @cindex @code{micromips} function attribute
4438 @cindex @code{nomicromips} function attribute
4440 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4441 function attributes to locally select or turn off microMIPS code generation.
4442 A function with the @code{micromips} attribute is emitted as microMIPS code,
4443 while microMIPS code generation is disabled for functions with the
4444 @code{nomicromips} attribute. These attributes override the
4445 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4446 (@pxref{MIPS Options}).
4448 When compiling files containing mixed microMIPS and non-microMIPS code, the
4449 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4451 not that within individual functions. Mixed microMIPS and non-microMIPS code
4452 may interact badly with some GCC extensions such as @code{__builtin_apply}
4453 (@pxref{Constructing Calls}).
4456 @cindex @code{nocompression} function attribute, MIPS
4457 On MIPS targets, you can use the @code{nocompression} function attribute
4458 to locally turn off MIPS16 and microMIPS code generation. This attribute
4459 overrides the @option{-mips16} and @option{-mmicromips} options on the
4460 command line (@pxref{MIPS Options}).
4463 @node MSP430 Function Attributes
4464 @subsection MSP430 Function Attributes
4466 These function attributes are supported by the MSP430 back end:
4470 @cindex @code{critical} function attribute, MSP430
4471 Critical functions disable interrupts upon entry and restore the
4472 previous interrupt state upon exit. Critical functions cannot also
4473 have the @code{naked} or @code{reentrant} attributes. They can have
4474 the @code{interrupt} attribute.
4477 @cindex @code{interrupt} function attribute, MSP430
4478 Use this attribute to indicate
4479 that the specified function is an interrupt handler. The compiler generates
4480 function entry and exit sequences suitable for use in an interrupt handler
4481 when this attribute is present.
4483 You can provide an argument to the interrupt
4484 attribute which specifies a name or number. If the argument is a
4485 number it indicates the slot in the interrupt vector table (0 - 31) to
4486 which this handler should be assigned. If the argument is a name it
4487 is treated as a symbolic name for the vector slot. These names should
4488 match up with appropriate entries in the linker script. By default
4489 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4490 @code{reset} for vector 31 are recognized.
4493 @cindex @code{naked} function attribute, MSP430
4494 This attribute allows the compiler to construct the
4495 requisite function declaration, while allowing the body of the
4496 function to be assembly code. The specified function will not have
4497 prologue/epilogue sequences generated by the compiler. Only basic
4498 @code{asm} statements can safely be included in naked functions
4499 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4500 basic @code{asm} and C code may appear to work, they cannot be
4501 depended upon to work reliably and are not supported.
4504 @cindex @code{reentrant} function attribute, MSP430
4505 Reentrant functions disable interrupts upon entry and enable them
4506 upon exit. Reentrant functions cannot also have the @code{naked}
4507 or @code{critical} attributes. They can have the @code{interrupt}
4511 @cindex @code{wakeup} function attribute, MSP430
4512 This attribute only applies to interrupt functions. It is silently
4513 ignored if applied to a non-interrupt function. A wakeup interrupt
4514 function will rouse the processor from any low-power state that it
4515 might be in when the function exits.
4520 @cindex @code{lower} function attribute, MSP430
4521 @cindex @code{upper} function attribute, MSP430
4522 @cindex @code{either} function attribute, MSP430
4523 On the MSP430 target these attributes can be used to specify whether
4524 the function or variable should be placed into low memory, high
4525 memory, or the placement should be left to the linker to decide. The
4526 attributes are only significant if compiling for the MSP430X
4529 The attributes work in conjunction with a linker script that has been
4530 augmented to specify where to place sections with a @code{.lower} and
4531 a @code{.upper} prefix. So, for example, as well as placing the
4532 @code{.data} section, the script also specifies the placement of a
4533 @code{.lower.data} and a @code{.upper.data} section. The intention
4534 is that @code{lower} sections are placed into a small but easier to
4535 access memory region and the upper sections are placed into a larger, but
4536 slower to access, region.
4538 The @code{either} attribute is special. It tells the linker to place
4539 the object into the corresponding @code{lower} section if there is
4540 room for it. If there is insufficient room then the object is placed
4541 into the corresponding @code{upper} section instead. Note that the
4542 placement algorithm is not very sophisticated. It does not attempt to
4543 find an optimal packing of the @code{lower} sections. It just makes
4544 one pass over the objects and does the best that it can. Using the
4545 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4546 options can help the packing, however, since they produce smaller,
4547 easier to pack regions.
4550 @node NDS32 Function Attributes
4551 @subsection NDS32 Function Attributes
4553 These function attributes are supported by the NDS32 back end:
4557 @cindex @code{exception} function attribute
4558 @cindex exception handler functions, NDS32
4559 Use this attribute on the NDS32 target to indicate that the specified function
4560 is an exception handler. The compiler will generate corresponding sections
4561 for use in an exception handler.
4564 @cindex @code{interrupt} function attribute, NDS32
4565 On NDS32 target, this attribute indicates that the specified function
4566 is an interrupt handler. The compiler generates corresponding sections
4567 for use in an interrupt handler. You can use the following attributes
4568 to modify the behavior:
4571 @cindex @code{nested} function attribute, NDS32
4572 This interrupt service routine is interruptible.
4574 @cindex @code{not_nested} function attribute, NDS32
4575 This interrupt service routine is not interruptible.
4577 @cindex @code{nested_ready} function attribute, NDS32
4578 This interrupt service routine is interruptible after @code{PSW.GIE}
4579 (global interrupt enable) is set. This allows interrupt service routine to
4580 finish some short critical code before enabling interrupts.
4582 @cindex @code{save_all} function attribute, NDS32
4583 The system will help save all registers into stack before entering
4586 @cindex @code{partial_save} function attribute, NDS32
4587 The system will help save caller registers into stack before entering
4592 @cindex @code{naked} function attribute, NDS32
4593 This attribute allows the compiler to construct the
4594 requisite function declaration, while allowing the body of the
4595 function to be assembly code. The specified function will not have
4596 prologue/epilogue sequences generated by the compiler. Only basic
4597 @code{asm} statements can safely be included in naked functions
4598 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4599 basic @code{asm} and C code may appear to work, they cannot be
4600 depended upon to work reliably and are not supported.
4603 @cindex @code{reset} function attribute, NDS32
4604 @cindex reset handler functions
4605 Use this attribute on the NDS32 target to indicate that the specified function
4606 is a reset handler. The compiler will generate corresponding sections
4607 for use in a reset handler. You can use the following attributes
4608 to provide extra exception handling:
4611 @cindex @code{nmi} function attribute, NDS32
4612 Provide a user-defined function to handle NMI exception.
4614 @cindex @code{warm} function attribute, NDS32
4615 Provide a user-defined function to handle warm reset exception.
4619 @node Nios II Function Attributes
4620 @subsection Nios II Function Attributes
4622 These function attributes are supported by the Nios II back end:
4625 @item target (@var{options})
4626 @cindex @code{target} function attribute
4627 As discussed in @ref{Common Function Attributes}, this attribute
4628 allows specification of target-specific compilation options.
4630 When compiling for Nios II, the following options are allowed:
4633 @item custom-@var{insn}=@var{N}
4634 @itemx no-custom-@var{insn}
4635 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4636 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4637 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4638 custom instruction with encoding @var{N} when generating code that uses
4639 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4640 the custom instruction @var{insn}.
4641 These target attributes correspond to the
4642 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4643 command-line options, and support the same set of @var{insn} keywords.
4644 @xref{Nios II Options}, for more information.
4646 @item custom-fpu-cfg=@var{name}
4647 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4648 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4649 command-line option, to select a predefined set of custom instructions
4651 @xref{Nios II Options}, for more information.
4655 @node Nvidia PTX Function Attributes
4656 @subsection Nvidia PTX Function Attributes
4658 These function attributes are supported by the Nvidia PTX back end:
4662 @cindex @code{kernel} attribute, Nvidia PTX
4663 This attribute indicates that the corresponding function should be compiled
4664 as a kernel function, which can be invoked from the host via the CUDA RT
4666 By default functions are only callable only from other PTX functions.
4668 Kernel functions must have @code{void} return type.
4671 @node PowerPC Function Attributes
4672 @subsection PowerPC Function Attributes
4674 These function attributes are supported by the PowerPC back end:
4679 @cindex indirect calls, PowerPC
4680 @cindex @code{longcall} function attribute, PowerPC
4681 @cindex @code{shortcall} function attribute, PowerPC
4682 The @code{longcall} attribute
4683 indicates that the function might be far away from the call site and
4684 require a different (more expensive) calling sequence. The
4685 @code{shortcall} attribute indicates that the function is always close
4686 enough for the shorter calling sequence to be used. These attributes
4687 override both the @option{-mlongcall} switch and
4688 the @code{#pragma longcall} setting.
4690 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4691 calls are necessary.
4693 @item target (@var{options})
4694 @cindex @code{target} function attribute
4695 As discussed in @ref{Common Function Attributes}, this attribute
4696 allows specification of target-specific compilation options.
4698 On the PowerPC, the following options are allowed:
4703 @cindex @code{target("altivec")} function attribute, PowerPC
4704 Generate code that uses (does not use) AltiVec instructions. In
4705 32-bit code, you cannot enable AltiVec instructions unless
4706 @option{-mabi=altivec} is used on the command line.
4710 @cindex @code{target("cmpb")} function attribute, PowerPC
4711 Generate code that uses (does not use) the compare bytes instruction
4712 implemented on the POWER6 processor and other processors that support
4713 the PowerPC V2.05 architecture.
4717 @cindex @code{target("dlmzb")} function attribute, PowerPC
4718 Generate code that uses (does not use) the string-search @samp{dlmzb}
4719 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4720 generated by default when targeting those processors.
4724 @cindex @code{target("fprnd")} function attribute, PowerPC
4725 Generate code that uses (does not use) the FP round to integer
4726 instructions implemented on the POWER5+ processor and other processors
4727 that support the PowerPC V2.03 architecture.
4731 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4732 Generate code that uses (does not use) the decimal floating-point
4733 instructions implemented on some POWER processors.
4737 @cindex @code{target("isel")} function attribute, PowerPC
4738 Generate code that uses (does not use) ISEL instruction.
4742 @cindex @code{target("mfcrf")} function attribute, PowerPC
4743 Generate code that uses (does not use) the move from condition
4744 register field instruction implemented on the POWER4 processor and
4745 other processors that support the PowerPC V2.01 architecture.
4749 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4750 Generate code that uses (does not use) the FP move to/from general
4751 purpose register instructions implemented on the POWER6X processor and
4752 other processors that support the extended PowerPC V2.05 architecture.
4756 @cindex @code{target("mulhw")} function attribute, PowerPC
4757 Generate code that uses (does not use) the half-word multiply and
4758 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4759 These instructions are generated by default when targeting those
4764 @cindex @code{target("multiple")} function attribute, PowerPC
4765 Generate code that uses (does not use) the load multiple word
4766 instructions and the store multiple word instructions.
4770 @cindex @code{target("update")} function attribute, PowerPC
4771 Generate code that uses (does not use) the load or store instructions
4772 that update the base register to the address of the calculated memory
4777 @cindex @code{target("popcntb")} function attribute, PowerPC
4778 Generate code that uses (does not use) the popcount and double-precision
4779 FP reciprocal estimate instruction implemented on the POWER5
4780 processor and other processors that support the PowerPC V2.02
4785 @cindex @code{target("popcntd")} function attribute, PowerPC
4786 Generate code that uses (does not use) the popcount instruction
4787 implemented on the POWER7 processor and other processors that support
4788 the PowerPC V2.06 architecture.
4790 @item powerpc-gfxopt
4791 @itemx no-powerpc-gfxopt
4792 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4793 Generate code that uses (does not use) the optional PowerPC
4794 architecture instructions in the Graphics group, including
4795 floating-point select.
4798 @itemx no-powerpc-gpopt
4799 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4800 Generate code that uses (does not use) the optional PowerPC
4801 architecture instructions in the General Purpose group, including
4802 floating-point square root.
4804 @item recip-precision
4805 @itemx no-recip-precision
4806 @cindex @code{target("recip-precision")} function attribute, PowerPC
4807 Assume (do not assume) that the reciprocal estimate instructions
4808 provide higher-precision estimates than is mandated by the PowerPC
4813 @cindex @code{target("string")} function attribute, PowerPC
4814 Generate code that uses (does not use) the load string instructions
4815 and the store string word instructions to save multiple registers and
4816 do small block moves.
4820 @cindex @code{target("vsx")} function attribute, PowerPC
4821 Generate code that uses (does not use) vector/scalar (VSX)
4822 instructions, and also enable the use of built-in functions that allow
4823 more direct access to the VSX instruction set. In 32-bit code, you
4824 cannot enable VSX or AltiVec instructions unless
4825 @option{-mabi=altivec} is used on the command line.
4829 @cindex @code{target("friz")} function attribute, PowerPC
4830 Generate (do not generate) the @code{friz} instruction when the
4831 @option{-funsafe-math-optimizations} option is used to optimize
4832 rounding a floating-point value to 64-bit integer and back to floating
4833 point. The @code{friz} instruction does not return the same value if
4834 the floating-point number is too large to fit in an integer.
4836 @item avoid-indexed-addresses
4837 @itemx no-avoid-indexed-addresses
4838 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4839 Generate code that tries to avoid (not avoid) the use of indexed load
4840 or store instructions.
4844 @cindex @code{target("paired")} function attribute, PowerPC
4845 Generate code that uses (does not use) the generation of PAIRED simd
4850 @cindex @code{target("longcall")} function attribute, PowerPC
4851 Generate code that assumes (does not assume) that all calls are far
4852 away so that a longer more expensive calling sequence is required.
4855 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4856 Specify the architecture to generate code for when compiling the
4857 function. If you select the @code{target("cpu=power7")} attribute when
4858 generating 32-bit code, VSX and AltiVec instructions are not generated
4859 unless you use the @option{-mabi=altivec} option on the command line.
4861 @item tune=@var{TUNE}
4862 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4863 Specify the architecture to tune for when compiling the function. If
4864 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4865 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4866 compilation tunes for the @var{CPU} architecture, and not the
4867 default tuning specified on the command line.
4870 On the PowerPC, the inliner does not inline a
4871 function that has different target options than the caller, unless the
4872 callee has a subset of the target options of the caller.
4875 @node RL78 Function Attributes
4876 @subsection RL78 Function Attributes
4878 These function attributes are supported by the RL78 back end:
4882 @itemx brk_interrupt
4883 @cindex @code{interrupt} function attribute, RL78
4884 @cindex @code{brk_interrupt} function attribute, RL78
4885 These attributes indicate
4886 that the specified function is an interrupt handler. The compiler generates
4887 function entry and exit sequences suitable for use in an interrupt handler
4888 when this attribute is present.
4890 Use @code{brk_interrupt} instead of @code{interrupt} for
4891 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4892 that must end with @code{RETB} instead of @code{RETI}).
4895 @cindex @code{naked} function attribute, RL78
4896 This attribute allows the compiler to construct the
4897 requisite function declaration, while allowing the body of the
4898 function to be assembly code. The specified function will not have
4899 prologue/epilogue sequences generated by the compiler. Only basic
4900 @code{asm} statements can safely be included in naked functions
4901 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4902 basic @code{asm} and C code may appear to work, they cannot be
4903 depended upon to work reliably and are not supported.
4906 @node RX Function Attributes
4907 @subsection RX Function Attributes
4909 These function attributes are supported by the RX back end:
4912 @item fast_interrupt
4913 @cindex @code{fast_interrupt} function attribute, RX
4914 Use this attribute on the RX port to indicate that the specified
4915 function is a fast interrupt handler. This is just like the
4916 @code{interrupt} attribute, except that @code{freit} is used to return
4917 instead of @code{reit}.
4920 @cindex @code{interrupt} function attribute, RX
4921 Use this attribute to indicate
4922 that the specified function is an interrupt handler. The compiler generates
4923 function entry and exit sequences suitable for use in an interrupt handler
4924 when this attribute is present.
4926 On RX targets, you may specify one or more vector numbers as arguments
4927 to the attribute, as well as naming an alternate table name.
4928 Parameters are handled sequentially, so one handler can be assigned to
4929 multiple entries in multiple tables. One may also pass the magic
4930 string @code{"$default"} which causes the function to be used for any
4931 unfilled slots in the current table.
4933 This example shows a simple assignment of a function to one vector in
4934 the default table (note that preprocessor macros may be used for
4935 chip-specific symbolic vector names):
4937 void __attribute__ ((interrupt (5))) txd1_handler ();
4940 This example assigns a function to two slots in the default table
4941 (using preprocessor macros defined elsewhere) and makes it the default
4942 for the @code{dct} table:
4944 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4949 @cindex @code{naked} function attribute, RX
4950 This attribute allows the compiler to construct the
4951 requisite function declaration, while allowing the body of the
4952 function to be assembly code. The specified function will not have
4953 prologue/epilogue sequences generated by the compiler. Only basic
4954 @code{asm} statements can safely be included in naked functions
4955 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4956 basic @code{asm} and C code may appear to work, they cannot be
4957 depended upon to work reliably and are not supported.
4960 @cindex @code{vector} function attribute, RX
4961 This RX attribute is similar to the @code{interrupt} attribute, including its
4962 parameters, but does not make the function an interrupt-handler type
4963 function (i.e. it retains the normal C function calling ABI). See the
4964 @code{interrupt} attribute for a description of its arguments.
4967 @node S/390 Function Attributes
4968 @subsection S/390 Function Attributes
4970 These function attributes are supported on the S/390:
4973 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4974 @cindex @code{hotpatch} function attribute, S/390
4976 On S/390 System z targets, you can use this function attribute to
4977 make GCC generate a ``hot-patching'' function prologue. If the
4978 @option{-mhotpatch=} command-line option is used at the same time,
4979 the @code{hotpatch} attribute takes precedence. The first of the
4980 two arguments specifies the number of halfwords to be added before
4981 the function label. A second argument can be used to specify the
4982 number of halfwords to be added after the function label. For
4983 both arguments the maximum allowed value is 1000000.
4985 If both arguments are zero, hotpatching is disabled.
4987 @item target (@var{options})
4988 @cindex @code{target} function attribute
4989 As discussed in @ref{Common Function Attributes}, this attribute
4990 allows specification of target-specific compilation options.
4992 On S/390, the following options are supported:
5000 @item warn-framesize=
5012 @itemx no-packed-stack
5014 @itemx no-small-exec
5017 @item warn-dynamicstack
5018 @itemx no-warn-dynamicstack
5021 The options work exactly like the S/390 specific command line
5022 options (without the prefix @option{-m}) except that they do not
5023 change any feature macros. For example,
5026 @code{target("no-vx")}
5029 does not undefine the @code{__VEC__} macro.
5032 @node SH Function Attributes
5033 @subsection SH Function Attributes
5035 These function attributes are supported on the SH family of processors:
5038 @item function_vector
5039 @cindex @code{function_vector} function attribute, SH
5040 @cindex calling functions through the function vector on SH2A
5041 On SH2A targets, this attribute declares a function to be called using the
5042 TBR relative addressing mode. The argument to this attribute is the entry
5043 number of the same function in a vector table containing all the TBR
5044 relative addressable functions. For correct operation the TBR must be setup
5045 accordingly to point to the start of the vector table before any functions with
5046 this attribute are invoked. Usually a good place to do the initialization is
5047 the startup routine. The TBR relative vector table can have at max 256 function
5048 entries. The jumps to these functions are generated using a SH2A specific,
5049 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5050 from GNU binutils version 2.7 or later for this attribute to work correctly.
5052 In an application, for a function being called once, this attribute
5053 saves at least 8 bytes of code; and if other successive calls are being
5054 made to the same function, it saves 2 bytes of code per each of these
5057 @item interrupt_handler
5058 @cindex @code{interrupt_handler} function attribute, SH
5059 Use this attribute to
5060 indicate that the specified function is an interrupt handler. The compiler
5061 generates function entry and exit sequences suitable for use in an
5062 interrupt handler when this attribute is present.
5064 @item nosave_low_regs
5065 @cindex @code{nosave_low_regs} function attribute, SH
5066 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5067 function should not save and restore registers R0..R7. This can be used on SH3*
5068 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5072 @cindex @code{renesas} function attribute, SH
5073 On SH targets this attribute specifies that the function or struct follows the
5077 @cindex @code{resbank} function attribute, SH
5078 On the SH2A target, this attribute enables the high-speed register
5079 saving and restoration using a register bank for @code{interrupt_handler}
5080 routines. Saving to the bank is performed automatically after the CPU
5081 accepts an interrupt that uses a register bank.
5083 The nineteen 32-bit registers comprising general register R0 to R14,
5084 control register GBR, and system registers MACH, MACL, and PR and the
5085 vector table address offset are saved into a register bank. Register
5086 banks are stacked in first-in last-out (FILO) sequence. Restoration
5087 from the bank is executed by issuing a RESBANK instruction.
5090 @cindex @code{sp_switch} function attribute, SH
5091 Use this attribute on the SH to indicate an @code{interrupt_handler}
5092 function should switch to an alternate stack. It expects a string
5093 argument that names a global variable holding the address of the
5098 void f () __attribute__ ((interrupt_handler,
5099 sp_switch ("alt_stack")));
5103 @cindex @code{trap_exit} function attribute, SH
5104 Use this attribute on the SH for an @code{interrupt_handler} to return using
5105 @code{trapa} instead of @code{rte}. This attribute expects an integer
5106 argument specifying the trap number to be used.
5109 @cindex @code{trapa_handler} function attribute, SH
5110 On SH targets this function attribute is similar to @code{interrupt_handler}
5111 but it does not save and restore all registers.
5114 @node SPU Function Attributes
5115 @subsection SPU Function Attributes
5117 These function attributes are supported by the SPU back end:
5121 @cindex @code{naked} function attribute, SPU
5122 This attribute allows the compiler to construct the
5123 requisite function declaration, while allowing the body of the
5124 function to be assembly code. The specified function will not have
5125 prologue/epilogue sequences generated by the compiler. Only basic
5126 @code{asm} statements can safely be included in naked functions
5127 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5128 basic @code{asm} and C code may appear to work, they cannot be
5129 depended upon to work reliably and are not supported.
5132 @node Symbian OS Function Attributes
5133 @subsection Symbian OS Function Attributes
5135 @xref{Microsoft Windows Function Attributes}, for discussion of the
5136 @code{dllexport} and @code{dllimport} attributes.
5138 @node V850 Function Attributes
5139 @subsection V850 Function Attributes
5141 The V850 back end supports these function attributes:
5145 @itemx interrupt_handler
5146 @cindex @code{interrupt} function attribute, V850
5147 @cindex @code{interrupt_handler} function attribute, V850
5148 Use these attributes to indicate
5149 that the specified function is an interrupt handler. The compiler generates
5150 function entry and exit sequences suitable for use in an interrupt handler
5151 when either attribute is present.
5154 @node Visium Function Attributes
5155 @subsection Visium Function Attributes
5157 These function attributes are supported by the Visium back end:
5161 @cindex @code{interrupt} function attribute, Visium
5162 Use this attribute to indicate
5163 that the specified function is an interrupt handler. The compiler generates
5164 function entry and exit sequences suitable for use in an interrupt handler
5165 when this attribute is present.
5168 @node x86 Function Attributes
5169 @subsection x86 Function Attributes
5171 These function attributes are supported by the x86 back end:
5175 @cindex @code{cdecl} function attribute, x86-32
5176 @cindex functions that pop the argument stack on x86-32
5178 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5179 assume that the calling function pops off the stack space used to
5180 pass arguments. This is
5181 useful to override the effects of the @option{-mrtd} switch.
5184 @cindex @code{fastcall} function attribute, x86-32
5185 @cindex functions that pop the argument stack on x86-32
5186 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5187 pass the first argument (if of integral type) in the register ECX and
5188 the second argument (if of integral type) in the register EDX@. Subsequent
5189 and other typed arguments are passed on the stack. The called function
5190 pops the arguments off the stack. If the number of arguments is variable all
5191 arguments are pushed on the stack.
5194 @cindex @code{thiscall} function attribute, x86-32
5195 @cindex functions that pop the argument stack on x86-32
5196 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5197 pass the first argument (if of integral type) in the register ECX.
5198 Subsequent and other typed arguments are passed on the stack. The called
5199 function pops the arguments off the stack.
5200 If the number of arguments is variable all arguments are pushed on the
5202 The @code{thiscall} attribute is intended for C++ non-static member functions.
5203 As a GCC extension, this calling convention can be used for C functions
5204 and for static member methods.
5208 @cindex @code{ms_abi} function attribute, x86
5209 @cindex @code{sysv_abi} function attribute, x86
5211 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5212 to indicate which calling convention should be used for a function. The
5213 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5214 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5215 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5216 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5218 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5219 requires the @option{-maccumulate-outgoing-args} option.
5221 @item callee_pop_aggregate_return (@var{number})
5222 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5224 On x86-32 targets, you can use this attribute to control how
5225 aggregates are returned in memory. If the caller is responsible for
5226 popping the hidden pointer together with the rest of the arguments, specify
5227 @var{number} equal to zero. If callee is responsible for popping the
5228 hidden pointer, specify @var{number} equal to one.
5230 The default x86-32 ABI assumes that the callee pops the
5231 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5232 the compiler assumes that the
5233 caller pops the stack for hidden pointer.
5235 @item ms_hook_prologue
5236 @cindex @code{ms_hook_prologue} function attribute, x86
5238 On 32-bit and 64-bit x86 targets, you can use
5239 this function attribute to make GCC generate the ``hot-patching'' function
5240 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5243 @item regparm (@var{number})
5244 @cindex @code{regparm} function attribute, x86
5245 @cindex functions that are passed arguments in registers on x86-32
5246 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5247 pass arguments number one to @var{number} if they are of integral type
5248 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5249 take a variable number of arguments continue to be passed all of their
5250 arguments on the stack.
5252 Beware that on some ELF systems this attribute is unsuitable for
5253 global functions in shared libraries with lazy binding (which is the
5254 default). Lazy binding sends the first call via resolving code in
5255 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5256 per the standard calling conventions. Solaris 8 is affected by this.
5257 Systems with the GNU C Library version 2.1 or higher
5258 and FreeBSD are believed to be
5259 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5260 disabled with the linker or the loader if desired, to avoid the
5264 @cindex @code{sseregparm} function attribute, x86
5265 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5266 causes the compiler to pass up to 3 floating-point arguments in
5267 SSE registers instead of on the stack. Functions that take a
5268 variable number of arguments continue to pass all of their
5269 floating-point arguments on the stack.
5271 @item force_align_arg_pointer
5272 @cindex @code{force_align_arg_pointer} function attribute, x86
5273 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5274 applied to individual function definitions, generating an alternate
5275 prologue and epilogue that realigns the run-time stack if necessary.
5276 This supports mixing legacy codes that run with a 4-byte aligned stack
5277 with modern codes that keep a 16-byte stack for SSE compatibility.
5280 @cindex @code{stdcall} function attribute, x86-32
5281 @cindex functions that pop the argument stack on x86-32
5282 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5283 assume that the called function pops off the stack space used to
5284 pass arguments, unless it takes a variable number of arguments.
5286 @item no_caller_saved_registers
5287 @cindex @code{no_caller_saved_registers} function attribute, x86
5288 Use this attribute to indicate that the specified function has no
5289 caller-saved registers. That is, all registers are callee-saved. For
5290 example, this attribute can be used for a function called from an
5291 interrupt handler. The compiler generates proper function entry and
5292 exit sequences to save and restore any modified registers, except for
5293 the EFLAGS register. Since GCC doesn't preserve MPX, SSE, MMX nor x87
5294 states, the GCC option @option{-mgeneral-regs-only} should be used to
5295 compile functions with @code{no_caller_saved_registers} attribute.
5298 @cindex @code{interrupt} function attribute, x86
5299 Use this attribute to indicate that the specified function is an
5300 interrupt handler or an exception handler (depending on parameters passed
5301 to the function, explained further). The compiler generates function
5302 entry and exit sequences suitable for use in an interrupt handler when
5303 this attribute is present. The @code{IRET} instruction, instead of the
5304 @code{RET} instruction, is used to return from interrupt handlers. All
5305 registers, except for the EFLAGS register which is restored by the
5306 @code{IRET} instruction, are preserved by the compiler. Since GCC
5307 doesn't preserve MPX, SSE, MMX nor x87 states, the GCC option
5308 @option{-mgeneral-regs-only} should be used to compile interrupt and
5311 Any interruptible-without-stack-switch code must be compiled with
5312 @option{-mno-red-zone} since interrupt handlers can and will, because
5313 of the hardware design, touch the red zone.
5315 An interrupt handler must be declared with a mandatory pointer
5319 struct interrupt_frame;
5321 __attribute__ ((interrupt))
5323 f (struct interrupt_frame *frame)
5329 and you must define @code{struct interrupt_frame} as described in the
5332 Exception handlers differ from interrupt handlers because the system
5333 pushes an error code on the stack. An exception handler declaration is
5334 similar to that for an interrupt handler, but with a different mandatory
5335 function signature. The compiler arranges to pop the error code off the
5336 stack before the @code{IRET} instruction.
5340 typedef unsigned long long int uword_t;
5342 typedef unsigned int uword_t;
5345 struct interrupt_frame;
5347 __attribute__ ((interrupt))
5349 f (struct interrupt_frame *frame, uword_t error_code)
5355 Exception handlers should only be used for exceptions that push an error
5356 code; you should use an interrupt handler in other cases. The system
5357 will crash if the wrong kind of handler is used.
5359 @item target (@var{options})
5360 @cindex @code{target} function attribute
5361 As discussed in @ref{Common Function Attributes}, this attribute
5362 allows specification of target-specific compilation options.
5364 On the x86, the following options are allowed:
5368 @cindex @code{target("abm")} function attribute, x86
5369 Enable/disable the generation of the advanced bit instructions.
5373 @cindex @code{target("aes")} function attribute, x86
5374 Enable/disable the generation of the AES instructions.
5377 @cindex @code{target("default")} function attribute, x86
5378 @xref{Function Multiversioning}, where it is used to specify the
5379 default function version.
5383 @cindex @code{target("mmx")} function attribute, x86
5384 Enable/disable the generation of the MMX instructions.
5388 @cindex @code{target("pclmul")} function attribute, x86
5389 Enable/disable the generation of the PCLMUL instructions.
5393 @cindex @code{target("popcnt")} function attribute, x86
5394 Enable/disable the generation of the POPCNT instruction.
5398 @cindex @code{target("sse")} function attribute, x86
5399 Enable/disable the generation of the SSE instructions.
5403 @cindex @code{target("sse2")} function attribute, x86
5404 Enable/disable the generation of the SSE2 instructions.
5408 @cindex @code{target("sse3")} function attribute, x86
5409 Enable/disable the generation of the SSE3 instructions.
5413 @cindex @code{target("sse4")} function attribute, x86
5414 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5419 @cindex @code{target("sse4.1")} function attribute, x86
5420 Enable/disable the generation of the sse4.1 instructions.
5424 @cindex @code{target("sse4.2")} function attribute, x86
5425 Enable/disable the generation of the sse4.2 instructions.
5429 @cindex @code{target("sse4a")} function attribute, x86
5430 Enable/disable the generation of the SSE4A instructions.
5434 @cindex @code{target("fma4")} function attribute, x86
5435 Enable/disable the generation of the FMA4 instructions.
5439 @cindex @code{target("xop")} function attribute, x86
5440 Enable/disable the generation of the XOP instructions.
5444 @cindex @code{target("lwp")} function attribute, x86
5445 Enable/disable the generation of the LWP instructions.
5449 @cindex @code{target("ssse3")} function attribute, x86
5450 Enable/disable the generation of the SSSE3 instructions.
5454 @cindex @code{target("cld")} function attribute, x86
5455 Enable/disable the generation of the CLD before string moves.
5457 @item fancy-math-387
5458 @itemx no-fancy-math-387
5459 @cindex @code{target("fancy-math-387")} function attribute, x86
5460 Enable/disable the generation of the @code{sin}, @code{cos}, and
5461 @code{sqrt} instructions on the 387 floating-point unit.
5464 @itemx no-fused-madd
5465 @cindex @code{target("fused-madd")} function attribute, x86
5466 Enable/disable the generation of the fused multiply/add instructions.
5470 @cindex @code{target("ieee-fp")} function attribute, x86
5471 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5473 @item inline-all-stringops
5474 @itemx no-inline-all-stringops
5475 @cindex @code{target("inline-all-stringops")} function attribute, x86
5476 Enable/disable inlining of string operations.
5478 @item inline-stringops-dynamically
5479 @itemx no-inline-stringops-dynamically
5480 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5481 Enable/disable the generation of the inline code to do small string
5482 operations and calling the library routines for large operations.
5484 @item align-stringops
5485 @itemx no-align-stringops
5486 @cindex @code{target("align-stringops")} function attribute, x86
5487 Do/do not align destination of inlined string operations.
5491 @cindex @code{target("recip")} function attribute, x86
5492 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5493 instructions followed an additional Newton-Raphson step instead of
5494 doing a floating-point division.
5496 @item arch=@var{ARCH}
5497 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5498 Specify the architecture to generate code for in compiling the function.
5500 @item tune=@var{TUNE}
5501 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5502 Specify the architecture to tune for in compiling the function.
5504 @item fpmath=@var{FPMATH}
5505 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5506 Specify which floating-point unit to use. You must specify the
5507 @code{target("fpmath=sse,387")} option as
5508 @code{target("fpmath=sse+387")} because the comma would separate
5512 On the x86, the inliner does not inline a
5513 function that has different target options than the caller, unless the
5514 callee has a subset of the target options of the caller. For example
5515 a function declared with @code{target("sse3")} can inline a function
5516 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5519 @node Xstormy16 Function Attributes
5520 @subsection Xstormy16 Function Attributes
5522 These function attributes are supported by the Xstormy16 back end:
5526 @cindex @code{interrupt} function attribute, Xstormy16
5527 Use this attribute to indicate
5528 that the specified function is an interrupt handler. The compiler generates
5529 function entry and exit sequences suitable for use in an interrupt handler
5530 when this attribute is present.
5533 @node Variable Attributes
5534 @section Specifying Attributes of Variables
5535 @cindex attribute of variables
5536 @cindex variable attributes
5538 The keyword @code{__attribute__} allows you to specify special
5539 attributes of variables or structure fields. This keyword is followed
5540 by an attribute specification inside double parentheses. Some
5541 attributes are currently defined generically for variables.
5542 Other attributes are defined for variables on particular target
5543 systems. Other attributes are available for functions
5544 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5545 enumerators (@pxref{Enumerator Attributes}), and for types
5546 (@pxref{Type Attributes}).
5547 Other front ends might define more attributes
5548 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5550 @xref{Attribute Syntax}, for details of the exact syntax for using
5554 * Common Variable Attributes::
5555 * AVR Variable Attributes::
5556 * Blackfin Variable Attributes::
5557 * H8/300 Variable Attributes::
5558 * IA-64 Variable Attributes::
5559 * M32R/D Variable Attributes::
5560 * MeP Variable Attributes::
5561 * Microsoft Windows Variable Attributes::
5562 * MSP430 Variable Attributes::
5563 * PowerPC Variable Attributes::
5564 * RL78 Variable Attributes::
5565 * SPU Variable Attributes::
5566 * V850 Variable Attributes::
5567 * x86 Variable Attributes::
5568 * Xstormy16 Variable Attributes::
5571 @node Common Variable Attributes
5572 @subsection Common Variable Attributes
5574 The following attributes are supported on most targets.
5577 @cindex @code{aligned} variable attribute
5578 @item aligned (@var{alignment})
5579 This attribute specifies a minimum alignment for the variable or
5580 structure field, measured in bytes. For example, the declaration:
5583 int x __attribute__ ((aligned (16))) = 0;
5587 causes the compiler to allocate the global variable @code{x} on a
5588 16-byte boundary. On a 68040, this could be used in conjunction with
5589 an @code{asm} expression to access the @code{move16} instruction which
5590 requires 16-byte aligned operands.
5592 You can also specify the alignment of structure fields. For example, to
5593 create a double-word aligned @code{int} pair, you could write:
5596 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5600 This is an alternative to creating a union with a @code{double} member,
5601 which forces the union to be double-word aligned.
5603 As in the preceding examples, you can explicitly specify the alignment
5604 (in bytes) that you wish the compiler to use for a given variable or
5605 structure field. Alternatively, you can leave out the alignment factor
5606 and just ask the compiler to align a variable or field to the
5607 default alignment for the target architecture you are compiling for.
5608 The default alignment is sufficient for all scalar types, but may not be
5609 enough for all vector types on a target that supports vector operations.
5610 The default alignment is fixed for a particular target ABI.
5612 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5613 which is the largest alignment ever used for any data type on the
5614 target machine you are compiling for. For example, you could write:
5617 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5620 The compiler automatically sets the alignment for the declared
5621 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5622 often make copy operations more efficient, because the compiler can
5623 use whatever instructions copy the biggest chunks of memory when
5624 performing copies to or from the variables or fields that you have
5625 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5626 may change depending on command-line options.
5628 When used on a struct, or struct member, the @code{aligned} attribute can
5629 only increase the alignment; in order to decrease it, the @code{packed}
5630 attribute must be specified as well. When used as part of a typedef, the
5631 @code{aligned} attribute can both increase and decrease alignment, and
5632 specifying the @code{packed} attribute generates a warning.
5634 Note that the effectiveness of @code{aligned} attributes may be limited
5635 by inherent limitations in your linker. On many systems, the linker is
5636 only able to arrange for variables to be aligned up to a certain maximum
5637 alignment. (For some linkers, the maximum supported alignment may
5638 be very very small.) If your linker is only able to align variables
5639 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5640 in an @code{__attribute__} still only provides you with 8-byte
5641 alignment. See your linker documentation for further information.
5643 The @code{aligned} attribute can also be used for functions
5644 (@pxref{Common Function Attributes}.)
5646 @item cleanup (@var{cleanup_function})
5647 @cindex @code{cleanup} variable attribute
5648 The @code{cleanup} attribute runs a function when the variable goes
5649 out of scope. This attribute can only be applied to auto function
5650 scope variables; it may not be applied to parameters or variables
5651 with static storage duration. The function must take one parameter,
5652 a pointer to a type compatible with the variable. The return value
5653 of the function (if any) is ignored.
5655 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5656 is run during the stack unwinding that happens during the
5657 processing of the exception. Note that the @code{cleanup} attribute
5658 does not allow the exception to be caught, only to perform an action.
5659 It is undefined what happens if @var{cleanup_function} does not
5664 @cindex @code{common} variable attribute
5665 @cindex @code{nocommon} variable attribute
5668 The @code{common} attribute requests GCC to place a variable in
5669 ``common'' storage. The @code{nocommon} attribute requests the
5670 opposite---to allocate space for it directly.
5672 These attributes override the default chosen by the
5673 @option{-fno-common} and @option{-fcommon} flags respectively.
5676 @itemx deprecated (@var{msg})
5677 @cindex @code{deprecated} variable attribute
5678 The @code{deprecated} attribute results in a warning if the variable
5679 is used anywhere in the source file. This is useful when identifying
5680 variables that are expected to be removed in a future version of a
5681 program. The warning also includes the location of the declaration
5682 of the deprecated variable, to enable users to easily find further
5683 information about why the variable is deprecated, or what they should
5684 do instead. Note that the warning only occurs for uses:
5687 extern int old_var __attribute__ ((deprecated));
5689 int new_fn () @{ return old_var; @}
5693 results in a warning on line 3 but not line 2. The optional @var{msg}
5694 argument, which must be a string, is printed in the warning if
5697 The @code{deprecated} attribute can also be used for functions and
5698 types (@pxref{Common Function Attributes},
5699 @pxref{Common Type Attributes}).
5701 @item mode (@var{mode})
5702 @cindex @code{mode} variable attribute
5703 This attribute specifies the data type for the declaration---whichever
5704 type corresponds to the mode @var{mode}. This in effect lets you
5705 request an integer or floating-point type according to its width.
5707 You may also specify a mode of @code{byte} or @code{__byte__} to
5708 indicate the mode corresponding to a one-byte integer, @code{word} or
5709 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5710 or @code{__pointer__} for the mode used to represent pointers.
5713 @cindex @code{packed} variable attribute
5714 The @code{packed} attribute specifies that a variable or structure field
5715 should have the smallest possible alignment---one byte for a variable,
5716 and one bit for a field, unless you specify a larger value with the
5717 @code{aligned} attribute.
5719 Here is a structure in which the field @code{x} is packed, so that it
5720 immediately follows @code{a}:
5726 int x[2] __attribute__ ((packed));
5730 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5731 @code{packed} attribute on bit-fields of type @code{char}. This has
5732 been fixed in GCC 4.4 but the change can lead to differences in the
5733 structure layout. See the documentation of
5734 @option{-Wpacked-bitfield-compat} for more information.
5736 @item section ("@var{section-name}")
5737 @cindex @code{section} variable attribute
5738 Normally, the compiler places the objects it generates in sections like
5739 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5740 or you need certain particular variables to appear in special sections,
5741 for example to map to special hardware. The @code{section}
5742 attribute specifies that a variable (or function) lives in a particular
5743 section. For example, this small program uses several specific section names:
5746 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5747 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5748 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5749 int init_data __attribute__ ((section ("INITDATA")));
5753 /* @r{Initialize stack pointer} */
5754 init_sp (stack + sizeof (stack));
5756 /* @r{Initialize initialized data} */
5757 memcpy (&init_data, &data, &edata - &data);
5759 /* @r{Turn on the serial ports} */
5766 Use the @code{section} attribute with
5767 @emph{global} variables and not @emph{local} variables,
5768 as shown in the example.
5770 You may use the @code{section} attribute with initialized or
5771 uninitialized global variables but the linker requires
5772 each object be defined once, with the exception that uninitialized
5773 variables tentatively go in the @code{common} (or @code{bss}) section
5774 and can be multiply ``defined''. Using the @code{section} attribute
5775 changes what section the variable goes into and may cause the
5776 linker to issue an error if an uninitialized variable has multiple
5777 definitions. You can force a variable to be initialized with the
5778 @option{-fno-common} flag or the @code{nocommon} attribute.
5780 Some file formats do not support arbitrary sections so the @code{section}
5781 attribute is not available on all platforms.
5782 If you need to map the entire contents of a module to a particular
5783 section, consider using the facilities of the linker instead.
5785 @item tls_model ("@var{tls_model}")
5786 @cindex @code{tls_model} variable attribute
5787 The @code{tls_model} attribute sets thread-local storage model
5788 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5789 overriding @option{-ftls-model=} command-line switch on a per-variable
5791 The @var{tls_model} argument should be one of @code{global-dynamic},
5792 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5794 Not all targets support this attribute.
5797 @cindex @code{unused} variable attribute
5798 This attribute, attached to a variable, means that the variable is meant
5799 to be possibly unused. GCC does not produce a warning for this
5803 @cindex @code{used} variable attribute
5804 This attribute, attached to a variable with static storage, means that
5805 the variable must be emitted even if it appears that the variable is not
5808 When applied to a static data member of a C++ class template, the
5809 attribute also means that the member is instantiated if the
5810 class itself is instantiated.
5812 @item vector_size (@var{bytes})
5813 @cindex @code{vector_size} variable attribute
5814 This attribute specifies the vector size for the variable, measured in
5815 bytes. For example, the declaration:
5818 int foo __attribute__ ((vector_size (16)));
5822 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5823 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5824 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5826 This attribute is only applicable to integral and float scalars,
5827 although arrays, pointers, and function return values are allowed in
5828 conjunction with this construct.
5830 Aggregates with this attribute are invalid, even if they are of the same
5831 size as a corresponding scalar. For example, the declaration:
5834 struct S @{ int a; @};
5835 struct S __attribute__ ((vector_size (16))) foo;
5839 is invalid even if the size of the structure is the same as the size of
5842 @item visibility ("@var{visibility_type}")
5843 @cindex @code{visibility} variable attribute
5844 This attribute affects the linkage of the declaration to which it is attached.
5845 The @code{visibility} attribute is described in
5846 @ref{Common Function Attributes}.
5849 @cindex @code{weak} variable attribute
5850 The @code{weak} attribute is described in
5851 @ref{Common Function Attributes}.
5855 @node AVR Variable Attributes
5856 @subsection AVR Variable Attributes
5860 @cindex @code{progmem} variable attribute, AVR
5861 The @code{progmem} attribute is used on the AVR to place read-only
5862 data in the non-volatile program memory (flash). The @code{progmem}
5863 attribute accomplishes this by putting respective variables into a
5864 section whose name starts with @code{.progmem}.
5866 This attribute works similar to the @code{section} attribute
5867 but adds additional checking.
5870 @item @bullet{}@tie{} Ordinary AVR cores with 32 general purpose registers:
5871 @code{progmem} affects the location
5872 of the data but not how this data is accessed.
5873 In order to read data located with the @code{progmem} attribute
5874 (inline) assembler must be used.
5876 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5877 #include <avr/pgmspace.h>
5879 /* Locate var in flash memory */
5880 const int var[2] PROGMEM = @{ 1, 2 @};
5882 int read_var (int i)
5884 /* Access var[] by accessor macro from avr/pgmspace.h */
5885 return (int) pgm_read_word (& var[i]);
5889 AVR is a Harvard architecture processor and data and read-only data
5890 normally resides in the data memory (RAM).
5892 See also the @ref{AVR Named Address Spaces} section for
5893 an alternate way to locate and access data in flash memory.
5895 @item @bullet{}@tie{}Reduced AVR Tiny cores like ATtiny40:
5896 The compiler adds @code{0x4000}
5897 to the addresses of objects and declarations in @code{progmem} and locates
5898 the objects in flash memory, namely in section @code{.progmem.data}.
5899 The offset is needed because the flash memory is visible in the RAM
5900 address space starting at address @code{0x4000}.
5902 Data in @code{progmem} can be accessed by means of ordinary C@tie{}code,
5903 no special functions or macros are needed.
5906 /* var is located in flash memory */
5907 extern const int var[2] __attribute__((progmem));
5909 int read_var (int i)
5918 @itemx io (@var{addr})
5919 @cindex @code{io} variable attribute, AVR
5920 Variables with the @code{io} attribute are used to address
5921 memory-mapped peripherals in the io address range.
5922 If an address is specified, the variable
5923 is assigned that address, and the value is interpreted as an
5924 address in the data address space.
5928 volatile int porta __attribute__((io (0x22)));
5931 The address specified in the address in the data address range.
5933 Otherwise, the variable it is not assigned an address, but the
5934 compiler will still use in/out instructions where applicable,
5935 assuming some other module assigns an address in the io address range.
5939 extern volatile int porta __attribute__((io));
5943 @itemx io_low (@var{addr})
5944 @cindex @code{io_low} variable attribute, AVR
5945 This is like the @code{io} attribute, but additionally it informs the
5946 compiler that the object lies in the lower half of the I/O area,
5947 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5951 @itemx address (@var{addr})
5952 @cindex @code{address} variable attribute, AVR
5953 Variables with the @code{address} attribute are used to address
5954 memory-mapped peripherals that may lie outside the io address range.
5957 volatile int porta __attribute__((address (0x600)));
5962 @node Blackfin Variable Attributes
5963 @subsection Blackfin Variable Attributes
5965 Three attributes are currently defined for the Blackfin.
5971 @cindex @code{l1_data} variable attribute, Blackfin
5972 @cindex @code{l1_data_A} variable attribute, Blackfin
5973 @cindex @code{l1_data_B} variable attribute, Blackfin
5974 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5975 Variables with @code{l1_data} attribute are put into the specific section
5976 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5977 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5978 attribute are put into the specific section named @code{.l1.data.B}.
5981 @cindex @code{l2} variable attribute, Blackfin
5982 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5983 Variables with @code{l2} attribute are put into the specific section
5984 named @code{.l2.data}.
5987 @node H8/300 Variable Attributes
5988 @subsection H8/300 Variable Attributes
5990 These variable attributes are available for H8/300 targets:
5994 @cindex @code{eightbit_data} variable attribute, H8/300
5995 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5996 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5997 variable should be placed into the eight-bit data section.
5998 The compiler generates more efficient code for certain operations
5999 on data in the eight-bit data area. Note the eight-bit data area is limited to
6002 You must use GAS and GLD from GNU binutils version 2.7 or later for
6003 this attribute to work correctly.
6006 @cindex @code{tiny_data} variable attribute, H8/300
6007 @cindex tiny data section on the H8/300H and H8S
6008 Use this attribute on the H8/300H and H8S to indicate that the specified
6009 variable should be placed into the tiny data section.
6010 The compiler generates more efficient code for loads and stores
6011 on data in the tiny data section. Note the tiny data area is limited to
6012 slightly under 32KB of data.
6016 @node IA-64 Variable Attributes
6017 @subsection IA-64 Variable Attributes
6019 The IA-64 back end supports the following variable attribute:
6022 @item model (@var{model-name})
6023 @cindex @code{model} variable attribute, IA-64
6025 On IA-64, use this attribute to set the addressability of an object.
6026 At present, the only supported identifier for @var{model-name} is
6027 @code{small}, indicating addressability via ``small'' (22-bit)
6028 addresses (so that their addresses can be loaded with the @code{addl}
6029 instruction). Caveat: such addressing is by definition not position
6030 independent and hence this attribute must not be used for objects
6031 defined by shared libraries.
6035 @node M32R/D Variable Attributes
6036 @subsection M32R/D Variable Attributes
6038 One attribute is currently defined for the M32R/D@.
6041 @item model (@var{model-name})
6042 @cindex @code{model-name} variable attribute, M32R/D
6043 @cindex variable addressability on the M32R/D
6044 Use this attribute on the M32R/D to set the addressability of an object.
6045 The identifier @var{model-name} is one of @code{small}, @code{medium},
6046 or @code{large}, representing each of the code models.
6048 Small model objects live in the lower 16MB of memory (so that their
6049 addresses can be loaded with the @code{ld24} instruction).
6051 Medium and large model objects may live anywhere in the 32-bit address space
6052 (the compiler generates @code{seth/add3} instructions to load their
6056 @node MeP Variable Attributes
6057 @subsection MeP Variable Attributes
6059 The MeP target has a number of addressing modes and busses. The
6060 @code{near} space spans the standard memory space's first 16 megabytes
6061 (24 bits). The @code{far} space spans the entire 32-bit memory space.
6062 The @code{based} space is a 128-byte region in the memory space that
6063 is addressed relative to the @code{$tp} register. The @code{tiny}
6064 space is a 65536-byte region relative to the @code{$gp} register. In
6065 addition to these memory regions, the MeP target has a separate 16-bit
6066 control bus which is specified with @code{cb} attributes.
6071 @cindex @code{based} variable attribute, MeP
6072 Any variable with the @code{based} attribute is assigned to the
6073 @code{.based} section, and is accessed with relative to the
6074 @code{$tp} register.
6077 @cindex @code{tiny} variable attribute, MeP
6078 Likewise, the @code{tiny} attribute assigned variables to the
6079 @code{.tiny} section, relative to the @code{$gp} register.
6082 @cindex @code{near} variable attribute, MeP
6083 Variables with the @code{near} attribute are assumed to have addresses
6084 that fit in a 24-bit addressing mode. This is the default for large
6085 variables (@code{-mtiny=4} is the default) but this attribute can
6086 override @code{-mtiny=} for small variables, or override @code{-ml}.
6089 @cindex @code{far} variable attribute, MeP
6090 Variables with the @code{far} attribute are addressed using a full
6091 32-bit address. Since this covers the entire memory space, this
6092 allows modules to make no assumptions about where variables might be
6096 @cindex @code{io} variable attribute, MeP
6097 @itemx io (@var{addr})
6098 Variables with the @code{io} attribute are used to address
6099 memory-mapped peripherals. If an address is specified, the variable
6100 is assigned that address, else it is not assigned an address (it is
6101 assumed some other module assigns an address). Example:
6104 int timer_count __attribute__((io(0x123)));
6108 @itemx cb (@var{addr})
6109 @cindex @code{cb} variable attribute, MeP
6110 Variables with the @code{cb} attribute are used to access the control
6111 bus, using special instructions. @code{addr} indicates the control bus
6115 int cpu_clock __attribute__((cb(0x123)));
6120 @node Microsoft Windows Variable Attributes
6121 @subsection Microsoft Windows Variable Attributes
6123 You can use these attributes on Microsoft Windows targets.
6124 @ref{x86 Variable Attributes} for additional Windows compatibility
6125 attributes available on all x86 targets.
6130 @cindex @code{dllimport} variable attribute
6131 @cindex @code{dllexport} variable attribute
6132 The @code{dllimport} and @code{dllexport} attributes are described in
6133 @ref{Microsoft Windows Function Attributes}.
6136 @cindex @code{selectany} variable attribute
6137 The @code{selectany} attribute causes an initialized global variable to
6138 have link-once semantics. When multiple definitions of the variable are
6139 encountered by the linker, the first is selected and the remainder are
6140 discarded. Following usage by the Microsoft compiler, the linker is told
6141 @emph{not} to warn about size or content differences of the multiple
6144 Although the primary usage of this attribute is for POD types, the
6145 attribute can also be applied to global C++ objects that are initialized
6146 by a constructor. In this case, the static initialization and destruction
6147 code for the object is emitted in each translation defining the object,
6148 but the calls to the constructor and destructor are protected by a
6149 link-once guard variable.
6151 The @code{selectany} attribute is only available on Microsoft Windows
6152 targets. You can use @code{__declspec (selectany)} as a synonym for
6153 @code{__attribute__ ((selectany))} for compatibility with other
6157 @cindex @code{shared} variable attribute
6158 On Microsoft Windows, in addition to putting variable definitions in a named
6159 section, the section can also be shared among all running copies of an
6160 executable or DLL@. For example, this small program defines shared data
6161 by putting it in a named section @code{shared} and marking the section
6165 int foo __attribute__((section ("shared"), shared)) = 0;
6170 /* @r{Read and write foo. All running
6171 copies see the same value.} */
6177 You may only use the @code{shared} attribute along with @code{section}
6178 attribute with a fully-initialized global definition because of the way
6179 linkers work. See @code{section} attribute for more information.
6181 The @code{shared} attribute is only available on Microsoft Windows@.
6185 @node MSP430 Variable Attributes
6186 @subsection MSP430 Variable Attributes
6190 @cindex @code{noinit} variable attribute, MSP430
6191 Any data with the @code{noinit} attribute will not be initialised by
6192 the C runtime startup code, or the program loader. Not initialising
6193 data in this way can reduce program startup times.
6196 @cindex @code{persistent} variable attribute, MSP430
6197 Any variable with the @code{persistent} attribute will not be
6198 initialised by the C runtime startup code. Instead its value will be
6199 set once, when the application is loaded, and then never initialised
6200 again, even if the processor is reset or the program restarts.
6201 Persistent data is intended to be placed into FLASH RAM, where its
6202 value will be retained across resets. The linker script being used to
6203 create the application should ensure that persistent data is correctly
6209 @cindex @code{lower} variable attribute, MSP430
6210 @cindex @code{upper} variable attribute, MSP430
6211 @cindex @code{either} variable attribute, MSP430
6212 These attributes are the same as the MSP430 function attributes of the
6213 same name (@pxref{MSP430 Function Attributes}).
6214 These attributes can be applied to both functions and variables.
6217 @node PowerPC Variable Attributes
6218 @subsection PowerPC Variable Attributes
6220 Three attributes currently are defined for PowerPC configurations:
6221 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6223 @cindex @code{ms_struct} variable attribute, PowerPC
6224 @cindex @code{gcc_struct} variable attribute, PowerPC
6225 For full documentation of the struct attributes please see the
6226 documentation in @ref{x86 Variable Attributes}.
6228 @cindex @code{altivec} variable attribute, PowerPC
6229 For documentation of @code{altivec} attribute please see the
6230 documentation in @ref{PowerPC Type Attributes}.
6232 @node RL78 Variable Attributes
6233 @subsection RL78 Variable Attributes
6235 @cindex @code{saddr} variable attribute, RL78
6236 The RL78 back end supports the @code{saddr} variable attribute. This
6237 specifies placement of the corresponding variable in the SADDR area,
6238 which can be accessed more efficiently than the default memory region.
6240 @node SPU Variable Attributes
6241 @subsection SPU Variable Attributes
6243 @cindex @code{spu_vector} variable attribute, SPU
6244 The SPU supports the @code{spu_vector} attribute for variables. For
6245 documentation of this attribute please see the documentation in
6246 @ref{SPU Type Attributes}.
6248 @node V850 Variable Attributes
6249 @subsection V850 Variable Attributes
6251 These variable attributes are supported by the V850 back end:
6256 @cindex @code{sda} variable attribute, V850
6257 Use this attribute to explicitly place a variable in the small data area,
6258 which can hold up to 64 kilobytes.
6261 @cindex @code{tda} variable attribute, V850
6262 Use this attribute to explicitly place a variable in the tiny data area,
6263 which can hold up to 256 bytes in total.
6266 @cindex @code{zda} variable attribute, V850
6267 Use this attribute to explicitly place a variable in the first 32 kilobytes
6271 @node x86 Variable Attributes
6272 @subsection x86 Variable Attributes
6274 Two attributes are currently defined for x86 configurations:
6275 @code{ms_struct} and @code{gcc_struct}.
6280 @cindex @code{ms_struct} variable attribute, x86
6281 @cindex @code{gcc_struct} variable attribute, x86
6283 If @code{packed} is used on a structure, or if bit-fields are used,
6284 it may be that the Microsoft ABI lays out the structure differently
6285 than the way GCC normally does. Particularly when moving packed
6286 data between functions compiled with GCC and the native Microsoft compiler
6287 (either via function call or as data in a file), it may be necessary to access
6290 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6291 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6292 command-line options, respectively;
6293 see @ref{x86 Options}, for details of how structure layout is affected.
6294 @xref{x86 Type Attributes}, for information about the corresponding
6295 attributes on types.
6299 @node Xstormy16 Variable Attributes
6300 @subsection Xstormy16 Variable Attributes
6302 One attribute is currently defined for xstormy16 configurations:
6307 @cindex @code{below100} variable attribute, Xstormy16
6309 If a variable has the @code{below100} attribute (@code{BELOW100} is
6310 allowed also), GCC places the variable in the first 0x100 bytes of
6311 memory and use special opcodes to access it. Such variables are
6312 placed in either the @code{.bss_below100} section or the
6313 @code{.data_below100} section.
6317 @node Type Attributes
6318 @section Specifying Attributes of Types
6319 @cindex attribute of types
6320 @cindex type attributes
6322 The keyword @code{__attribute__} allows you to specify special
6323 attributes of types. Some type attributes apply only to @code{struct}
6324 and @code{union} types, while others can apply to any type defined
6325 via a @code{typedef} declaration. Other attributes are defined for
6326 functions (@pxref{Function Attributes}), labels (@pxref{Label
6327 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6328 variables (@pxref{Variable Attributes}).
6330 The @code{__attribute__} keyword is followed by an attribute specification
6331 inside double parentheses.
6333 You may specify type attributes in an enum, struct or union type
6334 declaration or definition by placing them immediately after the
6335 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6336 syntax is to place them just past the closing curly brace of the
6339 You can also include type attributes in a @code{typedef} declaration.
6340 @xref{Attribute Syntax}, for details of the exact syntax for using
6344 * Common Type Attributes::
6345 * ARM Type Attributes::
6346 * MeP Type Attributes::
6347 * PowerPC Type Attributes::
6348 * SPU Type Attributes::
6349 * x86 Type Attributes::
6352 @node Common Type Attributes
6353 @subsection Common Type Attributes
6355 The following type attributes are supported on most targets.
6358 @cindex @code{aligned} type attribute
6359 @item aligned (@var{alignment})
6360 This attribute specifies a minimum alignment (in bytes) for variables
6361 of the specified type. For example, the declarations:
6364 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6365 typedef int more_aligned_int __attribute__ ((aligned (8)));
6369 force the compiler to ensure (as far as it can) that each variable whose
6370 type is @code{struct S} or @code{more_aligned_int} is allocated and
6371 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6372 variables of type @code{struct S} aligned to 8-byte boundaries allows
6373 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6374 store) instructions when copying one variable of type @code{struct S} to
6375 another, thus improving run-time efficiency.
6377 Note that the alignment of any given @code{struct} or @code{union} type
6378 is required by the ISO C standard to be at least a perfect multiple of
6379 the lowest common multiple of the alignments of all of the members of
6380 the @code{struct} or @code{union} in question. This means that you @emph{can}
6381 effectively adjust the alignment of a @code{struct} or @code{union}
6382 type by attaching an @code{aligned} attribute to any one of the members
6383 of such a type, but the notation illustrated in the example above is a
6384 more obvious, intuitive, and readable way to request the compiler to
6385 adjust the alignment of an entire @code{struct} or @code{union} type.
6387 As in the preceding example, you can explicitly specify the alignment
6388 (in bytes) that you wish the compiler to use for a given @code{struct}
6389 or @code{union} type. Alternatively, you can leave out the alignment factor
6390 and just ask the compiler to align a type to the maximum
6391 useful alignment for the target machine you are compiling for. For
6392 example, you could write:
6395 struct S @{ short f[3]; @} __attribute__ ((aligned));
6398 Whenever you leave out the alignment factor in an @code{aligned}
6399 attribute specification, the compiler automatically sets the alignment
6400 for the type to the largest alignment that is ever used for any data
6401 type on the target machine you are compiling for. Doing this can often
6402 make copy operations more efficient, because the compiler can use
6403 whatever instructions copy the biggest chunks of memory when performing
6404 copies to or from the variables that have types that you have aligned
6407 In the example above, if the size of each @code{short} is 2 bytes, then
6408 the size of the entire @code{struct S} type is 6 bytes. The smallest
6409 power of two that is greater than or equal to that is 8, so the
6410 compiler sets the alignment for the entire @code{struct S} type to 8
6413 Note that although you can ask the compiler to select a time-efficient
6414 alignment for a given type and then declare only individual stand-alone
6415 objects of that type, the compiler's ability to select a time-efficient
6416 alignment is primarily useful only when you plan to create arrays of
6417 variables having the relevant (efficiently aligned) type. If you
6418 declare or use arrays of variables of an efficiently-aligned type, then
6419 it is likely that your program also does pointer arithmetic (or
6420 subscripting, which amounts to the same thing) on pointers to the
6421 relevant type, and the code that the compiler generates for these
6422 pointer arithmetic operations is often more efficient for
6423 efficiently-aligned types than for other types.
6425 Note that the effectiveness of @code{aligned} attributes may be limited
6426 by inherent limitations in your linker. On many systems, the linker is
6427 only able to arrange for variables to be aligned up to a certain maximum
6428 alignment. (For some linkers, the maximum supported alignment may
6429 be very very small.) If your linker is only able to align variables
6430 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6431 in an @code{__attribute__} still only provides you with 8-byte
6432 alignment. See your linker documentation for further information.
6434 The @code{aligned} attribute can only increase alignment. Alignment
6435 can be decreased by specifying the @code{packed} attribute. See below.
6437 @item bnd_variable_size
6438 @cindex @code{bnd_variable_size} type attribute
6439 @cindex Pointer Bounds Checker attributes
6440 When applied to a structure field, this attribute tells Pointer
6441 Bounds Checker that the size of this field should not be computed
6442 using static type information. It may be used to mark variably-sized
6443 static array fields placed at the end of a structure.
6451 S *p = (S *)malloc (sizeof(S) + 100);
6452 p->data[10] = 0; //Bounds violation
6456 By using an attribute for the field we may avoid unwanted bound
6463 char data[1] __attribute__((bnd_variable_size));
6465 S *p = (S *)malloc (sizeof(S) + 100);
6466 p->data[10] = 0; //OK
6470 @itemx deprecated (@var{msg})
6471 @cindex @code{deprecated} type attribute
6472 The @code{deprecated} attribute results in a warning if the type
6473 is used anywhere in the source file. This is useful when identifying
6474 types that are expected to be removed in a future version of a program.
6475 If possible, the warning also includes the location of the declaration
6476 of the deprecated type, to enable users to easily find further
6477 information about why the type is deprecated, or what they should do
6478 instead. Note that the warnings only occur for uses and then only
6479 if the type is being applied to an identifier that itself is not being
6480 declared as deprecated.
6483 typedef int T1 __attribute__ ((deprecated));
6487 typedef T1 T3 __attribute__ ((deprecated));
6488 T3 z __attribute__ ((deprecated));
6492 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6493 warning is issued for line 4 because T2 is not explicitly
6494 deprecated. Line 5 has no warning because T3 is explicitly
6495 deprecated. Similarly for line 6. The optional @var{msg}
6496 argument, which must be a string, is printed in the warning if
6499 The @code{deprecated} attribute can also be used for functions and
6500 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6502 @item designated_init
6503 @cindex @code{designated_init} type attribute
6504 This attribute may only be applied to structure types. It indicates
6505 that any initialization of an object of this type must use designated
6506 initializers rather than positional initializers. The intent of this
6507 attribute is to allow the programmer to indicate that a structure's
6508 layout may change, and that therefore relying on positional
6509 initialization will result in future breakage.
6511 GCC emits warnings based on this attribute by default; use
6512 @option{-Wno-designated-init} to suppress them.
6515 @cindex @code{may_alias} type attribute
6516 Accesses through pointers to types with this attribute are not subject
6517 to type-based alias analysis, but are instead assumed to be able to alias
6518 any other type of objects.
6519 In the context of section 6.5 paragraph 7 of the C99 standard,
6520 an lvalue expression
6521 dereferencing such a pointer is treated like having a character type.
6522 See @option{-fstrict-aliasing} for more information on aliasing issues.
6523 This extension exists to support some vector APIs, in which pointers to
6524 one vector type are permitted to alias pointers to a different vector type.
6526 Note that an object of a type with this attribute does not have any
6532 typedef short __attribute__((__may_alias__)) short_a;
6538 short_a *b = (short_a *) &a;
6542 if (a == 0x12345678)
6550 If you replaced @code{short_a} with @code{short} in the variable
6551 declaration, the above program would abort when compiled with
6552 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6556 @cindex @code{packed} type attribute
6557 This attribute, attached to @code{struct} or @code{union} type
6558 definition, specifies that each member (other than zero-width bit-fields)
6559 of the structure or union is placed to minimize the memory required. When
6560 attached to an @code{enum} definition, it indicates that the smallest
6561 integral type should be used.
6563 @opindex fshort-enums
6564 Specifying the @code{packed} attribute for @code{struct} and @code{union}
6565 types is equivalent to specifying the @code{packed} attribute on each
6566 of the structure or union members. Specifying the @option{-fshort-enums}
6567 flag on the command line is equivalent to specifying the @code{packed}
6568 attribute on all @code{enum} definitions.
6570 In the following example @code{struct my_packed_struct}'s members are
6571 packed closely together, but the internal layout of its @code{s} member
6572 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6576 struct my_unpacked_struct
6582 struct __attribute__ ((__packed__)) my_packed_struct
6586 struct my_unpacked_struct s;
6590 You may only specify the @code{packed} attribute attribute on the definition
6591 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
6592 that does not also define the enumerated type, structure or union.
6594 @item scalar_storage_order ("@var{endianness}")
6595 @cindex @code{scalar_storage_order} type attribute
6596 When attached to a @code{union} or a @code{struct}, this attribute sets
6597 the storage order, aka endianness, of the scalar fields of the type, as
6598 well as the array fields whose component is scalar. The supported
6599 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
6600 has no effects on fields which are themselves a @code{union}, a @code{struct}
6601 or an array whose component is a @code{union} or a @code{struct}, and it is
6602 possible for these fields to have a different scalar storage order than the
6605 This attribute is supported only for targets that use a uniform default
6606 scalar storage order (fortunately, most of them), i.e. targets that store
6607 the scalars either all in big-endian or all in little-endian.
6609 Additional restrictions are enforced for types with the reverse scalar
6610 storage order with regard to the scalar storage order of the target:
6613 @item Taking the address of a scalar field of a @code{union} or a
6614 @code{struct} with reverse scalar storage order is not permitted and yields
6616 @item Taking the address of an array field, whose component is scalar, of
6617 a @code{union} or a @code{struct} with reverse scalar storage order is
6618 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
6620 @item Taking the address of a @code{union} or a @code{struct} with reverse
6621 scalar storage order is permitted.
6624 These restrictions exist because the storage order attribute is lost when
6625 the address of a scalar or the address of an array with scalar component is
6626 taken, so storing indirectly through this address generally does not work.
6627 The second case is nevertheless allowed to be able to perform a block copy
6628 from or to the array.
6630 Moreover, the use of type punning or aliasing to toggle the storage order
6631 is not supported; that is to say, a given scalar object cannot be accessed
6632 through distinct types that assign a different storage order to it.
6634 @item transparent_union
6635 @cindex @code{transparent_union} type attribute
6637 This attribute, attached to a @code{union} type definition, indicates
6638 that any function parameter having that union type causes calls to that
6639 function to be treated in a special way.
6641 First, the argument corresponding to a transparent union type can be of
6642 any type in the union; no cast is required. Also, if the union contains
6643 a pointer type, the corresponding argument can be a null pointer
6644 constant or a void pointer expression; and if the union contains a void
6645 pointer type, the corresponding argument can be any pointer expression.
6646 If the union member type is a pointer, qualifiers like @code{const} on
6647 the referenced type must be respected, just as with normal pointer
6650 Second, the argument is passed to the function using the calling
6651 conventions of the first member of the transparent union, not the calling
6652 conventions of the union itself. All members of the union must have the
6653 same machine representation; this is necessary for this argument passing
6656 Transparent unions are designed for library functions that have multiple
6657 interfaces for compatibility reasons. For example, suppose the
6658 @code{wait} function must accept either a value of type @code{int *} to
6659 comply with POSIX, or a value of type @code{union wait *} to comply with
6660 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6661 @code{wait} would accept both kinds of arguments, but it would also
6662 accept any other pointer type and this would make argument type checking
6663 less useful. Instead, @code{<sys/wait.h>} might define the interface
6667 typedef union __attribute__ ((__transparent_union__))
6671 @} wait_status_ptr_t;
6673 pid_t wait (wait_status_ptr_t);
6677 This interface allows either @code{int *} or @code{union wait *}
6678 arguments to be passed, using the @code{int *} calling convention.
6679 The program can call @code{wait} with arguments of either type:
6682 int w1 () @{ int w; return wait (&w); @}
6683 int w2 () @{ union wait w; return wait (&w); @}
6687 With this interface, @code{wait}'s implementation might look like this:
6690 pid_t wait (wait_status_ptr_t p)
6692 return waitpid (-1, p.__ip, 0);
6697 @cindex @code{unused} type attribute
6698 When attached to a type (including a @code{union} or a @code{struct}),
6699 this attribute means that variables of that type are meant to appear
6700 possibly unused. GCC does not produce a warning for any variables of
6701 that type, even if the variable appears to do nothing. This is often
6702 the case with lock or thread classes, which are usually defined and then
6703 not referenced, but contain constructors and destructors that have
6704 nontrivial bookkeeping functions.
6707 @cindex @code{visibility} type attribute
6708 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6709 applied to class, struct, union and enum types. Unlike other type
6710 attributes, the attribute must appear between the initial keyword and
6711 the name of the type; it cannot appear after the body of the type.
6713 Note that the type visibility is applied to vague linkage entities
6714 associated with the class (vtable, typeinfo node, etc.). In
6715 particular, if a class is thrown as an exception in one shared object
6716 and caught in another, the class must have default visibility.
6717 Otherwise the two shared objects are unable to use the same
6718 typeinfo node and exception handling will break.
6722 To specify multiple attributes, separate them by commas within the
6723 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6726 @node ARM Type Attributes
6727 @subsection ARM Type Attributes
6729 @cindex @code{notshared} type attribute, ARM
6730 On those ARM targets that support @code{dllimport} (such as Symbian
6731 OS), you can use the @code{notshared} attribute to indicate that the
6732 virtual table and other similar data for a class should not be
6733 exported from a DLL@. For example:
6736 class __declspec(notshared) C @{
6738 __declspec(dllimport) C();
6742 __declspec(dllexport)
6747 In this code, @code{C::C} is exported from the current DLL, but the
6748 virtual table for @code{C} is not exported. (You can use
6749 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6750 most Symbian OS code uses @code{__declspec}.)
6752 @node MeP Type Attributes
6753 @subsection MeP Type Attributes
6755 @cindex @code{based} type attribute, MeP
6756 @cindex @code{tiny} type attribute, MeP
6757 @cindex @code{near} type attribute, MeP
6758 @cindex @code{far} type attribute, MeP
6759 Many of the MeP variable attributes may be applied to types as well.
6760 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6761 @code{far} attributes may be applied to either. The @code{io} and
6762 @code{cb} attributes may not be applied to types.
6764 @node PowerPC Type Attributes
6765 @subsection PowerPC Type Attributes
6767 Three attributes currently are defined for PowerPC configurations:
6768 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6770 @cindex @code{ms_struct} type attribute, PowerPC
6771 @cindex @code{gcc_struct} type attribute, PowerPC
6772 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6773 attributes please see the documentation in @ref{x86 Type Attributes}.
6775 @cindex @code{altivec} type attribute, PowerPC
6776 The @code{altivec} attribute allows one to declare AltiVec vector data
6777 types supported by the AltiVec Programming Interface Manual. The
6778 attribute requires an argument to specify one of three vector types:
6779 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6780 and @code{bool__} (always followed by unsigned).
6783 __attribute__((altivec(vector__)))
6784 __attribute__((altivec(pixel__))) unsigned short
6785 __attribute__((altivec(bool__))) unsigned
6788 These attributes mainly are intended to support the @code{__vector},
6789 @code{__pixel}, and @code{__bool} AltiVec keywords.
6791 @node SPU Type Attributes
6792 @subsection SPU Type Attributes
6794 @cindex @code{spu_vector} type attribute, SPU
6795 The SPU supports the @code{spu_vector} attribute for types. This attribute
6796 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6797 Language Extensions Specification. It is intended to support the
6798 @code{__vector} keyword.
6800 @node x86 Type Attributes
6801 @subsection x86 Type Attributes
6803 Two attributes are currently defined for x86 configurations:
6804 @code{ms_struct} and @code{gcc_struct}.
6810 @cindex @code{ms_struct} type attribute, x86
6811 @cindex @code{gcc_struct} type attribute, x86
6813 If @code{packed} is used on a structure, or if bit-fields are used
6814 it may be that the Microsoft ABI packs them differently
6815 than GCC normally packs them. Particularly when moving packed
6816 data between functions compiled with GCC and the native Microsoft compiler
6817 (either via function call or as data in a file), it may be necessary to access
6820 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6821 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6822 command-line options, respectively;
6823 see @ref{x86 Options}, for details of how structure layout is affected.
6824 @xref{x86 Variable Attributes}, for information about the corresponding
6825 attributes on variables.
6829 @node Label Attributes
6830 @section Label Attributes
6831 @cindex Label Attributes
6833 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6834 details of the exact syntax for using attributes. Other attributes are
6835 available for functions (@pxref{Function Attributes}), variables
6836 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6837 and for types (@pxref{Type Attributes}).
6839 This example uses the @code{cold} label attribute to indicate the
6840 @code{ErrorHandling} branch is unlikely to be taken and that the
6841 @code{ErrorHandling} label is unused:
6845 asm goto ("some asm" : : : : NoError);
6847 /* This branch (the fall-through from the asm) is less commonly used */
6849 __attribute__((cold, unused)); /* Semi-colon is required here */
6854 printf("no error\n");
6860 @cindex @code{unused} label attribute
6861 This feature is intended for program-generated code that may contain
6862 unused labels, but which is compiled with @option{-Wall}. It is
6863 not normally appropriate to use in it human-written code, though it
6864 could be useful in cases where the code that jumps to the label is
6865 contained within an @code{#ifdef} conditional.
6868 @cindex @code{hot} label attribute
6869 The @code{hot} attribute on a label is used to inform the compiler that
6870 the path following the label is more likely than paths that are not so
6871 annotated. This attribute is used in cases where @code{__builtin_expect}
6872 cannot be used, for instance with computed goto or @code{asm goto}.
6875 @cindex @code{cold} label attribute
6876 The @code{cold} attribute on labels is used to inform the compiler that
6877 the path following the label is unlikely to be executed. This attribute
6878 is used in cases where @code{__builtin_expect} cannot be used, for instance
6879 with computed goto or @code{asm goto}.
6883 @node Enumerator Attributes
6884 @section Enumerator Attributes
6885 @cindex Enumerator Attributes
6887 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6888 details of the exact syntax for using attributes. Other attributes are
6889 available for functions (@pxref{Function Attributes}), variables
6890 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6891 and for types (@pxref{Type Attributes}).
6893 This example uses the @code{deprecated} enumerator attribute to indicate the
6894 @code{oldval} enumerator is deprecated:
6898 oldval __attribute__((deprecated)),
6911 @cindex @code{deprecated} enumerator attribute
6912 The @code{deprecated} attribute results in a warning if the enumerator
6913 is used anywhere in the source file. This is useful when identifying
6914 enumerators that are expected to be removed in a future version of a
6915 program. The warning also includes the location of the declaration
6916 of the deprecated enumerator, to enable users to easily find further
6917 information about why the enumerator is deprecated, or what they should
6918 do instead. Note that the warnings only occurs for uses.
6922 @node Attribute Syntax
6923 @section Attribute Syntax
6924 @cindex attribute syntax
6926 This section describes the syntax with which @code{__attribute__} may be
6927 used, and the constructs to which attribute specifiers bind, for the C
6928 language. Some details may vary for C++ and Objective-C@. Because of
6929 infelicities in the grammar for attributes, some forms described here
6930 may not be successfully parsed in all cases.
6932 There are some problems with the semantics of attributes in C++. For
6933 example, there are no manglings for attributes, although they may affect
6934 code generation, so problems may arise when attributed types are used in
6935 conjunction with templates or overloading. Similarly, @code{typeid}
6936 does not distinguish between types with different attributes. Support
6937 for attributes in C++ may be restricted in future to attributes on
6938 declarations only, but not on nested declarators.
6940 @xref{Function Attributes}, for details of the semantics of attributes
6941 applying to functions. @xref{Variable Attributes}, for details of the
6942 semantics of attributes applying to variables. @xref{Type Attributes},
6943 for details of the semantics of attributes applying to structure, union
6944 and enumerated types.
6945 @xref{Label Attributes}, for details of the semantics of attributes
6947 @xref{Enumerator Attributes}, for details of the semantics of attributes
6948 applying to enumerators.
6950 An @dfn{attribute specifier} is of the form
6951 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6952 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6953 each attribute is one of the following:
6957 Empty. Empty attributes are ignored.
6961 (which may be an identifier such as @code{unused}, or a reserved
6962 word such as @code{const}).
6965 An attribute name followed by a parenthesized list of
6966 parameters for the attribute.
6967 These parameters take one of the following forms:
6971 An identifier. For example, @code{mode} attributes use this form.
6974 An identifier followed by a comma and a non-empty comma-separated list
6975 of expressions. For example, @code{format} attributes use this form.
6978 A possibly empty comma-separated list of expressions. For example,
6979 @code{format_arg} attributes use this form with the list being a single
6980 integer constant expression, and @code{alias} attributes use this form
6981 with the list being a single string constant.
6985 An @dfn{attribute specifier list} is a sequence of one or more attribute
6986 specifiers, not separated by any other tokens.
6988 You may optionally specify attribute names with @samp{__}
6989 preceding and following the name.
6990 This allows you to use them in header files without
6991 being concerned about a possible macro of the same name. For example,
6992 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6995 @subsubheading Label Attributes
6997 In GNU C, an attribute specifier list may appear after the colon following a
6998 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6999 attributes on labels if the attribute specifier is immediately
7000 followed by a semicolon (i.e., the label applies to an empty
7001 statement). If the semicolon is missing, C++ label attributes are
7002 ambiguous, as it is permissible for a declaration, which could begin
7003 with an attribute list, to be labelled in C++. Declarations cannot be
7004 labelled in C90 or C99, so the ambiguity does not arise there.
7006 @subsubheading Enumerator Attributes
7008 In GNU C, an attribute specifier list may appear as part of an enumerator.
7009 The attribute goes after the enumeration constant, before @code{=}, if
7010 present. The optional attribute in the enumerator appertains to the
7011 enumeration constant. It is not possible to place the attribute after
7012 the constant expression, if present.
7014 @subsubheading Type Attributes
7016 An attribute specifier list may appear as part of a @code{struct},
7017 @code{union} or @code{enum} specifier. It may go either immediately
7018 after the @code{struct}, @code{union} or @code{enum} keyword, or after
7019 the closing brace. The former syntax is preferred.
7020 Where attribute specifiers follow the closing brace, they are considered
7021 to relate to the structure, union or enumerated type defined, not to any
7022 enclosing declaration the type specifier appears in, and the type
7023 defined is not complete until after the attribute specifiers.
7024 @c Otherwise, there would be the following problems: a shift/reduce
7025 @c conflict between attributes binding the struct/union/enum and
7026 @c binding to the list of specifiers/qualifiers; and "aligned"
7027 @c attributes could use sizeof for the structure, but the size could be
7028 @c changed later by "packed" attributes.
7031 @subsubheading All other attributes
7033 Otherwise, an attribute specifier appears as part of a declaration,
7034 counting declarations of unnamed parameters and type names, and relates
7035 to that declaration (which may be nested in another declaration, for
7036 example in the case of a parameter declaration), or to a particular declarator
7037 within a declaration. Where an
7038 attribute specifier is applied to a parameter declared as a function or
7039 an array, it should apply to the function or array rather than the
7040 pointer to which the parameter is implicitly converted, but this is not
7041 yet correctly implemented.
7043 Any list of specifiers and qualifiers at the start of a declaration may
7044 contain attribute specifiers, whether or not such a list may in that
7045 context contain storage class specifiers. (Some attributes, however,
7046 are essentially in the nature of storage class specifiers, and only make
7047 sense where storage class specifiers may be used; for example,
7048 @code{section}.) There is one necessary limitation to this syntax: the
7049 first old-style parameter declaration in a function definition cannot
7050 begin with an attribute specifier, because such an attribute applies to
7051 the function instead by syntax described below (which, however, is not
7052 yet implemented in this case). In some other cases, attribute
7053 specifiers are permitted by this grammar but not yet supported by the
7054 compiler. All attribute specifiers in this place relate to the
7055 declaration as a whole. In the obsolescent usage where a type of
7056 @code{int} is implied by the absence of type specifiers, such a list of
7057 specifiers and qualifiers may be an attribute specifier list with no
7058 other specifiers or qualifiers.
7060 At present, the first parameter in a function prototype must have some
7061 type specifier that is not an attribute specifier; this resolves an
7062 ambiguity in the interpretation of @code{void f(int
7063 (__attribute__((foo)) x))}, but is subject to change. At present, if
7064 the parentheses of a function declarator contain only attributes then
7065 those attributes are ignored, rather than yielding an error or warning
7066 or implying a single parameter of type int, but this is subject to
7069 An attribute specifier list may appear immediately before a declarator
7070 (other than the first) in a comma-separated list of declarators in a
7071 declaration of more than one identifier using a single list of
7072 specifiers and qualifiers. Such attribute specifiers apply
7073 only to the identifier before whose declarator they appear. For
7077 __attribute__((noreturn)) void d0 (void),
7078 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
7083 the @code{noreturn} attribute applies to all the functions
7084 declared; the @code{format} attribute only applies to @code{d1}.
7086 An attribute specifier list may appear immediately before the comma,
7087 @code{=} or semicolon terminating the declaration of an identifier other
7088 than a function definition. Such attribute specifiers apply
7089 to the declared object or function. Where an
7090 assembler name for an object or function is specified (@pxref{Asm
7091 Labels}), the attribute must follow the @code{asm}
7094 An attribute specifier list may, in future, be permitted to appear after
7095 the declarator in a function definition (before any old-style parameter
7096 declarations or the function body).
7098 Attribute specifiers may be mixed with type qualifiers appearing inside
7099 the @code{[]} of a parameter array declarator, in the C99 construct by
7100 which such qualifiers are applied to the pointer to which the array is
7101 implicitly converted. Such attribute specifiers apply to the pointer,
7102 not to the array, but at present this is not implemented and they are
7105 An attribute specifier list may appear at the start of a nested
7106 declarator. At present, there are some limitations in this usage: the
7107 attributes correctly apply to the declarator, but for most individual
7108 attributes the semantics this implies are not implemented.
7109 When attribute specifiers follow the @code{*} of a pointer
7110 declarator, they may be mixed with any type qualifiers present.
7111 The following describes the formal semantics of this syntax. It makes the
7112 most sense if you are familiar with the formal specification of
7113 declarators in the ISO C standard.
7115 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
7116 D1}, where @code{T} contains declaration specifiers that specify a type
7117 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
7118 contains an identifier @var{ident}. The type specified for @var{ident}
7119 for derived declarators whose type does not include an attribute
7120 specifier is as in the ISO C standard.
7122 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7123 and the declaration @code{T D} specifies the type
7124 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7125 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7126 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7128 If @code{D1} has the form @code{*
7129 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7130 declaration @code{T D} specifies the type
7131 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7132 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7133 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7139 void (__attribute__((noreturn)) ****f) (void);
7143 specifies the type ``pointer to pointer to pointer to pointer to
7144 non-returning function returning @code{void}''. As another example,
7147 char *__attribute__((aligned(8))) *f;
7151 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7152 Note again that this does not work with most attributes; for example,
7153 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7154 is not yet supported.
7156 For compatibility with existing code written for compiler versions that
7157 did not implement attributes on nested declarators, some laxity is
7158 allowed in the placing of attributes. If an attribute that only applies
7159 to types is applied to a declaration, it is treated as applying to
7160 the type of that declaration. If an attribute that only applies to
7161 declarations is applied to the type of a declaration, it is treated
7162 as applying to that declaration; and, for compatibility with code
7163 placing the attributes immediately before the identifier declared, such
7164 an attribute applied to a function return type is treated as
7165 applying to the function type, and such an attribute applied to an array
7166 element type is treated as applying to the array type. If an
7167 attribute that only applies to function types is applied to a
7168 pointer-to-function type, it is treated as applying to the pointer
7169 target type; if such an attribute is applied to a function return type
7170 that is not a pointer-to-function type, it is treated as applying
7171 to the function type.
7173 @node Function Prototypes
7174 @section Prototypes and Old-Style Function Definitions
7175 @cindex function prototype declarations
7176 @cindex old-style function definitions
7177 @cindex promotion of formal parameters
7179 GNU C extends ISO C to allow a function prototype to override a later
7180 old-style non-prototype definition. Consider the following example:
7183 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7190 /* @r{Prototype function declaration.} */
7191 int isroot P((uid_t));
7193 /* @r{Old-style function definition.} */
7195 isroot (x) /* @r{??? lossage here ???} */
7202 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7203 not allow this example, because subword arguments in old-style
7204 non-prototype definitions are promoted. Therefore in this example the
7205 function definition's argument is really an @code{int}, which does not
7206 match the prototype argument type of @code{short}.
7208 This restriction of ISO C makes it hard to write code that is portable
7209 to traditional C compilers, because the programmer does not know
7210 whether the @code{uid_t} type is @code{short}, @code{int}, or
7211 @code{long}. Therefore, in cases like these GNU C allows a prototype
7212 to override a later old-style definition. More precisely, in GNU C, a
7213 function prototype argument type overrides the argument type specified
7214 by a later old-style definition if the former type is the same as the
7215 latter type before promotion. Thus in GNU C the above example is
7216 equivalent to the following:
7229 GNU C++ does not support old-style function definitions, so this
7230 extension is irrelevant.
7233 @section C++ Style Comments
7235 @cindex C++ comments
7236 @cindex comments, C++ style
7238 In GNU C, you may use C++ style comments, which start with @samp{//} and
7239 continue until the end of the line. Many other C implementations allow
7240 such comments, and they are included in the 1999 C standard. However,
7241 C++ style comments are not recognized if you specify an @option{-std}
7242 option specifying a version of ISO C before C99, or @option{-ansi}
7243 (equivalent to @option{-std=c90}).
7246 @section Dollar Signs in Identifier Names
7248 @cindex dollar signs in identifier names
7249 @cindex identifier names, dollar signs in
7251 In GNU C, you may normally use dollar signs in identifier names.
7252 This is because many traditional C implementations allow such identifiers.
7253 However, dollar signs in identifiers are not supported on a few target
7254 machines, typically because the target assembler does not allow them.
7256 @node Character Escapes
7257 @section The Character @key{ESC} in Constants
7259 You can use the sequence @samp{\e} in a string or character constant to
7260 stand for the ASCII character @key{ESC}.
7263 @section Inquiring on Alignment of Types or Variables
7265 @cindex type alignment
7266 @cindex variable alignment
7268 The keyword @code{__alignof__} allows you to inquire about how an object
7269 is aligned, or the minimum alignment usually required by a type. Its
7270 syntax is just like @code{sizeof}.
7272 For example, if the target machine requires a @code{double} value to be
7273 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7274 This is true on many RISC machines. On more traditional machine
7275 designs, @code{__alignof__ (double)} is 4 or even 2.
7277 Some machines never actually require alignment; they allow reference to any
7278 data type even at an odd address. For these machines, @code{__alignof__}
7279 reports the smallest alignment that GCC gives the data type, usually as
7280 mandated by the target ABI.
7282 If the operand of @code{__alignof__} is an lvalue rather than a type,
7283 its value is the required alignment for its type, taking into account
7284 any minimum alignment specified with GCC's @code{__attribute__}
7285 extension (@pxref{Variable Attributes}). For example, after this
7289 struct foo @{ int x; char y; @} foo1;
7293 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7294 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7296 It is an error to ask for the alignment of an incomplete type.
7300 @section An Inline Function is As Fast As a Macro
7301 @cindex inline functions
7302 @cindex integrating function code
7304 @cindex macros, inline alternative
7306 By declaring a function inline, you can direct GCC to make
7307 calls to that function faster. One way GCC can achieve this is to
7308 integrate that function's code into the code for its callers. This
7309 makes execution faster by eliminating the function-call overhead; in
7310 addition, if any of the actual argument values are constant, their
7311 known values may permit simplifications at compile time so that not
7312 all of the inline function's code needs to be included. The effect on
7313 code size is less predictable; object code may be larger or smaller
7314 with function inlining, depending on the particular case. You can
7315 also direct GCC to try to integrate all ``simple enough'' functions
7316 into their callers with the option @option{-finline-functions}.
7318 GCC implements three different semantics of declaring a function
7319 inline. One is available with @option{-std=gnu89} or
7320 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7321 on all inline declarations, another when
7322 @option{-std=c99}, @option{-std=c11},
7323 @option{-std=gnu99} or @option{-std=gnu11}
7324 (without @option{-fgnu89-inline}), and the third
7325 is used when compiling C++.
7327 To declare a function inline, use the @code{inline} keyword in its
7328 declaration, like this:
7338 If you are writing a header file to be included in ISO C90 programs, write
7339 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7341 The three types of inlining behave similarly in two important cases:
7342 when the @code{inline} keyword is used on a @code{static} function,
7343 like the example above, and when a function is first declared without
7344 using the @code{inline} keyword and then is defined with
7345 @code{inline}, like this:
7348 extern int inc (int *a);
7356 In both of these common cases, the program behaves the same as if you
7357 had not used the @code{inline} keyword, except for its speed.
7359 @cindex inline functions, omission of
7360 @opindex fkeep-inline-functions
7361 When a function is both inline and @code{static}, if all calls to the
7362 function are integrated into the caller, and the function's address is
7363 never used, then the function's own assembler code is never referenced.
7364 In this case, GCC does not actually output assembler code for the
7365 function, unless you specify the option @option{-fkeep-inline-functions}.
7366 If there is a nonintegrated call, then the function is compiled to
7367 assembler code as usual. The function must also be compiled as usual if
7368 the program refers to its address, because that can't be inlined.
7371 Note that certain usages in a function definition can make it unsuitable
7372 for inline substitution. Among these usages are: variadic functions,
7373 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7374 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7375 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7376 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7377 function marked @code{inline} could not be substituted, and gives the
7378 reason for the failure.
7380 @cindex automatic @code{inline} for C++ member fns
7381 @cindex @code{inline} automatic for C++ member fns
7382 @cindex member fns, automatically @code{inline}
7383 @cindex C++ member fns, automatically @code{inline}
7384 @opindex fno-default-inline
7385 As required by ISO C++, GCC considers member functions defined within
7386 the body of a class to be marked inline even if they are
7387 not explicitly declared with the @code{inline} keyword. You can
7388 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7389 Options,,Options Controlling C++ Dialect}.
7391 GCC does not inline any functions when not optimizing unless you specify
7392 the @samp{always_inline} attribute for the function, like this:
7395 /* @r{Prototype.} */
7396 inline void foo (const char) __attribute__((always_inline));
7399 The remainder of this section is specific to GNU C90 inlining.
7401 @cindex non-static inline function
7402 When an inline function is not @code{static}, then the compiler must assume
7403 that there may be calls from other source files; since a global symbol can
7404 be defined only once in any program, the function must not be defined in
7405 the other source files, so the calls therein cannot be integrated.
7406 Therefore, a non-@code{static} inline function is always compiled on its
7407 own in the usual fashion.
7409 If you specify both @code{inline} and @code{extern} in the function
7410 definition, then the definition is used only for inlining. In no case
7411 is the function compiled on its own, not even if you refer to its
7412 address explicitly. Such an address becomes an external reference, as
7413 if you had only declared the function, and had not defined it.
7415 This combination of @code{inline} and @code{extern} has almost the
7416 effect of a macro. The way to use it is to put a function definition in
7417 a header file with these keywords, and put another copy of the
7418 definition (lacking @code{inline} and @code{extern}) in a library file.
7419 The definition in the header file causes most calls to the function
7420 to be inlined. If any uses of the function remain, they refer to
7421 the single copy in the library.
7424 @section When is a Volatile Object Accessed?
7425 @cindex accessing volatiles
7426 @cindex volatile read
7427 @cindex volatile write
7428 @cindex volatile access
7430 C has the concept of volatile objects. These are normally accessed by
7431 pointers and used for accessing hardware or inter-thread
7432 communication. The standard encourages compilers to refrain from
7433 optimizations concerning accesses to volatile objects, but leaves it
7434 implementation defined as to what constitutes a volatile access. The
7435 minimum requirement is that at a sequence point all previous accesses
7436 to volatile objects have stabilized and no subsequent accesses have
7437 occurred. Thus an implementation is free to reorder and combine
7438 volatile accesses that occur between sequence points, but cannot do
7439 so for accesses across a sequence point. The use of volatile does
7440 not allow you to violate the restriction on updating objects multiple
7441 times between two sequence points.
7443 Accesses to non-volatile objects are not ordered with respect to
7444 volatile accesses. You cannot use a volatile object as a memory
7445 barrier to order a sequence of writes to non-volatile memory. For
7449 int *ptr = @var{something};
7451 *ptr = @var{something};
7456 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7457 that the write to @var{*ptr} occurs by the time the update
7458 of @var{vobj} happens. If you need this guarantee, you must use
7459 a stronger memory barrier such as:
7462 int *ptr = @var{something};
7464 *ptr = @var{something};
7465 asm volatile ("" : : : "memory");
7469 A scalar volatile object is read when it is accessed in a void context:
7472 volatile int *src = @var{somevalue};
7476 Such expressions are rvalues, and GCC implements this as a
7477 read of the volatile object being pointed to.
7479 Assignments are also expressions and have an rvalue. However when
7480 assigning to a scalar volatile, the volatile object is not reread,
7481 regardless of whether the assignment expression's rvalue is used or
7482 not. If the assignment's rvalue is used, the value is that assigned
7483 to the volatile object. For instance, there is no read of @var{vobj}
7484 in all the following cases:
7489 vobj = @var{something};
7490 obj = vobj = @var{something};
7491 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7492 obj = (@var{something}, vobj = @var{anotherthing});
7495 If you need to read the volatile object after an assignment has
7496 occurred, you must use a separate expression with an intervening
7499 As bit-fields are not individually addressable, volatile bit-fields may
7500 be implicitly read when written to, or when adjacent bit-fields are
7501 accessed. Bit-field operations may be optimized such that adjacent
7502 bit-fields are only partially accessed, if they straddle a storage unit
7503 boundary. For these reasons it is unwise to use volatile bit-fields to
7506 @node Using Assembly Language with C
7507 @section How to Use Inline Assembly Language in C Code
7508 @cindex @code{asm} keyword
7509 @cindex assembly language in C
7510 @cindex inline assembly language
7511 @cindex mixing assembly language and C
7513 The @code{asm} keyword allows you to embed assembler instructions
7514 within C code. GCC provides two forms of inline @code{asm}
7515 statements. A @dfn{basic @code{asm}} statement is one with no
7516 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7517 statement (@pxref{Extended Asm}) includes one or more operands.
7518 The extended form is preferred for mixing C and assembly language
7519 within a function, but to include assembly language at
7520 top level you must use basic @code{asm}.
7522 You can also use the @code{asm} keyword to override the assembler name
7523 for a C symbol, or to place a C variable in a specific register.
7526 * Basic Asm:: Inline assembler without operands.
7527 * Extended Asm:: Inline assembler with operands.
7528 * Constraints:: Constraints for @code{asm} operands
7529 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7530 * Explicit Register Variables:: Defining variables residing in specified
7532 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7536 @subsection Basic Asm --- Assembler Instructions Without Operands
7537 @cindex basic @code{asm}
7538 @cindex assembly language in C, basic
7540 A basic @code{asm} statement has the following syntax:
7543 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7546 The @code{asm} keyword is a GNU extension.
7547 When writing code that can be compiled with @option{-ansi} and the
7548 various @option{-std} options, use @code{__asm__} instead of
7549 @code{asm} (@pxref{Alternate Keywords}).
7551 @subsubheading Qualifiers
7554 The optional @code{volatile} qualifier has no effect.
7555 All basic @code{asm} blocks are implicitly volatile.
7558 @subsubheading Parameters
7561 @item AssemblerInstructions
7562 This is a literal string that specifies the assembler code. The string can
7563 contain any instructions recognized by the assembler, including directives.
7564 GCC does not parse the assembler instructions themselves and
7565 does not know what they mean or even whether they are valid assembler input.
7567 You may place multiple assembler instructions together in a single @code{asm}
7568 string, separated by the characters normally used in assembly code for the
7569 system. A combination that works in most places is a newline to break the
7570 line, plus a tab character (written as @samp{\n\t}).
7571 Some assemblers allow semicolons as a line separator. However,
7572 note that some assembler dialects use semicolons to start a comment.
7575 @subsubheading Remarks
7576 Using extended @code{asm} (@pxref{Extended Asm}) typically produces
7577 smaller, safer, and more efficient code, and in most cases it is a
7578 better solution than basic @code{asm}. However, there are two
7579 situations where only basic @code{asm} can be used:
7583 Extended @code{asm} statements have to be inside a C
7584 function, so to write inline assembly language at file scope (``top-level''),
7585 outside of C functions, you must use basic @code{asm}.
7586 You can use this technique to emit assembler directives,
7587 define assembly language macros that can be invoked elsewhere in the file,
7588 or write entire functions in assembly language.
7592 with the @code{naked} attribute also require basic @code{asm}
7593 (@pxref{Function Attributes}).
7596 Safely accessing C data and calling functions from basic @code{asm} is more
7597 complex than it may appear. To access C data, it is better to use extended
7600 Do not expect a sequence of @code{asm} statements to remain perfectly
7601 consecutive after compilation. If certain instructions need to remain
7602 consecutive in the output, put them in a single multi-instruction @code{asm}
7603 statement. Note that GCC's optimizers can move @code{asm} statements
7604 relative to other code, including across jumps.
7606 @code{asm} statements may not perform jumps into other @code{asm} statements.
7607 GCC does not know about these jumps, and therefore cannot take
7608 account of them when deciding how to optimize. Jumps from @code{asm} to C
7609 labels are only supported in extended @code{asm}.
7611 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7612 assembly code when optimizing. This can lead to unexpected duplicate
7613 symbol errors during compilation if your assembly code defines symbols or
7616 @strong{Warning:} The C standards do not specify semantics for @code{asm},
7617 making it a potential source of incompatibilities between compilers. These
7618 incompatibilities may not produce compiler warnings/errors.
7620 GCC does not parse basic @code{asm}'s @var{AssemblerInstructions}, which
7621 means there is no way to communicate to the compiler what is happening
7622 inside them. GCC has no visibility of symbols in the @code{asm} and may
7623 discard them as unreferenced. It also does not know about side effects of
7624 the assembler code, such as modifications to memory or registers. Unlike
7625 some compilers, GCC assumes that no changes to general purpose registers
7626 occur. This assumption may change in a future release.
7628 To avoid complications from future changes to the semantics and the
7629 compatibility issues between compilers, consider replacing basic @code{asm}
7630 with extended @code{asm}. See
7631 @uref{https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended, How to convert
7632 from basic asm to extended asm} for information about how to perform this
7635 The compiler copies the assembler instructions in a basic @code{asm}
7636 verbatim to the assembly language output file, without
7637 processing dialects or any of the @samp{%} operators that are available with
7638 extended @code{asm}. This results in minor differences between basic
7639 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7640 registers you might use @samp{%eax} in basic @code{asm} and
7641 @samp{%%eax} in extended @code{asm}.
7643 On targets such as x86 that support multiple assembler dialects,
7644 all basic @code{asm} blocks use the assembler dialect specified by the
7645 @option{-masm} command-line option (@pxref{x86 Options}).
7646 Basic @code{asm} provides no
7647 mechanism to provide different assembler strings for different dialects.
7649 For basic @code{asm} with non-empty assembler string GCC assumes
7650 the assembler block does not change any general purpose registers,
7651 but it may read or write any globally accessible variable.
7653 Here is an example of basic @code{asm} for i386:
7656 /* Note that this code will not compile with -masm=intel */
7657 #define DebugBreak() asm("int $3")
7661 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7662 @cindex extended @code{asm}
7663 @cindex assembly language in C, extended
7665 With extended @code{asm} you can read and write C variables from
7666 assembler and perform jumps from assembler code to C labels.
7667 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7668 the operand parameters after the assembler template:
7671 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7672 : @var{OutputOperands}
7673 @r{[} : @var{InputOperands}
7674 @r{[} : @var{Clobbers} @r{]} @r{]})
7676 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7678 : @var{InputOperands}
7683 The @code{asm} keyword is a GNU extension.
7684 When writing code that can be compiled with @option{-ansi} and the
7685 various @option{-std} options, use @code{__asm__} instead of
7686 @code{asm} (@pxref{Alternate Keywords}).
7688 @subsubheading Qualifiers
7692 The typical use of extended @code{asm} statements is to manipulate input
7693 values to produce output values. However, your @code{asm} statements may
7694 also produce side effects. If so, you may need to use the @code{volatile}
7695 qualifier to disable certain optimizations. @xref{Volatile}.
7698 This qualifier informs the compiler that the @code{asm} statement may
7699 perform a jump to one of the labels listed in the @var{GotoLabels}.
7703 @subsubheading Parameters
7705 @item AssemblerTemplate
7706 This is a literal string that is the template for the assembler code. It is a
7707 combination of fixed text and tokens that refer to the input, output,
7708 and goto parameters. @xref{AssemblerTemplate}.
7710 @item OutputOperands
7711 A comma-separated list of the C variables modified by the instructions in the
7712 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7715 A comma-separated list of C expressions read by the instructions in the
7716 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7719 A comma-separated list of registers or other values changed by the
7720 @var{AssemblerTemplate}, beyond those listed as outputs.
7721 An empty list is permitted. @xref{Clobbers}.
7724 When you are using the @code{goto} form of @code{asm}, this section contains
7725 the list of all C labels to which the code in the
7726 @var{AssemblerTemplate} may jump.
7729 @code{asm} statements may not perform jumps into other @code{asm} statements,
7730 only to the listed @var{GotoLabels}.
7731 GCC's optimizers do not know about other jumps; therefore they cannot take
7732 account of them when deciding how to optimize.
7735 The total number of input + output + goto operands is limited to 30.
7737 @subsubheading Remarks
7738 The @code{asm} statement allows you to include assembly instructions directly
7739 within C code. This may help you to maximize performance in time-sensitive
7740 code or to access assembly instructions that are not readily available to C
7743 Note that extended @code{asm} statements must be inside a function. Only
7744 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7745 Functions declared with the @code{naked} attribute also require basic
7746 @code{asm} (@pxref{Function Attributes}).
7748 While the uses of @code{asm} are many and varied, it may help to think of an
7749 @code{asm} statement as a series of low-level instructions that convert input
7750 parameters to output parameters. So a simple (if not particularly useful)
7751 example for i386 using @code{asm} might look like this:
7757 asm ("mov %1, %0\n\t"
7762 printf("%d\n", dst);
7765 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7768 @subsubsection Volatile
7769 @cindex volatile @code{asm}
7770 @cindex @code{asm} volatile
7772 GCC's optimizers sometimes discard @code{asm} statements if they determine
7773 there is no need for the output variables. Also, the optimizers may move
7774 code out of loops if they believe that the code will always return the same
7775 result (i.e. none of its input values change between calls). Using the
7776 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7777 that have no output operands, including @code{asm goto} statements,
7778 are implicitly volatile.
7780 This i386 code demonstrates a case that does not use (or require) the
7781 @code{volatile} qualifier. If it is performing assertion checking, this code
7782 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7783 unreferenced by any code. As a result, the optimizers can discard the
7784 @code{asm} statement, which in turn removes the need for the entire
7785 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7786 isn't needed you allow the optimizers to produce the most efficient code
7790 void DoCheck(uint32_t dwSomeValue)
7794 // Assumes dwSomeValue is not zero.
7804 The next example shows a case where the optimizers can recognize that the input
7805 (@code{dwSomeValue}) never changes during the execution of the function and can
7806 therefore move the @code{asm} outside the loop to produce more efficient code.
7807 Again, using @code{volatile} disables this type of optimization.
7810 void do_print(uint32_t dwSomeValue)
7814 for (uint32_t x=0; x < 5; x++)
7816 // Assumes dwSomeValue is not zero.
7822 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7827 The following example demonstrates a case where you need to use the
7828 @code{volatile} qualifier.
7829 It uses the x86 @code{rdtsc} instruction, which reads
7830 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7831 the optimizers might assume that the @code{asm} block will always return the
7832 same value and therefore optimize away the second call.
7837 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7838 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7839 "or %%rdx, %0" // 'Or' in the lower bits.
7844 printf("msr: %llx\n", msr);
7848 // Reprint the timestamp
7849 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7850 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7851 "or %%rdx, %0" // 'Or' in the lower bits.
7856 printf("msr: %llx\n", msr);
7859 GCC's optimizers do not treat this code like the non-volatile code in the
7860 earlier examples. They do not move it out of loops or omit it on the
7861 assumption that the result from a previous call is still valid.
7863 Note that the compiler can move even volatile @code{asm} instructions relative
7864 to other code, including across jump instructions. For example, on many
7865 targets there is a system register that controls the rounding mode of
7866 floating-point operations. Setting it with a volatile @code{asm}, as in the
7867 following PowerPC example, does not work reliably.
7870 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7874 The compiler may move the addition back before the volatile @code{asm}. To
7875 make it work as expected, add an artificial dependency to the @code{asm} by
7876 referencing a variable in the subsequent code, for example:
7879 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7883 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7884 assembly code when optimizing. This can lead to unexpected duplicate symbol
7885 errors during compilation if your asm code defines symbols or labels.
7887 (@pxref{AssemblerTemplate}) may help resolve this problem.
7889 @anchor{AssemblerTemplate}
7890 @subsubsection Assembler Template
7891 @cindex @code{asm} assembler template
7893 An assembler template is a literal string containing assembler instructions.
7894 The compiler replaces tokens in the template that refer
7895 to inputs, outputs, and goto labels,
7896 and then outputs the resulting string to the assembler. The
7897 string can contain any instructions recognized by the assembler, including
7898 directives. GCC does not parse the assembler instructions
7899 themselves and does not know what they mean or even whether they are valid
7900 assembler input. However, it does count the statements
7901 (@pxref{Size of an asm}).
7903 You may place multiple assembler instructions together in a single @code{asm}
7904 string, separated by the characters normally used in assembly code for the
7905 system. A combination that works in most places is a newline to break the
7906 line, plus a tab character to move to the instruction field (written as
7908 Some assemblers allow semicolons as a line separator. However, note
7909 that some assembler dialects use semicolons to start a comment.
7911 Do not expect a sequence of @code{asm} statements to remain perfectly
7912 consecutive after compilation, even when you are using the @code{volatile}
7913 qualifier. If certain instructions need to remain consecutive in the output,
7914 put them in a single multi-instruction asm statement.
7916 Accessing data from C programs without using input/output operands (such as
7917 by using global symbols directly from the assembler template) may not work as
7918 expected. Similarly, calling functions directly from an assembler template
7919 requires a detailed understanding of the target assembler and ABI.
7921 Since GCC does not parse the assembler template,
7922 it has no visibility of any
7923 symbols it references. This may result in GCC discarding those symbols as
7924 unreferenced unless they are also listed as input, output, or goto operands.
7926 @subsubheading Special format strings
7928 In addition to the tokens described by the input, output, and goto operands,
7929 these tokens have special meanings in the assembler template:
7933 Outputs a single @samp{%} into the assembler code.
7936 Outputs a number that is unique to each instance of the @code{asm}
7937 statement in the entire compilation. This option is useful when creating local
7938 labels and referring to them multiple times in a single template that
7939 generates multiple assembler instructions.
7944 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7945 into the assembler code. When unescaped, these characters have special
7946 meaning to indicate multiple assembler dialects, as described below.
7949 @subsubheading Multiple assembler dialects in @code{asm} templates
7951 On targets such as x86, GCC supports multiple assembler dialects.
7952 The @option{-masm} option controls which dialect GCC uses as its
7953 default for inline assembler. The target-specific documentation for the
7954 @option{-masm} option contains the list of supported dialects, as well as the
7955 default dialect if the option is not specified. This information may be
7956 important to understand, since assembler code that works correctly when
7957 compiled using one dialect will likely fail if compiled using another.
7960 If your code needs to support multiple assembler dialects (for example, if
7961 you are writing public headers that need to support a variety of compilation
7962 options), use constructs of this form:
7965 @{ dialect0 | dialect1 | dialect2... @}
7968 This construct outputs @code{dialect0}
7969 when using dialect #0 to compile the code,
7970 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7971 braces than the number of dialects the compiler supports, the construct
7974 For example, if an x86 compiler supports two dialects
7975 (@samp{att}, @samp{intel}), an
7976 assembler template such as this:
7979 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7983 is equivalent to one of
7986 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7987 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7990 Using that same compiler, this code:
7993 "xchg@{l@}\t@{%%@}ebx, %1"
7997 corresponds to either
8000 "xchgl\t%%ebx, %1" @r{/* att dialect */}
8001 "xchg\tebx, %1" @r{/* intel dialect */}
8004 There is no support for nesting dialect alternatives.
8006 @anchor{OutputOperands}
8007 @subsubsection Output Operands
8008 @cindex @code{asm} output operands
8010 An @code{asm} statement has zero or more output operands indicating the names
8011 of C variables modified by the assembler code.
8013 In this i386 example, @code{old} (referred to in the template string as
8014 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
8015 (@code{%2}) is an input:
8020 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
8021 "sbb %0,%0" // Use the CF to calculate old.
8022 : "=r" (old), "+rm" (*Base)
8029 Operands are separated by commas. Each operand has this format:
8032 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
8036 @item asmSymbolicName
8037 Specifies a symbolic name for the operand.
8038 Reference the name in the assembler template
8039 by enclosing it in square brackets
8040 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8041 that contains the definition. Any valid C variable name is acceptable,
8042 including names already defined in the surrounding code. No two operands
8043 within the same @code{asm} statement can use the same symbolic name.
8045 When not using an @var{asmSymbolicName}, use the (zero-based) position
8047 in the list of operands in the assembler template. For example if there are
8048 three output operands, use @samp{%0} in the template to refer to the first,
8049 @samp{%1} for the second, and @samp{%2} for the third.
8052 A string constant specifying constraints on the placement of the operand;
8053 @xref{Constraints}, for details.
8055 Output constraints must begin with either @samp{=} (a variable overwriting an
8056 existing value) or @samp{+} (when reading and writing). When using
8057 @samp{=}, do not assume the location contains the existing value
8058 on entry to the @code{asm}, except
8059 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
8061 After the prefix, there must be one or more additional constraints
8062 (@pxref{Constraints}) that describe where the value resides. Common
8063 constraints include @samp{r} for register and @samp{m} for memory.
8064 When you list more than one possible location (for example, @code{"=rm"}),
8065 the compiler chooses the most efficient one based on the current context.
8066 If you list as many alternates as the @code{asm} statement allows, you permit
8067 the optimizers to produce the best possible code.
8068 If you must use a specific register, but your Machine Constraints do not
8069 provide sufficient control to select the specific register you want,
8070 local register variables may provide a solution (@pxref{Local Register
8074 Specifies a C lvalue expression to hold the output, typically a variable name.
8075 The enclosing parentheses are a required part of the syntax.
8079 When the compiler selects the registers to use to
8080 represent the output operands, it does not use any of the clobbered registers
8083 Output operand expressions must be lvalues. The compiler cannot check whether
8084 the operands have data types that are reasonable for the instruction being
8085 executed. For output expressions that are not directly addressable (for
8086 example a bit-field), the constraint must allow a register. In that case, GCC
8087 uses the register as the output of the @code{asm}, and then stores that
8088 register into the output.
8090 Operands using the @samp{+} constraint modifier count as two operands
8091 (that is, both as input and output) towards the total maximum of 30 operands
8092 per @code{asm} statement.
8094 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
8095 operands that must not overlap an input. Otherwise,
8096 GCC may allocate the output operand in the same register as an unrelated
8097 input operand, on the assumption that the assembler code consumes its
8098 inputs before producing outputs. This assumption may be false if the assembler
8099 code actually consists of more than one instruction.
8101 The same problem can occur if one output parameter (@var{a}) allows a register
8102 constraint and another output parameter (@var{b}) allows a memory constraint.
8103 The code generated by GCC to access the memory address in @var{b} can contain
8104 registers which @emph{might} be shared by @var{a}, and GCC considers those
8105 registers to be inputs to the asm. As above, GCC assumes that such input
8106 registers are consumed before any outputs are written. This assumption may
8107 result in incorrect behavior if the asm writes to @var{a} before using
8108 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
8109 ensures that modifying @var{a} does not affect the address referenced by
8110 @var{b}. Otherwise, the location of @var{b}
8111 is undefined if @var{a} is modified before using @var{b}.
8113 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8114 instead of simply @samp{%2}). Typically these qualifiers are hardware
8115 dependent. The list of supported modifiers for x86 is found at
8116 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8118 If the C code that follows the @code{asm} makes no use of any of the output
8119 operands, use @code{volatile} for the @code{asm} statement to prevent the
8120 optimizers from discarding the @code{asm} statement as unneeded
8121 (see @ref{Volatile}).
8123 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
8124 references the first output operand as @code{%0} (were there a second, it
8125 would be @code{%1}, etc). The number of the first input operand is one greater
8126 than that of the last output operand. In this i386 example, that makes
8127 @code{Mask} referenced as @code{%1}:
8130 uint32_t Mask = 1234;
8139 That code overwrites the variable @code{Index} (@samp{=}),
8140 placing the value in a register (@samp{r}).
8141 Using the generic @samp{r} constraint instead of a constraint for a specific
8142 register allows the compiler to pick the register to use, which can result
8143 in more efficient code. This may not be possible if an assembler instruction
8144 requires a specific register.
8146 The following i386 example uses the @var{asmSymbolicName} syntax.
8148 same result as the code above, but some may consider it more readable or more
8149 maintainable since reordering index numbers is not necessary when adding or
8150 removing operands. The names @code{aIndex} and @code{aMask}
8151 are only used in this example to emphasize which
8152 names get used where.
8153 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8156 uint32_t Mask = 1234;
8159 asm ("bsfl %[aMask], %[aIndex]"
8160 : [aIndex] "=r" (Index)
8161 : [aMask] "r" (Mask)
8165 Here are some more examples of output operands.
8172 asm ("mov %[e], %[d]"
8177 Here, @code{d} may either be in a register or in memory. Since the compiler
8178 might already have the current value of the @code{uint32_t} location
8179 pointed to by @code{e}
8180 in a register, you can enable it to choose the best location
8181 for @code{d} by specifying both constraints.
8183 @anchor{FlagOutputOperands}
8184 @subsubsection Flag Output Operands
8185 @cindex @code{asm} flag output operands
8187 Some targets have a special register that holds the ``flags'' for the
8188 result of an operation or comparison. Normally, the contents of that
8189 register are either unmodifed by the asm, or the asm is considered to
8190 clobber the contents.
8192 On some targets, a special form of output operand exists by which
8193 conditions in the flags register may be outputs of the asm. The set of
8194 conditions supported are target specific, but the general rule is that
8195 the output variable must be a scalar integer, and the value is boolean.
8196 When supported, the target defines the preprocessor symbol
8197 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8199 Because of the special nature of the flag output operands, the constraint
8200 may not include alternatives.
8202 Most often, the target has only one flags register, and thus is an implied
8203 operand of many instructions. In this case, the operand should not be
8204 referenced within the assembler template via @code{%0} etc, as there's
8205 no corresponding text in the assembly language.
8209 The flag output constraints for the x86 family are of the form
8210 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8211 conditions defined in the ISA manual for @code{j@var{cc}} or
8216 ``above'' or unsigned greater than
8218 ``above or equal'' or unsigned greater than or equal
8220 ``below'' or unsigned less than
8222 ``below or equal'' or unsigned less than or equal
8227 ``equal'' or zero flag set
8231 signed greater than or equal
8235 signed less than or equal
8256 ``not'' @var{flag}, or inverted versions of those above
8261 @anchor{InputOperands}
8262 @subsubsection Input Operands
8263 @cindex @code{asm} input operands
8264 @cindex @code{asm} expressions
8266 Input operands make values from C variables and expressions available to the
8269 Operands are separated by commas. Each operand has this format:
8272 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8276 @item asmSymbolicName
8277 Specifies a symbolic name for the operand.
8278 Reference the name in the assembler template
8279 by enclosing it in square brackets
8280 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8281 that contains the definition. Any valid C variable name is acceptable,
8282 including names already defined in the surrounding code. No two operands
8283 within the same @code{asm} statement can use the same symbolic name.
8285 When not using an @var{asmSymbolicName}, use the (zero-based) position
8287 in the list of operands in the assembler template. For example if there are
8288 two output operands and three inputs,
8289 use @samp{%2} in the template to refer to the first input operand,
8290 @samp{%3} for the second, and @samp{%4} for the third.
8293 A string constant specifying constraints on the placement of the operand;
8294 @xref{Constraints}, for details.
8296 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8297 When you list more than one possible location (for example, @samp{"irm"}),
8298 the compiler chooses the most efficient one based on the current context.
8299 If you must use a specific register, but your Machine Constraints do not
8300 provide sufficient control to select the specific register you want,
8301 local register variables may provide a solution (@pxref{Local Register
8304 Input constraints can also be digits (for example, @code{"0"}). This indicates
8305 that the specified input must be in the same place as the output constraint
8306 at the (zero-based) index in the output constraint list.
8307 When using @var{asmSymbolicName} syntax for the output operands,
8308 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8311 This is the C variable or expression being passed to the @code{asm} statement
8312 as input. The enclosing parentheses are a required part of the syntax.
8316 When the compiler selects the registers to use to represent the input
8317 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8319 If there are no output operands but there are input operands, place two
8320 consecutive colons where the output operands would go:
8323 __asm__ ("some instructions"
8325 : "r" (Offset / 8));
8328 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8329 (except for inputs tied to outputs). The compiler assumes that on exit from
8330 the @code{asm} statement these operands contain the same values as they
8331 had before executing the statement.
8332 It is @emph{not} possible to use clobbers
8333 to inform the compiler that the values in these inputs are changing. One
8334 common work-around is to tie the changing input variable to an output variable
8335 that never gets used. Note, however, that if the code that follows the
8336 @code{asm} statement makes no use of any of the output operands, the GCC
8337 optimizers may discard the @code{asm} statement as unneeded
8338 (see @ref{Volatile}).
8340 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8341 instead of simply @samp{%2}). Typically these qualifiers are hardware
8342 dependent. The list of supported modifiers for x86 is found at
8343 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8345 In this example using the fictitious @code{combine} instruction, the
8346 constraint @code{"0"} for input operand 1 says that it must occupy the same
8347 location as output operand 0. Only input operands may use numbers in
8348 constraints, and they must each refer to an output operand. Only a number (or
8349 the symbolic assembler name) in the constraint can guarantee that one operand
8350 is in the same place as another. The mere fact that @code{foo} is the value of
8351 both operands is not enough to guarantee that they are in the same place in
8352 the generated assembler code.
8355 asm ("combine %2, %0"
8357 : "0" (foo), "g" (bar));
8360 Here is an example using symbolic names.
8363 asm ("cmoveq %1, %2, %[result]"
8364 : [result] "=r"(result)
8365 : "r" (test), "r" (new), "[result]" (old));
8369 @subsubsection Clobbers
8370 @cindex @code{asm} clobbers
8372 While the compiler is aware of changes to entries listed in the output
8373 operands, the inline @code{asm} code may modify more than just the outputs. For
8374 example, calculations may require additional registers, or the processor may
8375 overwrite a register as a side effect of a particular assembler instruction.
8376 In order to inform the compiler of these changes, list them in the clobber
8377 list. Clobber list items are either register names or the special clobbers
8378 (listed below). Each clobber list item is a string constant
8379 enclosed in double quotes and separated by commas.
8381 Clobber descriptions may not in any way overlap with an input or output
8382 operand. For example, you may not have an operand describing a register class
8383 with one member when listing that register in the clobber list. Variables
8384 declared to live in specific registers (@pxref{Explicit Register
8385 Variables}) and used
8386 as @code{asm} input or output operands must have no part mentioned in the
8387 clobber description. In particular, there is no way to specify that input
8388 operands get modified without also specifying them as output operands.
8390 When the compiler selects which registers to use to represent input and output
8391 operands, it does not use any of the clobbered registers. As a result,
8392 clobbered registers are available for any use in the assembler code.
8394 Here is a realistic example for the VAX showing the use of clobbered
8398 asm volatile ("movc3 %0, %1, %2"
8400 : "g" (from), "g" (to), "g" (count)
8401 : "r0", "r1", "r2", "r3", "r4", "r5");
8404 Also, there are two special clobber arguments:
8408 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8409 register. On some machines, GCC represents the condition codes as a specific
8410 hardware register; @code{"cc"} serves to name this register.
8411 On other machines, condition code handling is different,
8412 and specifying @code{"cc"} has no effect. But
8413 it is valid no matter what the target.
8416 The @code{"memory"} clobber tells the compiler that the assembly code
8418 reads or writes to items other than those listed in the input and output
8419 operands (for example, accessing the memory pointed to by one of the input
8420 parameters). To ensure memory contains correct values, GCC may need to flush
8421 specific register values to memory before executing the @code{asm}. Further,
8422 the compiler does not assume that any values read from memory before an
8423 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8425 Using the @code{"memory"} clobber effectively forms a read/write
8426 memory barrier for the compiler.
8428 Note that this clobber does not prevent the @emph{processor} from doing
8429 speculative reads past the @code{asm} statement. To prevent that, you need
8430 processor-specific fence instructions.
8432 Flushing registers to memory has performance implications and may be an issue
8433 for time-sensitive code. You can use a trick to avoid this if the size of
8434 the memory being accessed is known at compile time. For example, if accessing
8435 ten bytes of a string, use a memory input like:
8437 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8442 @subsubsection Goto Labels
8443 @cindex @code{asm} goto labels
8445 @code{asm goto} allows assembly code to jump to one or more C labels. The
8446 @var{GotoLabels} section in an @code{asm goto} statement contains
8448 list of all C labels to which the assembler code may jump. GCC assumes that
8449 @code{asm} execution falls through to the next statement (if this is not the
8450 case, consider using the @code{__builtin_unreachable} intrinsic after the
8451 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8452 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8455 An @code{asm goto} statement cannot have outputs.
8456 This is due to an internal restriction of
8457 the compiler: control transfer instructions cannot have outputs.
8458 If the assembler code does modify anything, use the @code{"memory"} clobber
8460 optimizers to flush all register values to memory and reload them if
8461 necessary after the @code{asm} statement.
8463 Also note that an @code{asm goto} statement is always implicitly
8464 considered volatile.
8466 To reference a label in the assembler template,
8467 prefix it with @samp{%l} (lowercase @samp{L}) followed
8468 by its (zero-based) position in @var{GotoLabels} plus the number of input
8469 operands. For example, if the @code{asm} has three inputs and references two
8470 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8472 Alternately, you can reference labels using the actual C label name enclosed
8473 in brackets. For example, to reference a label named @code{carry}, you can
8474 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8475 section when using this approach.
8477 Here is an example of @code{asm goto} for i386:
8484 : "r" (p1), "r" (p2)
8494 The following example shows an @code{asm goto} that uses a memory clobber.
8500 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8511 @anchor{x86Operandmodifiers}
8512 @subsubsection x86 Operand Modifiers
8514 References to input, output, and goto operands in the assembler template
8515 of extended @code{asm} statements can use
8516 modifiers to affect the way the operands are formatted in
8517 the code output to the assembler. For example, the
8518 following code uses the @samp{h} and @samp{b} modifiers for x86:
8522 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8526 These modifiers generate this assembler code:
8532 The rest of this discussion uses the following code for illustrative purposes.
8541 asm volatile goto ("some assembler instructions here"
8543 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8544 : /* No clobbers. */
8549 With no modifiers, this is what the output from the operands would be for the
8550 @samp{att} and @samp{intel} dialects of assembler:
8552 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8553 @headitem Operand @tab masm=att @tab masm=intel
8562 @tab @code{OFFSET FLAT:.L2}
8565 The table below shows the list of supported modifiers and their effects.
8567 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8568 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8570 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8575 @tab Print the QImode name of the register.
8580 @tab Print the QImode name for a ``high'' register.
8585 @tab Print the HImode name of the register.
8590 @tab Print the SImode name of the register.
8595 @tab Print the DImode name of the register.
8600 @tab Print the label name with no punctuation.
8605 @tab Require a constant operand and print the constant expression with no punctuation.
8611 @anchor{x86floatingpointasmoperands}
8612 @subsubsection x86 Floating-Point @code{asm} Operands
8614 On x86 targets, there are several rules on the usage of stack-like registers
8615 in the operands of an @code{asm}. These rules apply only to the operands
8616 that are stack-like registers:
8620 Given a set of input registers that die in an @code{asm}, it is
8621 necessary to know which are implicitly popped by the @code{asm}, and
8622 which must be explicitly popped by GCC@.
8624 An input register that is implicitly popped by the @code{asm} must be
8625 explicitly clobbered, unless it is constrained to match an
8629 For any input register that is implicitly popped by an @code{asm}, it is
8630 necessary to know how to adjust the stack to compensate for the pop.
8631 If any non-popped input is closer to the top of the reg-stack than
8632 the implicitly popped register, it would not be possible to know what the
8633 stack looked like---it's not clear how the rest of the stack ``slides
8636 All implicitly popped input registers must be closer to the top of
8637 the reg-stack than any input that is not implicitly popped.
8639 It is possible that if an input dies in an @code{asm}, the compiler might
8640 use the input register for an output reload. Consider this example:
8643 asm ("foo" : "=t" (a) : "f" (b));
8647 This code says that input @code{b} is not popped by the @code{asm}, and that
8648 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8649 deeper after the @code{asm} than it was before. But, it is possible that
8650 reload may think that it can use the same register for both the input and
8653 To prevent this from happening,
8654 if any input operand uses the @samp{f} constraint, all output register
8655 constraints must use the @samp{&} early-clobber modifier.
8657 The example above is correctly written as:
8660 asm ("foo" : "=&t" (a) : "f" (b));
8664 Some operands need to be in particular places on the stack. All
8665 output operands fall in this category---GCC has no other way to
8666 know which registers the outputs appear in unless you indicate
8667 this in the constraints.
8669 Output operands must specifically indicate which register an output
8670 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8671 constraints must select a class with a single register.
8674 Output operands may not be ``inserted'' between existing stack registers.
8675 Since no 387 opcode uses a read/write operand, all output operands
8676 are dead before the @code{asm}, and are pushed by the @code{asm}.
8677 It makes no sense to push anywhere but the top of the reg-stack.
8679 Output operands must start at the top of the reg-stack: output
8680 operands may not ``skip'' a register.
8683 Some @code{asm} statements may need extra stack space for internal
8684 calculations. This can be guaranteed by clobbering stack registers
8685 unrelated to the inputs and outputs.
8690 takes one input, which is internally popped, and produces two outputs.
8693 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8697 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8698 and replaces them with one output. The @code{st(1)} clobber is necessary
8699 for the compiler to know that @code{fyl2xp1} pops both inputs.
8702 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8710 @subsection Controlling Names Used in Assembler Code
8711 @cindex assembler names for identifiers
8712 @cindex names used in assembler code
8713 @cindex identifiers, names in assembler code
8715 You can specify the name to be used in the assembler code for a C
8716 function or variable by writing the @code{asm} (or @code{__asm__})
8717 keyword after the declarator.
8718 It is up to you to make sure that the assembler names you choose do not
8719 conflict with any other assembler symbols, or reference registers.
8721 @subsubheading Assembler names for data:
8723 This sample shows how to specify the assembler name for data:
8726 int foo asm ("myfoo") = 2;
8730 This specifies that the name to be used for the variable @code{foo} in
8731 the assembler code should be @samp{myfoo} rather than the usual
8734 On systems where an underscore is normally prepended to the name of a C
8735 variable, this feature allows you to define names for the
8736 linker that do not start with an underscore.
8738 GCC does not support using this feature with a non-static local variable
8739 since such variables do not have assembler names. If you are
8740 trying to put the variable in a particular register, see
8741 @ref{Explicit Register Variables}.
8743 @subsubheading Assembler names for functions:
8745 To specify the assembler name for functions, write a declaration for the
8746 function before its definition and put @code{asm} there, like this:
8749 int func (int x, int y) asm ("MYFUNC");
8751 int func (int x, int y)
8757 This specifies that the name to be used for the function @code{func} in
8758 the assembler code should be @code{MYFUNC}.
8760 @node Explicit Register Variables
8761 @subsection Variables in Specified Registers
8762 @anchor{Explicit Reg Vars}
8763 @cindex explicit register variables
8764 @cindex variables in specified registers
8765 @cindex specified registers
8767 GNU C allows you to associate specific hardware registers with C
8768 variables. In almost all cases, allowing the compiler to assign
8769 registers produces the best code. However under certain unusual
8770 circumstances, more precise control over the variable storage is
8773 Both global and local variables can be associated with a register. The
8774 consequences of performing this association are very different between
8775 the two, as explained in the sections below.
8778 * Global Register Variables:: Variables declared at global scope.
8779 * Local Register Variables:: Variables declared within a function.
8782 @node Global Register Variables
8783 @subsubsection Defining Global Register Variables
8784 @anchor{Global Reg Vars}
8785 @cindex global register variables
8786 @cindex registers, global variables in
8787 @cindex registers, global allocation
8789 You can define a global register variable and associate it with a specified
8793 register int *foo asm ("r12");
8797 Here @code{r12} is the name of the register that should be used. Note that
8798 this is the same syntax used for defining local register variables, but for
8799 a global variable the declaration appears outside a function. The
8800 @code{register} keyword is required, and cannot be combined with
8801 @code{static}. The register name must be a valid register name for the
8804 Registers are a scarce resource on most systems and allowing the
8805 compiler to manage their usage usually results in the best code. However,
8806 under special circumstances it can make sense to reserve some globally.
8807 For example this may be useful in programs such as programming language
8808 interpreters that have a couple of global variables that are accessed
8811 After defining a global register variable, for the current compilation
8815 @item The register is reserved entirely for this use, and will not be
8816 allocated for any other purpose.
8817 @item The register is not saved and restored by any functions.
8818 @item Stores into this register are never deleted even if they appear to be
8819 dead, but references may be deleted, moved or simplified.
8822 Note that these points @emph{only} apply to code that is compiled with the
8823 definition. The behavior of code that is merely linked in (for example
8824 code from libraries) is not affected.
8826 If you want to recompile source files that do not actually use your global
8827 register variable so they do not use the specified register for any other
8828 purpose, you need not actually add the global register declaration to
8829 their source code. It suffices to specify the compiler option
8830 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8833 @subsubheading Declaring the variable
8835 Global register variables can not have initial values, because an
8836 executable file has no means to supply initial contents for a register.
8838 When selecting a register, choose one that is normally saved and
8839 restored by function calls on your machine. This ensures that code
8840 which is unaware of this reservation (such as library routines) will
8841 restore it before returning.
8843 On machines with register windows, be sure to choose a global
8844 register that is not affected magically by the function call mechanism.
8846 @subsubheading Using the variable
8848 @cindex @code{qsort}, and global register variables
8849 When calling routines that are not aware of the reservation, be
8850 cautious if those routines call back into code which uses them. As an
8851 example, if you call the system library version of @code{qsort}, it may
8852 clobber your registers during execution, but (if you have selected
8853 appropriate registers) it will restore them before returning. However
8854 it will @emph{not} restore them before calling @code{qsort}'s comparison
8855 function. As a result, global values will not reliably be available to
8856 the comparison function unless the @code{qsort} function itself is rebuilt.
8858 Similarly, it is not safe to access the global register variables from signal
8859 handlers or from more than one thread of control. Unless you recompile
8860 them specially for the task at hand, the system library routines may
8861 temporarily use the register for other things.
8863 @cindex register variable after @code{longjmp}
8864 @cindex global register after @code{longjmp}
8865 @cindex value after @code{longjmp}
8868 On most machines, @code{longjmp} restores to each global register
8869 variable the value it had at the time of the @code{setjmp}. On some
8870 machines, however, @code{longjmp} does not change the value of global
8871 register variables. To be portable, the function that called @code{setjmp}
8872 should make other arrangements to save the values of the global register
8873 variables, and to restore them in a @code{longjmp}. This way, the same
8874 thing happens regardless of what @code{longjmp} does.
8876 Eventually there may be a way of asking the compiler to choose a register
8877 automatically, but first we need to figure out how it should choose and
8878 how to enable you to guide the choice. No solution is evident.
8880 @node Local Register Variables
8881 @subsubsection Specifying Registers for Local Variables
8882 @anchor{Local Reg Vars}
8883 @cindex local variables, specifying registers
8884 @cindex specifying registers for local variables
8885 @cindex registers for local variables
8887 You can define a local register variable and associate it with a specified
8891 register int *foo asm ("r12");
8895 Here @code{r12} is the name of the register that should be used. Note
8896 that this is the same syntax used for defining global register variables,
8897 but for a local variable the declaration appears within a function. The
8898 @code{register} keyword is required, and cannot be combined with
8899 @code{static}. The register name must be a valid register name for the
8902 As with global register variables, it is recommended that you choose
8903 a register that is normally saved and restored by function calls on your
8904 machine, so that calls to library routines will not clobber it.
8906 The only supported use for this feature is to specify registers
8907 for input and output operands when calling Extended @code{asm}
8908 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8909 particular machine don't provide sufficient control to select the desired
8910 register. To force an operand into a register, create a local variable
8911 and specify the register name after the variable's declaration. Then use
8912 the local variable for the @code{asm} operand and specify any constraint
8913 letter that matches the register:
8916 register int *p1 asm ("r0") = @dots{};
8917 register int *p2 asm ("r1") = @dots{};
8918 register int *result asm ("r0");
8919 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8922 @emph{Warning:} In the above example, be aware that a register (for example
8923 @code{r0}) can be call-clobbered by subsequent code, including function
8924 calls and library calls for arithmetic operators on other variables (for
8925 example the initialization of @code{p2}). In this case, use temporary
8926 variables for expressions between the register assignments:
8930 register int *p1 asm ("r0") = @dots{};
8931 register int *p2 asm ("r1") = t1;
8932 register int *result asm ("r0");
8933 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8936 Defining a register variable does not reserve the register. Other than
8937 when invoking the Extended @code{asm}, the contents of the specified
8938 register are not guaranteed. For this reason, the following uses
8939 are explicitly @emph{not} supported. If they appear to work, it is only
8940 happenstance, and may stop working as intended due to (seemingly)
8941 unrelated changes in surrounding code, or even minor changes in the
8942 optimization of a future version of gcc:
8945 @item Passing parameters to or from Basic @code{asm}
8946 @item Passing parameters to or from Extended @code{asm} without using input
8948 @item Passing parameters to or from routines written in assembler (or
8949 other languages) using non-standard calling conventions.
8952 Some developers use Local Register Variables in an attempt to improve
8953 gcc's allocation of registers, especially in large functions. In this
8954 case the register name is essentially a hint to the register allocator.
8955 While in some instances this can generate better code, improvements are
8956 subject to the whims of the allocator/optimizers. Since there are no
8957 guarantees that your improvements won't be lost, this usage of Local
8958 Register Variables is discouraged.
8960 On the MIPS platform, there is related use for local register variables
8961 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8962 Defining coprocessor specifics for MIPS targets, gccint,
8963 GNU Compiler Collection (GCC) Internals}).
8965 @node Size of an asm
8966 @subsection Size of an @code{asm}
8968 Some targets require that GCC track the size of each instruction used
8969 in order to generate correct code. Because the final length of the
8970 code produced by an @code{asm} statement is only known by the
8971 assembler, GCC must make an estimate as to how big it will be. It
8972 does this by counting the number of instructions in the pattern of the
8973 @code{asm} and multiplying that by the length of the longest
8974 instruction supported by that processor. (When working out the number
8975 of instructions, it assumes that any occurrence of a newline or of
8976 whatever statement separator character is supported by the assembler --
8977 typically @samp{;} --- indicates the end of an instruction.)
8979 Normally, GCC's estimate is adequate to ensure that correct
8980 code is generated, but it is possible to confuse the compiler if you use
8981 pseudo instructions or assembler macros that expand into multiple real
8982 instructions, or if you use assembler directives that expand to more
8983 space in the object file than is needed for a single instruction.
8984 If this happens then the assembler may produce a diagnostic saying that
8985 a label is unreachable.
8987 @node Alternate Keywords
8988 @section Alternate Keywords
8989 @cindex alternate keywords
8990 @cindex keywords, alternate
8992 @option{-ansi} and the various @option{-std} options disable certain
8993 keywords. This causes trouble when you want to use GNU C extensions, or
8994 a general-purpose header file that should be usable by all programs,
8995 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8996 @code{inline} are not available in programs compiled with
8997 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8998 program compiled with @option{-std=c99} or @option{-std=c11}). The
9000 @code{restrict} is only available when @option{-std=gnu99} (which will
9001 eventually be the default) or @option{-std=c99} (or the equivalent
9002 @option{-std=iso9899:1999}), or an option for a later standard
9005 The way to solve these problems is to put @samp{__} at the beginning and
9006 end of each problematical keyword. For example, use @code{__asm__}
9007 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
9009 Other C compilers won't accept these alternative keywords; if you want to
9010 compile with another compiler, you can define the alternate keywords as
9011 macros to replace them with the customary keywords. It looks like this:
9019 @findex __extension__
9021 @option{-pedantic} and other options cause warnings for many GNU C extensions.
9023 prevent such warnings within one expression by writing
9024 @code{__extension__} before the expression. @code{__extension__} has no
9025 effect aside from this.
9027 @node Incomplete Enums
9028 @section Incomplete @code{enum} Types
9030 You can define an @code{enum} tag without specifying its possible values.
9031 This results in an incomplete type, much like what you get if you write
9032 @code{struct foo} without describing the elements. A later declaration
9033 that does specify the possible values completes the type.
9035 You can't allocate variables or storage using the type while it is
9036 incomplete. However, you can work with pointers to that type.
9038 This extension may not be very useful, but it makes the handling of
9039 @code{enum} more consistent with the way @code{struct} and @code{union}
9042 This extension is not supported by GNU C++.
9044 @node Function Names
9045 @section Function Names as Strings
9046 @cindex @code{__func__} identifier
9047 @cindex @code{__FUNCTION__} identifier
9048 @cindex @code{__PRETTY_FUNCTION__} identifier
9050 GCC provides three magic constants that hold the name of the current
9051 function as a string. In C++11 and later modes, all three are treated
9052 as constant expressions and can be used in @code{constexpr} constexts.
9053 The first of these constants is @code{__func__}, which is part of
9056 The identifier @code{__func__} is implicitly declared by the translator
9057 as if, immediately following the opening brace of each function
9058 definition, the declaration
9061 static const char __func__[] = "function-name";
9065 appeared, where function-name is the name of the lexically-enclosing
9066 function. This name is the unadorned name of the function. As an
9067 extension, at file (or, in C++, namespace scope), @code{__func__}
9068 evaluates to the empty string.
9070 @code{__FUNCTION__} is another name for @code{__func__}, provided for
9071 backward compatibility with old versions of GCC.
9073 In C, @code{__PRETTY_FUNCTION__} is yet another name for
9074 @code{__func__}, except that at file (or, in C++, namespace scope),
9075 it evaluates to the string @code{"top level"}. In addition, in C++,
9076 @code{__PRETTY_FUNCTION__} contains the signature of the function as
9077 well as its bare name. For example, this program:
9080 extern "C" int printf (const char *, ...);
9086 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
9087 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
9105 __PRETTY_FUNCTION__ = void a::sub(int)
9108 These identifiers are variables, not preprocessor macros, and may not
9109 be used to initialize @code{char} arrays or be concatenated with string
9112 @node Return Address
9113 @section Getting the Return or Frame Address of a Function
9115 These functions may be used to get information about the callers of a
9118 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
9119 This function returns the return address of the current function, or of
9120 one of its callers. The @var{level} argument is number of frames to
9121 scan up the call stack. A value of @code{0} yields the return address
9122 of the current function, a value of @code{1} yields the return address
9123 of the caller of the current function, and so forth. When inlining
9124 the expected behavior is that the function returns the address of
9125 the function that is returned to. To work around this behavior use
9126 the @code{noinline} function attribute.
9128 The @var{level} argument must be a constant integer.
9130 On some machines it may be impossible to determine the return address of
9131 any function other than the current one; in such cases, or when the top
9132 of the stack has been reached, this function returns @code{0} or a
9133 random value. In addition, @code{__builtin_frame_address} may be used
9134 to determine if the top of the stack has been reached.
9136 Additional post-processing of the returned value may be needed, see
9137 @code{__builtin_extract_return_addr}.
9139 Calling this function with a nonzero argument can have unpredictable
9140 effects, including crashing the calling program. As a result, calls
9141 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9142 option is in effect. Such calls should only be made in debugging
9146 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9147 The address as returned by @code{__builtin_return_address} may have to be fed
9148 through this function to get the actual encoded address. For example, on the
9149 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9150 platforms an offset has to be added for the true next instruction to be
9153 If no fixup is needed, this function simply passes through @var{addr}.
9156 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9157 This function does the reverse of @code{__builtin_extract_return_addr}.
9160 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9161 This function is similar to @code{__builtin_return_address}, but it
9162 returns the address of the function frame rather than the return address
9163 of the function. Calling @code{__builtin_frame_address} with a value of
9164 @code{0} yields the frame address of the current function, a value of
9165 @code{1} yields the frame address of the caller of the current function,
9168 The frame is the area on the stack that holds local variables and saved
9169 registers. The frame address is normally the address of the first word
9170 pushed on to the stack by the function. However, the exact definition
9171 depends upon the processor and the calling convention. If the processor
9172 has a dedicated frame pointer register, and the function has a frame,
9173 then @code{__builtin_frame_address} returns the value of the frame
9176 On some machines it may be impossible to determine the frame address of
9177 any function other than the current one; in such cases, or when the top
9178 of the stack has been reached, this function returns @code{0} if
9179 the first frame pointer is properly initialized by the startup code.
9181 Calling this function with a nonzero argument can have unpredictable
9182 effects, including crashing the calling program. As a result, calls
9183 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9184 option is in effect. Such calls should only be made in debugging
9188 @node Vector Extensions
9189 @section Using Vector Instructions through Built-in Functions
9191 On some targets, the instruction set contains SIMD vector instructions which
9192 operate on multiple values contained in one large register at the same time.
9193 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9196 The first step in using these extensions is to provide the necessary data
9197 types. This should be done using an appropriate @code{typedef}:
9200 typedef int v4si __attribute__ ((vector_size (16)));
9204 The @code{int} type specifies the base type, while the attribute specifies
9205 the vector size for the variable, measured in bytes. For example, the
9206 declaration above causes the compiler to set the mode for the @code{v4si}
9207 type to be 16 bytes wide and divided into @code{int} sized units. For
9208 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9209 corresponding mode of @code{foo} is @acronym{V4SI}.
9211 The @code{vector_size} attribute is only applicable to integral and
9212 float scalars, although arrays, pointers, and function return values
9213 are allowed in conjunction with this construct. Only sizes that are
9214 a power of two are currently allowed.
9216 All the basic integer types can be used as base types, both as signed
9217 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9218 @code{long long}. In addition, @code{float} and @code{double} can be
9219 used to build floating-point vector types.
9221 Specifying a combination that is not valid for the current architecture
9222 causes GCC to synthesize the instructions using a narrower mode.
9223 For example, if you specify a variable of type @code{V4SI} and your
9224 architecture does not allow for this specific SIMD type, GCC
9225 produces code that uses 4 @code{SIs}.
9227 The types defined in this manner can be used with a subset of normal C
9228 operations. Currently, GCC allows using the following operators
9229 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9231 The operations behave like C++ @code{valarrays}. Addition is defined as
9232 the addition of the corresponding elements of the operands. For
9233 example, in the code below, each of the 4 elements in @var{a} is
9234 added to the corresponding 4 elements in @var{b} and the resulting
9235 vector is stored in @var{c}.
9238 typedef int v4si __attribute__ ((vector_size (16)));
9245 Subtraction, multiplication, division, and the logical operations
9246 operate in a similar manner. Likewise, the result of using the unary
9247 minus or complement operators on a vector type is a vector whose
9248 elements are the negative or complemented values of the corresponding
9249 elements in the operand.
9251 It is possible to use shifting operators @code{<<}, @code{>>} on
9252 integer-type vectors. The operation is defined as following: @code{@{a0,
9253 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9254 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9257 For convenience, it is allowed to use a binary vector operation
9258 where one operand is a scalar. In that case the compiler transforms
9259 the scalar operand into a vector where each element is the scalar from
9260 the operation. The transformation happens only if the scalar could be
9261 safely converted to the vector-element type.
9262 Consider the following code.
9265 typedef int v4si __attribute__ ((vector_size (16)));
9270 a = b + 1; /* a = b + @{1,1,1,1@}; */
9271 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9273 a = l + a; /* Error, cannot convert long to int. */
9276 Vectors can be subscripted as if the vector were an array with
9277 the same number of elements and base type. Out of bound accesses
9278 invoke undefined behavior at run time. Warnings for out of bound
9279 accesses for vector subscription can be enabled with
9280 @option{-Warray-bounds}.
9282 Vector comparison is supported with standard comparison
9283 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9284 vector expressions of integer-type or real-type. Comparison between
9285 integer-type vectors and real-type vectors are not supported. The
9286 result of the comparison is a vector of the same width and number of
9287 elements as the comparison operands with a signed integral element
9290 Vectors are compared element-wise producing 0 when comparison is false
9291 and -1 (constant of the appropriate type where all bits are set)
9292 otherwise. Consider the following example.
9295 typedef int v4si __attribute__ ((vector_size (16)));
9297 v4si a = @{1,2,3,4@};
9298 v4si b = @{3,2,1,4@};
9301 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9302 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9305 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9306 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9307 integer vector with the same number of elements of the same size as @code{b}
9308 and @code{c}, computes all three arguments and creates a vector
9309 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9310 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9311 As in the case of binary operations, this syntax is also accepted when
9312 one of @code{b} or @code{c} is a scalar that is then transformed into a
9313 vector. If both @code{b} and @code{c} are scalars and the type of
9314 @code{true?b:c} has the same size as the element type of @code{a}, then
9315 @code{b} and @code{c} are converted to a vector type whose elements have
9316 this type and with the same number of elements as @code{a}.
9318 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9319 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9320 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9321 For mixed operations between a scalar @code{s} and a vector @code{v},
9322 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9323 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9325 Vector shuffling is available using functions
9326 @code{__builtin_shuffle (vec, mask)} and
9327 @code{__builtin_shuffle (vec0, vec1, mask)}.
9328 Both functions construct a permutation of elements from one or two
9329 vectors and return a vector of the same type as the input vector(s).
9330 The @var{mask} is an integral vector with the same width (@var{W})
9331 and element count (@var{N}) as the output vector.
9333 The elements of the input vectors are numbered in memory ordering of
9334 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9335 elements of @var{mask} are considered modulo @var{N} in the single-operand
9336 case and modulo @math{2*@var{N}} in the two-operand case.
9338 Consider the following example,
9341 typedef int v4si __attribute__ ((vector_size (16)));
9343 v4si a = @{1,2,3,4@};
9344 v4si b = @{5,6,7,8@};
9345 v4si mask1 = @{0,1,1,3@};
9346 v4si mask2 = @{0,4,2,5@};
9349 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9350 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9353 Note that @code{__builtin_shuffle} is intentionally semantically
9354 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9356 You can declare variables and use them in function calls and returns, as
9357 well as in assignments and some casts. You can specify a vector type as
9358 a return type for a function. Vector types can also be used as function
9359 arguments. It is possible to cast from one vector type to another,
9360 provided they are of the same size (in fact, you can also cast vectors
9361 to and from other datatypes of the same size).
9363 You cannot operate between vectors of different lengths or different
9364 signedness without a cast.
9367 @section Support for @code{offsetof}
9368 @findex __builtin_offsetof
9370 GCC implements for both C and C++ a syntactic extension to implement
9371 the @code{offsetof} macro.
9375 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9377 offsetof_member_designator:
9379 | offsetof_member_designator "." @code{identifier}
9380 | offsetof_member_designator "[" @code{expr} "]"
9383 This extension is sufficient such that
9386 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9390 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9391 may be dependent. In either case, @var{member} may consist of a single
9392 identifier, or a sequence of member accesses and array references.
9394 @node __sync Builtins
9395 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9397 The following built-in functions
9398 are intended to be compatible with those described
9399 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9400 section 7.4. As such, they depart from normal GCC practice by not using
9401 the @samp{__builtin_} prefix and also by being overloaded so that they
9402 work on multiple types.
9404 The definition given in the Intel documentation allows only for the use of
9405 the types @code{int}, @code{long}, @code{long long} or their unsigned
9406 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
9407 size other than the C type @code{_Bool} or the C++ type @code{bool}.
9408 Operations on pointer arguments are performed as if the operands were
9409 of the @code{uintptr_t} type. That is, they are not scaled by the size
9410 of the type to which the pointer points.
9412 These functions are implemented in terms of the @samp{__atomic}
9413 builtins (@pxref{__atomic Builtins}). They should not be used for new
9414 code which should use the @samp{__atomic} builtins instead.
9416 Not all operations are supported by all target processors. If a particular
9417 operation cannot be implemented on the target processor, a warning is
9418 generated and a call to an external function is generated. The external
9419 function carries the same name as the built-in version,
9420 with an additional suffix
9421 @samp{_@var{n}} where @var{n} is the size of the data type.
9423 @c ??? Should we have a mechanism to suppress this warning? This is almost
9424 @c useful for implementing the operation under the control of an external
9427 In most cases, these built-in functions are considered a @dfn{full barrier}.
9429 no memory operand is moved across the operation, either forward or
9430 backward. Further, instructions are issued as necessary to prevent the
9431 processor from speculating loads across the operation and from queuing stores
9432 after the operation.
9434 All of the routines are described in the Intel documentation to take
9435 ``an optional list of variables protected by the memory barrier''. It's
9436 not clear what is meant by that; it could mean that @emph{only} the
9437 listed variables are protected, or it could mean a list of additional
9438 variables to be protected. The list is ignored by GCC which treats it as
9439 empty. GCC interprets an empty list as meaning that all globally
9440 accessible variables should be protected.
9443 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9444 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9445 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9446 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9447 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9448 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9449 @findex __sync_fetch_and_add
9450 @findex __sync_fetch_and_sub
9451 @findex __sync_fetch_and_or
9452 @findex __sync_fetch_and_and
9453 @findex __sync_fetch_and_xor
9454 @findex __sync_fetch_and_nand
9455 These built-in functions perform the operation suggested by the name, and
9456 returns the value that had previously been in memory. That is, operations
9457 on integer operands have the following semantics. Operations on pointer
9458 arguments are performed as if the operands were of the @code{uintptr_t}
9459 type. That is, they are not scaled by the size of the type to which
9463 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9464 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9467 The object pointed to by the first argument must be of integer or pointer
9468 type. It must not be a Boolean type.
9470 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9471 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9473 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9474 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9475 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9476 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9477 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9478 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9479 @findex __sync_add_and_fetch
9480 @findex __sync_sub_and_fetch
9481 @findex __sync_or_and_fetch
9482 @findex __sync_and_and_fetch
9483 @findex __sync_xor_and_fetch
9484 @findex __sync_nand_and_fetch
9485 These built-in functions perform the operation suggested by the name, and
9486 return the new value. That is, operations on integer operands have
9487 the following semantics. Operations on pointer operands are performed as
9488 if the operand's type were @code{uintptr_t}.
9491 @{ *ptr @var{op}= value; return *ptr; @}
9492 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9495 The same constraints on arguments apply as for the corresponding
9496 @code{__sync_op_and_fetch} built-in functions.
9498 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9499 as @code{*ptr = ~(*ptr & value)} instead of
9500 @code{*ptr = ~*ptr & value}.
9502 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9503 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9504 @findex __sync_bool_compare_and_swap
9505 @findex __sync_val_compare_and_swap
9506 These built-in functions perform an atomic compare and swap.
9507 That is, if the current
9508 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9511 The ``bool'' version returns true if the comparison is successful and
9512 @var{newval} is written. The ``val'' version returns the contents
9513 of @code{*@var{ptr}} before the operation.
9515 @item __sync_synchronize (...)
9516 @findex __sync_synchronize
9517 This built-in function issues a full memory barrier.
9519 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9520 @findex __sync_lock_test_and_set
9521 This built-in function, as described by Intel, is not a traditional test-and-set
9522 operation, but rather an atomic exchange operation. It writes @var{value}
9523 into @code{*@var{ptr}}, and returns the previous contents of
9526 Many targets have only minimal support for such locks, and do not support
9527 a full exchange operation. In this case, a target may support reduced
9528 functionality here by which the @emph{only} valid value to store is the
9529 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9530 is implementation defined.
9532 This built-in function is not a full barrier,
9533 but rather an @dfn{acquire barrier}.
9534 This means that references after the operation cannot move to (or be
9535 speculated to) before the operation, but previous memory stores may not
9536 be globally visible yet, and previous memory loads may not yet be
9539 @item void __sync_lock_release (@var{type} *ptr, ...)
9540 @findex __sync_lock_release
9541 This built-in function releases the lock acquired by
9542 @code{__sync_lock_test_and_set}.
9543 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9545 This built-in function is not a full barrier,
9546 but rather a @dfn{release barrier}.
9547 This means that all previous memory stores are globally visible, and all
9548 previous memory loads have been satisfied, but following memory reads
9549 are not prevented from being speculated to before the barrier.
9552 @node __atomic Builtins
9553 @section Built-in Functions for Memory Model Aware Atomic Operations
9555 The following built-in functions approximately match the requirements
9556 for the C++11 memory model. They are all
9557 identified by being prefixed with @samp{__atomic} and most are
9558 overloaded so that they work with multiple types.
9560 These functions are intended to replace the legacy @samp{__sync}
9561 builtins. The main difference is that the memory order that is requested
9562 is a parameter to the functions. New code should always use the
9563 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9565 Note that the @samp{__atomic} builtins assume that programs will
9566 conform to the C++11 memory model. In particular, they assume
9567 that programs are free of data races. See the C++11 standard for
9568 detailed requirements.
9570 The @samp{__atomic} builtins can be used with any integral scalar or
9571 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9572 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9573 supported by the architecture.
9575 The four non-arithmetic functions (load, store, exchange, and
9576 compare_exchange) all have a generic version as well. This generic
9577 version works on any data type. It uses the lock-free built-in function
9578 if the specific data type size makes that possible; otherwise, an
9579 external call is left to be resolved at run time. This external call is
9580 the same format with the addition of a @samp{size_t} parameter inserted
9581 as the first parameter indicating the size of the object being pointed to.
9582 All objects must be the same size.
9584 There are 6 different memory orders that can be specified. These map
9585 to the C++11 memory orders with the same names, see the C++11 standard
9586 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9587 on atomic synchronization} for detailed definitions. Individual
9588 targets may also support additional memory orders for use on specific
9589 architectures. Refer to the target documentation for details of
9592 An atomic operation can both constrain code motion and
9593 be mapped to hardware instructions for synchronization between threads
9594 (e.g., a fence). To which extent this happens is controlled by the
9595 memory orders, which are listed here in approximately ascending order of
9596 strength. The description of each memory order is only meant to roughly
9597 illustrate the effects and is not a specification; see the C++11
9598 memory model for precise semantics.
9601 @item __ATOMIC_RELAXED
9602 Implies no inter-thread ordering constraints.
9603 @item __ATOMIC_CONSUME
9604 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9605 memory order because of a deficiency in C++11's semantics for
9606 @code{memory_order_consume}.
9607 @item __ATOMIC_ACQUIRE
9608 Creates an inter-thread happens-before constraint from the release (or
9609 stronger) semantic store to this acquire load. Can prevent hoisting
9610 of code to before the operation.
9611 @item __ATOMIC_RELEASE
9612 Creates an inter-thread happens-before constraint to acquire (or stronger)
9613 semantic loads that read from this release store. Can prevent sinking
9614 of code to after the operation.
9615 @item __ATOMIC_ACQ_REL
9616 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9617 @code{__ATOMIC_RELEASE}.
9618 @item __ATOMIC_SEQ_CST
9619 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9622 Note that in the C++11 memory model, @emph{fences} (e.g.,
9623 @samp{__atomic_thread_fence}) take effect in combination with other
9624 atomic operations on specific memory locations (e.g., atomic loads);
9625 operations on specific memory locations do not necessarily affect other
9626 operations in the same way.
9628 Target architectures are encouraged to provide their own patterns for
9629 each of the atomic built-in functions. If no target is provided, the original
9630 non-memory model set of @samp{__sync} atomic built-in functions are
9631 used, along with any required synchronization fences surrounding it in
9632 order to achieve the proper behavior. Execution in this case is subject
9633 to the same restrictions as those built-in functions.
9635 If there is no pattern or mechanism to provide a lock-free instruction
9636 sequence, a call is made to an external routine with the same parameters
9637 to be resolved at run time.
9639 When implementing patterns for these built-in functions, the memory order
9640 parameter can be ignored as long as the pattern implements the most
9641 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9642 orders execute correctly with this memory order but they may not execute as
9643 efficiently as they could with a more appropriate implementation of the
9644 relaxed requirements.
9646 Note that the C++11 standard allows for the memory order parameter to be
9647 determined at run time rather than at compile time. These built-in
9648 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9649 than invoke a runtime library call or inline a switch statement. This is
9650 standard compliant, safe, and the simplest approach for now.
9652 The memory order parameter is a signed int, but only the lower 16 bits are
9653 reserved for the memory order. The remainder of the signed int is reserved
9654 for target use and should be 0. Use of the predefined atomic values
9655 ensures proper usage.
9657 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9658 This built-in function implements an atomic load operation. It returns the
9659 contents of @code{*@var{ptr}}.
9661 The valid memory order variants are
9662 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9663 and @code{__ATOMIC_CONSUME}.
9667 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9668 This is the generic version of an atomic load. It returns the
9669 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9673 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9674 This built-in function implements an atomic store operation. It writes
9675 @code{@var{val}} into @code{*@var{ptr}}.
9677 The valid memory order variants are
9678 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9682 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9683 This is the generic version of an atomic store. It stores the value
9684 of @code{*@var{val}} into @code{*@var{ptr}}.
9688 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9689 This built-in function implements an atomic exchange operation. It writes
9690 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9693 The valid memory order variants are
9694 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9695 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9699 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9700 This is the generic version of an atomic exchange. It stores the
9701 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9702 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9706 @deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memorder, int failure_memorder)
9707 This built-in function implements an atomic compare and exchange operation.
9708 This compares the contents of @code{*@var{ptr}} with the contents of
9709 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9710 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9711 equal, the operation is a @emph{read} and the current contents of
9712 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
9713 for weak compare_exchange, which may fail spuriously, and false for
9714 the strong variation, which never fails spuriously. Many targets
9715 only offer the strong variation and ignore the parameter. When in doubt, use
9716 the strong variation.
9718 If @var{desired} is written into @code{*@var{ptr}} then true is returned
9719 and memory is affected according to the
9720 memory order specified by @var{success_memorder}. There are no
9721 restrictions on what memory order can be used here.
9723 Otherwise, false is returned and memory is affected according
9724 to @var{failure_memorder}. This memory order cannot be
9725 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9726 stronger order than that specified by @var{success_memorder}.
9730 @deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memorder, int failure_memorder)
9731 This built-in function implements the generic version of
9732 @code{__atomic_compare_exchange}. The function is virtually identical to
9733 @code{__atomic_compare_exchange_n}, except the desired value is also a
9738 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9739 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9740 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9741 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9742 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9743 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9744 These built-in functions perform the operation suggested by the name, and
9745 return the result of the operation. Operations on pointer arguments are
9746 performed as if the operands were of the @code{uintptr_t} type. That is,
9747 they are not scaled by the size of the type to which the pointer points.
9750 @{ *ptr @var{op}= val; return *ptr; @}
9753 The object pointed to by the first argument must be of integer or pointer
9754 type. It must not be a Boolean type. All memory orders are valid.
9758 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9759 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9760 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9761 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9762 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9763 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9764 These built-in functions perform the operation suggested by the name, and
9765 return the value that had previously been in @code{*@var{ptr}}. Operations
9766 on pointer arguments are performed as if the operands were of
9767 the @code{uintptr_t} type. That is, they are not scaled by the size of
9768 the type to which the pointer points.
9771 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9774 The same constraints on arguments apply as for the corresponding
9775 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
9779 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9781 This built-in function performs an atomic test-and-set operation on
9782 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9783 defined nonzero ``set'' value and the return value is @code{true} if and only
9784 if the previous contents were ``set''.
9785 It should be only used for operands of type @code{bool} or @code{char}. For
9786 other types only part of the value may be set.
9788 All memory orders are valid.
9792 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9794 This built-in function performs an atomic clear operation on
9795 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9796 It should be only used for operands of type @code{bool} or @code{char} and
9797 in conjunction with @code{__atomic_test_and_set}.
9798 For other types it may only clear partially. If the type is not @code{bool}
9799 prefer using @code{__atomic_store}.
9801 The valid memory order variants are
9802 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9803 @code{__ATOMIC_RELEASE}.
9807 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9809 This built-in function acts as a synchronization fence between threads
9810 based on the specified memory order.
9812 All memory orders are valid.
9816 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9818 This built-in function acts as a synchronization fence between a thread
9819 and signal handlers based in the same thread.
9821 All memory orders are valid.
9825 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9827 This built-in function returns true if objects of @var{size} bytes always
9828 generate lock-free atomic instructions for the target architecture.
9829 @var{size} must resolve to a compile-time constant and the result also
9830 resolves to a compile-time constant.
9832 @var{ptr} is an optional pointer to the object that may be used to determine
9833 alignment. A value of 0 indicates typical alignment should be used. The
9834 compiler may also ignore this parameter.
9837 if (__atomic_always_lock_free (sizeof (long long), 0))
9842 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9844 This built-in function returns true if objects of @var{size} bytes always
9845 generate lock-free atomic instructions for the target architecture. If
9846 the built-in function is not known to be lock-free, a call is made to a
9847 runtime routine named @code{__atomic_is_lock_free}.
9849 @var{ptr} is an optional pointer to the object that may be used to determine
9850 alignment. A value of 0 indicates typical alignment should be used. The
9851 compiler may also ignore this parameter.
9854 @node Integer Overflow Builtins
9855 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9857 The following built-in functions allow performing simple arithmetic operations
9858 together with checking whether the operations overflowed.
9860 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9861 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9862 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9863 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9864 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9865 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9866 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9868 These built-in functions promote the first two operands into infinite precision signed
9869 type and perform addition on those promoted operands. The result is then
9870 cast to the type the third pointer argument points to and stored there.
9871 If the stored result is equal to the infinite precision result, the built-in
9872 functions return false, otherwise they return true. As the addition is
9873 performed in infinite signed precision, these built-in functions have fully defined
9874 behavior for all argument values.
9876 The first built-in function allows arbitrary integral types for operands and
9877 the result type must be pointer to some integral type other than enumerated or
9878 Boolean type, the rest of the built-in functions have explicit integer types.
9880 The compiler will attempt to use hardware instructions to implement
9881 these built-in functions where possible, like conditional jump on overflow
9882 after addition, conditional jump on carry etc.
9886 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9887 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9888 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9889 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9890 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9891 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9892 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9894 These built-in functions are similar to the add overflow checking built-in
9895 functions above, except they perform subtraction, subtract the second argument
9896 from the first one, instead of addition.
9900 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9901 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9902 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9903 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9904 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9905 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9906 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9908 These built-in functions are similar to the add overflow checking built-in
9909 functions above, except they perform multiplication, instead of addition.
9913 The following built-in functions allow checking if simple arithmetic operation
9916 @deftypefn {Built-in Function} bool __builtin_add_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9917 @deftypefnx {Built-in Function} bool __builtin_sub_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9918 @deftypefnx {Built-in Function} bool __builtin_mul_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9920 These built-in functions are similar to @code{__builtin_add_overflow},
9921 @code{__builtin_sub_overflow}, or @code{__builtin_mul_overflow}, except that
9922 they don't store the result of the arithmetic operation anywhere and the
9923 last argument is not a pointer, but some expression with integral type other
9924 than enumerated or Boolean type.
9926 The built-in functions promote the first two operands into infinite precision signed type
9927 and perform addition on those promoted operands. The result is then
9928 cast to the type of the third argument. If the cast result is equal to the infinite
9929 precision result, the built-in functions return false, otherwise they return true.
9930 The value of the third argument is ignored, just the side-effects in the third argument
9931 are evaluated, and no integral argument promotions are performed on the last argument.
9932 If the third argument is a bit-field, the type used for the result cast has the
9933 precision and signedness of the given bit-field, rather than precision and signedness
9934 of the underlying type.
9936 For example, the following macro can be used to portably check, at
9937 compile-time, whether or not adding two constant integers will overflow,
9938 and perform the addition only when it is known to be safe and not to trigger
9939 a @option{-Woverflow} warning.
9942 #define INT_ADD_OVERFLOW_P(a, b) \
9943 __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0)
9947 C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B,
9948 D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0)
9952 The compiler will attempt to use hardware instructions to implement
9953 these built-in functions where possible, like conditional jump on overflow
9954 after addition, conditional jump on carry etc.
9958 @node x86 specific memory model extensions for transactional memory
9959 @section x86-Specific Memory Model Extensions for Transactional Memory
9961 The x86 architecture supports additional memory ordering flags
9962 to mark lock critical sections for hardware lock elision.
9963 These must be specified in addition to an existing memory order to
9967 @item __ATOMIC_HLE_ACQUIRE
9968 Start lock elision on a lock variable.
9969 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9970 @item __ATOMIC_HLE_RELEASE
9971 End lock elision on a lock variable.
9972 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9975 When a lock acquire fails, it is required for good performance to abort
9976 the transaction quickly. This can be done with a @code{_mm_pause}.
9979 #include <immintrin.h> // For _mm_pause
9983 /* Acquire lock with lock elision */
9984 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9985 _mm_pause(); /* Abort failed transaction */
9987 /* Free lock with lock elision */
9988 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9991 @node Object Size Checking
9992 @section Object Size Checking Built-in Functions
9993 @findex __builtin_object_size
9994 @findex __builtin___memcpy_chk
9995 @findex __builtin___mempcpy_chk
9996 @findex __builtin___memmove_chk
9997 @findex __builtin___memset_chk
9998 @findex __builtin___strcpy_chk
9999 @findex __builtin___stpcpy_chk
10000 @findex __builtin___strncpy_chk
10001 @findex __builtin___strcat_chk
10002 @findex __builtin___strncat_chk
10003 @findex __builtin___sprintf_chk
10004 @findex __builtin___snprintf_chk
10005 @findex __builtin___vsprintf_chk
10006 @findex __builtin___vsnprintf_chk
10007 @findex __builtin___printf_chk
10008 @findex __builtin___vprintf_chk
10009 @findex __builtin___fprintf_chk
10010 @findex __builtin___vfprintf_chk
10012 GCC implements a limited buffer overflow protection mechanism
10013 that can prevent some buffer overflow attacks.
10015 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
10016 is a built-in construct that returns a constant number of bytes from
10017 @var{ptr} to the end of the object @var{ptr} pointer points to
10018 (if known at compile time). @code{__builtin_object_size} never evaluates
10019 its arguments for side-effects. If there are any side-effects in them, it
10020 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10021 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
10022 point to and all of them are known at compile time, the returned number
10023 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
10024 0 and minimum if nonzero. If it is not possible to determine which objects
10025 @var{ptr} points to at compile time, @code{__builtin_object_size} should
10026 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10027 for @var{type} 2 or 3.
10029 @var{type} is an integer constant from 0 to 3. If the least significant
10030 bit is clear, objects are whole variables, if it is set, a closest
10031 surrounding subobject is considered the object a pointer points to.
10032 The second bit determines if maximum or minimum of remaining bytes
10036 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
10037 char *p = &var.buf1[1], *q = &var.b;
10039 /* Here the object p points to is var. */
10040 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
10041 /* The subobject p points to is var.buf1. */
10042 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
10043 /* The object q points to is var. */
10044 assert (__builtin_object_size (q, 0)
10045 == (char *) (&var + 1) - (char *) &var.b);
10046 /* The subobject q points to is var.b. */
10047 assert (__builtin_object_size (q, 1) == sizeof (var.b));
10051 There are built-in functions added for many common string operation
10052 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
10053 built-in is provided. This built-in has an additional last argument,
10054 which is the number of bytes remaining in object the @var{dest}
10055 argument points to or @code{(size_t) -1} if the size is not known.
10057 The built-in functions are optimized into the normal string functions
10058 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
10059 it is known at compile time that the destination object will not
10060 be overflown. If the compiler can determine at compile time the
10061 object will be always overflown, it issues a warning.
10063 The intended use can be e.g.@:
10067 #define bos0(dest) __builtin_object_size (dest, 0)
10068 #define memcpy(dest, src, n) \
10069 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
10073 /* It is unknown what object p points to, so this is optimized
10074 into plain memcpy - no checking is possible. */
10075 memcpy (p, "abcde", n);
10076 /* Destination is known and length too. It is known at compile
10077 time there will be no overflow. */
10078 memcpy (&buf[5], "abcde", 5);
10079 /* Destination is known, but the length is not known at compile time.
10080 This will result in __memcpy_chk call that can check for overflow
10082 memcpy (&buf[5], "abcde", n);
10083 /* Destination is known and it is known at compile time there will
10084 be overflow. There will be a warning and __memcpy_chk call that
10085 will abort the program at run time. */
10086 memcpy (&buf[6], "abcde", 5);
10089 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
10090 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
10091 @code{strcat} and @code{strncat}.
10093 There are also checking built-in functions for formatted output functions.
10095 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
10096 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10097 const char *fmt, ...);
10098 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
10100 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10101 const char *fmt, va_list ap);
10104 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
10105 etc.@: functions and can contain implementation specific flags on what
10106 additional security measures the checking function might take, such as
10107 handling @code{%n} differently.
10109 The @var{os} argument is the object size @var{s} points to, like in the
10110 other built-in functions. There is a small difference in the behavior
10111 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
10112 optimized into the non-checking functions only if @var{flag} is 0, otherwise
10113 the checking function is called with @var{os} argument set to
10114 @code{(size_t) -1}.
10116 In addition to this, there are checking built-in functions
10117 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
10118 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
10119 These have just one additional argument, @var{flag}, right before
10120 format string @var{fmt}. If the compiler is able to optimize them to
10121 @code{fputc} etc.@: functions, it does, otherwise the checking function
10122 is called and the @var{flag} argument passed to it.
10124 @node Pointer Bounds Checker builtins
10125 @section Pointer Bounds Checker Built-in Functions
10126 @cindex Pointer Bounds Checker builtins
10127 @findex __builtin___bnd_set_ptr_bounds
10128 @findex __builtin___bnd_narrow_ptr_bounds
10129 @findex __builtin___bnd_copy_ptr_bounds
10130 @findex __builtin___bnd_init_ptr_bounds
10131 @findex __builtin___bnd_null_ptr_bounds
10132 @findex __builtin___bnd_store_ptr_bounds
10133 @findex __builtin___bnd_chk_ptr_lbounds
10134 @findex __builtin___bnd_chk_ptr_ubounds
10135 @findex __builtin___bnd_chk_ptr_bounds
10136 @findex __builtin___bnd_get_ptr_lbound
10137 @findex __builtin___bnd_get_ptr_ubound
10139 GCC provides a set of built-in functions to control Pointer Bounds Checker
10140 instrumentation. Note that all Pointer Bounds Checker builtins can be used
10141 even if you compile with Pointer Bounds Checker off
10142 (@option{-fno-check-pointer-bounds}).
10143 The behavior may differ in such case as documented below.
10145 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
10147 This built-in function returns a new pointer with the value of @var{q}, and
10148 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
10149 Bounds Checker off, the built-in function just returns the first argument.
10152 extern void *__wrap_malloc (size_t n)
10154 void *p = (void *)__real_malloc (n);
10155 if (!p) return __builtin___bnd_null_ptr_bounds (p);
10156 return __builtin___bnd_set_ptr_bounds (p, n);
10162 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
10164 This built-in function returns a new pointer with the value of @var{p}
10165 and associates it with the narrowed bounds formed by the intersection
10166 of bounds associated with @var{q} and the bounds
10167 [@var{p}, @var{p} + @var{size} - 1].
10168 With Pointer Bounds Checker off, the built-in function just returns the first
10172 void init_objects (object *objs, size_t size)
10175 /* Initialize objects one-by-one passing pointers with bounds of
10176 an object, not the full array of objects. */
10177 for (i = 0; i < size; i++)
10178 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
10185 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
10187 This built-in function returns a new pointer with the value of @var{q},
10188 and associates it with the bounds already associated with pointer @var{r}.
10189 With Pointer Bounds Checker off, the built-in function just returns the first
10193 /* Here is a way to get pointer to object's field but
10194 still with the full object's bounds. */
10195 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10201 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10203 This built-in function returns a new pointer with the value of @var{q}, and
10204 associates it with INIT (allowing full memory access) bounds. With Pointer
10205 Bounds Checker off, the built-in function just returns the first argument.
10209 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10211 This built-in function returns a new pointer with the value of @var{q}, and
10212 associates it with NULL (allowing no memory access) bounds. With Pointer
10213 Bounds Checker off, the built-in function just returns the first argument.
10217 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10219 This built-in function stores the bounds associated with pointer @var{ptr_val}
10220 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10221 bounds from legacy code without touching the associated pointer's memory when
10222 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10223 function call is ignored.
10227 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10229 This built-in function checks if the pointer @var{q} is within the lower
10230 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10231 function call is ignored.
10234 extern void *__wrap_memset (void *dst, int c, size_t len)
10238 __builtin___bnd_chk_ptr_lbounds (dst);
10239 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10240 __real_memset (dst, c, len);
10248 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10250 This built-in function checks if the pointer @var{q} is within the upper
10251 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10252 function call is ignored.
10256 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10258 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10259 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10260 off, the built-in function call is ignored.
10263 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10267 __bnd_chk_ptr_bounds (dst, n);
10268 __bnd_chk_ptr_bounds (src, n);
10269 __real_memcpy (dst, src, n);
10277 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10279 This built-in function returns the lower bound associated
10280 with the pointer @var{q}, as a pointer value.
10281 This is useful for debugging using @code{printf}.
10282 With Pointer Bounds Checker off, the built-in function returns 0.
10285 void *lb = __builtin___bnd_get_ptr_lbound (q);
10286 void *ub = __builtin___bnd_get_ptr_ubound (q);
10287 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10292 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10294 This built-in function returns the upper bound (which is a pointer) associated
10295 with the pointer @var{q}. With Pointer Bounds Checker off,
10296 the built-in function returns -1.
10300 @node Cilk Plus Builtins
10301 @section Cilk Plus C/C++ Language Extension Built-in Functions
10303 GCC provides support for the following built-in reduction functions if Cilk Plus
10304 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10307 @item @code{__sec_implicit_index}
10308 @item @code{__sec_reduce}
10309 @item @code{__sec_reduce_add}
10310 @item @code{__sec_reduce_all_nonzero}
10311 @item @code{__sec_reduce_all_zero}
10312 @item @code{__sec_reduce_any_nonzero}
10313 @item @code{__sec_reduce_any_zero}
10314 @item @code{__sec_reduce_max}
10315 @item @code{__sec_reduce_min}
10316 @item @code{__sec_reduce_max_ind}
10317 @item @code{__sec_reduce_min_ind}
10318 @item @code{__sec_reduce_mul}
10319 @item @code{__sec_reduce_mutating}
10322 Further details and examples about these built-in functions are described
10323 in the Cilk Plus language manual which can be found at
10324 @uref{http://www.cilkplus.org}.
10326 @node Other Builtins
10327 @section Other Built-in Functions Provided by GCC
10328 @cindex built-in functions
10329 @findex __builtin_alloca
10330 @findex __builtin_alloca_with_align
10331 @findex __builtin_call_with_static_chain
10332 @findex __builtin_fpclassify
10333 @findex __builtin_isfinite
10334 @findex __builtin_isnormal
10335 @findex __builtin_isgreater
10336 @findex __builtin_isgreaterequal
10337 @findex __builtin_isinf_sign
10338 @findex __builtin_isless
10339 @findex __builtin_islessequal
10340 @findex __builtin_islessgreater
10341 @findex __builtin_isunordered
10342 @findex __builtin_powi
10343 @findex __builtin_powif
10344 @findex __builtin_powil
10505 @findex fprintf_unlocked
10507 @findex fputs_unlocked
10615 @findex nexttowardf
10616 @findex nexttowardl
10624 @findex printf_unlocked
10654 @findex signbitd128
10655 @findex significand
10656 @findex significandf
10657 @findex significandl
10685 @findex strncasecmp
10728 GCC provides a large number of built-in functions other than the ones
10729 mentioned above. Some of these are for internal use in the processing
10730 of exceptions or variable-length argument lists and are not
10731 documented here because they may change from time to time; we do not
10732 recommend general use of these functions.
10734 The remaining functions are provided for optimization purposes.
10736 With the exception of built-ins that have library equivalents such as
10737 the standard C library functions discussed below, or that expand to
10738 library calls, GCC built-in functions are always expanded inline and
10739 thus do not have corresponding entry points and their address cannot
10740 be obtained. Attempting to use them in an expression other than
10741 a function call results in a compile-time error.
10743 @opindex fno-builtin
10744 GCC includes built-in versions of many of the functions in the standard
10745 C library. These functions come in two forms: one whose names start with
10746 the @code{__builtin_} prefix, and the other without. Both forms have the
10747 same type (including prototype), the same address (when their address is
10748 taken), and the same meaning as the C library functions even if you specify
10749 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10750 functions are only optimized in certain cases; if they are not optimized in
10751 a particular case, a call to the library function is emitted.
10755 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10756 @option{-std=c99} or @option{-std=c11}), the functions
10757 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10758 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10759 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10760 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10761 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10762 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10763 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10764 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10765 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10766 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10767 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10768 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10769 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10770 @code{significandl}, @code{significand}, @code{sincosf},
10771 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10772 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10773 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10774 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10776 may be handled as built-in functions.
10777 All these functions have corresponding versions
10778 prefixed with @code{__builtin_}, which may be used even in strict C90
10781 The ISO C99 functions
10782 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10783 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10784 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10785 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10786 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10787 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10788 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10789 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10790 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10791 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10792 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10793 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10794 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10795 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10796 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10797 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10798 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10799 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10800 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10801 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10802 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10803 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10804 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10805 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10806 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10807 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10808 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10809 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10810 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10811 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10812 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10813 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10814 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10815 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10816 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10817 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10818 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10819 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10820 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10821 are handled as built-in functions
10822 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10824 There are also built-in versions of the ISO C99 functions
10825 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10826 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10827 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10828 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10829 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10830 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10831 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10832 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10833 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10834 that are recognized in any mode since ISO C90 reserves these names for
10835 the purpose to which ISO C99 puts them. All these functions have
10836 corresponding versions prefixed with @code{__builtin_}.
10838 There are also GNU extension functions @code{clog10}, @code{clog10f} and
10839 @code{clog10l} which names are reserved by ISO C99 for future use.
10840 All these functions have versions prefixed with @code{__builtin_}.
10842 The ISO C94 functions
10843 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10844 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10845 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10847 are handled as built-in functions
10848 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10850 The ISO C90 functions
10851 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10852 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10853 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10854 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10855 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10856 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10857 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10858 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10859 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10860 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10861 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10862 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10863 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10864 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10865 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10866 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10867 are all recognized as built-in functions unless
10868 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10869 is specified for an individual function). All of these functions have
10870 corresponding versions prefixed with @code{__builtin_}.
10872 GCC provides built-in versions of the ISO C99 floating-point comparison
10873 macros that avoid raising exceptions for unordered operands. They have
10874 the same names as the standard macros ( @code{isgreater},
10875 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10876 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10877 prefixed. We intend for a library implementor to be able to simply
10878 @code{#define} each standard macro to its built-in equivalent.
10879 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10880 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10881 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10882 built-in functions appear both with and without the @code{__builtin_} prefix.
10884 @deftypefn {Built-in Function} void *__builtin_alloca (size_t size)
10885 The @code{__builtin_alloca} function must be called at block scope.
10886 The function allocates an object @var{size} bytes large on the stack
10887 of the calling function. The object is aligned on the default stack
10888 alignment boundary for the target determined by the
10889 @code{__BIGGEST_ALIGNMENT__} macro. The @code{__builtin_alloca}
10890 function returns a pointer to the first byte of the allocated object.
10891 The lifetime of the allocated object ends just before the calling
10892 function returns to its caller. This is so even when
10893 @code{__builtin_alloca} is called within a nested block.
10895 For example, the following function allocates eight objects of @code{n}
10896 bytes each on the stack, storing a pointer to each in consecutive elements
10897 of the array @code{a}. It then passes the array to function @code{g}
10898 which can safely use the storage pointed to by each of the array elements.
10901 void f (unsigned n)
10904 for (int i = 0; i != 8; ++i)
10905 a [i] = __builtin_alloca (n);
10907 g (a, n); // @r{safe}
10911 Since the @code{__builtin_alloca} function doesn't validate its argument
10912 it is the responsibility of its caller to make sure the argument doesn't
10913 cause it to exceed the stack size limit.
10914 The @code{__builtin_alloca} function is provided to make it possible to
10915 allocate on the stack arrays of bytes with an upper bound that may be
10916 computed at run time. Since C99 Variable Length Arrays offer
10917 similar functionality under a portable, more convenient, and safer
10918 interface they are recommended instead, in both C99 and C++ programs
10919 where GCC provides them as an extension.
10920 @xref{Variable Length}, for details.
10924 @deftypefn {Built-in Function} void *__builtin_alloca_with_align (size_t size, size_t alignment)
10925 The @code{__builtin_alloca_with_align} function must be called at block
10926 scope. The function allocates an object @var{size} bytes large on
10927 the stack of the calling function. The allocated object is aligned on
10928 the boundary specified by the argument @var{alignment} whose unit is given
10929 in bits (not bytes). The @var{size} argument must be positive and not
10930 exceed the stack size limit. The @var{alignment} argument must be a constant
10931 integer expression that evaluates to a power of 2 greater than or equal to
10932 @code{CHAR_BIT} and less than some unspecified maximum. Invocations
10933 with other values are rejected with an error indicating the valid bounds.
10934 The function returns a pointer to the first byte of the allocated object.
10935 The lifetime of the allocated object ends at the end of the block in which
10936 the function was called. The allocated storage is released no later than
10937 just before the calling function returns to its caller, but may be released
10938 at the end of the block in which the function was called.
10940 For example, in the following function the call to @code{g} is unsafe
10941 because when @code{overalign} is non-zero, the space allocated by
10942 @code{__builtin_alloca_with_align} may have been released at the end
10943 of the @code{if} statement in which it was called.
10946 void f (unsigned n, bool overalign)
10950 p = __builtin_alloca_with_align (n, 64 /* bits */);
10952 p = __builtin_alloc (n);
10954 g (p, n); // @r{unsafe}
10958 Since the @code{__builtin_alloca_with_align} function doesn't validate its
10959 @var{size} argument it is the responsibility of its caller to make sure
10960 the argument doesn't cause it to exceed the stack size limit.
10961 The @code{__builtin_alloca_with_align} function is provided to make
10962 it possible to allocate on the stack overaligned arrays of bytes with
10963 an upper bound that may be computed at run time. Since C99
10964 Variable Length Arrays offer the same functionality under
10965 a portable, more convenient, and safer interface they are recommended
10966 instead, in both C99 and C++ programs where GCC provides them as
10967 an extension. @xref{Variable Length}, for details.
10971 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10973 You can use the built-in function @code{__builtin_types_compatible_p} to
10974 determine whether two types are the same.
10976 This built-in function returns 1 if the unqualified versions of the
10977 types @var{type1} and @var{type2} (which are types, not expressions) are
10978 compatible, 0 otherwise. The result of this built-in function can be
10979 used in integer constant expressions.
10981 This built-in function ignores top level qualifiers (e.g., @code{const},
10982 @code{volatile}). For example, @code{int} is equivalent to @code{const
10985 The type @code{int[]} and @code{int[5]} are compatible. On the other
10986 hand, @code{int} and @code{char *} are not compatible, even if the size
10987 of their types, on the particular architecture are the same. Also, the
10988 amount of pointer indirection is taken into account when determining
10989 similarity. Consequently, @code{short *} is not similar to
10990 @code{short **}. Furthermore, two types that are typedefed are
10991 considered compatible if their underlying types are compatible.
10993 An @code{enum} type is not considered to be compatible with another
10994 @code{enum} type even if both are compatible with the same integer
10995 type; this is what the C standard specifies.
10996 For example, @code{enum @{foo, bar@}} is not similar to
10997 @code{enum @{hot, dog@}}.
10999 You typically use this function in code whose execution varies
11000 depending on the arguments' types. For example:
11005 typeof (x) tmp = (x); \
11006 if (__builtin_types_compatible_p (typeof (x), long double)) \
11007 tmp = foo_long_double (tmp); \
11008 else if (__builtin_types_compatible_p (typeof (x), double)) \
11009 tmp = foo_double (tmp); \
11010 else if (__builtin_types_compatible_p (typeof (x), float)) \
11011 tmp = foo_float (tmp); \
11018 @emph{Note:} This construct is only available for C@.
11022 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
11024 The @var{call_exp} expression must be a function call, and the
11025 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
11026 is passed to the function call in the target's static chain location.
11027 The result of builtin is the result of the function call.
11029 @emph{Note:} This builtin is only available for C@.
11030 This builtin can be used to call Go closures from C.
11034 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
11036 You can use the built-in function @code{__builtin_choose_expr} to
11037 evaluate code depending on the value of a constant expression. This
11038 built-in function returns @var{exp1} if @var{const_exp}, which is an
11039 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
11041 This built-in function is analogous to the @samp{? :} operator in C,
11042 except that the expression returned has its type unaltered by promotion
11043 rules. Also, the built-in function does not evaluate the expression
11044 that is not chosen. For example, if @var{const_exp} evaluates to true,
11045 @var{exp2} is not evaluated even if it has side-effects.
11047 This built-in function can return an lvalue if the chosen argument is an
11050 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
11051 type. Similarly, if @var{exp2} is returned, its return type is the same
11058 __builtin_choose_expr ( \
11059 __builtin_types_compatible_p (typeof (x), double), \
11061 __builtin_choose_expr ( \
11062 __builtin_types_compatible_p (typeof (x), float), \
11064 /* @r{The void expression results in a compile-time error} \
11065 @r{when assigning the result to something.} */ \
11069 @emph{Note:} This construct is only available for C@. Furthermore, the
11070 unused expression (@var{exp1} or @var{exp2} depending on the value of
11071 @var{const_exp}) may still generate syntax errors. This may change in
11076 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
11078 The built-in function @code{__builtin_complex} is provided for use in
11079 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
11080 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
11081 real binary floating-point type, and the result has the corresponding
11082 complex type with real and imaginary parts @var{real} and @var{imag}.
11083 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
11084 infinities, NaNs and negative zeros are involved.
11088 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
11089 You can use the built-in function @code{__builtin_constant_p} to
11090 determine if a value is known to be constant at compile time and hence
11091 that GCC can perform constant-folding on expressions involving that
11092 value. The argument of the function is the value to test. The function
11093 returns the integer 1 if the argument is known to be a compile-time
11094 constant and 0 if it is not known to be a compile-time constant. A
11095 return of 0 does not indicate that the value is @emph{not} a constant,
11096 but merely that GCC cannot prove it is a constant with the specified
11097 value of the @option{-O} option.
11099 You typically use this function in an embedded application where
11100 memory is a critical resource. If you have some complex calculation,
11101 you may want it to be folded if it involves constants, but need to call
11102 a function if it does not. For example:
11105 #define Scale_Value(X) \
11106 (__builtin_constant_p (X) \
11107 ? ((X) * SCALE + OFFSET) : Scale (X))
11110 You may use this built-in function in either a macro or an inline
11111 function. However, if you use it in an inlined function and pass an
11112 argument of the function as the argument to the built-in, GCC
11113 never returns 1 when you call the inline function with a string constant
11114 or compound literal (@pxref{Compound Literals}) and does not return 1
11115 when you pass a constant numeric value to the inline function unless you
11116 specify the @option{-O} option.
11118 You may also use @code{__builtin_constant_p} in initializers for static
11119 data. For instance, you can write
11122 static const int table[] = @{
11123 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
11129 This is an acceptable initializer even if @var{EXPRESSION} is not a
11130 constant expression, including the case where
11131 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
11132 folded to a constant but @var{EXPRESSION} contains operands that are
11133 not otherwise permitted in a static initializer (for example,
11134 @code{0 && foo ()}). GCC must be more conservative about evaluating the
11135 built-in in this case, because it has no opportunity to perform
11139 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
11140 @opindex fprofile-arcs
11141 You may use @code{__builtin_expect} to provide the compiler with
11142 branch prediction information. In general, you should prefer to
11143 use actual profile feedback for this (@option{-fprofile-arcs}), as
11144 programmers are notoriously bad at predicting how their programs
11145 actually perform. However, there are applications in which this
11146 data is hard to collect.
11148 The return value is the value of @var{exp}, which should be an integral
11149 expression. The semantics of the built-in are that it is expected that
11150 @var{exp} == @var{c}. For example:
11153 if (__builtin_expect (x, 0))
11158 indicates that we do not expect to call @code{foo}, since
11159 we expect @code{x} to be zero. Since you are limited to integral
11160 expressions for @var{exp}, you should use constructions such as
11163 if (__builtin_expect (ptr != NULL, 1))
11168 when testing pointer or floating-point values.
11171 @deftypefn {Built-in Function} void __builtin_trap (void)
11172 This function causes the program to exit abnormally. GCC implements
11173 this function by using a target-dependent mechanism (such as
11174 intentionally executing an illegal instruction) or by calling
11175 @code{abort}. The mechanism used may vary from release to release so
11176 you should not rely on any particular implementation.
11179 @deftypefn {Built-in Function} void __builtin_unreachable (void)
11180 If control flow reaches the point of the @code{__builtin_unreachable},
11181 the program is undefined. It is useful in situations where the
11182 compiler cannot deduce the unreachability of the code.
11184 One such case is immediately following an @code{asm} statement that
11185 either never terminates, or one that transfers control elsewhere
11186 and never returns. In this example, without the
11187 @code{__builtin_unreachable}, GCC issues a warning that control
11188 reaches the end of a non-void function. It also generates code
11189 to return after the @code{asm}.
11192 int f (int c, int v)
11200 asm("jmp error_handler");
11201 __builtin_unreachable ();
11207 Because the @code{asm} statement unconditionally transfers control out
11208 of the function, control never reaches the end of the function
11209 body. The @code{__builtin_unreachable} is in fact unreachable and
11210 communicates this fact to the compiler.
11212 Another use for @code{__builtin_unreachable} is following a call a
11213 function that never returns but that is not declared
11214 @code{__attribute__((noreturn))}, as in this example:
11217 void function_that_never_returns (void);
11227 function_that_never_returns ();
11228 __builtin_unreachable ();
11235 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
11236 This function returns its first argument, and allows the compiler
11237 to assume that the returned pointer is at least @var{align} bytes
11238 aligned. This built-in can have either two or three arguments,
11239 if it has three, the third argument should have integer type, and
11240 if it is nonzero means misalignment offset. For example:
11243 void *x = __builtin_assume_aligned (arg, 16);
11247 means that the compiler can assume @code{x}, set to @code{arg}, is at least
11248 16-byte aligned, while:
11251 void *x = __builtin_assume_aligned (arg, 32, 8);
11255 means that the compiler can assume for @code{x}, set to @code{arg}, that
11256 @code{(char *) x - 8} is 32-byte aligned.
11259 @deftypefn {Built-in Function} int __builtin_LINE ()
11260 This function is the equivalent of the preprocessor @code{__LINE__}
11261 macro and returns a constant integer expression that evaluates to
11262 the line number of the invocation of the built-in. When used as a C++
11263 default argument for a function @var{F}, it returns the line number
11264 of the call to @var{F}.
11267 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
11268 This function is the equivalent of the @code{__FUNCTION__} symbol
11269 and returns an address constant pointing to the name of the function
11270 from which the built-in was invoked, or the empty string if
11271 the invocation is not at function scope. When used as a C++ default
11272 argument for a function @var{F}, it returns the name of @var{F}'s
11273 caller or the empty string if the call was not made at function
11277 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
11278 This function is the equivalent of the preprocessor @code{__FILE__}
11279 macro and returns an address constant pointing to the file name
11280 containing the invocation of the built-in, or the empty string if
11281 the invocation is not at function scope. When used as a C++ default
11282 argument for a function @var{F}, it returns the file name of the call
11283 to @var{F} or the empty string if the call was not made at function
11286 For example, in the following, each call to function @code{foo} will
11287 print a line similar to @code{"file.c:123: foo: message"} with the name
11288 of the file and the line number of the @code{printf} call, the name of
11289 the function @code{foo}, followed by the word @code{message}.
11293 function (const char *func = __builtin_FUNCTION ())
11300 printf ("%s:%i: %s: message\n", file (), line (), function ());
11306 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
11307 This function is used to flush the processor's instruction cache for
11308 the region of memory between @var{begin} inclusive and @var{end}
11309 exclusive. Some targets require that the instruction cache be
11310 flushed, after modifying memory containing code, in order to obtain
11311 deterministic behavior.
11313 If the target does not require instruction cache flushes,
11314 @code{__builtin___clear_cache} has no effect. Otherwise either
11315 instructions are emitted in-line to clear the instruction cache or a
11316 call to the @code{__clear_cache} function in libgcc is made.
11319 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
11320 This function is used to minimize cache-miss latency by moving data into
11321 a cache before it is accessed.
11322 You can insert calls to @code{__builtin_prefetch} into code for which
11323 you know addresses of data in memory that is likely to be accessed soon.
11324 If the target supports them, data prefetch instructions are generated.
11325 If the prefetch is done early enough before the access then the data will
11326 be in the cache by the time it is accessed.
11328 The value of @var{addr} is the address of the memory to prefetch.
11329 There are two optional arguments, @var{rw} and @var{locality}.
11330 The value of @var{rw} is a compile-time constant one or zero; one
11331 means that the prefetch is preparing for a write to the memory address
11332 and zero, the default, means that the prefetch is preparing for a read.
11333 The value @var{locality} must be a compile-time constant integer between
11334 zero and three. A value of zero means that the data has no temporal
11335 locality, so it need not be left in the cache after the access. A value
11336 of three means that the data has a high degree of temporal locality and
11337 should be left in all levels of cache possible. Values of one and two
11338 mean, respectively, a low or moderate degree of temporal locality. The
11342 for (i = 0; i < n; i++)
11344 a[i] = a[i] + b[i];
11345 __builtin_prefetch (&a[i+j], 1, 1);
11346 __builtin_prefetch (&b[i+j], 0, 1);
11351 Data prefetch does not generate faults if @var{addr} is invalid, but
11352 the address expression itself must be valid. For example, a prefetch
11353 of @code{p->next} does not fault if @code{p->next} is not a valid
11354 address, but evaluation faults if @code{p} is not a valid address.
11356 If the target does not support data prefetch, the address expression
11357 is evaluated if it includes side effects but no other code is generated
11358 and GCC does not issue a warning.
11361 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11362 Returns a positive infinity, if supported by the floating-point format,
11363 else @code{DBL_MAX}. This function is suitable for implementing the
11364 ISO C macro @code{HUGE_VAL}.
11367 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11368 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11371 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11372 Similar to @code{__builtin_huge_val}, except the return
11373 type is @code{long double}.
11376 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11377 This built-in implements the C99 fpclassify functionality. The first
11378 five int arguments should be the target library's notion of the
11379 possible FP classes and are used for return values. They must be
11380 constant values and they must appear in this order: @code{FP_NAN},
11381 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11382 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11383 to classify. GCC treats the last argument as type-generic, which
11384 means it does not do default promotion from float to double.
11387 @deftypefn {Built-in Function} double __builtin_inf (void)
11388 Similar to @code{__builtin_huge_val}, except a warning is generated
11389 if the target floating-point format does not support infinities.
11392 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11393 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11396 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11397 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11400 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11401 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11404 @deftypefn {Built-in Function} float __builtin_inff (void)
11405 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11406 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11409 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11410 Similar to @code{__builtin_inf}, except the return
11411 type is @code{long double}.
11414 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11415 Similar to @code{isinf}, except the return value is -1 for
11416 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11417 Note while the parameter list is an
11418 ellipsis, this function only accepts exactly one floating-point
11419 argument. GCC treats this parameter as type-generic, which means it
11420 does not do default promotion from float to double.
11423 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11424 This is an implementation of the ISO C99 function @code{nan}.
11426 Since ISO C99 defines this function in terms of @code{strtod}, which we
11427 do not implement, a description of the parsing is in order. The string
11428 is parsed as by @code{strtol}; that is, the base is recognized by
11429 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11430 in the significand such that the least significant bit of the number
11431 is at the least significant bit of the significand. The number is
11432 truncated to fit the significand field provided. The significand is
11433 forced to be a quiet NaN@.
11435 This function, if given a string literal all of which would have been
11436 consumed by @code{strtol}, is evaluated early enough that it is considered a
11437 compile-time constant.
11440 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11441 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11444 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11445 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11448 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11449 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11452 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11453 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11456 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11457 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11460 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11461 Similar to @code{__builtin_nan}, except the significand is forced
11462 to be a signaling NaN@. The @code{nans} function is proposed by
11463 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11466 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11467 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11470 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11471 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11474 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11475 Returns one plus the index of the least significant 1-bit of @var{x}, or
11476 if @var{x} is zero, returns zero.
11479 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11480 Returns the number of leading 0-bits in @var{x}, starting at the most
11481 significant bit position. If @var{x} is 0, the result is undefined.
11484 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11485 Returns the number of trailing 0-bits in @var{x}, starting at the least
11486 significant bit position. If @var{x} is 0, the result is undefined.
11489 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11490 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11491 number of bits following the most significant bit that are identical
11492 to it. There are no special cases for 0 or other values.
11495 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11496 Returns the number of 1-bits in @var{x}.
11499 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11500 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11504 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11505 Similar to @code{__builtin_ffs}, except the argument type is
11509 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11510 Similar to @code{__builtin_clz}, except the argument type is
11511 @code{unsigned long}.
11514 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11515 Similar to @code{__builtin_ctz}, except the argument type is
11516 @code{unsigned long}.
11519 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11520 Similar to @code{__builtin_clrsb}, except the argument type is
11524 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11525 Similar to @code{__builtin_popcount}, except the argument type is
11526 @code{unsigned long}.
11529 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11530 Similar to @code{__builtin_parity}, except the argument type is
11531 @code{unsigned long}.
11534 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11535 Similar to @code{__builtin_ffs}, except the argument type is
11539 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11540 Similar to @code{__builtin_clz}, except the argument type is
11541 @code{unsigned long long}.
11544 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11545 Similar to @code{__builtin_ctz}, except the argument type is
11546 @code{unsigned long long}.
11549 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11550 Similar to @code{__builtin_clrsb}, except the argument type is
11554 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11555 Similar to @code{__builtin_popcount}, except the argument type is
11556 @code{unsigned long long}.
11559 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11560 Similar to @code{__builtin_parity}, except the argument type is
11561 @code{unsigned long long}.
11564 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11565 Returns the first argument raised to the power of the second. Unlike the
11566 @code{pow} function no guarantees about precision and rounding are made.
11569 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11570 Similar to @code{__builtin_powi}, except the argument and return types
11574 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11575 Similar to @code{__builtin_powi}, except the argument and return types
11576 are @code{long double}.
11579 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11580 Returns @var{x} with the order of the bytes reversed; for example,
11581 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11585 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11586 Similar to @code{__builtin_bswap16}, except the argument and return types
11590 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11591 Similar to @code{__builtin_bswap32}, except the argument and return types
11595 @node Target Builtins
11596 @section Built-in Functions Specific to Particular Target Machines
11598 On some target machines, GCC supports many built-in functions specific
11599 to those machines. Generally these generate calls to specific machine
11600 instructions, but allow the compiler to schedule those calls.
11603 * AArch64 Built-in Functions::
11604 * Alpha Built-in Functions::
11605 * Altera Nios II Built-in Functions::
11606 * ARC Built-in Functions::
11607 * ARC SIMD Built-in Functions::
11608 * ARM iWMMXt Built-in Functions::
11609 * ARM C Language Extensions (ACLE)::
11610 * ARM Floating Point Status and Control Intrinsics::
11611 * AVR Built-in Functions::
11612 * Blackfin Built-in Functions::
11613 * FR-V Built-in Functions::
11614 * MIPS DSP Built-in Functions::
11615 * MIPS Paired-Single Support::
11616 * MIPS Loongson Built-in Functions::
11617 * MIPS SIMD Architecture (MSA) Support::
11618 * Other MIPS Built-in Functions::
11619 * MSP430 Built-in Functions::
11620 * NDS32 Built-in Functions::
11621 * picoChip Built-in Functions::
11622 * PowerPC Built-in Functions::
11623 * PowerPC AltiVec/VSX Built-in Functions::
11624 * PowerPC Hardware Transactional Memory Built-in Functions::
11625 * RX Built-in Functions::
11626 * S/390 System z Built-in Functions::
11627 * SH Built-in Functions::
11628 * SPARC VIS Built-in Functions::
11629 * SPU Built-in Functions::
11630 * TI C6X Built-in Functions::
11631 * TILE-Gx Built-in Functions::
11632 * TILEPro Built-in Functions::
11633 * x86 Built-in Functions::
11634 * x86 transactional memory intrinsics::
11637 @node AArch64 Built-in Functions
11638 @subsection AArch64 Built-in Functions
11640 These built-in functions are available for the AArch64 family of
11643 unsigned int __builtin_aarch64_get_fpcr ()
11644 void __builtin_aarch64_set_fpcr (unsigned int)
11645 unsigned int __builtin_aarch64_get_fpsr ()
11646 void __builtin_aarch64_set_fpsr (unsigned int)
11649 @node Alpha Built-in Functions
11650 @subsection Alpha Built-in Functions
11652 These built-in functions are available for the Alpha family of
11653 processors, depending on the command-line switches used.
11655 The following built-in functions are always available. They
11656 all generate the machine instruction that is part of the name.
11659 long __builtin_alpha_implver (void)
11660 long __builtin_alpha_rpcc (void)
11661 long __builtin_alpha_amask (long)
11662 long __builtin_alpha_cmpbge (long, long)
11663 long __builtin_alpha_extbl (long, long)
11664 long __builtin_alpha_extwl (long, long)
11665 long __builtin_alpha_extll (long, long)
11666 long __builtin_alpha_extql (long, long)
11667 long __builtin_alpha_extwh (long, long)
11668 long __builtin_alpha_extlh (long, long)
11669 long __builtin_alpha_extqh (long, long)
11670 long __builtin_alpha_insbl (long, long)
11671 long __builtin_alpha_inswl (long, long)
11672 long __builtin_alpha_insll (long, long)
11673 long __builtin_alpha_insql (long, long)
11674 long __builtin_alpha_inswh (long, long)
11675 long __builtin_alpha_inslh (long, long)
11676 long __builtin_alpha_insqh (long, long)
11677 long __builtin_alpha_mskbl (long, long)
11678 long __builtin_alpha_mskwl (long, long)
11679 long __builtin_alpha_mskll (long, long)
11680 long __builtin_alpha_mskql (long, long)
11681 long __builtin_alpha_mskwh (long, long)
11682 long __builtin_alpha_msklh (long, long)
11683 long __builtin_alpha_mskqh (long, long)
11684 long __builtin_alpha_umulh (long, long)
11685 long __builtin_alpha_zap (long, long)
11686 long __builtin_alpha_zapnot (long, long)
11689 The following built-in functions are always with @option{-mmax}
11690 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11691 later. They all generate the machine instruction that is part
11695 long __builtin_alpha_pklb (long)
11696 long __builtin_alpha_pkwb (long)
11697 long __builtin_alpha_unpkbl (long)
11698 long __builtin_alpha_unpkbw (long)
11699 long __builtin_alpha_minub8 (long, long)
11700 long __builtin_alpha_minsb8 (long, long)
11701 long __builtin_alpha_minuw4 (long, long)
11702 long __builtin_alpha_minsw4 (long, long)
11703 long __builtin_alpha_maxub8 (long, long)
11704 long __builtin_alpha_maxsb8 (long, long)
11705 long __builtin_alpha_maxuw4 (long, long)
11706 long __builtin_alpha_maxsw4 (long, long)
11707 long __builtin_alpha_perr (long, long)
11710 The following built-in functions are always with @option{-mcix}
11711 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11712 later. They all generate the machine instruction that is part
11716 long __builtin_alpha_cttz (long)
11717 long __builtin_alpha_ctlz (long)
11718 long __builtin_alpha_ctpop (long)
11721 The following built-in functions are available on systems that use the OSF/1
11722 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11723 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11724 @code{rdval} and @code{wrval}.
11727 void *__builtin_thread_pointer (void)
11728 void __builtin_set_thread_pointer (void *)
11731 @node Altera Nios II Built-in Functions
11732 @subsection Altera Nios II Built-in Functions
11734 These built-in functions are available for the Altera Nios II
11735 family of processors.
11737 The following built-in functions are always available. They
11738 all generate the machine instruction that is part of the name.
11741 int __builtin_ldbio (volatile const void *)
11742 int __builtin_ldbuio (volatile const void *)
11743 int __builtin_ldhio (volatile const void *)
11744 int __builtin_ldhuio (volatile const void *)
11745 int __builtin_ldwio (volatile const void *)
11746 void __builtin_stbio (volatile void *, int)
11747 void __builtin_sthio (volatile void *, int)
11748 void __builtin_stwio (volatile void *, int)
11749 void __builtin_sync (void)
11750 int __builtin_rdctl (int)
11751 int __builtin_rdprs (int, int)
11752 void __builtin_wrctl (int, int)
11753 void __builtin_flushd (volatile void *)
11754 void __builtin_flushda (volatile void *)
11755 int __builtin_wrpie (int);
11756 void __builtin_eni (int);
11757 int __builtin_ldex (volatile const void *)
11758 int __builtin_stex (volatile void *, int)
11759 int __builtin_ldsex (volatile const void *)
11760 int __builtin_stsex (volatile void *, int)
11763 The following built-in functions are always available. They
11764 all generate a Nios II Custom Instruction. The name of the
11765 function represents the types that the function takes and
11766 returns. The letter before the @code{n} is the return type
11767 or void if absent. The @code{n} represents the first parameter
11768 to all the custom instructions, the custom instruction number.
11769 The two letters after the @code{n} represent the up to two
11770 parameters to the function.
11772 The letters represent the following data types:
11775 @code{void} for return type and no parameter for parameter types.
11778 @code{int} for return type and parameter type
11781 @code{float} for return type and parameter type
11784 @code{void *} for return type and parameter type
11788 And the function names are:
11790 void __builtin_custom_n (void)
11791 void __builtin_custom_ni (int)
11792 void __builtin_custom_nf (float)
11793 void __builtin_custom_np (void *)
11794 void __builtin_custom_nii (int, int)
11795 void __builtin_custom_nif (int, float)
11796 void __builtin_custom_nip (int, void *)
11797 void __builtin_custom_nfi (float, int)
11798 void __builtin_custom_nff (float, float)
11799 void __builtin_custom_nfp (float, void *)
11800 void __builtin_custom_npi (void *, int)
11801 void __builtin_custom_npf (void *, float)
11802 void __builtin_custom_npp (void *, void *)
11803 int __builtin_custom_in (void)
11804 int __builtin_custom_ini (int)
11805 int __builtin_custom_inf (float)
11806 int __builtin_custom_inp (void *)
11807 int __builtin_custom_inii (int, int)
11808 int __builtin_custom_inif (int, float)
11809 int __builtin_custom_inip (int, void *)
11810 int __builtin_custom_infi (float, int)
11811 int __builtin_custom_inff (float, float)
11812 int __builtin_custom_infp (float, void *)
11813 int __builtin_custom_inpi (void *, int)
11814 int __builtin_custom_inpf (void *, float)
11815 int __builtin_custom_inpp (void *, void *)
11816 float __builtin_custom_fn (void)
11817 float __builtin_custom_fni (int)
11818 float __builtin_custom_fnf (float)
11819 float __builtin_custom_fnp (void *)
11820 float __builtin_custom_fnii (int, int)
11821 float __builtin_custom_fnif (int, float)
11822 float __builtin_custom_fnip (int, void *)
11823 float __builtin_custom_fnfi (float, int)
11824 float __builtin_custom_fnff (float, float)
11825 float __builtin_custom_fnfp (float, void *)
11826 float __builtin_custom_fnpi (void *, int)
11827 float __builtin_custom_fnpf (void *, float)
11828 float __builtin_custom_fnpp (void *, void *)
11829 void * __builtin_custom_pn (void)
11830 void * __builtin_custom_pni (int)
11831 void * __builtin_custom_pnf (float)
11832 void * __builtin_custom_pnp (void *)
11833 void * __builtin_custom_pnii (int, int)
11834 void * __builtin_custom_pnif (int, float)
11835 void * __builtin_custom_pnip (int, void *)
11836 void * __builtin_custom_pnfi (float, int)
11837 void * __builtin_custom_pnff (float, float)
11838 void * __builtin_custom_pnfp (float, void *)
11839 void * __builtin_custom_pnpi (void *, int)
11840 void * __builtin_custom_pnpf (void *, float)
11841 void * __builtin_custom_pnpp (void *, void *)
11844 @node ARC Built-in Functions
11845 @subsection ARC Built-in Functions
11847 The following built-in functions are provided for ARC targets. The
11848 built-ins generate the corresponding assembly instructions. In the
11849 examples given below, the generated code often requires an operand or
11850 result to be in a register. Where necessary further code will be
11851 generated to ensure this is true, but for brevity this is not
11852 described in each case.
11854 @emph{Note:} Using a built-in to generate an instruction not supported
11855 by a target may cause problems. At present the compiler is not
11856 guaranteed to detect such misuse, and as a result an internal compiler
11857 error may be generated.
11859 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11860 Return 1 if @var{val} is known to have the byte alignment given
11861 by @var{alignval}, otherwise return 0.
11862 Note that this is different from
11864 __alignof__(*(char *)@var{val}) >= alignval
11866 because __alignof__ sees only the type of the dereference, whereas
11867 __builtin_arc_align uses alignment information from the pointer
11868 as well as from the pointed-to type.
11869 The information available will depend on optimization level.
11872 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11879 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11880 The operand is the number of a register to be read. Generates:
11882 mov @var{dest}, r@var{regno}
11884 where the value in @var{dest} will be the result returned from the
11888 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11889 The first operand is the number of a register to be written, the
11890 second operand is a compile time constant to write into that
11891 register. Generates:
11893 mov r@var{regno}, @var{val}
11897 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11898 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11901 divaw @var{dest}, @var{a}, @var{b}
11903 where the value in @var{dest} will be the result returned from the
11907 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11914 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11915 The operand, @var{auxv}, is the address of an auxiliary register and
11916 must be a compile time constant. Generates:
11918 lr @var{dest}, [@var{auxr}]
11920 Where the value in @var{dest} will be the result returned from the
11924 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11925 Only available with @option{-mmul64}. Generates:
11927 mul64 @var{a}, @var{b}
11931 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11932 Only available with @option{-mmul64}. Generates:
11934 mulu64 @var{a}, @var{b}
11938 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11945 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11946 Only valid if the @samp{norm} instruction is available through the
11947 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11950 norm @var{dest}, @var{src}
11952 Where the value in @var{dest} will be the result returned from the
11956 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11957 Only valid if the @samp{normw} instruction is available through the
11958 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11961 normw @var{dest}, @var{src}
11963 Where the value in @var{dest} will be the result returned from the
11967 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11974 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11981 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11982 The first argument, @var{auxv}, is the address of an auxiliary
11983 register, the second argument, @var{val}, is a compile time constant
11984 to be written to the register. Generates:
11986 sr @var{auxr}, [@var{val}]
11990 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11991 Only valid with @option{-mswap}. Generates:
11993 swap @var{dest}, @var{src}
11995 Where the value in @var{dest} will be the result returned from the
11999 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
12006 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
12007 Only available with @option{-mcpu=ARC700}. Generates:
12013 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
12014 Only available with @option{-mcpu=ARC700}. Generates:
12020 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
12021 Only available with @option{-mcpu=ARC700}. Generates:
12027 The instructions generated by the following builtins are not
12028 considered as candidates for scheduling. They are not moved around by
12029 the compiler during scheduling, and thus can be expected to appear
12030 where they are put in the C code:
12032 __builtin_arc_brk()
12033 __builtin_arc_core_read()
12034 __builtin_arc_core_write()
12035 __builtin_arc_flag()
12037 __builtin_arc_sleep()
12039 __builtin_arc_swi()
12042 @node ARC SIMD Built-in Functions
12043 @subsection ARC SIMD Built-in Functions
12045 SIMD builtins provided by the compiler can be used to generate the
12046 vector instructions. This section describes the available builtins
12047 and their usage in programs. With the @option{-msimd} option, the
12048 compiler provides 128-bit vector types, which can be specified using
12049 the @code{vector_size} attribute. The header file @file{arc-simd.h}
12050 can be included to use the following predefined types:
12052 typedef int __v4si __attribute__((vector_size(16)));
12053 typedef short __v8hi __attribute__((vector_size(16)));
12056 These types can be used to define 128-bit variables. The built-in
12057 functions listed in the following section can be used on these
12058 variables to generate the vector operations.
12060 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
12061 @file{arc-simd.h} also provides equivalent macros called
12062 @code{_@var{someinsn}} that can be used for programming ease and
12063 improved readability. The following macros for DMA control are also
12066 #define _setup_dma_in_channel_reg _vdiwr
12067 #define _setup_dma_out_channel_reg _vdowr
12070 The following is a complete list of all the SIMD built-ins provided
12071 for ARC, grouped by calling signature.
12073 The following take two @code{__v8hi} arguments and return a
12074 @code{__v8hi} result:
12076 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
12077 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
12078 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
12079 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
12080 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
12081 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
12082 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
12083 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
12084 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
12085 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
12086 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
12087 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
12088 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
12089 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
12090 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
12091 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
12092 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
12093 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
12094 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
12095 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
12096 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
12097 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
12098 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
12099 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
12100 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
12101 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
12102 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
12103 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
12104 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
12105 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
12106 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
12107 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
12108 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
12109 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
12110 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
12111 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
12112 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
12113 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
12114 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
12115 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
12116 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
12117 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
12118 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
12119 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
12120 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
12121 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
12122 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
12123 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
12126 The following take one @code{__v8hi} and one @code{int} argument and return a
12127 @code{__v8hi} result:
12130 __v8hi __builtin_arc_vbaddw (__v8hi, int)
12131 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
12132 __v8hi __builtin_arc_vbminw (__v8hi, int)
12133 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
12134 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
12135 __v8hi __builtin_arc_vbmulw (__v8hi, int)
12136 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
12137 __v8hi __builtin_arc_vbsubw (__v8hi, int)
12140 The following take one @code{__v8hi} argument and one @code{int} argument which
12141 must be a 3-bit compile time constant indicating a register number
12142 I0-I7. They return a @code{__v8hi} result.
12144 __v8hi __builtin_arc_vasrw (__v8hi, const int)
12145 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
12146 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
12149 The following take one @code{__v8hi} argument and one @code{int}
12150 argument which must be a 6-bit compile time constant. They return a
12151 @code{__v8hi} result.
12153 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
12154 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
12155 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
12156 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
12157 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
12158 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
12159 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
12162 The following take one @code{__v8hi} argument and one @code{int} argument which
12163 must be a 8-bit compile time constant. They return a @code{__v8hi}
12166 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
12167 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
12168 __v8hi __builtin_arc_vmvw (__v8hi, const int)
12169 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
12172 The following take two @code{int} arguments, the second of which which
12173 must be a 8-bit compile time constant. They return a @code{__v8hi}
12176 __v8hi __builtin_arc_vmovaw (int, const int)
12177 __v8hi __builtin_arc_vmovw (int, const int)
12178 __v8hi __builtin_arc_vmovzw (int, const int)
12181 The following take a single @code{__v8hi} argument and return a
12182 @code{__v8hi} result:
12184 __v8hi __builtin_arc_vabsaw (__v8hi)
12185 __v8hi __builtin_arc_vabsw (__v8hi)
12186 __v8hi __builtin_arc_vaddsuw (__v8hi)
12187 __v8hi __builtin_arc_vexch1 (__v8hi)
12188 __v8hi __builtin_arc_vexch2 (__v8hi)
12189 __v8hi __builtin_arc_vexch4 (__v8hi)
12190 __v8hi __builtin_arc_vsignw (__v8hi)
12191 __v8hi __builtin_arc_vupbaw (__v8hi)
12192 __v8hi __builtin_arc_vupbw (__v8hi)
12193 __v8hi __builtin_arc_vupsbaw (__v8hi)
12194 __v8hi __builtin_arc_vupsbw (__v8hi)
12197 The following take two @code{int} arguments and return no result:
12199 void __builtin_arc_vdirun (int, int)
12200 void __builtin_arc_vdorun (int, int)
12203 The following take two @code{int} arguments and return no result. The
12204 first argument must a 3-bit compile time constant indicating one of
12205 the DR0-DR7 DMA setup channels:
12207 void __builtin_arc_vdiwr (const int, int)
12208 void __builtin_arc_vdowr (const int, int)
12211 The following take an @code{int} argument and return no result:
12213 void __builtin_arc_vendrec (int)
12214 void __builtin_arc_vrec (int)
12215 void __builtin_arc_vrecrun (int)
12216 void __builtin_arc_vrun (int)
12219 The following take a @code{__v8hi} argument and two @code{int}
12220 arguments and return a @code{__v8hi} result. The second argument must
12221 be a 3-bit compile time constants, indicating one the registers I0-I7,
12222 and the third argument must be an 8-bit compile time constant.
12224 @emph{Note:} Although the equivalent hardware instructions do not take
12225 an SIMD register as an operand, these builtins overwrite the relevant
12226 bits of the @code{__v8hi} register provided as the first argument with
12227 the value loaded from the @code{[Ib, u8]} location in the SDM.
12230 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
12231 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
12232 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
12233 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
12236 The following take two @code{int} arguments and return a @code{__v8hi}
12237 result. The first argument must be a 3-bit compile time constants,
12238 indicating one the registers I0-I7, and the second argument must be an
12239 8-bit compile time constant.
12242 __v8hi __builtin_arc_vld128 (const int, const int)
12243 __v8hi __builtin_arc_vld64w (const int, const int)
12246 The following take a @code{__v8hi} argument and two @code{int}
12247 arguments and return no result. The second argument must be a 3-bit
12248 compile time constants, indicating one the registers I0-I7, and the
12249 third argument must be an 8-bit compile time constant.
12252 void __builtin_arc_vst128 (__v8hi, const int, const int)
12253 void __builtin_arc_vst64 (__v8hi, const int, const int)
12256 The following take a @code{__v8hi} argument and three @code{int}
12257 arguments and return no result. The second argument must be a 3-bit
12258 compile-time constant, identifying the 16-bit sub-register to be
12259 stored, the third argument must be a 3-bit compile time constants,
12260 indicating one the registers I0-I7, and the fourth argument must be an
12261 8-bit compile time constant.
12264 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
12265 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
12268 @node ARM iWMMXt Built-in Functions
12269 @subsection ARM iWMMXt Built-in Functions
12271 These built-in functions are available for the ARM family of
12272 processors when the @option{-mcpu=iwmmxt} switch is used:
12275 typedef int v2si __attribute__ ((vector_size (8)));
12276 typedef short v4hi __attribute__ ((vector_size (8)));
12277 typedef char v8qi __attribute__ ((vector_size (8)));
12279 int __builtin_arm_getwcgr0 (void)
12280 void __builtin_arm_setwcgr0 (int)
12281 int __builtin_arm_getwcgr1 (void)
12282 void __builtin_arm_setwcgr1 (int)
12283 int __builtin_arm_getwcgr2 (void)
12284 void __builtin_arm_setwcgr2 (int)
12285 int __builtin_arm_getwcgr3 (void)
12286 void __builtin_arm_setwcgr3 (int)
12287 int __builtin_arm_textrmsb (v8qi, int)
12288 int __builtin_arm_textrmsh (v4hi, int)
12289 int __builtin_arm_textrmsw (v2si, int)
12290 int __builtin_arm_textrmub (v8qi, int)
12291 int __builtin_arm_textrmuh (v4hi, int)
12292 int __builtin_arm_textrmuw (v2si, int)
12293 v8qi __builtin_arm_tinsrb (v8qi, int, int)
12294 v4hi __builtin_arm_tinsrh (v4hi, int, int)
12295 v2si __builtin_arm_tinsrw (v2si, int, int)
12296 long long __builtin_arm_tmia (long long, int, int)
12297 long long __builtin_arm_tmiabb (long long, int, int)
12298 long long __builtin_arm_tmiabt (long long, int, int)
12299 long long __builtin_arm_tmiaph (long long, int, int)
12300 long long __builtin_arm_tmiatb (long long, int, int)
12301 long long __builtin_arm_tmiatt (long long, int, int)
12302 int __builtin_arm_tmovmskb (v8qi)
12303 int __builtin_arm_tmovmskh (v4hi)
12304 int __builtin_arm_tmovmskw (v2si)
12305 long long __builtin_arm_waccb (v8qi)
12306 long long __builtin_arm_wacch (v4hi)
12307 long long __builtin_arm_waccw (v2si)
12308 v8qi __builtin_arm_waddb (v8qi, v8qi)
12309 v8qi __builtin_arm_waddbss (v8qi, v8qi)
12310 v8qi __builtin_arm_waddbus (v8qi, v8qi)
12311 v4hi __builtin_arm_waddh (v4hi, v4hi)
12312 v4hi __builtin_arm_waddhss (v4hi, v4hi)
12313 v4hi __builtin_arm_waddhus (v4hi, v4hi)
12314 v2si __builtin_arm_waddw (v2si, v2si)
12315 v2si __builtin_arm_waddwss (v2si, v2si)
12316 v2si __builtin_arm_waddwus (v2si, v2si)
12317 v8qi __builtin_arm_walign (v8qi, v8qi, int)
12318 long long __builtin_arm_wand(long long, long long)
12319 long long __builtin_arm_wandn (long long, long long)
12320 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
12321 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
12322 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
12323 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
12324 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
12325 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
12326 v2si __builtin_arm_wcmpeqw (v2si, v2si)
12327 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
12328 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
12329 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
12330 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
12331 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
12332 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
12333 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
12334 long long __builtin_arm_wmacsz (v4hi, v4hi)
12335 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
12336 long long __builtin_arm_wmacuz (v4hi, v4hi)
12337 v4hi __builtin_arm_wmadds (v4hi, v4hi)
12338 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
12339 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
12340 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
12341 v2si __builtin_arm_wmaxsw (v2si, v2si)
12342 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
12343 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
12344 v2si __builtin_arm_wmaxuw (v2si, v2si)
12345 v8qi __builtin_arm_wminsb (v8qi, v8qi)
12346 v4hi __builtin_arm_wminsh (v4hi, v4hi)
12347 v2si __builtin_arm_wminsw (v2si, v2si)
12348 v8qi __builtin_arm_wminub (v8qi, v8qi)
12349 v4hi __builtin_arm_wminuh (v4hi, v4hi)
12350 v2si __builtin_arm_wminuw (v2si, v2si)
12351 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
12352 v4hi __builtin_arm_wmulul (v4hi, v4hi)
12353 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12354 long long __builtin_arm_wor (long long, long long)
12355 v2si __builtin_arm_wpackdss (long long, long long)
12356 v2si __builtin_arm_wpackdus (long long, long long)
12357 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12358 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12359 v4hi __builtin_arm_wpackwss (v2si, v2si)
12360 v4hi __builtin_arm_wpackwus (v2si, v2si)
12361 long long __builtin_arm_wrord (long long, long long)
12362 long long __builtin_arm_wrordi (long long, int)
12363 v4hi __builtin_arm_wrorh (v4hi, long long)
12364 v4hi __builtin_arm_wrorhi (v4hi, int)
12365 v2si __builtin_arm_wrorw (v2si, long long)
12366 v2si __builtin_arm_wrorwi (v2si, int)
12367 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12368 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12369 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12370 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12371 v4hi __builtin_arm_wshufh (v4hi, int)
12372 long long __builtin_arm_wslld (long long, long long)
12373 long long __builtin_arm_wslldi (long long, int)
12374 v4hi __builtin_arm_wsllh (v4hi, long long)
12375 v4hi __builtin_arm_wsllhi (v4hi, int)
12376 v2si __builtin_arm_wsllw (v2si, long long)
12377 v2si __builtin_arm_wsllwi (v2si, int)
12378 long long __builtin_arm_wsrad (long long, long long)
12379 long long __builtin_arm_wsradi (long long, int)
12380 v4hi __builtin_arm_wsrah (v4hi, long long)
12381 v4hi __builtin_arm_wsrahi (v4hi, int)
12382 v2si __builtin_arm_wsraw (v2si, long long)
12383 v2si __builtin_arm_wsrawi (v2si, int)
12384 long long __builtin_arm_wsrld (long long, long long)
12385 long long __builtin_arm_wsrldi (long long, int)
12386 v4hi __builtin_arm_wsrlh (v4hi, long long)
12387 v4hi __builtin_arm_wsrlhi (v4hi, int)
12388 v2si __builtin_arm_wsrlw (v2si, long long)
12389 v2si __builtin_arm_wsrlwi (v2si, int)
12390 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12391 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12392 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12393 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12394 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12395 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12396 v2si __builtin_arm_wsubw (v2si, v2si)
12397 v2si __builtin_arm_wsubwss (v2si, v2si)
12398 v2si __builtin_arm_wsubwus (v2si, v2si)
12399 v4hi __builtin_arm_wunpckehsb (v8qi)
12400 v2si __builtin_arm_wunpckehsh (v4hi)
12401 long long __builtin_arm_wunpckehsw (v2si)
12402 v4hi __builtin_arm_wunpckehub (v8qi)
12403 v2si __builtin_arm_wunpckehuh (v4hi)
12404 long long __builtin_arm_wunpckehuw (v2si)
12405 v4hi __builtin_arm_wunpckelsb (v8qi)
12406 v2si __builtin_arm_wunpckelsh (v4hi)
12407 long long __builtin_arm_wunpckelsw (v2si)
12408 v4hi __builtin_arm_wunpckelub (v8qi)
12409 v2si __builtin_arm_wunpckeluh (v4hi)
12410 long long __builtin_arm_wunpckeluw (v2si)
12411 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12412 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12413 v2si __builtin_arm_wunpckihw (v2si, v2si)
12414 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12415 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12416 v2si __builtin_arm_wunpckilw (v2si, v2si)
12417 long long __builtin_arm_wxor (long long, long long)
12418 long long __builtin_arm_wzero ()
12422 @node ARM C Language Extensions (ACLE)
12423 @subsection ARM C Language Extensions (ACLE)
12425 GCC implements extensions for C as described in the ARM C Language
12426 Extensions (ACLE) specification, which can be found at
12427 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12429 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12430 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12431 intrinsics can be found at
12432 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12433 The built-in intrinsics for the Advanced SIMD extension are available when
12436 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12437 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12438 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12439 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12442 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12443 availability of extensions.
12445 @node ARM Floating Point Status and Control Intrinsics
12446 @subsection ARM Floating Point Status and Control Intrinsics
12448 These built-in functions are available for the ARM family of
12449 processors with floating-point unit.
12452 unsigned int __builtin_arm_get_fpscr ()
12453 void __builtin_arm_set_fpscr (unsigned int)
12456 @node AVR Built-in Functions
12457 @subsection AVR Built-in Functions
12459 For each built-in function for AVR, there is an equally named,
12460 uppercase built-in macro defined. That way users can easily query if
12461 or if not a specific built-in is implemented or not. For example, if
12462 @code{__builtin_avr_nop} is available the macro
12463 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12465 The following built-in functions map to the respective machine
12466 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12467 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12468 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12469 as library call if no hardware multiplier is available.
12472 void __builtin_avr_nop (void)
12473 void __builtin_avr_sei (void)
12474 void __builtin_avr_cli (void)
12475 void __builtin_avr_sleep (void)
12476 void __builtin_avr_wdr (void)
12477 unsigned char __builtin_avr_swap (unsigned char)
12478 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12479 int __builtin_avr_fmuls (char, char)
12480 int __builtin_avr_fmulsu (char, unsigned char)
12483 In order to delay execution for a specific number of cycles, GCC
12486 void __builtin_avr_delay_cycles (unsigned long ticks)
12490 @code{ticks} is the number of ticks to delay execution. Note that this
12491 built-in does not take into account the effect of interrupts that
12492 might increase delay time. @code{ticks} must be a compile-time
12493 integer constant; delays with a variable number of cycles are not supported.
12496 char __builtin_avr_flash_segment (const __memx void*)
12500 This built-in takes a byte address to the 24-bit
12501 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12502 the number of the flash segment (the 64 KiB chunk) where the address
12503 points to. Counting starts at @code{0}.
12504 If the address does not point to flash memory, return @code{-1}.
12507 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12511 Insert bits from @var{bits} into @var{val} and return the resulting
12512 value. The nibbles of @var{map} determine how the insertion is
12513 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12515 @item If @var{X} is @code{0xf},
12516 then the @var{n}-th bit of @var{val} is returned unaltered.
12518 @item If X is in the range 0@dots{}7,
12519 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12521 @item If X is in the range 8@dots{}@code{0xe},
12522 then the @var{n}-th result bit is undefined.
12526 One typical use case for this built-in is adjusting input and
12527 output values to non-contiguous port layouts. Some examples:
12530 // same as val, bits is unused
12531 __builtin_avr_insert_bits (0xffffffff, bits, val)
12535 // same as bits, val is unused
12536 __builtin_avr_insert_bits (0x76543210, bits, val)
12540 // same as rotating bits by 4
12541 __builtin_avr_insert_bits (0x32107654, bits, 0)
12545 // high nibble of result is the high nibble of val
12546 // low nibble of result is the low nibble of bits
12547 __builtin_avr_insert_bits (0xffff3210, bits, val)
12551 // reverse the bit order of bits
12552 __builtin_avr_insert_bits (0x01234567, bits, 0)
12556 void __builtin_avr_nops (unsigned count)
12560 Insert @code{count} @code{NOP} instructions.
12561 The number of instructions must be a compile-time integer constant.
12563 @node Blackfin Built-in Functions
12564 @subsection Blackfin Built-in Functions
12566 Currently, there are two Blackfin-specific built-in functions. These are
12567 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12568 using inline assembly; by using these built-in functions the compiler can
12569 automatically add workarounds for hardware errata involving these
12570 instructions. These functions are named as follows:
12573 void __builtin_bfin_csync (void)
12574 void __builtin_bfin_ssync (void)
12577 @node FR-V Built-in Functions
12578 @subsection FR-V Built-in Functions
12580 GCC provides many FR-V-specific built-in functions. In general,
12581 these functions are intended to be compatible with those described
12582 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12583 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12584 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12585 pointer rather than by value.
12587 Most of the functions are named after specific FR-V instructions.
12588 Such functions are said to be ``directly mapped'' and are summarized
12589 here in tabular form.
12593 * Directly-mapped Integer Functions::
12594 * Directly-mapped Media Functions::
12595 * Raw read/write Functions::
12596 * Other Built-in Functions::
12599 @node Argument Types
12600 @subsubsection Argument Types
12602 The arguments to the built-in functions can be divided into three groups:
12603 register numbers, compile-time constants and run-time values. In order
12604 to make this classification clear at a glance, the arguments and return
12605 values are given the following pseudo types:
12607 @multitable @columnfractions .20 .30 .15 .35
12608 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12609 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12610 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12611 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12612 @item @code{uw2} @tab @code{unsigned long long} @tab No
12613 @tab an unsigned doubleword
12614 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12615 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12616 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12617 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12620 These pseudo types are not defined by GCC, they are simply a notational
12621 convenience used in this manual.
12623 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12624 and @code{sw2} are evaluated at run time. They correspond to
12625 register operands in the underlying FR-V instructions.
12627 @code{const} arguments represent immediate operands in the underlying
12628 FR-V instructions. They must be compile-time constants.
12630 @code{acc} arguments are evaluated at compile time and specify the number
12631 of an accumulator register. For example, an @code{acc} argument of 2
12632 selects the ACC2 register.
12634 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12635 number of an IACC register. See @pxref{Other Built-in Functions}
12638 @node Directly-mapped Integer Functions
12639 @subsubsection Directly-Mapped Integer Functions
12641 The functions listed below map directly to FR-V I-type instructions.
12643 @multitable @columnfractions .45 .32 .23
12644 @item Function prototype @tab Example usage @tab Assembly output
12645 @item @code{sw1 __ADDSS (sw1, sw1)}
12646 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12647 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12648 @item @code{sw1 __SCAN (sw1, sw1)}
12649 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12650 @tab @code{SCAN @var{a},@var{b},@var{c}}
12651 @item @code{sw1 __SCUTSS (sw1)}
12652 @tab @code{@var{b} = __SCUTSS (@var{a})}
12653 @tab @code{SCUTSS @var{a},@var{b}}
12654 @item @code{sw1 __SLASS (sw1, sw1)}
12655 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12656 @tab @code{SLASS @var{a},@var{b},@var{c}}
12657 @item @code{void __SMASS (sw1, sw1)}
12658 @tab @code{__SMASS (@var{a}, @var{b})}
12659 @tab @code{SMASS @var{a},@var{b}}
12660 @item @code{void __SMSSS (sw1, sw1)}
12661 @tab @code{__SMSSS (@var{a}, @var{b})}
12662 @tab @code{SMSSS @var{a},@var{b}}
12663 @item @code{void __SMU (sw1, sw1)}
12664 @tab @code{__SMU (@var{a}, @var{b})}
12665 @tab @code{SMU @var{a},@var{b}}
12666 @item @code{sw2 __SMUL (sw1, sw1)}
12667 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12668 @tab @code{SMUL @var{a},@var{b},@var{c}}
12669 @item @code{sw1 __SUBSS (sw1, sw1)}
12670 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12671 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12672 @item @code{uw2 __UMUL (uw1, uw1)}
12673 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12674 @tab @code{UMUL @var{a},@var{b},@var{c}}
12677 @node Directly-mapped Media Functions
12678 @subsubsection Directly-Mapped Media Functions
12680 The functions listed below map directly to FR-V M-type instructions.
12682 @multitable @columnfractions .45 .32 .23
12683 @item Function prototype @tab Example usage @tab Assembly output
12684 @item @code{uw1 __MABSHS (sw1)}
12685 @tab @code{@var{b} = __MABSHS (@var{a})}
12686 @tab @code{MABSHS @var{a},@var{b}}
12687 @item @code{void __MADDACCS (acc, acc)}
12688 @tab @code{__MADDACCS (@var{b}, @var{a})}
12689 @tab @code{MADDACCS @var{a},@var{b}}
12690 @item @code{sw1 __MADDHSS (sw1, sw1)}
12691 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12692 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12693 @item @code{uw1 __MADDHUS (uw1, uw1)}
12694 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12695 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12696 @item @code{uw1 __MAND (uw1, uw1)}
12697 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12698 @tab @code{MAND @var{a},@var{b},@var{c}}
12699 @item @code{void __MASACCS (acc, acc)}
12700 @tab @code{__MASACCS (@var{b}, @var{a})}
12701 @tab @code{MASACCS @var{a},@var{b}}
12702 @item @code{uw1 __MAVEH (uw1, uw1)}
12703 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12704 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12705 @item @code{uw2 __MBTOH (uw1)}
12706 @tab @code{@var{b} = __MBTOH (@var{a})}
12707 @tab @code{MBTOH @var{a},@var{b}}
12708 @item @code{void __MBTOHE (uw1 *, uw1)}
12709 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12710 @tab @code{MBTOHE @var{a},@var{b}}
12711 @item @code{void __MCLRACC (acc)}
12712 @tab @code{__MCLRACC (@var{a})}
12713 @tab @code{MCLRACC @var{a}}
12714 @item @code{void __MCLRACCA (void)}
12715 @tab @code{__MCLRACCA ()}
12716 @tab @code{MCLRACCA}
12717 @item @code{uw1 __Mcop1 (uw1, uw1)}
12718 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12719 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12720 @item @code{uw1 __Mcop2 (uw1, uw1)}
12721 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12722 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12723 @item @code{uw1 __MCPLHI (uw2, const)}
12724 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12725 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12726 @item @code{uw1 __MCPLI (uw2, const)}
12727 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12728 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12729 @item @code{void __MCPXIS (acc, sw1, sw1)}
12730 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12731 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12732 @item @code{void __MCPXIU (acc, uw1, uw1)}
12733 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12734 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12735 @item @code{void __MCPXRS (acc, sw1, sw1)}
12736 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12737 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12738 @item @code{void __MCPXRU (acc, uw1, uw1)}
12739 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12740 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12741 @item @code{uw1 __MCUT (acc, uw1)}
12742 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12743 @tab @code{MCUT @var{a},@var{b},@var{c}}
12744 @item @code{uw1 __MCUTSS (acc, sw1)}
12745 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12746 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12747 @item @code{void __MDADDACCS (acc, acc)}
12748 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12749 @tab @code{MDADDACCS @var{a},@var{b}}
12750 @item @code{void __MDASACCS (acc, acc)}
12751 @tab @code{__MDASACCS (@var{b}, @var{a})}
12752 @tab @code{MDASACCS @var{a},@var{b}}
12753 @item @code{uw2 __MDCUTSSI (acc, const)}
12754 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12755 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12756 @item @code{uw2 __MDPACKH (uw2, uw2)}
12757 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12758 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12759 @item @code{uw2 __MDROTLI (uw2, const)}
12760 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12761 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12762 @item @code{void __MDSUBACCS (acc, acc)}
12763 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12764 @tab @code{MDSUBACCS @var{a},@var{b}}
12765 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12766 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12767 @tab @code{MDUNPACKH @var{a},@var{b}}
12768 @item @code{uw2 __MEXPDHD (uw1, const)}
12769 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12770 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12771 @item @code{uw1 __MEXPDHW (uw1, const)}
12772 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12773 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12774 @item @code{uw1 __MHDSETH (uw1, const)}
12775 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12776 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12777 @item @code{sw1 __MHDSETS (const)}
12778 @tab @code{@var{b} = __MHDSETS (@var{a})}
12779 @tab @code{MHDSETS #@var{a},@var{b}}
12780 @item @code{uw1 __MHSETHIH (uw1, const)}
12781 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12782 @tab @code{MHSETHIH #@var{a},@var{b}}
12783 @item @code{sw1 __MHSETHIS (sw1, const)}
12784 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12785 @tab @code{MHSETHIS #@var{a},@var{b}}
12786 @item @code{uw1 __MHSETLOH (uw1, const)}
12787 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12788 @tab @code{MHSETLOH #@var{a},@var{b}}
12789 @item @code{sw1 __MHSETLOS (sw1, const)}
12790 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12791 @tab @code{MHSETLOS #@var{a},@var{b}}
12792 @item @code{uw1 __MHTOB (uw2)}
12793 @tab @code{@var{b} = __MHTOB (@var{a})}
12794 @tab @code{MHTOB @var{a},@var{b}}
12795 @item @code{void __MMACHS (acc, sw1, sw1)}
12796 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12797 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12798 @item @code{void __MMACHU (acc, uw1, uw1)}
12799 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12800 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12801 @item @code{void __MMRDHS (acc, sw1, sw1)}
12802 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12803 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12804 @item @code{void __MMRDHU (acc, uw1, uw1)}
12805 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12806 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12807 @item @code{void __MMULHS (acc, sw1, sw1)}
12808 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12809 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12810 @item @code{void __MMULHU (acc, uw1, uw1)}
12811 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12812 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12813 @item @code{void __MMULXHS (acc, sw1, sw1)}
12814 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12815 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12816 @item @code{void __MMULXHU (acc, uw1, uw1)}
12817 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12818 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12819 @item @code{uw1 __MNOT (uw1)}
12820 @tab @code{@var{b} = __MNOT (@var{a})}
12821 @tab @code{MNOT @var{a},@var{b}}
12822 @item @code{uw1 __MOR (uw1, uw1)}
12823 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12824 @tab @code{MOR @var{a},@var{b},@var{c}}
12825 @item @code{uw1 __MPACKH (uh, uh)}
12826 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12827 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12828 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12829 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12830 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12831 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12832 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12833 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12834 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12835 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12836 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12837 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12838 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12839 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12840 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12841 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12842 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12843 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12844 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12845 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12846 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12847 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12848 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12849 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12850 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12851 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12852 @item @code{void __MQMACHS (acc, sw2, sw2)}
12853 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12854 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12855 @item @code{void __MQMACHU (acc, uw2, uw2)}
12856 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12857 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12858 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12859 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12860 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12861 @item @code{void __MQMULHS (acc, sw2, sw2)}
12862 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12863 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12864 @item @code{void __MQMULHU (acc, uw2, uw2)}
12865 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12866 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12867 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12868 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12869 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12870 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12871 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12872 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12873 @item @code{sw2 __MQSATHS (sw2, sw2)}
12874 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12875 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12876 @item @code{uw2 __MQSLLHI (uw2, int)}
12877 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12878 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12879 @item @code{sw2 __MQSRAHI (sw2, int)}
12880 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12881 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12882 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12883 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12884 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12885 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12886 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12887 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12888 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12889 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12890 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12891 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12892 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12893 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12894 @item @code{uw1 __MRDACC (acc)}
12895 @tab @code{@var{b} = __MRDACC (@var{a})}
12896 @tab @code{MRDACC @var{a},@var{b}}
12897 @item @code{uw1 __MRDACCG (acc)}
12898 @tab @code{@var{b} = __MRDACCG (@var{a})}
12899 @tab @code{MRDACCG @var{a},@var{b}}
12900 @item @code{uw1 __MROTLI (uw1, const)}
12901 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12902 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12903 @item @code{uw1 __MROTRI (uw1, const)}
12904 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12905 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12906 @item @code{sw1 __MSATHS (sw1, sw1)}
12907 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12908 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12909 @item @code{uw1 __MSATHU (uw1, uw1)}
12910 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12911 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12912 @item @code{uw1 __MSLLHI (uw1, const)}
12913 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12914 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12915 @item @code{sw1 __MSRAHI (sw1, const)}
12916 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12917 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12918 @item @code{uw1 __MSRLHI (uw1, const)}
12919 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12920 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12921 @item @code{void __MSUBACCS (acc, acc)}
12922 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12923 @tab @code{MSUBACCS @var{a},@var{b}}
12924 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12925 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12926 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12927 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12928 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12929 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12930 @item @code{void __MTRAP (void)}
12931 @tab @code{__MTRAP ()}
12933 @item @code{uw2 __MUNPACKH (uw1)}
12934 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12935 @tab @code{MUNPACKH @var{a},@var{b}}
12936 @item @code{uw1 __MWCUT (uw2, uw1)}
12937 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12938 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12939 @item @code{void __MWTACC (acc, uw1)}
12940 @tab @code{__MWTACC (@var{b}, @var{a})}
12941 @tab @code{MWTACC @var{a},@var{b}}
12942 @item @code{void __MWTACCG (acc, uw1)}
12943 @tab @code{__MWTACCG (@var{b}, @var{a})}
12944 @tab @code{MWTACCG @var{a},@var{b}}
12945 @item @code{uw1 __MXOR (uw1, uw1)}
12946 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12947 @tab @code{MXOR @var{a},@var{b},@var{c}}
12950 @node Raw read/write Functions
12951 @subsubsection Raw Read/Write Functions
12953 This sections describes built-in functions related to read and write
12954 instructions to access memory. These functions generate
12955 @code{membar} instructions to flush the I/O load and stores where
12956 appropriate, as described in Fujitsu's manual described above.
12960 @item unsigned char __builtin_read8 (void *@var{data})
12961 @item unsigned short __builtin_read16 (void *@var{data})
12962 @item unsigned long __builtin_read32 (void *@var{data})
12963 @item unsigned long long __builtin_read64 (void *@var{data})
12965 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12966 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12967 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12968 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12971 @node Other Built-in Functions
12972 @subsubsection Other Built-in Functions
12974 This section describes built-in functions that are not named after
12975 a specific FR-V instruction.
12978 @item sw2 __IACCreadll (iacc @var{reg})
12979 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12980 for future expansion and must be 0.
12982 @item sw1 __IACCreadl (iacc @var{reg})
12983 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12984 Other values of @var{reg} are rejected as invalid.
12986 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12987 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12988 is reserved for future expansion and must be 0.
12990 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12991 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12992 is 1. Other values of @var{reg} are rejected as invalid.
12994 @item void __data_prefetch0 (const void *@var{x})
12995 Use the @code{dcpl} instruction to load the contents of address @var{x}
12996 into the data cache.
12998 @item void __data_prefetch (const void *@var{x})
12999 Use the @code{nldub} instruction to load the contents of address @var{x}
13000 into the data cache. The instruction is issued in slot I1@.
13003 @node MIPS DSP Built-in Functions
13004 @subsection MIPS DSP Built-in Functions
13006 The MIPS DSP Application-Specific Extension (ASE) includes new
13007 instructions that are designed to improve the performance of DSP and
13008 media applications. It provides instructions that operate on packed
13009 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
13011 GCC supports MIPS DSP operations using both the generic
13012 vector extensions (@pxref{Vector Extensions}) and a collection of
13013 MIPS-specific built-in functions. Both kinds of support are
13014 enabled by the @option{-mdsp} command-line option.
13016 Revision 2 of the ASE was introduced in the second half of 2006.
13017 This revision adds extra instructions to the original ASE, but is
13018 otherwise backwards-compatible with it. You can select revision 2
13019 using the command-line option @option{-mdspr2}; this option implies
13022 The SCOUNT and POS bits of the DSP control register are global. The
13023 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
13024 POS bits. During optimization, the compiler does not delete these
13025 instructions and it does not delete calls to functions containing
13026 these instructions.
13028 At present, GCC only provides support for operations on 32-bit
13029 vectors. The vector type associated with 8-bit integer data is
13030 usually called @code{v4i8}, the vector type associated with Q7
13031 is usually called @code{v4q7}, the vector type associated with 16-bit
13032 integer data is usually called @code{v2i16}, and the vector type
13033 associated with Q15 is usually called @code{v2q15}. They can be
13034 defined in C as follows:
13037 typedef signed char v4i8 __attribute__ ((vector_size(4)));
13038 typedef signed char v4q7 __attribute__ ((vector_size(4)));
13039 typedef short v2i16 __attribute__ ((vector_size(4)));
13040 typedef short v2q15 __attribute__ ((vector_size(4)));
13043 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
13044 initialized in the same way as aggregates. For example:
13047 v4i8 a = @{1, 2, 3, 4@};
13049 b = (v4i8) @{5, 6, 7, 8@};
13051 v2q15 c = @{0x0fcb, 0x3a75@};
13053 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
13056 @emph{Note:} The CPU's endianness determines the order in which values
13057 are packed. On little-endian targets, the first value is the least
13058 significant and the last value is the most significant. The opposite
13059 order applies to big-endian targets. For example, the code above
13060 sets the lowest byte of @code{a} to @code{1} on little-endian targets
13061 and @code{4} on big-endian targets.
13063 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
13064 representation. As shown in this example, the integer representation
13065 of a Q7 value can be obtained by multiplying the fractional value by
13066 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
13067 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
13070 The table below lists the @code{v4i8} and @code{v2q15} operations for which
13071 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
13072 and @code{c} and @code{d} are @code{v2q15} values.
13074 @multitable @columnfractions .50 .50
13075 @item C code @tab MIPS instruction
13076 @item @code{a + b} @tab @code{addu.qb}
13077 @item @code{c + d} @tab @code{addq.ph}
13078 @item @code{a - b} @tab @code{subu.qb}
13079 @item @code{c - d} @tab @code{subq.ph}
13082 The table below lists the @code{v2i16} operation for which
13083 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
13084 @code{v2i16} values.
13086 @multitable @columnfractions .50 .50
13087 @item C code @tab MIPS instruction
13088 @item @code{e * f} @tab @code{mul.ph}
13091 It is easier to describe the DSP built-in functions if we first define
13092 the following types:
13097 typedef unsigned int ui32;
13098 typedef long long a64;
13101 @code{q31} and @code{i32} are actually the same as @code{int}, but we
13102 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
13103 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
13104 @code{long long}, but we use @code{a64} to indicate values that are
13105 placed in one of the four DSP accumulators (@code{$ac0},
13106 @code{$ac1}, @code{$ac2} or @code{$ac3}).
13108 Also, some built-in functions prefer or require immediate numbers as
13109 parameters, because the corresponding DSP instructions accept both immediate
13110 numbers and register operands, or accept immediate numbers only. The
13111 immediate parameters are listed as follows.
13119 imm0_255: 0 to 255.
13120 imm_n32_31: -32 to 31.
13121 imm_n512_511: -512 to 511.
13124 The following built-in functions map directly to a particular MIPS DSP
13125 instruction. Please refer to the architecture specification
13126 for details on what each instruction does.
13129 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
13130 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
13131 q31 __builtin_mips_addq_s_w (q31, q31)
13132 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
13133 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
13134 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
13135 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
13136 q31 __builtin_mips_subq_s_w (q31, q31)
13137 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
13138 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
13139 i32 __builtin_mips_addsc (i32, i32)
13140 i32 __builtin_mips_addwc (i32, i32)
13141 i32 __builtin_mips_modsub (i32, i32)
13142 i32 __builtin_mips_raddu_w_qb (v4i8)
13143 v2q15 __builtin_mips_absq_s_ph (v2q15)
13144 q31 __builtin_mips_absq_s_w (q31)
13145 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
13146 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
13147 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
13148 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
13149 q31 __builtin_mips_preceq_w_phl (v2q15)
13150 q31 __builtin_mips_preceq_w_phr (v2q15)
13151 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
13152 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
13153 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
13154 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
13155 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
13156 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
13157 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
13158 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
13159 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
13160 v4i8 __builtin_mips_shll_qb (v4i8, i32)
13161 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
13162 v2q15 __builtin_mips_shll_ph (v2q15, i32)
13163 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
13164 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
13165 q31 __builtin_mips_shll_s_w (q31, imm0_31)
13166 q31 __builtin_mips_shll_s_w (q31, i32)
13167 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
13168 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
13169 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
13170 v2q15 __builtin_mips_shra_ph (v2q15, i32)
13171 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
13172 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
13173 q31 __builtin_mips_shra_r_w (q31, imm0_31)
13174 q31 __builtin_mips_shra_r_w (q31, i32)
13175 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
13176 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
13177 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
13178 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
13179 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
13180 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
13181 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
13182 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
13183 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
13184 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
13185 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
13186 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
13187 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
13188 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
13189 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
13190 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
13191 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
13192 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
13193 i32 __builtin_mips_bitrev (i32)
13194 i32 __builtin_mips_insv (i32, i32)
13195 v4i8 __builtin_mips_repl_qb (imm0_255)
13196 v4i8 __builtin_mips_repl_qb (i32)
13197 v2q15 __builtin_mips_repl_ph (imm_n512_511)
13198 v2q15 __builtin_mips_repl_ph (i32)
13199 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
13200 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
13201 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
13202 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
13203 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
13204 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
13205 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
13206 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
13207 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
13208 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
13209 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
13210 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
13211 i32 __builtin_mips_extr_w (a64, imm0_31)
13212 i32 __builtin_mips_extr_w (a64, i32)
13213 i32 __builtin_mips_extr_r_w (a64, imm0_31)
13214 i32 __builtin_mips_extr_s_h (a64, i32)
13215 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
13216 i32 __builtin_mips_extr_rs_w (a64, i32)
13217 i32 __builtin_mips_extr_s_h (a64, imm0_31)
13218 i32 __builtin_mips_extr_r_w (a64, i32)
13219 i32 __builtin_mips_extp (a64, imm0_31)
13220 i32 __builtin_mips_extp (a64, i32)
13221 i32 __builtin_mips_extpdp (a64, imm0_31)
13222 i32 __builtin_mips_extpdp (a64, i32)
13223 a64 __builtin_mips_shilo (a64, imm_n32_31)
13224 a64 __builtin_mips_shilo (a64, i32)
13225 a64 __builtin_mips_mthlip (a64, i32)
13226 void __builtin_mips_wrdsp (i32, imm0_63)
13227 i32 __builtin_mips_rddsp (imm0_63)
13228 i32 __builtin_mips_lbux (void *, i32)
13229 i32 __builtin_mips_lhx (void *, i32)
13230 i32 __builtin_mips_lwx (void *, i32)
13231 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
13232 i32 __builtin_mips_bposge32 (void)
13233 a64 __builtin_mips_madd (a64, i32, i32);
13234 a64 __builtin_mips_maddu (a64, ui32, ui32);
13235 a64 __builtin_mips_msub (a64, i32, i32);
13236 a64 __builtin_mips_msubu (a64, ui32, ui32);
13237 a64 __builtin_mips_mult (i32, i32);
13238 a64 __builtin_mips_multu (ui32, ui32);
13241 The following built-in functions map directly to a particular MIPS DSP REV 2
13242 instruction. Please refer to the architecture specification
13243 for details on what each instruction does.
13246 v4q7 __builtin_mips_absq_s_qb (v4q7);
13247 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
13248 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
13249 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
13250 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
13251 i32 __builtin_mips_append (i32, i32, imm0_31);
13252 i32 __builtin_mips_balign (i32, i32, imm0_3);
13253 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
13254 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
13255 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
13256 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
13257 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
13258 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
13259 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
13260 q31 __builtin_mips_mulq_rs_w (q31, q31);
13261 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
13262 q31 __builtin_mips_mulq_s_w (q31, q31);
13263 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
13264 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
13265 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
13266 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
13267 i32 __builtin_mips_prepend (i32, i32, imm0_31);
13268 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
13269 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
13270 v4i8 __builtin_mips_shra_qb (v4i8, i32);
13271 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
13272 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
13273 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
13274 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
13275 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
13276 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
13277 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
13278 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
13279 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
13280 q31 __builtin_mips_addqh_w (q31, q31);
13281 q31 __builtin_mips_addqh_r_w (q31, q31);
13282 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
13283 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
13284 q31 __builtin_mips_subqh_w (q31, q31);
13285 q31 __builtin_mips_subqh_r_w (q31, q31);
13286 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
13287 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
13288 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
13289 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
13290 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
13291 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
13295 @node MIPS Paired-Single Support
13296 @subsection MIPS Paired-Single Support
13298 The MIPS64 architecture includes a number of instructions that
13299 operate on pairs of single-precision floating-point values.
13300 Each pair is packed into a 64-bit floating-point register,
13301 with one element being designated the ``upper half'' and
13302 the other being designated the ``lower half''.
13304 GCC supports paired-single operations using both the generic
13305 vector extensions (@pxref{Vector Extensions}) and a collection of
13306 MIPS-specific built-in functions. Both kinds of support are
13307 enabled by the @option{-mpaired-single} command-line option.
13309 The vector type associated with paired-single values is usually
13310 called @code{v2sf}. It can be defined in C as follows:
13313 typedef float v2sf __attribute__ ((vector_size (8)));
13316 @code{v2sf} values are initialized in the same way as aggregates.
13320 v2sf a = @{1.5, 9.1@};
13323 b = (v2sf) @{e, f@};
13326 @emph{Note:} The CPU's endianness determines which value is stored in
13327 the upper half of a register and which value is stored in the lower half.
13328 On little-endian targets, the first value is the lower one and the second
13329 value is the upper one. The opposite order applies to big-endian targets.
13330 For example, the code above sets the lower half of @code{a} to
13331 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
13333 @node MIPS Loongson Built-in Functions
13334 @subsection MIPS Loongson Built-in Functions
13336 GCC provides intrinsics to access the SIMD instructions provided by the
13337 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
13338 available after inclusion of the @code{loongson.h} header file,
13339 operate on the following 64-bit vector types:
13342 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
13343 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
13344 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
13345 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
13346 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
13347 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
13350 The intrinsics provided are listed below; each is named after the
13351 machine instruction to which it corresponds, with suffixes added as
13352 appropriate to distinguish intrinsics that expand to the same machine
13353 instruction yet have different argument types. Refer to the architecture
13354 documentation for a description of the functionality of each
13358 int16x4_t packsswh (int32x2_t s, int32x2_t t);
13359 int8x8_t packsshb (int16x4_t s, int16x4_t t);
13360 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
13361 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13362 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13363 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13364 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13365 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13366 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13367 uint64_t paddd_u (uint64_t s, uint64_t t);
13368 int64_t paddd_s (int64_t s, int64_t t);
13369 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13370 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13371 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13372 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13373 uint64_t pandn_ud (uint64_t s, uint64_t t);
13374 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13375 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13376 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13377 int64_t pandn_sd (int64_t s, int64_t t);
13378 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13379 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13380 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13381 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13382 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13383 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13384 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13385 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13386 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13387 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13388 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13389 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13390 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13391 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13392 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13393 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13394 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13395 uint16x4_t pextrh_u (uint16x4_t s, int field);
13396 int16x4_t pextrh_s (int16x4_t s, int field);
13397 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13398 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13399 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13400 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13401 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13402 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13403 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13404 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13405 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13406 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13407 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13408 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13409 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13410 uint8x8_t pmovmskb_u (uint8x8_t s);
13411 int8x8_t pmovmskb_s (int8x8_t s);
13412 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13413 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13414 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13415 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13416 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13417 uint16x4_t biadd (uint8x8_t s);
13418 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13419 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13420 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13421 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13422 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13423 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13424 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13425 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13426 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13427 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13428 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13429 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13430 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13431 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13432 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13433 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13434 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13435 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13436 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13437 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13438 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13439 uint64_t psubd_u (uint64_t s, uint64_t t);
13440 int64_t psubd_s (int64_t s, int64_t t);
13441 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13442 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13443 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13444 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13445 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13446 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13447 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13448 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13449 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13450 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13451 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13452 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13453 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13454 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13455 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13456 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13460 * Paired-Single Arithmetic::
13461 * Paired-Single Built-in Functions::
13462 * MIPS-3D Built-in Functions::
13465 @node Paired-Single Arithmetic
13466 @subsubsection Paired-Single Arithmetic
13468 The table below lists the @code{v2sf} operations for which hardware
13469 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13470 values and @code{x} is an integral value.
13472 @multitable @columnfractions .50 .50
13473 @item C code @tab MIPS instruction
13474 @item @code{a + b} @tab @code{add.ps}
13475 @item @code{a - b} @tab @code{sub.ps}
13476 @item @code{-a} @tab @code{neg.ps}
13477 @item @code{a * b} @tab @code{mul.ps}
13478 @item @code{a * b + c} @tab @code{madd.ps}
13479 @item @code{a * b - c} @tab @code{msub.ps}
13480 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13481 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13482 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13485 Note that the multiply-accumulate instructions can be disabled
13486 using the command-line option @code{-mno-fused-madd}.
13488 @node Paired-Single Built-in Functions
13489 @subsubsection Paired-Single Built-in Functions
13491 The following paired-single functions map directly to a particular
13492 MIPS instruction. Please refer to the architecture specification
13493 for details on what each instruction does.
13496 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13497 Pair lower lower (@code{pll.ps}).
13499 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13500 Pair upper lower (@code{pul.ps}).
13502 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13503 Pair lower upper (@code{plu.ps}).
13505 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13506 Pair upper upper (@code{puu.ps}).
13508 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13509 Convert pair to paired single (@code{cvt.ps.s}).
13511 @item float __builtin_mips_cvt_s_pl (v2sf)
13512 Convert pair lower to single (@code{cvt.s.pl}).
13514 @item float __builtin_mips_cvt_s_pu (v2sf)
13515 Convert pair upper to single (@code{cvt.s.pu}).
13517 @item v2sf __builtin_mips_abs_ps (v2sf)
13518 Absolute value (@code{abs.ps}).
13520 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13521 Align variable (@code{alnv.ps}).
13523 @emph{Note:} The value of the third parameter must be 0 or 4
13524 modulo 8, otherwise the result is unpredictable. Please read the
13525 instruction description for details.
13528 The following multi-instruction functions are also available.
13529 In each case, @var{cond} can be any of the 16 floating-point conditions:
13530 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13531 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13532 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13535 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13536 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13537 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13538 @code{movt.ps}/@code{movf.ps}).
13540 The @code{movt} functions return the value @var{x} computed by:
13543 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13544 mov.ps @var{x},@var{c}
13545 movt.ps @var{x},@var{d},@var{cc}
13548 The @code{movf} functions are similar but use @code{movf.ps} instead
13551 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13552 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13553 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13554 @code{bc1t}/@code{bc1f}).
13556 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13557 and return either the upper or lower half of the result. For example:
13561 if (__builtin_mips_upper_c_eq_ps (a, b))
13562 upper_halves_are_equal ();
13564 upper_halves_are_unequal ();
13566 if (__builtin_mips_lower_c_eq_ps (a, b))
13567 lower_halves_are_equal ();
13569 lower_halves_are_unequal ();
13573 @node MIPS-3D Built-in Functions
13574 @subsubsection MIPS-3D Built-in Functions
13576 The MIPS-3D Application-Specific Extension (ASE) includes additional
13577 paired-single instructions that are designed to improve the performance
13578 of 3D graphics operations. Support for these instructions is controlled
13579 by the @option{-mips3d} command-line option.
13581 The functions listed below map directly to a particular MIPS-3D
13582 instruction. Please refer to the architecture specification for
13583 more details on what each instruction does.
13586 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13587 Reduction add (@code{addr.ps}).
13589 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13590 Reduction multiply (@code{mulr.ps}).
13592 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13593 Convert paired single to paired word (@code{cvt.pw.ps}).
13595 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13596 Convert paired word to paired single (@code{cvt.ps.pw}).
13598 @item float __builtin_mips_recip1_s (float)
13599 @itemx double __builtin_mips_recip1_d (double)
13600 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13601 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13603 @item float __builtin_mips_recip2_s (float, float)
13604 @itemx double __builtin_mips_recip2_d (double, double)
13605 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13606 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13608 @item float __builtin_mips_rsqrt1_s (float)
13609 @itemx double __builtin_mips_rsqrt1_d (double)
13610 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13611 Reduced-precision reciprocal square root (sequence step 1)
13612 (@code{rsqrt1.@var{fmt}}).
13614 @item float __builtin_mips_rsqrt2_s (float, float)
13615 @itemx double __builtin_mips_rsqrt2_d (double, double)
13616 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13617 Reduced-precision reciprocal square root (sequence step 2)
13618 (@code{rsqrt2.@var{fmt}}).
13621 The following multi-instruction functions are also available.
13622 In each case, @var{cond} can be any of the 16 floating-point conditions:
13623 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13624 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13625 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13628 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13629 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13630 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13631 @code{bc1t}/@code{bc1f}).
13633 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13634 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13639 if (__builtin_mips_cabs_eq_s (a, b))
13645 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13646 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13647 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13648 @code{bc1t}/@code{bc1f}).
13650 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13651 and return either the upper or lower half of the result. For example:
13655 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13656 upper_halves_are_equal ();
13658 upper_halves_are_unequal ();
13660 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13661 lower_halves_are_equal ();
13663 lower_halves_are_unequal ();
13666 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13667 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13668 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13669 @code{movt.ps}/@code{movf.ps}).
13671 The @code{movt} functions return the value @var{x} computed by:
13674 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13675 mov.ps @var{x},@var{c}
13676 movt.ps @var{x},@var{d},@var{cc}
13679 The @code{movf} functions are similar but use @code{movf.ps} instead
13682 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13683 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13684 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13685 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13686 Comparison of two paired-single values
13687 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13688 @code{bc1any2t}/@code{bc1any2f}).
13690 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13691 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13692 result is true and the @code{all} forms return true if both results are true.
13697 if (__builtin_mips_any_c_eq_ps (a, b))
13702 if (__builtin_mips_all_c_eq_ps (a, b))
13708 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13709 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13710 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13711 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13712 Comparison of four paired-single values
13713 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13714 @code{bc1any4t}/@code{bc1any4f}).
13716 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13717 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13718 The @code{any} forms return true if any of the four results are true
13719 and the @code{all} forms return true if all four results are true.
13724 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13729 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13736 @node MIPS SIMD Architecture (MSA) Support
13737 @subsection MIPS SIMD Architecture (MSA) Support
13740 * MIPS SIMD Architecture Built-in Functions::
13743 GCC provides intrinsics to access the SIMD instructions provided by the
13744 MSA MIPS SIMD Architecture. The interface is made available by including
13745 @code{<msa.h>} and using @option{-mmsa -mhard-float -mfp64 -mnan=2008}.
13746 For each @code{__builtin_msa_*}, there is a shortened name of the intrinsic,
13749 MSA implements 128-bit wide vector registers, operating on 8-, 16-, 32- and
13750 64-bit integer, 16- and 32-bit fixed-point, or 32- and 64-bit floating point
13751 data elements. The following vectors typedefs are included in @code{msa.h}:
13753 @item @code{v16i8}, a vector of sixteen signed 8-bit integers;
13754 @item @code{v16u8}, a vector of sixteen unsigned 8-bit integers;
13755 @item @code{v8i16}, a vector of eight signed 16-bit integers;
13756 @item @code{v8u16}, a vector of eight unsigned 16-bit integers;
13757 @item @code{v4i32}, a vector of four signed 32-bit integers;
13758 @item @code{v4u32}, a vector of four unsigned 32-bit integers;
13759 @item @code{v2i64}, a vector of two signed 64-bit integers;
13760 @item @code{v2u64}, a vector of two unsigned 64-bit integers;
13761 @item @code{v4f32}, a vector of four 32-bit floats;
13762 @item @code{v2f64}, a vector of two 64-bit doubles.
13765 Intructions and corresponding built-ins may have additional restrictions and/or
13766 input/output values manipulated:
13768 @item @code{imm0_1}, an integer literal in range 0 to 1;
13769 @item @code{imm0_3}, an integer literal in range 0 to 3;
13770 @item @code{imm0_7}, an integer literal in range 0 to 7;
13771 @item @code{imm0_15}, an integer literal in range 0 to 15;
13772 @item @code{imm0_31}, an integer literal in range 0 to 31;
13773 @item @code{imm0_63}, an integer literal in range 0 to 63;
13774 @item @code{imm0_255}, an integer literal in range 0 to 255;
13775 @item @code{imm_n16_15}, an integer literal in range -16 to 15;
13776 @item @code{imm_n512_511}, an integer literal in range -512 to 511;
13777 @item @code{imm_n1024_1022}, an integer literal in range -512 to 511 left
13778 shifted by 1 bit, i.e., -1024, -1022, @dots{}, 1020, 1022;
13779 @item @code{imm_n2048_2044}, an integer literal in range -512 to 511 left
13780 shifted by 2 bits, i.e., -2048, -2044, @dots{}, 2040, 2044;
13781 @item @code{imm_n4096_4088}, an integer literal in range -512 to 511 left
13782 shifted by 3 bits, i.e., -4096, -4088, @dots{}, 4080, 4088;
13783 @item @code{imm1_4}, an integer literal in range 1 to 4;
13784 @item @code{i32, i64, u32, u64, f32, f64}, defined as follows:
13790 #if __LONG_MAX__ == __LONG_LONG_MAX__
13793 typedef long long i64;
13796 typedef unsigned int u32;
13797 #if __LONG_MAX__ == __LONG_LONG_MAX__
13798 typedef unsigned long u64;
13800 typedef unsigned long long u64;
13803 typedef double f64;
13808 @node MIPS SIMD Architecture Built-in Functions
13809 @subsubsection MIPS SIMD Architecture Built-in Functions
13811 The intrinsics provided are listed below; each is named after the
13812 machine instruction.
13815 v16i8 __builtin_msa_add_a_b (v16i8, v16i8);
13816 v8i16 __builtin_msa_add_a_h (v8i16, v8i16);
13817 v4i32 __builtin_msa_add_a_w (v4i32, v4i32);
13818 v2i64 __builtin_msa_add_a_d (v2i64, v2i64);
13820 v16i8 __builtin_msa_adds_a_b (v16i8, v16i8);
13821 v8i16 __builtin_msa_adds_a_h (v8i16, v8i16);
13822 v4i32 __builtin_msa_adds_a_w (v4i32, v4i32);
13823 v2i64 __builtin_msa_adds_a_d (v2i64, v2i64);
13825 v16i8 __builtin_msa_adds_s_b (v16i8, v16i8);
13826 v8i16 __builtin_msa_adds_s_h (v8i16, v8i16);
13827 v4i32 __builtin_msa_adds_s_w (v4i32, v4i32);
13828 v2i64 __builtin_msa_adds_s_d (v2i64, v2i64);
13830 v16u8 __builtin_msa_adds_u_b (v16u8, v16u8);
13831 v8u16 __builtin_msa_adds_u_h (v8u16, v8u16);
13832 v4u32 __builtin_msa_adds_u_w (v4u32, v4u32);
13833 v2u64 __builtin_msa_adds_u_d (v2u64, v2u64);
13835 v16i8 __builtin_msa_addv_b (v16i8, v16i8);
13836 v8i16 __builtin_msa_addv_h (v8i16, v8i16);
13837 v4i32 __builtin_msa_addv_w (v4i32, v4i32);
13838 v2i64 __builtin_msa_addv_d (v2i64, v2i64);
13840 v16i8 __builtin_msa_addvi_b (v16i8, imm0_31);
13841 v8i16 __builtin_msa_addvi_h (v8i16, imm0_31);
13842 v4i32 __builtin_msa_addvi_w (v4i32, imm0_31);
13843 v2i64 __builtin_msa_addvi_d (v2i64, imm0_31);
13845 v16u8 __builtin_msa_and_v (v16u8, v16u8);
13847 v16u8 __builtin_msa_andi_b (v16u8, imm0_255);
13849 v16i8 __builtin_msa_asub_s_b (v16i8, v16i8);
13850 v8i16 __builtin_msa_asub_s_h (v8i16, v8i16);
13851 v4i32 __builtin_msa_asub_s_w (v4i32, v4i32);
13852 v2i64 __builtin_msa_asub_s_d (v2i64, v2i64);
13854 v16u8 __builtin_msa_asub_u_b (v16u8, v16u8);
13855 v8u16 __builtin_msa_asub_u_h (v8u16, v8u16);
13856 v4u32 __builtin_msa_asub_u_w (v4u32, v4u32);
13857 v2u64 __builtin_msa_asub_u_d (v2u64, v2u64);
13859 v16i8 __builtin_msa_ave_s_b (v16i8, v16i8);
13860 v8i16 __builtin_msa_ave_s_h (v8i16, v8i16);
13861 v4i32 __builtin_msa_ave_s_w (v4i32, v4i32);
13862 v2i64 __builtin_msa_ave_s_d (v2i64, v2i64);
13864 v16u8 __builtin_msa_ave_u_b (v16u8, v16u8);
13865 v8u16 __builtin_msa_ave_u_h (v8u16, v8u16);
13866 v4u32 __builtin_msa_ave_u_w (v4u32, v4u32);
13867 v2u64 __builtin_msa_ave_u_d (v2u64, v2u64);
13869 v16i8 __builtin_msa_aver_s_b (v16i8, v16i8);
13870 v8i16 __builtin_msa_aver_s_h (v8i16, v8i16);
13871 v4i32 __builtin_msa_aver_s_w (v4i32, v4i32);
13872 v2i64 __builtin_msa_aver_s_d (v2i64, v2i64);
13874 v16u8 __builtin_msa_aver_u_b (v16u8, v16u8);
13875 v8u16 __builtin_msa_aver_u_h (v8u16, v8u16);
13876 v4u32 __builtin_msa_aver_u_w (v4u32, v4u32);
13877 v2u64 __builtin_msa_aver_u_d (v2u64, v2u64);
13879 v16u8 __builtin_msa_bclr_b (v16u8, v16u8);
13880 v8u16 __builtin_msa_bclr_h (v8u16, v8u16);
13881 v4u32 __builtin_msa_bclr_w (v4u32, v4u32);
13882 v2u64 __builtin_msa_bclr_d (v2u64, v2u64);
13884 v16u8 __builtin_msa_bclri_b (v16u8, imm0_7);
13885 v8u16 __builtin_msa_bclri_h (v8u16, imm0_15);
13886 v4u32 __builtin_msa_bclri_w (v4u32, imm0_31);
13887 v2u64 __builtin_msa_bclri_d (v2u64, imm0_63);
13889 v16u8 __builtin_msa_binsl_b (v16u8, v16u8, v16u8);
13890 v8u16 __builtin_msa_binsl_h (v8u16, v8u16, v8u16);
13891 v4u32 __builtin_msa_binsl_w (v4u32, v4u32, v4u32);
13892 v2u64 __builtin_msa_binsl_d (v2u64, v2u64, v2u64);
13894 v16u8 __builtin_msa_binsli_b (v16u8, v16u8, imm0_7);
13895 v8u16 __builtin_msa_binsli_h (v8u16, v8u16, imm0_15);
13896 v4u32 __builtin_msa_binsli_w (v4u32, v4u32, imm0_31);
13897 v2u64 __builtin_msa_binsli_d (v2u64, v2u64, imm0_63);
13899 v16u8 __builtin_msa_binsr_b (v16u8, v16u8, v16u8);
13900 v8u16 __builtin_msa_binsr_h (v8u16, v8u16, v8u16);
13901 v4u32 __builtin_msa_binsr_w (v4u32, v4u32, v4u32);
13902 v2u64 __builtin_msa_binsr_d (v2u64, v2u64, v2u64);
13904 v16u8 __builtin_msa_binsri_b (v16u8, v16u8, imm0_7);
13905 v8u16 __builtin_msa_binsri_h (v8u16, v8u16, imm0_15);
13906 v4u32 __builtin_msa_binsri_w (v4u32, v4u32, imm0_31);
13907 v2u64 __builtin_msa_binsri_d (v2u64, v2u64, imm0_63);
13909 v16u8 __builtin_msa_bmnz_v (v16u8, v16u8, v16u8);
13911 v16u8 __builtin_msa_bmnzi_b (v16u8, v16u8, imm0_255);
13913 v16u8 __builtin_msa_bmz_v (v16u8, v16u8, v16u8);
13915 v16u8 __builtin_msa_bmzi_b (v16u8, v16u8, imm0_255);
13917 v16u8 __builtin_msa_bneg_b (v16u8, v16u8);
13918 v8u16 __builtin_msa_bneg_h (v8u16, v8u16);
13919 v4u32 __builtin_msa_bneg_w (v4u32, v4u32);
13920 v2u64 __builtin_msa_bneg_d (v2u64, v2u64);
13922 v16u8 __builtin_msa_bnegi_b (v16u8, imm0_7);
13923 v8u16 __builtin_msa_bnegi_h (v8u16, imm0_15);
13924 v4u32 __builtin_msa_bnegi_w (v4u32, imm0_31);
13925 v2u64 __builtin_msa_bnegi_d (v2u64, imm0_63);
13927 i32 __builtin_msa_bnz_b (v16u8);
13928 i32 __builtin_msa_bnz_h (v8u16);
13929 i32 __builtin_msa_bnz_w (v4u32);
13930 i32 __builtin_msa_bnz_d (v2u64);
13932 i32 __builtin_msa_bnz_v (v16u8);
13934 v16u8 __builtin_msa_bsel_v (v16u8, v16u8, v16u8);
13936 v16u8 __builtin_msa_bseli_b (v16u8, v16u8, imm0_255);
13938 v16u8 __builtin_msa_bset_b (v16u8, v16u8);
13939 v8u16 __builtin_msa_bset_h (v8u16, v8u16);
13940 v4u32 __builtin_msa_bset_w (v4u32, v4u32);
13941 v2u64 __builtin_msa_bset_d (v2u64, v2u64);
13943 v16u8 __builtin_msa_bseti_b (v16u8, imm0_7);
13944 v8u16 __builtin_msa_bseti_h (v8u16, imm0_15);
13945 v4u32 __builtin_msa_bseti_w (v4u32, imm0_31);
13946 v2u64 __builtin_msa_bseti_d (v2u64, imm0_63);
13948 i32 __builtin_msa_bz_b (v16u8);
13949 i32 __builtin_msa_bz_h (v8u16);
13950 i32 __builtin_msa_bz_w (v4u32);
13951 i32 __builtin_msa_bz_d (v2u64);
13953 i32 __builtin_msa_bz_v (v16u8);
13955 v16i8 __builtin_msa_ceq_b (v16i8, v16i8);
13956 v8i16 __builtin_msa_ceq_h (v8i16, v8i16);
13957 v4i32 __builtin_msa_ceq_w (v4i32, v4i32);
13958 v2i64 __builtin_msa_ceq_d (v2i64, v2i64);
13960 v16i8 __builtin_msa_ceqi_b (v16i8, imm_n16_15);
13961 v8i16 __builtin_msa_ceqi_h (v8i16, imm_n16_15);
13962 v4i32 __builtin_msa_ceqi_w (v4i32, imm_n16_15);
13963 v2i64 __builtin_msa_ceqi_d (v2i64, imm_n16_15);
13965 i32 __builtin_msa_cfcmsa (imm0_31);
13967 v16i8 __builtin_msa_cle_s_b (v16i8, v16i8);
13968 v8i16 __builtin_msa_cle_s_h (v8i16, v8i16);
13969 v4i32 __builtin_msa_cle_s_w (v4i32, v4i32);
13970 v2i64 __builtin_msa_cle_s_d (v2i64, v2i64);
13972 v16i8 __builtin_msa_cle_u_b (v16u8, v16u8);
13973 v8i16 __builtin_msa_cle_u_h (v8u16, v8u16);
13974 v4i32 __builtin_msa_cle_u_w (v4u32, v4u32);
13975 v2i64 __builtin_msa_cle_u_d (v2u64, v2u64);
13977 v16i8 __builtin_msa_clei_s_b (v16i8, imm_n16_15);
13978 v8i16 __builtin_msa_clei_s_h (v8i16, imm_n16_15);
13979 v4i32 __builtin_msa_clei_s_w (v4i32, imm_n16_15);
13980 v2i64 __builtin_msa_clei_s_d (v2i64, imm_n16_15);
13982 v16i8 __builtin_msa_clei_u_b (v16u8, imm0_31);
13983 v8i16 __builtin_msa_clei_u_h (v8u16, imm0_31);
13984 v4i32 __builtin_msa_clei_u_w (v4u32, imm0_31);
13985 v2i64 __builtin_msa_clei_u_d (v2u64, imm0_31);
13987 v16i8 __builtin_msa_clt_s_b (v16i8, v16i8);
13988 v8i16 __builtin_msa_clt_s_h (v8i16, v8i16);
13989 v4i32 __builtin_msa_clt_s_w (v4i32, v4i32);
13990 v2i64 __builtin_msa_clt_s_d (v2i64, v2i64);
13992 v16i8 __builtin_msa_clt_u_b (v16u8, v16u8);
13993 v8i16 __builtin_msa_clt_u_h (v8u16, v8u16);
13994 v4i32 __builtin_msa_clt_u_w (v4u32, v4u32);
13995 v2i64 __builtin_msa_clt_u_d (v2u64, v2u64);
13997 v16i8 __builtin_msa_clti_s_b (v16i8, imm_n16_15);
13998 v8i16 __builtin_msa_clti_s_h (v8i16, imm_n16_15);
13999 v4i32 __builtin_msa_clti_s_w (v4i32, imm_n16_15);
14000 v2i64 __builtin_msa_clti_s_d (v2i64, imm_n16_15);
14002 v16i8 __builtin_msa_clti_u_b (v16u8, imm0_31);
14003 v8i16 __builtin_msa_clti_u_h (v8u16, imm0_31);
14004 v4i32 __builtin_msa_clti_u_w (v4u32, imm0_31);
14005 v2i64 __builtin_msa_clti_u_d (v2u64, imm0_31);
14007 i32 __builtin_msa_copy_s_b (v16i8, imm0_15);
14008 i32 __builtin_msa_copy_s_h (v8i16, imm0_7);
14009 i32 __builtin_msa_copy_s_w (v4i32, imm0_3);
14010 i64 __builtin_msa_copy_s_d (v2i64, imm0_1);
14012 u32 __builtin_msa_copy_u_b (v16i8, imm0_15);
14013 u32 __builtin_msa_copy_u_h (v8i16, imm0_7);
14014 u32 __builtin_msa_copy_u_w (v4i32, imm0_3);
14015 u64 __builtin_msa_copy_u_d (v2i64, imm0_1);
14017 void __builtin_msa_ctcmsa (imm0_31, i32);
14019 v16i8 __builtin_msa_div_s_b (v16i8, v16i8);
14020 v8i16 __builtin_msa_div_s_h (v8i16, v8i16);
14021 v4i32 __builtin_msa_div_s_w (v4i32, v4i32);
14022 v2i64 __builtin_msa_div_s_d (v2i64, v2i64);
14024 v16u8 __builtin_msa_div_u_b (v16u8, v16u8);
14025 v8u16 __builtin_msa_div_u_h (v8u16, v8u16);
14026 v4u32 __builtin_msa_div_u_w (v4u32, v4u32);
14027 v2u64 __builtin_msa_div_u_d (v2u64, v2u64);
14029 v8i16 __builtin_msa_dotp_s_h (v16i8, v16i8);
14030 v4i32 __builtin_msa_dotp_s_w (v8i16, v8i16);
14031 v2i64 __builtin_msa_dotp_s_d (v4i32, v4i32);
14033 v8u16 __builtin_msa_dotp_u_h (v16u8, v16u8);
14034 v4u32 __builtin_msa_dotp_u_w (v8u16, v8u16);
14035 v2u64 __builtin_msa_dotp_u_d (v4u32, v4u32);
14037 v8i16 __builtin_msa_dpadd_s_h (v8i16, v16i8, v16i8);
14038 v4i32 __builtin_msa_dpadd_s_w (v4i32, v8i16, v8i16);
14039 v2i64 __builtin_msa_dpadd_s_d (v2i64, v4i32, v4i32);
14041 v8u16 __builtin_msa_dpadd_u_h (v8u16, v16u8, v16u8);
14042 v4u32 __builtin_msa_dpadd_u_w (v4u32, v8u16, v8u16);
14043 v2u64 __builtin_msa_dpadd_u_d (v2u64, v4u32, v4u32);
14045 v8i16 __builtin_msa_dpsub_s_h (v8i16, v16i8, v16i8);
14046 v4i32 __builtin_msa_dpsub_s_w (v4i32, v8i16, v8i16);
14047 v2i64 __builtin_msa_dpsub_s_d (v2i64, v4i32, v4i32);
14049 v8i16 __builtin_msa_dpsub_u_h (v8i16, v16u8, v16u8);
14050 v4i32 __builtin_msa_dpsub_u_w (v4i32, v8u16, v8u16);
14051 v2i64 __builtin_msa_dpsub_u_d (v2i64, v4u32, v4u32);
14053 v4f32 __builtin_msa_fadd_w (v4f32, v4f32);
14054 v2f64 __builtin_msa_fadd_d (v2f64, v2f64);
14056 v4i32 __builtin_msa_fcaf_w (v4f32, v4f32);
14057 v2i64 __builtin_msa_fcaf_d (v2f64, v2f64);
14059 v4i32 __builtin_msa_fceq_w (v4f32, v4f32);
14060 v2i64 __builtin_msa_fceq_d (v2f64, v2f64);
14062 v4i32 __builtin_msa_fclass_w (v4f32);
14063 v2i64 __builtin_msa_fclass_d (v2f64);
14065 v4i32 __builtin_msa_fcle_w (v4f32, v4f32);
14066 v2i64 __builtin_msa_fcle_d (v2f64, v2f64);
14068 v4i32 __builtin_msa_fclt_w (v4f32, v4f32);
14069 v2i64 __builtin_msa_fclt_d (v2f64, v2f64);
14071 v4i32 __builtin_msa_fcne_w (v4f32, v4f32);
14072 v2i64 __builtin_msa_fcne_d (v2f64, v2f64);
14074 v4i32 __builtin_msa_fcor_w (v4f32, v4f32);
14075 v2i64 __builtin_msa_fcor_d (v2f64, v2f64);
14077 v4i32 __builtin_msa_fcueq_w (v4f32, v4f32);
14078 v2i64 __builtin_msa_fcueq_d (v2f64, v2f64);
14080 v4i32 __builtin_msa_fcule_w (v4f32, v4f32);
14081 v2i64 __builtin_msa_fcule_d (v2f64, v2f64);
14083 v4i32 __builtin_msa_fcult_w (v4f32, v4f32);
14084 v2i64 __builtin_msa_fcult_d (v2f64, v2f64);
14086 v4i32 __builtin_msa_fcun_w (v4f32, v4f32);
14087 v2i64 __builtin_msa_fcun_d (v2f64, v2f64);
14089 v4i32 __builtin_msa_fcune_w (v4f32, v4f32);
14090 v2i64 __builtin_msa_fcune_d (v2f64, v2f64);
14092 v4f32 __builtin_msa_fdiv_w (v4f32, v4f32);
14093 v2f64 __builtin_msa_fdiv_d (v2f64, v2f64);
14095 v8i16 __builtin_msa_fexdo_h (v4f32, v4f32);
14096 v4f32 __builtin_msa_fexdo_w (v2f64, v2f64);
14098 v4f32 __builtin_msa_fexp2_w (v4f32, v4i32);
14099 v2f64 __builtin_msa_fexp2_d (v2f64, v2i64);
14101 v4f32 __builtin_msa_fexupl_w (v8i16);
14102 v2f64 __builtin_msa_fexupl_d (v4f32);
14104 v4f32 __builtin_msa_fexupr_w (v8i16);
14105 v2f64 __builtin_msa_fexupr_d (v4f32);
14107 v4f32 __builtin_msa_ffint_s_w (v4i32);
14108 v2f64 __builtin_msa_ffint_s_d (v2i64);
14110 v4f32 __builtin_msa_ffint_u_w (v4u32);
14111 v2f64 __builtin_msa_ffint_u_d (v2u64);
14113 v4f32 __builtin_msa_ffql_w (v8i16);
14114 v2f64 __builtin_msa_ffql_d (v4i32);
14116 v4f32 __builtin_msa_ffqr_w (v8i16);
14117 v2f64 __builtin_msa_ffqr_d (v4i32);
14119 v16i8 __builtin_msa_fill_b (i32);
14120 v8i16 __builtin_msa_fill_h (i32);
14121 v4i32 __builtin_msa_fill_w (i32);
14122 v2i64 __builtin_msa_fill_d (i64);
14124 v4f32 __builtin_msa_flog2_w (v4f32);
14125 v2f64 __builtin_msa_flog2_d (v2f64);
14127 v4f32 __builtin_msa_fmadd_w (v4f32, v4f32, v4f32);
14128 v2f64 __builtin_msa_fmadd_d (v2f64, v2f64, v2f64);
14130 v4f32 __builtin_msa_fmax_w (v4f32, v4f32);
14131 v2f64 __builtin_msa_fmax_d (v2f64, v2f64);
14133 v4f32 __builtin_msa_fmax_a_w (v4f32, v4f32);
14134 v2f64 __builtin_msa_fmax_a_d (v2f64, v2f64);
14136 v4f32 __builtin_msa_fmin_w (v4f32, v4f32);
14137 v2f64 __builtin_msa_fmin_d (v2f64, v2f64);
14139 v4f32 __builtin_msa_fmin_a_w (v4f32, v4f32);
14140 v2f64 __builtin_msa_fmin_a_d (v2f64, v2f64);
14142 v4f32 __builtin_msa_fmsub_w (v4f32, v4f32, v4f32);
14143 v2f64 __builtin_msa_fmsub_d (v2f64, v2f64, v2f64);
14145 v4f32 __builtin_msa_fmul_w (v4f32, v4f32);
14146 v2f64 __builtin_msa_fmul_d (v2f64, v2f64);
14148 v4f32 __builtin_msa_frint_w (v4f32);
14149 v2f64 __builtin_msa_frint_d (v2f64);
14151 v4f32 __builtin_msa_frcp_w (v4f32);
14152 v2f64 __builtin_msa_frcp_d (v2f64);
14154 v4f32 __builtin_msa_frsqrt_w (v4f32);
14155 v2f64 __builtin_msa_frsqrt_d (v2f64);
14157 v4i32 __builtin_msa_fsaf_w (v4f32, v4f32);
14158 v2i64 __builtin_msa_fsaf_d (v2f64, v2f64);
14160 v4i32 __builtin_msa_fseq_w (v4f32, v4f32);
14161 v2i64 __builtin_msa_fseq_d (v2f64, v2f64);
14163 v4i32 __builtin_msa_fsle_w (v4f32, v4f32);
14164 v2i64 __builtin_msa_fsle_d (v2f64, v2f64);
14166 v4i32 __builtin_msa_fslt_w (v4f32, v4f32);
14167 v2i64 __builtin_msa_fslt_d (v2f64, v2f64);
14169 v4i32 __builtin_msa_fsne_w (v4f32, v4f32);
14170 v2i64 __builtin_msa_fsne_d (v2f64, v2f64);
14172 v4i32 __builtin_msa_fsor_w (v4f32, v4f32);
14173 v2i64 __builtin_msa_fsor_d (v2f64, v2f64);
14175 v4f32 __builtin_msa_fsqrt_w (v4f32);
14176 v2f64 __builtin_msa_fsqrt_d (v2f64);
14178 v4f32 __builtin_msa_fsub_w (v4f32, v4f32);
14179 v2f64 __builtin_msa_fsub_d (v2f64, v2f64);
14181 v4i32 __builtin_msa_fsueq_w (v4f32, v4f32);
14182 v2i64 __builtin_msa_fsueq_d (v2f64, v2f64);
14184 v4i32 __builtin_msa_fsule_w (v4f32, v4f32);
14185 v2i64 __builtin_msa_fsule_d (v2f64, v2f64);
14187 v4i32 __builtin_msa_fsult_w (v4f32, v4f32);
14188 v2i64 __builtin_msa_fsult_d (v2f64, v2f64);
14190 v4i32 __builtin_msa_fsun_w (v4f32, v4f32);
14191 v2i64 __builtin_msa_fsun_d (v2f64, v2f64);
14193 v4i32 __builtin_msa_fsune_w (v4f32, v4f32);
14194 v2i64 __builtin_msa_fsune_d (v2f64, v2f64);
14196 v4i32 __builtin_msa_ftint_s_w (v4f32);
14197 v2i64 __builtin_msa_ftint_s_d (v2f64);
14199 v4u32 __builtin_msa_ftint_u_w (v4f32);
14200 v2u64 __builtin_msa_ftint_u_d (v2f64);
14202 v8i16 __builtin_msa_ftq_h (v4f32, v4f32);
14203 v4i32 __builtin_msa_ftq_w (v2f64, v2f64);
14205 v4i32 __builtin_msa_ftrunc_s_w (v4f32);
14206 v2i64 __builtin_msa_ftrunc_s_d (v2f64);
14208 v4u32 __builtin_msa_ftrunc_u_w (v4f32);
14209 v2u64 __builtin_msa_ftrunc_u_d (v2f64);
14211 v8i16 __builtin_msa_hadd_s_h (v16i8, v16i8);
14212 v4i32 __builtin_msa_hadd_s_w (v8i16, v8i16);
14213 v2i64 __builtin_msa_hadd_s_d (v4i32, v4i32);
14215 v8u16 __builtin_msa_hadd_u_h (v16u8, v16u8);
14216 v4u32 __builtin_msa_hadd_u_w (v8u16, v8u16);
14217 v2u64 __builtin_msa_hadd_u_d (v4u32, v4u32);
14219 v8i16 __builtin_msa_hsub_s_h (v16i8, v16i8);
14220 v4i32 __builtin_msa_hsub_s_w (v8i16, v8i16);
14221 v2i64 __builtin_msa_hsub_s_d (v4i32, v4i32);
14223 v8i16 __builtin_msa_hsub_u_h (v16u8, v16u8);
14224 v4i32 __builtin_msa_hsub_u_w (v8u16, v8u16);
14225 v2i64 __builtin_msa_hsub_u_d (v4u32, v4u32);
14227 v16i8 __builtin_msa_ilvev_b (v16i8, v16i8);
14228 v8i16 __builtin_msa_ilvev_h (v8i16, v8i16);
14229 v4i32 __builtin_msa_ilvev_w (v4i32, v4i32);
14230 v2i64 __builtin_msa_ilvev_d (v2i64, v2i64);
14232 v16i8 __builtin_msa_ilvl_b (v16i8, v16i8);
14233 v8i16 __builtin_msa_ilvl_h (v8i16, v8i16);
14234 v4i32 __builtin_msa_ilvl_w (v4i32, v4i32);
14235 v2i64 __builtin_msa_ilvl_d (v2i64, v2i64);
14237 v16i8 __builtin_msa_ilvod_b (v16i8, v16i8);
14238 v8i16 __builtin_msa_ilvod_h (v8i16, v8i16);
14239 v4i32 __builtin_msa_ilvod_w (v4i32, v4i32);
14240 v2i64 __builtin_msa_ilvod_d (v2i64, v2i64);
14242 v16i8 __builtin_msa_ilvr_b (v16i8, v16i8);
14243 v8i16 __builtin_msa_ilvr_h (v8i16, v8i16);
14244 v4i32 __builtin_msa_ilvr_w (v4i32, v4i32);
14245 v2i64 __builtin_msa_ilvr_d (v2i64, v2i64);
14247 v16i8 __builtin_msa_insert_b (v16i8, imm0_15, i32);
14248 v8i16 __builtin_msa_insert_h (v8i16, imm0_7, i32);
14249 v4i32 __builtin_msa_insert_w (v4i32, imm0_3, i32);
14250 v2i64 __builtin_msa_insert_d (v2i64, imm0_1, i64);
14252 v16i8 __builtin_msa_insve_b (v16i8, imm0_15, v16i8);
14253 v8i16 __builtin_msa_insve_h (v8i16, imm0_7, v8i16);
14254 v4i32 __builtin_msa_insve_w (v4i32, imm0_3, v4i32);
14255 v2i64 __builtin_msa_insve_d (v2i64, imm0_1, v2i64);
14257 v16i8 __builtin_msa_ld_b (void *, imm_n512_511);
14258 v8i16 __builtin_msa_ld_h (void *, imm_n1024_1022);
14259 v4i32 __builtin_msa_ld_w (void *, imm_n2048_2044);
14260 v2i64 __builtin_msa_ld_d (void *, imm_n4096_4088);
14262 v16i8 __builtin_msa_ldi_b (imm_n512_511);
14263 v8i16 __builtin_msa_ldi_h (imm_n512_511);
14264 v4i32 __builtin_msa_ldi_w (imm_n512_511);
14265 v2i64 __builtin_msa_ldi_d (imm_n512_511);
14267 v8i16 __builtin_msa_madd_q_h (v8i16, v8i16, v8i16);
14268 v4i32 __builtin_msa_madd_q_w (v4i32, v4i32, v4i32);
14270 v8i16 __builtin_msa_maddr_q_h (v8i16, v8i16, v8i16);
14271 v4i32 __builtin_msa_maddr_q_w (v4i32, v4i32, v4i32);
14273 v16i8 __builtin_msa_maddv_b (v16i8, v16i8, v16i8);
14274 v8i16 __builtin_msa_maddv_h (v8i16, v8i16, v8i16);
14275 v4i32 __builtin_msa_maddv_w (v4i32, v4i32, v4i32);
14276 v2i64 __builtin_msa_maddv_d (v2i64, v2i64, v2i64);
14278 v16i8 __builtin_msa_max_a_b (v16i8, v16i8);
14279 v8i16 __builtin_msa_max_a_h (v8i16, v8i16);
14280 v4i32 __builtin_msa_max_a_w (v4i32, v4i32);
14281 v2i64 __builtin_msa_max_a_d (v2i64, v2i64);
14283 v16i8 __builtin_msa_max_s_b (v16i8, v16i8);
14284 v8i16 __builtin_msa_max_s_h (v8i16, v8i16);
14285 v4i32 __builtin_msa_max_s_w (v4i32, v4i32);
14286 v2i64 __builtin_msa_max_s_d (v2i64, v2i64);
14288 v16u8 __builtin_msa_max_u_b (v16u8, v16u8);
14289 v8u16 __builtin_msa_max_u_h (v8u16, v8u16);
14290 v4u32 __builtin_msa_max_u_w (v4u32, v4u32);
14291 v2u64 __builtin_msa_max_u_d (v2u64, v2u64);
14293 v16i8 __builtin_msa_maxi_s_b (v16i8, imm_n16_15);
14294 v8i16 __builtin_msa_maxi_s_h (v8i16, imm_n16_15);
14295 v4i32 __builtin_msa_maxi_s_w (v4i32, imm_n16_15);
14296 v2i64 __builtin_msa_maxi_s_d (v2i64, imm_n16_15);
14298 v16u8 __builtin_msa_maxi_u_b (v16u8, imm0_31);
14299 v8u16 __builtin_msa_maxi_u_h (v8u16, imm0_31);
14300 v4u32 __builtin_msa_maxi_u_w (v4u32, imm0_31);
14301 v2u64 __builtin_msa_maxi_u_d (v2u64, imm0_31);
14303 v16i8 __builtin_msa_min_a_b (v16i8, v16i8);
14304 v8i16 __builtin_msa_min_a_h (v8i16, v8i16);
14305 v4i32 __builtin_msa_min_a_w (v4i32, v4i32);
14306 v2i64 __builtin_msa_min_a_d (v2i64, v2i64);
14308 v16i8 __builtin_msa_min_s_b (v16i8, v16i8);
14309 v8i16 __builtin_msa_min_s_h (v8i16, v8i16);
14310 v4i32 __builtin_msa_min_s_w (v4i32, v4i32);
14311 v2i64 __builtin_msa_min_s_d (v2i64, v2i64);
14313 v16u8 __builtin_msa_min_u_b (v16u8, v16u8);
14314 v8u16 __builtin_msa_min_u_h (v8u16, v8u16);
14315 v4u32 __builtin_msa_min_u_w (v4u32, v4u32);
14316 v2u64 __builtin_msa_min_u_d (v2u64, v2u64);
14318 v16i8 __builtin_msa_mini_s_b (v16i8, imm_n16_15);
14319 v8i16 __builtin_msa_mini_s_h (v8i16, imm_n16_15);
14320 v4i32 __builtin_msa_mini_s_w (v4i32, imm_n16_15);
14321 v2i64 __builtin_msa_mini_s_d (v2i64, imm_n16_15);
14323 v16u8 __builtin_msa_mini_u_b (v16u8, imm0_31);
14324 v8u16 __builtin_msa_mini_u_h (v8u16, imm0_31);
14325 v4u32 __builtin_msa_mini_u_w (v4u32, imm0_31);
14326 v2u64 __builtin_msa_mini_u_d (v2u64, imm0_31);
14328 v16i8 __builtin_msa_mod_s_b (v16i8, v16i8);
14329 v8i16 __builtin_msa_mod_s_h (v8i16, v8i16);
14330 v4i32 __builtin_msa_mod_s_w (v4i32, v4i32);
14331 v2i64 __builtin_msa_mod_s_d (v2i64, v2i64);
14333 v16u8 __builtin_msa_mod_u_b (v16u8, v16u8);
14334 v8u16 __builtin_msa_mod_u_h (v8u16, v8u16);
14335 v4u32 __builtin_msa_mod_u_w (v4u32, v4u32);
14336 v2u64 __builtin_msa_mod_u_d (v2u64, v2u64);
14338 v16i8 __builtin_msa_move_v (v16i8);
14340 v8i16 __builtin_msa_msub_q_h (v8i16, v8i16, v8i16);
14341 v4i32 __builtin_msa_msub_q_w (v4i32, v4i32, v4i32);
14343 v8i16 __builtin_msa_msubr_q_h (v8i16, v8i16, v8i16);
14344 v4i32 __builtin_msa_msubr_q_w (v4i32, v4i32, v4i32);
14346 v16i8 __builtin_msa_msubv_b (v16i8, v16i8, v16i8);
14347 v8i16 __builtin_msa_msubv_h (v8i16, v8i16, v8i16);
14348 v4i32 __builtin_msa_msubv_w (v4i32, v4i32, v4i32);
14349 v2i64 __builtin_msa_msubv_d (v2i64, v2i64, v2i64);
14351 v8i16 __builtin_msa_mul_q_h (v8i16, v8i16);
14352 v4i32 __builtin_msa_mul_q_w (v4i32, v4i32);
14354 v8i16 __builtin_msa_mulr_q_h (v8i16, v8i16);
14355 v4i32 __builtin_msa_mulr_q_w (v4i32, v4i32);
14357 v16i8 __builtin_msa_mulv_b (v16i8, v16i8);
14358 v8i16 __builtin_msa_mulv_h (v8i16, v8i16);
14359 v4i32 __builtin_msa_mulv_w (v4i32, v4i32);
14360 v2i64 __builtin_msa_mulv_d (v2i64, v2i64);
14362 v16i8 __builtin_msa_nloc_b (v16i8);
14363 v8i16 __builtin_msa_nloc_h (v8i16);
14364 v4i32 __builtin_msa_nloc_w (v4i32);
14365 v2i64 __builtin_msa_nloc_d (v2i64);
14367 v16i8 __builtin_msa_nlzc_b (v16i8);
14368 v8i16 __builtin_msa_nlzc_h (v8i16);
14369 v4i32 __builtin_msa_nlzc_w (v4i32);
14370 v2i64 __builtin_msa_nlzc_d (v2i64);
14372 v16u8 __builtin_msa_nor_v (v16u8, v16u8);
14374 v16u8 __builtin_msa_nori_b (v16u8, imm0_255);
14376 v16u8 __builtin_msa_or_v (v16u8, v16u8);
14378 v16u8 __builtin_msa_ori_b (v16u8, imm0_255);
14380 v16i8 __builtin_msa_pckev_b (v16i8, v16i8);
14381 v8i16 __builtin_msa_pckev_h (v8i16, v8i16);
14382 v4i32 __builtin_msa_pckev_w (v4i32, v4i32);
14383 v2i64 __builtin_msa_pckev_d (v2i64, v2i64);
14385 v16i8 __builtin_msa_pckod_b (v16i8, v16i8);
14386 v8i16 __builtin_msa_pckod_h (v8i16, v8i16);
14387 v4i32 __builtin_msa_pckod_w (v4i32, v4i32);
14388 v2i64 __builtin_msa_pckod_d (v2i64, v2i64);
14390 v16i8 __builtin_msa_pcnt_b (v16i8);
14391 v8i16 __builtin_msa_pcnt_h (v8i16);
14392 v4i32 __builtin_msa_pcnt_w (v4i32);
14393 v2i64 __builtin_msa_pcnt_d (v2i64);
14395 v16i8 __builtin_msa_sat_s_b (v16i8, imm0_7);
14396 v8i16 __builtin_msa_sat_s_h (v8i16, imm0_15);
14397 v4i32 __builtin_msa_sat_s_w (v4i32, imm0_31);
14398 v2i64 __builtin_msa_sat_s_d (v2i64, imm0_63);
14400 v16u8 __builtin_msa_sat_u_b (v16u8, imm0_7);
14401 v8u16 __builtin_msa_sat_u_h (v8u16, imm0_15);
14402 v4u32 __builtin_msa_sat_u_w (v4u32, imm0_31);
14403 v2u64 __builtin_msa_sat_u_d (v2u64, imm0_63);
14405 v16i8 __builtin_msa_shf_b (v16i8, imm0_255);
14406 v8i16 __builtin_msa_shf_h (v8i16, imm0_255);
14407 v4i32 __builtin_msa_shf_w (v4i32, imm0_255);
14409 v16i8 __builtin_msa_sld_b (v16i8, v16i8, i32);
14410 v8i16 __builtin_msa_sld_h (v8i16, v8i16, i32);
14411 v4i32 __builtin_msa_sld_w (v4i32, v4i32, i32);
14412 v2i64 __builtin_msa_sld_d (v2i64, v2i64, i32);
14414 v16i8 __builtin_msa_sldi_b (v16i8, v16i8, imm0_15);
14415 v8i16 __builtin_msa_sldi_h (v8i16, v8i16, imm0_7);
14416 v4i32 __builtin_msa_sldi_w (v4i32, v4i32, imm0_3);
14417 v2i64 __builtin_msa_sldi_d (v2i64, v2i64, imm0_1);
14419 v16i8 __builtin_msa_sll_b (v16i8, v16i8);
14420 v8i16 __builtin_msa_sll_h (v8i16, v8i16);
14421 v4i32 __builtin_msa_sll_w (v4i32, v4i32);
14422 v2i64 __builtin_msa_sll_d (v2i64, v2i64);
14424 v16i8 __builtin_msa_slli_b (v16i8, imm0_7);
14425 v8i16 __builtin_msa_slli_h (v8i16, imm0_15);
14426 v4i32 __builtin_msa_slli_w (v4i32, imm0_31);
14427 v2i64 __builtin_msa_slli_d (v2i64, imm0_63);
14429 v16i8 __builtin_msa_splat_b (v16i8, i32);
14430 v8i16 __builtin_msa_splat_h (v8i16, i32);
14431 v4i32 __builtin_msa_splat_w (v4i32, i32);
14432 v2i64 __builtin_msa_splat_d (v2i64, i32);
14434 v16i8 __builtin_msa_splati_b (v16i8, imm0_15);
14435 v8i16 __builtin_msa_splati_h (v8i16, imm0_7);
14436 v4i32 __builtin_msa_splati_w (v4i32, imm0_3);
14437 v2i64 __builtin_msa_splati_d (v2i64, imm0_1);
14439 v16i8 __builtin_msa_sra_b (v16i8, v16i8);
14440 v8i16 __builtin_msa_sra_h (v8i16, v8i16);
14441 v4i32 __builtin_msa_sra_w (v4i32, v4i32);
14442 v2i64 __builtin_msa_sra_d (v2i64, v2i64);
14444 v16i8 __builtin_msa_srai_b (v16i8, imm0_7);
14445 v8i16 __builtin_msa_srai_h (v8i16, imm0_15);
14446 v4i32 __builtin_msa_srai_w (v4i32, imm0_31);
14447 v2i64 __builtin_msa_srai_d (v2i64, imm0_63);
14449 v16i8 __builtin_msa_srar_b (v16i8, v16i8);
14450 v8i16 __builtin_msa_srar_h (v8i16, v8i16);
14451 v4i32 __builtin_msa_srar_w (v4i32, v4i32);
14452 v2i64 __builtin_msa_srar_d (v2i64, v2i64);
14454 v16i8 __builtin_msa_srari_b (v16i8, imm0_7);
14455 v8i16 __builtin_msa_srari_h (v8i16, imm0_15);
14456 v4i32 __builtin_msa_srari_w (v4i32, imm0_31);
14457 v2i64 __builtin_msa_srari_d (v2i64, imm0_63);
14459 v16i8 __builtin_msa_srl_b (v16i8, v16i8);
14460 v8i16 __builtin_msa_srl_h (v8i16, v8i16);
14461 v4i32 __builtin_msa_srl_w (v4i32, v4i32);
14462 v2i64 __builtin_msa_srl_d (v2i64, v2i64);
14464 v16i8 __builtin_msa_srli_b (v16i8, imm0_7);
14465 v8i16 __builtin_msa_srli_h (v8i16, imm0_15);
14466 v4i32 __builtin_msa_srli_w (v4i32, imm0_31);
14467 v2i64 __builtin_msa_srli_d (v2i64, imm0_63);
14469 v16i8 __builtin_msa_srlr_b (v16i8, v16i8);
14470 v8i16 __builtin_msa_srlr_h (v8i16, v8i16);
14471 v4i32 __builtin_msa_srlr_w (v4i32, v4i32);
14472 v2i64 __builtin_msa_srlr_d (v2i64, v2i64);
14474 v16i8 __builtin_msa_srlri_b (v16i8, imm0_7);
14475 v8i16 __builtin_msa_srlri_h (v8i16, imm0_15);
14476 v4i32 __builtin_msa_srlri_w (v4i32, imm0_31);
14477 v2i64 __builtin_msa_srlri_d (v2i64, imm0_63);
14479 void __builtin_msa_st_b (v16i8, void *, imm_n512_511);
14480 void __builtin_msa_st_h (v8i16, void *, imm_n1024_1022);
14481 void __builtin_msa_st_w (v4i32, void *, imm_n2048_2044);
14482 void __builtin_msa_st_d (v2i64, void *, imm_n4096_4088);
14484 v16i8 __builtin_msa_subs_s_b (v16i8, v16i8);
14485 v8i16 __builtin_msa_subs_s_h (v8i16, v8i16);
14486 v4i32 __builtin_msa_subs_s_w (v4i32, v4i32);
14487 v2i64 __builtin_msa_subs_s_d (v2i64, v2i64);
14489 v16u8 __builtin_msa_subs_u_b (v16u8, v16u8);
14490 v8u16 __builtin_msa_subs_u_h (v8u16, v8u16);
14491 v4u32 __builtin_msa_subs_u_w (v4u32, v4u32);
14492 v2u64 __builtin_msa_subs_u_d (v2u64, v2u64);
14494 v16u8 __builtin_msa_subsus_u_b (v16u8, v16i8);
14495 v8u16 __builtin_msa_subsus_u_h (v8u16, v8i16);
14496 v4u32 __builtin_msa_subsus_u_w (v4u32, v4i32);
14497 v2u64 __builtin_msa_subsus_u_d (v2u64, v2i64);
14499 v16i8 __builtin_msa_subsuu_s_b (v16u8, v16u8);
14500 v8i16 __builtin_msa_subsuu_s_h (v8u16, v8u16);
14501 v4i32 __builtin_msa_subsuu_s_w (v4u32, v4u32);
14502 v2i64 __builtin_msa_subsuu_s_d (v2u64, v2u64);
14504 v16i8 __builtin_msa_subv_b (v16i8, v16i8);
14505 v8i16 __builtin_msa_subv_h (v8i16, v8i16);
14506 v4i32 __builtin_msa_subv_w (v4i32, v4i32);
14507 v2i64 __builtin_msa_subv_d (v2i64, v2i64);
14509 v16i8 __builtin_msa_subvi_b (v16i8, imm0_31);
14510 v8i16 __builtin_msa_subvi_h (v8i16, imm0_31);
14511 v4i32 __builtin_msa_subvi_w (v4i32, imm0_31);
14512 v2i64 __builtin_msa_subvi_d (v2i64, imm0_31);
14514 v16i8 __builtin_msa_vshf_b (v16i8, v16i8, v16i8);
14515 v8i16 __builtin_msa_vshf_h (v8i16, v8i16, v8i16);
14516 v4i32 __builtin_msa_vshf_w (v4i32, v4i32, v4i32);
14517 v2i64 __builtin_msa_vshf_d (v2i64, v2i64, v2i64);
14519 v16u8 __builtin_msa_xor_v (v16u8, v16u8);
14521 v16u8 __builtin_msa_xori_b (v16u8, imm0_255);
14524 @node Other MIPS Built-in Functions
14525 @subsection Other MIPS Built-in Functions
14527 GCC provides other MIPS-specific built-in functions:
14530 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
14531 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
14532 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
14533 when this function is available.
14535 @item unsigned int __builtin_mips_get_fcsr (void)
14536 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
14537 Get and set the contents of the floating-point control and status register
14538 (FPU control register 31). These functions are only available in hard-float
14539 code but can be called in both MIPS16 and non-MIPS16 contexts.
14541 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
14542 register except the condition codes, which GCC assumes are preserved.
14545 @node MSP430 Built-in Functions
14546 @subsection MSP430 Built-in Functions
14548 GCC provides a couple of special builtin functions to aid in the
14549 writing of interrupt handlers in C.
14552 @item __bic_SR_register_on_exit (int @var{mask})
14553 This clears the indicated bits in the saved copy of the status register
14554 currently residing on the stack. This only works inside interrupt
14555 handlers and the changes to the status register will only take affect
14556 once the handler returns.
14558 @item __bis_SR_register_on_exit (int @var{mask})
14559 This sets the indicated bits in the saved copy of the status register
14560 currently residing on the stack. This only works inside interrupt
14561 handlers and the changes to the status register will only take affect
14562 once the handler returns.
14564 @item __delay_cycles (long long @var{cycles})
14565 This inserts an instruction sequence that takes exactly @var{cycles}
14566 cycles (between 0 and about 17E9) to complete. The inserted sequence
14567 may use jumps, loops, or no-ops, and does not interfere with any other
14568 instructions. Note that @var{cycles} must be a compile-time constant
14569 integer - that is, you must pass a number, not a variable that may be
14570 optimized to a constant later. The number of cycles delayed by this
14574 @node NDS32 Built-in Functions
14575 @subsection NDS32 Built-in Functions
14577 These built-in functions are available for the NDS32 target:
14579 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
14580 Insert an ISYNC instruction into the instruction stream where
14581 @var{addr} is an instruction address for serialization.
14584 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
14585 Insert an ISB instruction into the instruction stream.
14588 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
14589 Return the content of a system register which is mapped by @var{sr}.
14592 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
14593 Return the content of a user space register which is mapped by @var{usr}.
14596 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
14597 Move the @var{value} to a system register which is mapped by @var{sr}.
14600 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
14601 Move the @var{value} to a user space register which is mapped by @var{usr}.
14604 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
14605 Enable global interrupt.
14608 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
14609 Disable global interrupt.
14612 @node picoChip Built-in Functions
14613 @subsection picoChip Built-in Functions
14615 GCC provides an interface to selected machine instructions from the
14616 picoChip instruction set.
14619 @item int __builtin_sbc (int @var{value})
14620 Sign bit count. Return the number of consecutive bits in @var{value}
14621 that have the same value as the sign bit. The result is the number of
14622 leading sign bits minus one, giving the number of redundant sign bits in
14625 @item int __builtin_byteswap (int @var{value})
14626 Byte swap. Return the result of swapping the upper and lower bytes of
14629 @item int __builtin_brev (int @var{value})
14630 Bit reversal. Return the result of reversing the bits in
14631 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
14634 @item int __builtin_adds (int @var{x}, int @var{y})
14635 Saturating addition. Return the result of adding @var{x} and @var{y},
14636 storing the value 32767 if the result overflows.
14638 @item int __builtin_subs (int @var{x}, int @var{y})
14639 Saturating subtraction. Return the result of subtracting @var{y} from
14640 @var{x}, storing the value @minus{}32768 if the result overflows.
14642 @item void __builtin_halt (void)
14643 Halt. The processor stops execution. This built-in is useful for
14644 implementing assertions.
14648 @node PowerPC Built-in Functions
14649 @subsection PowerPC Built-in Functions
14651 The following built-in functions are always available and can be used to
14652 check the PowerPC target platform type:
14654 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
14655 This function is a @code{nop} on the PowerPC platform and is included solely
14656 to maintain API compatibility with the x86 builtins.
14659 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
14660 This function returns a value of @code{1} if the run-time CPU is of type
14661 @var{cpuname} and returns @code{0} otherwise. The following CPU names can be
14666 IBM POWER9 Server CPU.
14668 IBM POWER8 Server CPU.
14670 IBM POWER7 Server CPU.
14672 IBM POWER6 Server CPU (RAW mode).
14674 IBM POWER6 Server CPU (Architected mode).
14676 IBM POWER5+ Server CPU.
14678 IBM POWER5 Server CPU.
14680 IBM 970 Server CPU (ie, Apple G5).
14682 IBM POWER4 Server CPU.
14684 IBM A2 64-bit Embedded CPU
14686 IBM PowerPC 476FP 32-bit Embedded CPU.
14688 IBM PowerPC 464 32-bit Embedded CPU.
14690 PowerPC 440 32-bit Embedded CPU.
14692 PowerPC 405 32-bit Embedded CPU.
14694 IBM PowerPC Cell Broadband Engine Architecture CPU.
14697 Here is an example:
14699 if (__builtin_cpu_is ("power8"))
14701 do_power8 (); // POWER8 specific implementation.
14705 do_generic (); // Generic implementation.
14710 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
14711 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
14712 feature @var{feature} and returns @code{0} otherwise. The following features can be
14717 4xx CPU has a Multiply Accumulator.
14719 CPU has a SIMD/Vector Unit.
14721 CPU supports ISA 2.05 (eg, POWER6)
14723 CPU supports ISA 2.06 (eg, POWER7)
14725 CPU supports ISA 2.07 (eg, POWER8)
14727 CPU supports ISA 3.0 (eg, POWER9)
14729 CPU supports the set of compatible performance monitoring events.
14731 CPU supports the Embedded ISA category.
14733 CPU has a CELL broadband engine.
14735 CPU has a decimal floating point unit.
14737 CPU supports the data stream control register.
14739 CPU supports event base branching.
14741 CPU has a SPE double precision floating point unit.
14743 CPU has a SPE single precision floating point unit.
14745 CPU has a floating point unit.
14747 CPU has hardware transaction memory instructions.
14749 Kernel aborts hardware transactions when a syscall is made.
14751 CPU supports icache snooping capabilities.
14753 CPU supports 128-bit IEEE binary floating point instructions.
14755 CPU supports the integer select instruction.
14757 CPU has a memory management unit.
14759 CPU does not have a timebase (eg, 601 and 403gx).
14761 CPU supports the PA Semi 6T CORE ISA.
14763 CPU supports ISA 2.00 (eg, POWER4)
14765 CPU supports ISA 2.02 (eg, POWER5)
14767 CPU supports ISA 2.03 (eg, POWER5+)
14769 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
14771 CPU supports 32-bit mode execution.
14773 CPU supports the old POWER ISA (eg, 601)
14775 CPU supports 64-bit mode execution.
14777 CPU supports a little-endian mode that uses address swizzling.
14779 CPU support simultaneous multi-threading.
14781 CPU has a signal processing extension unit.
14783 CPU supports the target address register.
14785 CPU supports true little-endian mode.
14787 CPU has unified I/D cache.
14789 CPU supports the vector cryptography instructions.
14791 CPU supports the vector-scalar extension.
14794 Here is an example:
14796 if (__builtin_cpu_supports ("fpu"))
14798 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
14802 dst = __fadd (src1, src2); // Software FP addition function.
14807 These built-in functions are available for the PowerPC family of
14810 float __builtin_recipdivf (float, float);
14811 float __builtin_rsqrtf (float);
14812 double __builtin_recipdiv (double, double);
14813 double __builtin_rsqrt (double);
14814 uint64_t __builtin_ppc_get_timebase ();
14815 unsigned long __builtin_ppc_mftb ();
14816 double __builtin_unpack_longdouble (long double, int);
14817 long double __builtin_pack_longdouble (double, double);
14820 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
14821 @code{__builtin_rsqrtf} functions generate multiple instructions to
14822 implement the reciprocal sqrt functionality using reciprocal sqrt
14823 estimate instructions.
14825 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
14826 functions generate multiple instructions to implement division using
14827 the reciprocal estimate instructions.
14829 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
14830 functions generate instructions to read the Time Base Register. The
14831 @code{__builtin_ppc_get_timebase} function may generate multiple
14832 instructions and always returns the 64 bits of the Time Base Register.
14833 The @code{__builtin_ppc_mftb} function always generates one instruction and
14834 returns the Time Base Register value as an unsigned long, throwing away
14835 the most significant word on 32-bit environments.
14837 Additional built-in functions are available for the 64-bit PowerPC
14838 family of processors, for efficient use of 128-bit floating point
14839 (@code{__float128}) values.
14841 The following floating-point built-in functions are available with
14842 @code{-mfloat128} and Altivec support. All of them implement the
14843 function that is part of the name.
14846 __float128 __builtin_fabsq (__float128)
14847 __float128 __builtin_copysignq (__float128, __float128)
14850 The following built-in functions are available with @code{-mfloat128}
14851 and Altivec support.
14854 @item __float128 __builtin_infq (void)
14855 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
14856 @findex __builtin_infq
14858 @item __float128 __builtin_huge_valq (void)
14859 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
14860 @findex __builtin_huge_valq
14862 @item __float128 __builtin_nanq (void)
14863 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
14864 @findex __builtin_nanq
14866 @item __float128 __builtin_nansq (void)
14867 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
14868 @findex __builtin_nansq
14871 The following built-in functions are available for the PowerPC family
14872 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
14873 or @option{-mpopcntd}):
14875 long __builtin_bpermd (long, long);
14876 int __builtin_divwe (int, int);
14877 int __builtin_divweo (int, int);
14878 unsigned int __builtin_divweu (unsigned int, unsigned int);
14879 unsigned int __builtin_divweuo (unsigned int, unsigned int);
14880 long __builtin_divde (long, long);
14881 long __builtin_divdeo (long, long);
14882 unsigned long __builtin_divdeu (unsigned long, unsigned long);
14883 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
14884 unsigned int cdtbcd (unsigned int);
14885 unsigned int cbcdtd (unsigned int);
14886 unsigned int addg6s (unsigned int, unsigned int);
14889 The @code{__builtin_divde}, @code{__builtin_divdeo},
14890 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
14891 64-bit environment support ISA 2.06 or later.
14893 The following built-in functions are available for the PowerPC family
14894 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
14896 long long __builtin_darn (void);
14897 long long __builtin_darn_raw (void);
14898 int __builtin_darn_32 (void);
14900 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal64 value);
14901 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal128 value);
14902 int __builtin_dfp_dtstsfi_lt_dd (unsigned int comparison, _Decimal64 value);
14903 int __builtin_dfp_dtstsfi_lt_td (unsigned int comparison, _Decimal128 value);
14905 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal64 value);
14906 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal128 value);
14907 int __builtin_dfp_dtstsfi_gt_dd (unsigned int comparison, _Decimal64 value);
14908 int __builtin_dfp_dtstsfi_gt_td (unsigned int comparison, _Decimal128 value);
14910 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal64 value);
14911 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal128 value);
14912 int __builtin_dfp_dtstsfi_eq_dd (unsigned int comparison, _Decimal64 value);
14913 int __builtin_dfp_dtstsfi_eq_td (unsigned int comparison, _Decimal128 value);
14915 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal64 value);
14916 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal128 value);
14917 int __builtin_dfp_dtstsfi_ov_dd (unsigned int comparison, _Decimal64 value);
14918 int __builtin_dfp_dtstsfi_ov_td (unsigned int comparison, _Decimal128 value);
14920 unsigned int scalar_extract_exp (double source);
14921 unsigned long long int scalar_extract_sig (double source);
14924 scalar_insert_exp (unsigned long long int significand, unsigned long long int exponent);
14926 int scalar_cmp_exp_gt (double arg1, double arg2);
14927 int scalar_cmp_exp_lt (double arg1, double arg2);
14928 int scalar_cmp_exp_eq (double arg1, double arg2);
14929 int scalar_cmp_exp_unordered (double arg1, double arg2);
14931 int scalar_test_data_class (float source, unsigned int condition);
14932 int scalar_test_data_class (double source, unsigned int condition);
14934 int scalar_test_neg (float source);
14935 int scalar_test_neg (double source);
14938 The @code{__builtin_darn} and @code{__builtin_darn_raw}
14939 functions require a
14940 64-bit environment supporting ISA 3.0 or later.
14941 The @code{__builtin_darn} function provides a 64-bit conditioned
14942 random number. The @code{__builtin_darn_raw} function provides a
14943 64-bit raw random number. The @code{__builtin_darn_32} function
14944 provides a 32-bit random number.
14946 The @code{scalar_extract_sig} and @code{scalar_insert_exp}
14947 functions require a 64-bit environment supporting ISA 3.0 or later.
14948 The @code{scalar_extract_exp} and @code{vec_extract_sig} built-in
14949 functions return the significand and exponent respectively of their
14950 @code{source} arguments. The
14951 @code{scalar_insert_exp} built-in function returns a double-precision
14952 floating point value that is constructed by assembling the values of its
14953 @code{significand} and @code{exponent} arguments. The sign of the
14954 result is copied from the most significant bit of the
14955 @code{significand} argument. The significand and exponent components
14956 of the result are composed of the least significant 11 bits of the
14957 @code{significand} argument and the least significant 52 bits of the
14958 @code{exponent} argument.
14960 The @code{scalar_cmp_exp_gt}, @code{scalar_cmp_exp_lt},
14961 @code{scalar_cmp_exp_eq}, and @code{scalar_cmp_exp_unordered} built-in
14962 functions return a non-zero value if @code{arg1} is greater than, less
14963 than, equal to, or not comparable to @code{arg2} respectively. The
14964 arguments are not comparable if one or the other equals NaN (not a
14967 The @code{scalar_test_data_class} built-in functions return a non-zero
14968 value if any of the condition tests enabled by the value of the
14969 @code{condition} variable are true. The
14970 @code{condition} argument must be an unsigned integer with value not
14972 @code{condition} argument is encoded as a bitmask with each bit
14973 enabling the testing of a different condition, as characterized by the
14977 0x20 Test for +Infinity
14978 0x10 Test for -Infinity
14979 0x08 Test for +Zero
14980 0x04 Test for -Zero
14981 0x02 Test for +Denormal
14982 0x01 Test for -Denormal
14985 If all of the enabled test conditions are false, the return value is 0.
14987 The @code{scalar_test_neg} built-in functions return a non-zero value
14988 if their @code{source} argument holds a negative value.
14990 The @code{__builtin_dfp_dtstsfi_lt} function returns a non-zero value
14991 if and only if the number of signficant digits of its @code{value} argument
14992 is less than its @code{comparison} argument. The
14993 @code{__builtin_dfp_dtstsfi_lt_dd} and
14994 @code{__builtin_dfp_dtstsfi_lt_td} functions behave similarly, but
14995 require that the type of the @code{value} argument be
14996 @code{__Decimal64} and @code{__Decimal128} respectively.
14998 The @code{__builtin_dfp_dtstsfi_gt} function returns a non-zero value
14999 if and only if the number of signficant digits of its @code{value} argument
15000 is greater than its @code{comparison} argument. The
15001 @code{__builtin_dfp_dtstsfi_gt_dd} and
15002 @code{__builtin_dfp_dtstsfi_gt_td} functions behave similarly, but
15003 require that the type of the @code{value} argument be
15004 @code{__Decimal64} and @code{__Decimal128} respectively.
15006 The @code{__builtin_dfp_dtstsfi_eq} function returns a non-zero value
15007 if and only if the number of signficant digits of its @code{value} argument
15008 equals its @code{comparison} argument. The
15009 @code{__builtin_dfp_dtstsfi_eq_dd} and
15010 @code{__builtin_dfp_dtstsfi_eq_td} functions behave similarly, but
15011 require that the type of the @code{value} argument be
15012 @code{__Decimal64} and @code{__Decimal128} respectively.
15014 The @code{__builtin_dfp_dtstsfi_ov} function returns a non-zero value
15015 if and only if its @code{value} argument has an undefined number of
15016 significant digits, such as when @code{value} is an encoding of @code{NaN}.
15017 The @code{__builtin_dfp_dtstsfi_ov_dd} and
15018 @code{__builtin_dfp_dtstsfi_ov_td} functions behave similarly, but
15019 require that the type of the @code{value} argument be
15020 @code{__Decimal64} and @code{__Decimal128} respectively.
15022 The following built-in functions are available for the PowerPC family
15023 of processors when hardware decimal floating point
15024 (@option{-mhard-dfp}) is available:
15026 _Decimal64 __builtin_dxex (_Decimal64);
15027 _Decimal128 __builtin_dxexq (_Decimal128);
15028 _Decimal64 __builtin_ddedpd (int, _Decimal64);
15029 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
15030 _Decimal64 __builtin_denbcd (int, _Decimal64);
15031 _Decimal128 __builtin_denbcdq (int, _Decimal128);
15032 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
15033 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
15034 _Decimal64 __builtin_dscli (_Decimal64, int);
15035 _Decimal128 __builtin_dscliq (_Decimal128, int);
15036 _Decimal64 __builtin_dscri (_Decimal64, int);
15037 _Decimal128 __builtin_dscriq (_Decimal128, int);
15038 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
15039 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
15042 The following built-in functions are available for the PowerPC family
15043 of processors when the Vector Scalar (vsx) instruction set is
15046 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
15047 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
15048 unsigned long long);
15051 @node PowerPC AltiVec/VSX Built-in Functions
15052 @subsection PowerPC AltiVec Built-in Functions
15054 GCC provides an interface for the PowerPC family of processors to access
15055 the AltiVec operations described in Motorola's AltiVec Programming
15056 Interface Manual. The interface is made available by including
15057 @code{<altivec.h>} and using @option{-maltivec} and
15058 @option{-mabi=altivec}. The interface supports the following vector
15062 vector unsigned char
15066 vector unsigned short
15067 vector signed short
15071 vector unsigned int
15077 If @option{-mvsx} is used the following additional vector types are
15081 vector unsigned long
15086 The long types are only implemented for 64-bit code generation, and
15087 the long type is only used in the floating point/integer conversion
15090 GCC's implementation of the high-level language interface available from
15091 C and C++ code differs from Motorola's documentation in several ways.
15096 A vector constant is a list of constant expressions within curly braces.
15099 A vector initializer requires no cast if the vector constant is of the
15100 same type as the variable it is initializing.
15103 If @code{signed} or @code{unsigned} is omitted, the signedness of the
15104 vector type is the default signedness of the base type. The default
15105 varies depending on the operating system, so a portable program should
15106 always specify the signedness.
15109 Compiling with @option{-maltivec} adds keywords @code{__vector},
15110 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
15111 @code{bool}. When compiling ISO C, the context-sensitive substitution
15112 of the keywords @code{vector}, @code{pixel} and @code{bool} is
15113 disabled. To use them, you must include @code{<altivec.h>} instead.
15116 GCC allows using a @code{typedef} name as the type specifier for a
15120 For C, overloaded functions are implemented with macros so the following
15124 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
15128 Since @code{vec_add} is a macro, the vector constant in the example
15129 is treated as four separate arguments. Wrap the entire argument in
15130 parentheses for this to work.
15133 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
15134 Internally, GCC uses built-in functions to achieve the functionality in
15135 the aforementioned header file, but they are not supported and are
15136 subject to change without notice.
15138 The following interfaces are supported for the generic and specific
15139 AltiVec operations and the AltiVec predicates. In cases where there
15140 is a direct mapping between generic and specific operations, only the
15141 generic names are shown here, although the specific operations can also
15144 Arguments that are documented as @code{const int} require literal
15145 integral values within the range required for that operation.
15148 vector signed char vec_abs (vector signed char);
15149 vector signed short vec_abs (vector signed short);
15150 vector signed int vec_abs (vector signed int);
15151 vector float vec_abs (vector float);
15153 vector signed char vec_abss (vector signed char);
15154 vector signed short vec_abss (vector signed short);
15155 vector signed int vec_abss (vector signed int);
15157 vector signed char vec_add (vector bool char, vector signed char);
15158 vector signed char vec_add (vector signed char, vector bool char);
15159 vector signed char vec_add (vector signed char, vector signed char);
15160 vector unsigned char vec_add (vector bool char, vector unsigned char);
15161 vector unsigned char vec_add (vector unsigned char, vector bool char);
15162 vector unsigned char vec_add (vector unsigned char,
15163 vector unsigned char);
15164 vector signed short vec_add (vector bool short, vector signed short);
15165 vector signed short vec_add (vector signed short, vector bool short);
15166 vector signed short vec_add (vector signed short, vector signed short);
15167 vector unsigned short vec_add (vector bool short,
15168 vector unsigned short);
15169 vector unsigned short vec_add (vector unsigned short,
15170 vector bool short);
15171 vector unsigned short vec_add (vector unsigned short,
15172 vector unsigned short);
15173 vector signed int vec_add (vector bool int, vector signed int);
15174 vector signed int vec_add (vector signed int, vector bool int);
15175 vector signed int vec_add (vector signed int, vector signed int);
15176 vector unsigned int vec_add (vector bool int, vector unsigned int);
15177 vector unsigned int vec_add (vector unsigned int, vector bool int);
15178 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
15179 vector float vec_add (vector float, vector float);
15181 vector float vec_vaddfp (vector float, vector float);
15183 vector signed int vec_vadduwm (vector bool int, vector signed int);
15184 vector signed int vec_vadduwm (vector signed int, vector bool int);
15185 vector signed int vec_vadduwm (vector signed int, vector signed int);
15186 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
15187 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
15188 vector unsigned int vec_vadduwm (vector unsigned int,
15189 vector unsigned int);
15191 vector signed short vec_vadduhm (vector bool short,
15192 vector signed short);
15193 vector signed short vec_vadduhm (vector signed short,
15194 vector bool short);
15195 vector signed short vec_vadduhm (vector signed short,
15196 vector signed short);
15197 vector unsigned short vec_vadduhm (vector bool short,
15198 vector unsigned short);
15199 vector unsigned short vec_vadduhm (vector unsigned short,
15200 vector bool short);
15201 vector unsigned short vec_vadduhm (vector unsigned short,
15202 vector unsigned short);
15204 vector signed char vec_vaddubm (vector bool char, vector signed char);
15205 vector signed char vec_vaddubm (vector signed char, vector bool char);
15206 vector signed char vec_vaddubm (vector signed char, vector signed char);
15207 vector unsigned char vec_vaddubm (vector bool char,
15208 vector unsigned char);
15209 vector unsigned char vec_vaddubm (vector unsigned char,
15211 vector unsigned char vec_vaddubm (vector unsigned char,
15212 vector unsigned char);
15214 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
15216 vector unsigned char vec_adds (vector bool char, vector unsigned char);
15217 vector unsigned char vec_adds (vector unsigned char, vector bool char);
15218 vector unsigned char vec_adds (vector unsigned char,
15219 vector unsigned char);
15220 vector signed char vec_adds (vector bool char, vector signed char);
15221 vector signed char vec_adds (vector signed char, vector bool char);
15222 vector signed char vec_adds (vector signed char, vector signed char);
15223 vector unsigned short vec_adds (vector bool short,
15224 vector unsigned short);
15225 vector unsigned short vec_adds (vector unsigned short,
15226 vector bool short);
15227 vector unsigned short vec_adds (vector unsigned short,
15228 vector unsigned short);
15229 vector signed short vec_adds (vector bool short, vector signed short);
15230 vector signed short vec_adds (vector signed short, vector bool short);
15231 vector signed short vec_adds (vector signed short, vector signed short);
15232 vector unsigned int vec_adds (vector bool int, vector unsigned int);
15233 vector unsigned int vec_adds (vector unsigned int, vector bool int);
15234 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
15235 vector signed int vec_adds (vector bool int, vector signed int);
15236 vector signed int vec_adds (vector signed int, vector bool int);
15237 vector signed int vec_adds (vector signed int, vector signed int);
15239 vector signed int vec_vaddsws (vector bool int, vector signed int);
15240 vector signed int vec_vaddsws (vector signed int, vector bool int);
15241 vector signed int vec_vaddsws (vector signed int, vector signed int);
15243 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
15244 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
15245 vector unsigned int vec_vadduws (vector unsigned int,
15246 vector unsigned int);
15248 vector signed short vec_vaddshs (vector bool short,
15249 vector signed short);
15250 vector signed short vec_vaddshs (vector signed short,
15251 vector bool short);
15252 vector signed short vec_vaddshs (vector signed short,
15253 vector signed short);
15255 vector unsigned short vec_vadduhs (vector bool short,
15256 vector unsigned short);
15257 vector unsigned short vec_vadduhs (vector unsigned short,
15258 vector bool short);
15259 vector unsigned short vec_vadduhs (vector unsigned short,
15260 vector unsigned short);
15262 vector signed char vec_vaddsbs (vector bool char, vector signed char);
15263 vector signed char vec_vaddsbs (vector signed char, vector bool char);
15264 vector signed char vec_vaddsbs (vector signed char, vector signed char);
15266 vector unsigned char vec_vaddubs (vector bool char,
15267 vector unsigned char);
15268 vector unsigned char vec_vaddubs (vector unsigned char,
15270 vector unsigned char vec_vaddubs (vector unsigned char,
15271 vector unsigned char);
15273 vector float vec_and (vector float, vector float);
15274 vector float vec_and (vector float, vector bool int);
15275 vector float vec_and (vector bool int, vector float);
15276 vector bool int vec_and (vector bool int, vector bool int);
15277 vector signed int vec_and (vector bool int, vector signed int);
15278 vector signed int vec_and (vector signed int, vector bool int);
15279 vector signed int vec_and (vector signed int, vector signed int);
15280 vector unsigned int vec_and (vector bool int, vector unsigned int);
15281 vector unsigned int vec_and (vector unsigned int, vector bool int);
15282 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
15283 vector bool short vec_and (vector bool short, vector bool short);
15284 vector signed short vec_and (vector bool short, vector signed short);
15285 vector signed short vec_and (vector signed short, vector bool short);
15286 vector signed short vec_and (vector signed short, vector signed short);
15287 vector unsigned short vec_and (vector bool short,
15288 vector unsigned short);
15289 vector unsigned short vec_and (vector unsigned short,
15290 vector bool short);
15291 vector unsigned short vec_and (vector unsigned short,
15292 vector unsigned short);
15293 vector signed char vec_and (vector bool char, vector signed char);
15294 vector bool char vec_and (vector bool char, vector bool char);
15295 vector signed char vec_and (vector signed char, vector bool char);
15296 vector signed char vec_and (vector signed char, vector signed char);
15297 vector unsigned char vec_and (vector bool char, vector unsigned char);
15298 vector unsigned char vec_and (vector unsigned char, vector bool char);
15299 vector unsigned char vec_and (vector unsigned char,
15300 vector unsigned char);
15302 vector float vec_andc (vector float, vector float);
15303 vector float vec_andc (vector float, vector bool int);
15304 vector float vec_andc (vector bool int, vector float);
15305 vector bool int vec_andc (vector bool int, vector bool int);
15306 vector signed int vec_andc (vector bool int, vector signed int);
15307 vector signed int vec_andc (vector signed int, vector bool int);
15308 vector signed int vec_andc (vector signed int, vector signed int);
15309 vector unsigned int vec_andc (vector bool int, vector unsigned int);
15310 vector unsigned int vec_andc (vector unsigned int, vector bool int);
15311 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
15312 vector bool short vec_andc (vector bool short, vector bool short);
15313 vector signed short vec_andc (vector bool short, vector signed short);
15314 vector signed short vec_andc (vector signed short, vector bool short);
15315 vector signed short vec_andc (vector signed short, vector signed short);
15316 vector unsigned short vec_andc (vector bool short,
15317 vector unsigned short);
15318 vector unsigned short vec_andc (vector unsigned short,
15319 vector bool short);
15320 vector unsigned short vec_andc (vector unsigned short,
15321 vector unsigned short);
15322 vector signed char vec_andc (vector bool char, vector signed char);
15323 vector bool char vec_andc (vector bool char, vector bool char);
15324 vector signed char vec_andc (vector signed char, vector bool char);
15325 vector signed char vec_andc (vector signed char, vector signed char);
15326 vector unsigned char vec_andc (vector bool char, vector unsigned char);
15327 vector unsigned char vec_andc (vector unsigned char, vector bool char);
15328 vector unsigned char vec_andc (vector unsigned char,
15329 vector unsigned char);
15331 vector unsigned char vec_avg (vector unsigned char,
15332 vector unsigned char);
15333 vector signed char vec_avg (vector signed char, vector signed char);
15334 vector unsigned short vec_avg (vector unsigned short,
15335 vector unsigned short);
15336 vector signed short vec_avg (vector signed short, vector signed short);
15337 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
15338 vector signed int vec_avg (vector signed int, vector signed int);
15340 vector signed int vec_vavgsw (vector signed int, vector signed int);
15342 vector unsigned int vec_vavguw (vector unsigned int,
15343 vector unsigned int);
15345 vector signed short vec_vavgsh (vector signed short,
15346 vector signed short);
15348 vector unsigned short vec_vavguh (vector unsigned short,
15349 vector unsigned short);
15351 vector signed char vec_vavgsb (vector signed char, vector signed char);
15353 vector unsigned char vec_vavgub (vector unsigned char,
15354 vector unsigned char);
15356 vector float vec_copysign (vector float);
15358 vector float vec_ceil (vector float);
15360 vector signed int vec_cmpb (vector float, vector float);
15362 vector bool char vec_cmpeq (vector signed char, vector signed char);
15363 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
15364 vector bool short vec_cmpeq (vector signed short, vector signed short);
15365 vector bool short vec_cmpeq (vector unsigned short,
15366 vector unsigned short);
15367 vector bool int vec_cmpeq (vector signed int, vector signed int);
15368 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
15369 vector bool int vec_cmpeq (vector float, vector float);
15371 vector bool int vec_vcmpeqfp (vector float, vector float);
15373 vector bool int vec_vcmpequw (vector signed int, vector signed int);
15374 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
15376 vector bool short vec_vcmpequh (vector signed short,
15377 vector signed short);
15378 vector bool short vec_vcmpequh (vector unsigned short,
15379 vector unsigned short);
15381 vector bool char vec_vcmpequb (vector signed char, vector signed char);
15382 vector bool char vec_vcmpequb (vector unsigned char,
15383 vector unsigned char);
15385 vector bool int vec_cmpge (vector float, vector float);
15387 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
15388 vector bool char vec_cmpgt (vector signed char, vector signed char);
15389 vector bool short vec_cmpgt (vector unsigned short,
15390 vector unsigned short);
15391 vector bool short vec_cmpgt (vector signed short, vector signed short);
15392 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
15393 vector bool int vec_cmpgt (vector signed int, vector signed int);
15394 vector bool int vec_cmpgt (vector float, vector float);
15396 vector bool int vec_vcmpgtfp (vector float, vector float);
15398 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
15400 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
15402 vector bool short vec_vcmpgtsh (vector signed short,
15403 vector signed short);
15405 vector bool short vec_vcmpgtuh (vector unsigned short,
15406 vector unsigned short);
15408 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
15410 vector bool char vec_vcmpgtub (vector unsigned char,
15411 vector unsigned char);
15413 vector bool int vec_cmple (vector float, vector float);
15415 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
15416 vector bool char vec_cmplt (vector signed char, vector signed char);
15417 vector bool short vec_cmplt (vector unsigned short,
15418 vector unsigned short);
15419 vector bool short vec_cmplt (vector signed short, vector signed short);
15420 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
15421 vector bool int vec_cmplt (vector signed int, vector signed int);
15422 vector bool int vec_cmplt (vector float, vector float);
15424 vector float vec_cpsgn (vector float, vector float);
15426 vector float vec_ctf (vector unsigned int, const int);
15427 vector float vec_ctf (vector signed int, const int);
15428 vector double vec_ctf (vector unsigned long, const int);
15429 vector double vec_ctf (vector signed long, const int);
15431 vector float vec_vcfsx (vector signed int, const int);
15433 vector float vec_vcfux (vector unsigned int, const int);
15435 vector signed int vec_cts (vector float, const int);
15436 vector signed long vec_cts (vector double, const int);
15438 vector unsigned int vec_ctu (vector float, const int);
15439 vector unsigned long vec_ctu (vector double, const int);
15441 void vec_dss (const int);
15443 void vec_dssall (void);
15445 void vec_dst (const vector unsigned char *, int, const int);
15446 void vec_dst (const vector signed char *, int, const int);
15447 void vec_dst (const vector bool char *, int, const int);
15448 void vec_dst (const vector unsigned short *, int, const int);
15449 void vec_dst (const vector signed short *, int, const int);
15450 void vec_dst (const vector bool short *, int, const int);
15451 void vec_dst (const vector pixel *, int, const int);
15452 void vec_dst (const vector unsigned int *, int, const int);
15453 void vec_dst (const vector signed int *, int, const int);
15454 void vec_dst (const vector bool int *, int, const int);
15455 void vec_dst (const vector float *, int, const int);
15456 void vec_dst (const unsigned char *, int, const int);
15457 void vec_dst (const signed char *, int, const int);
15458 void vec_dst (const unsigned short *, int, const int);
15459 void vec_dst (const short *, int, const int);
15460 void vec_dst (const unsigned int *, int, const int);
15461 void vec_dst (const int *, int, const int);
15462 void vec_dst (const unsigned long *, int, const int);
15463 void vec_dst (const long *, int, const int);
15464 void vec_dst (const float *, int, const int);
15466 void vec_dstst (const vector unsigned char *, int, const int);
15467 void vec_dstst (const vector signed char *, int, const int);
15468 void vec_dstst (const vector bool char *, int, const int);
15469 void vec_dstst (const vector unsigned short *, int, const int);
15470 void vec_dstst (const vector signed short *, int, const int);
15471 void vec_dstst (const vector bool short *, int, const int);
15472 void vec_dstst (const vector pixel *, int, const int);
15473 void vec_dstst (const vector unsigned int *, int, const int);
15474 void vec_dstst (const vector signed int *, int, const int);
15475 void vec_dstst (const vector bool int *, int, const int);
15476 void vec_dstst (const vector float *, int, const int);
15477 void vec_dstst (const unsigned char *, int, const int);
15478 void vec_dstst (const signed char *, int, const int);
15479 void vec_dstst (const unsigned short *, int, const int);
15480 void vec_dstst (const short *, int, const int);
15481 void vec_dstst (const unsigned int *, int, const int);
15482 void vec_dstst (const int *, int, const int);
15483 void vec_dstst (const unsigned long *, int, const int);
15484 void vec_dstst (const long *, int, const int);
15485 void vec_dstst (const float *, int, const int);
15487 void vec_dststt (const vector unsigned char *, int, const int);
15488 void vec_dststt (const vector signed char *, int, const int);
15489 void vec_dststt (const vector bool char *, int, const int);
15490 void vec_dststt (const vector unsigned short *, int, const int);
15491 void vec_dststt (const vector signed short *, int, const int);
15492 void vec_dststt (const vector bool short *, int, const int);
15493 void vec_dststt (const vector pixel *, int, const int);
15494 void vec_dststt (const vector unsigned int *, int, const int);
15495 void vec_dststt (const vector signed int *, int, const int);
15496 void vec_dststt (const vector bool int *, int, const int);
15497 void vec_dststt (const vector float *, int, const int);
15498 void vec_dststt (const unsigned char *, int, const int);
15499 void vec_dststt (const signed char *, int, const int);
15500 void vec_dststt (const unsigned short *, int, const int);
15501 void vec_dststt (const short *, int, const int);
15502 void vec_dststt (const unsigned int *, int, const int);
15503 void vec_dststt (const int *, int, const int);
15504 void vec_dststt (const unsigned long *, int, const int);
15505 void vec_dststt (const long *, int, const int);
15506 void vec_dststt (const float *, int, const int);
15508 void vec_dstt (const vector unsigned char *, int, const int);
15509 void vec_dstt (const vector signed char *, int, const int);
15510 void vec_dstt (const vector bool char *, int, const int);
15511 void vec_dstt (const vector unsigned short *, int, const int);
15512 void vec_dstt (const vector signed short *, int, const int);
15513 void vec_dstt (const vector bool short *, int, const int);
15514 void vec_dstt (const vector pixel *, int, const int);
15515 void vec_dstt (const vector unsigned int *, int, const int);
15516 void vec_dstt (const vector signed int *, int, const int);
15517 void vec_dstt (const vector bool int *, int, const int);
15518 void vec_dstt (const vector float *, int, const int);
15519 void vec_dstt (const unsigned char *, int, const int);
15520 void vec_dstt (const signed char *, int, const int);
15521 void vec_dstt (const unsigned short *, int, const int);
15522 void vec_dstt (const short *, int, const int);
15523 void vec_dstt (const unsigned int *, int, const int);
15524 void vec_dstt (const int *, int, const int);
15525 void vec_dstt (const unsigned long *, int, const int);
15526 void vec_dstt (const long *, int, const int);
15527 void vec_dstt (const float *, int, const int);
15529 vector float vec_expte (vector float);
15531 vector float vec_floor (vector float);
15533 vector float vec_ld (int, const vector float *);
15534 vector float vec_ld (int, const float *);
15535 vector bool int vec_ld (int, const vector bool int *);
15536 vector signed int vec_ld (int, const vector signed int *);
15537 vector signed int vec_ld (int, const int *);
15538 vector signed int vec_ld (int, const long *);
15539 vector unsigned int vec_ld (int, const vector unsigned int *);
15540 vector unsigned int vec_ld (int, const unsigned int *);
15541 vector unsigned int vec_ld (int, const unsigned long *);
15542 vector bool short vec_ld (int, const vector bool short *);
15543 vector pixel vec_ld (int, const vector pixel *);
15544 vector signed short vec_ld (int, const vector signed short *);
15545 vector signed short vec_ld (int, const short *);
15546 vector unsigned short vec_ld (int, const vector unsigned short *);
15547 vector unsigned short vec_ld (int, const unsigned short *);
15548 vector bool char vec_ld (int, const vector bool char *);
15549 vector signed char vec_ld (int, const vector signed char *);
15550 vector signed char vec_ld (int, const signed char *);
15551 vector unsigned char vec_ld (int, const vector unsigned char *);
15552 vector unsigned char vec_ld (int, const unsigned char *);
15554 vector signed char vec_lde (int, const signed char *);
15555 vector unsigned char vec_lde (int, const unsigned char *);
15556 vector signed short vec_lde (int, const short *);
15557 vector unsigned short vec_lde (int, const unsigned short *);
15558 vector float vec_lde (int, const float *);
15559 vector signed int vec_lde (int, const int *);
15560 vector unsigned int vec_lde (int, const unsigned int *);
15561 vector signed int vec_lde (int, const long *);
15562 vector unsigned int vec_lde (int, const unsigned long *);
15564 vector float vec_lvewx (int, float *);
15565 vector signed int vec_lvewx (int, int *);
15566 vector unsigned int vec_lvewx (int, unsigned int *);
15567 vector signed int vec_lvewx (int, long *);
15568 vector unsigned int vec_lvewx (int, unsigned long *);
15570 vector signed short vec_lvehx (int, short *);
15571 vector unsigned short vec_lvehx (int, unsigned short *);
15573 vector signed char vec_lvebx (int, char *);
15574 vector unsigned char vec_lvebx (int, unsigned char *);
15576 vector float vec_ldl (int, const vector float *);
15577 vector float vec_ldl (int, const float *);
15578 vector bool int vec_ldl (int, const vector bool int *);
15579 vector signed int vec_ldl (int, const vector signed int *);
15580 vector signed int vec_ldl (int, const int *);
15581 vector signed int vec_ldl (int, const long *);
15582 vector unsigned int vec_ldl (int, const vector unsigned int *);
15583 vector unsigned int vec_ldl (int, const unsigned int *);
15584 vector unsigned int vec_ldl (int, const unsigned long *);
15585 vector bool short vec_ldl (int, const vector bool short *);
15586 vector pixel vec_ldl (int, const vector pixel *);
15587 vector signed short vec_ldl (int, const vector signed short *);
15588 vector signed short vec_ldl (int, const short *);
15589 vector unsigned short vec_ldl (int, const vector unsigned short *);
15590 vector unsigned short vec_ldl (int, const unsigned short *);
15591 vector bool char vec_ldl (int, const vector bool char *);
15592 vector signed char vec_ldl (int, const vector signed char *);
15593 vector signed char vec_ldl (int, const signed char *);
15594 vector unsigned char vec_ldl (int, const vector unsigned char *);
15595 vector unsigned char vec_ldl (int, const unsigned char *);
15597 vector float vec_loge (vector float);
15599 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
15600 vector unsigned char vec_lvsl (int, const volatile signed char *);
15601 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
15602 vector unsigned char vec_lvsl (int, const volatile short *);
15603 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
15604 vector unsigned char vec_lvsl (int, const volatile int *);
15605 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
15606 vector unsigned char vec_lvsl (int, const volatile long *);
15607 vector unsigned char vec_lvsl (int, const volatile float *);
15609 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
15610 vector unsigned char vec_lvsr (int, const volatile signed char *);
15611 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
15612 vector unsigned char vec_lvsr (int, const volatile short *);
15613 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
15614 vector unsigned char vec_lvsr (int, const volatile int *);
15615 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
15616 vector unsigned char vec_lvsr (int, const volatile long *);
15617 vector unsigned char vec_lvsr (int, const volatile float *);
15619 vector float vec_madd (vector float, vector float, vector float);
15621 vector signed short vec_madds (vector signed short,
15622 vector signed short,
15623 vector signed short);
15625 vector unsigned char vec_max (vector bool char, vector unsigned char);
15626 vector unsigned char vec_max (vector unsigned char, vector bool char);
15627 vector unsigned char vec_max (vector unsigned char,
15628 vector unsigned char);
15629 vector signed char vec_max (vector bool char, vector signed char);
15630 vector signed char vec_max (vector signed char, vector bool char);
15631 vector signed char vec_max (vector signed char, vector signed char);
15632 vector unsigned short vec_max (vector bool short,
15633 vector unsigned short);
15634 vector unsigned short vec_max (vector unsigned short,
15635 vector bool short);
15636 vector unsigned short vec_max (vector unsigned short,
15637 vector unsigned short);
15638 vector signed short vec_max (vector bool short, vector signed short);
15639 vector signed short vec_max (vector signed short, vector bool short);
15640 vector signed short vec_max (vector signed short, vector signed short);
15641 vector unsigned int vec_max (vector bool int, vector unsigned int);
15642 vector unsigned int vec_max (vector unsigned int, vector bool int);
15643 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
15644 vector signed int vec_max (vector bool int, vector signed int);
15645 vector signed int vec_max (vector signed int, vector bool int);
15646 vector signed int vec_max (vector signed int, vector signed int);
15647 vector float vec_max (vector float, vector float);
15649 vector float vec_vmaxfp (vector float, vector float);
15651 vector signed int vec_vmaxsw (vector bool int, vector signed int);
15652 vector signed int vec_vmaxsw (vector signed int, vector bool int);
15653 vector signed int vec_vmaxsw (vector signed int, vector signed int);
15655 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
15656 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
15657 vector unsigned int vec_vmaxuw (vector unsigned int,
15658 vector unsigned int);
15660 vector signed short vec_vmaxsh (vector bool short, vector signed short);
15661 vector signed short vec_vmaxsh (vector signed short, vector bool short);
15662 vector signed short vec_vmaxsh (vector signed short,
15663 vector signed short);
15665 vector unsigned short vec_vmaxuh (vector bool short,
15666 vector unsigned short);
15667 vector unsigned short vec_vmaxuh (vector unsigned short,
15668 vector bool short);
15669 vector unsigned short vec_vmaxuh (vector unsigned short,
15670 vector unsigned short);
15672 vector signed char vec_vmaxsb (vector bool char, vector signed char);
15673 vector signed char vec_vmaxsb (vector signed char, vector bool char);
15674 vector signed char vec_vmaxsb (vector signed char, vector signed char);
15676 vector unsigned char vec_vmaxub (vector bool char,
15677 vector unsigned char);
15678 vector unsigned char vec_vmaxub (vector unsigned char,
15680 vector unsigned char vec_vmaxub (vector unsigned char,
15681 vector unsigned char);
15683 vector bool char vec_mergeh (vector bool char, vector bool char);
15684 vector signed char vec_mergeh (vector signed char, vector signed char);
15685 vector unsigned char vec_mergeh (vector unsigned char,
15686 vector unsigned char);
15687 vector bool short vec_mergeh (vector bool short, vector bool short);
15688 vector pixel vec_mergeh (vector pixel, vector pixel);
15689 vector signed short vec_mergeh (vector signed short,
15690 vector signed short);
15691 vector unsigned short vec_mergeh (vector unsigned short,
15692 vector unsigned short);
15693 vector float vec_mergeh (vector float, vector float);
15694 vector bool int vec_mergeh (vector bool int, vector bool int);
15695 vector signed int vec_mergeh (vector signed int, vector signed int);
15696 vector unsigned int vec_mergeh (vector unsigned int,
15697 vector unsigned int);
15699 vector float vec_vmrghw (vector float, vector float);
15700 vector bool int vec_vmrghw (vector bool int, vector bool int);
15701 vector signed int vec_vmrghw (vector signed int, vector signed int);
15702 vector unsigned int vec_vmrghw (vector unsigned int,
15703 vector unsigned int);
15705 vector bool short vec_vmrghh (vector bool short, vector bool short);
15706 vector signed short vec_vmrghh (vector signed short,
15707 vector signed short);
15708 vector unsigned short vec_vmrghh (vector unsigned short,
15709 vector unsigned short);
15710 vector pixel vec_vmrghh (vector pixel, vector pixel);
15712 vector bool char vec_vmrghb (vector bool char, vector bool char);
15713 vector signed char vec_vmrghb (vector signed char, vector signed char);
15714 vector unsigned char vec_vmrghb (vector unsigned char,
15715 vector unsigned char);
15717 vector bool char vec_mergel (vector bool char, vector bool char);
15718 vector signed char vec_mergel (vector signed char, vector signed char);
15719 vector unsigned char vec_mergel (vector unsigned char,
15720 vector unsigned char);
15721 vector bool short vec_mergel (vector bool short, vector bool short);
15722 vector pixel vec_mergel (vector pixel, vector pixel);
15723 vector signed short vec_mergel (vector signed short,
15724 vector signed short);
15725 vector unsigned short vec_mergel (vector unsigned short,
15726 vector unsigned short);
15727 vector float vec_mergel (vector float, vector float);
15728 vector bool int vec_mergel (vector bool int, vector bool int);
15729 vector signed int vec_mergel (vector signed int, vector signed int);
15730 vector unsigned int vec_mergel (vector unsigned int,
15731 vector unsigned int);
15733 vector float vec_vmrglw (vector float, vector float);
15734 vector signed int vec_vmrglw (vector signed int, vector signed int);
15735 vector unsigned int vec_vmrglw (vector unsigned int,
15736 vector unsigned int);
15737 vector bool int vec_vmrglw (vector bool int, vector bool int);
15739 vector bool short vec_vmrglh (vector bool short, vector bool short);
15740 vector signed short vec_vmrglh (vector signed short,
15741 vector signed short);
15742 vector unsigned short vec_vmrglh (vector unsigned short,
15743 vector unsigned short);
15744 vector pixel vec_vmrglh (vector pixel, vector pixel);
15746 vector bool char vec_vmrglb (vector bool char, vector bool char);
15747 vector signed char vec_vmrglb (vector signed char, vector signed char);
15748 vector unsigned char vec_vmrglb (vector unsigned char,
15749 vector unsigned char);
15751 vector unsigned short vec_mfvscr (void);
15753 vector unsigned char vec_min (vector bool char, vector unsigned char);
15754 vector unsigned char vec_min (vector unsigned char, vector bool char);
15755 vector unsigned char vec_min (vector unsigned char,
15756 vector unsigned char);
15757 vector signed char vec_min (vector bool char, vector signed char);
15758 vector signed char vec_min (vector signed char, vector bool char);
15759 vector signed char vec_min (vector signed char, vector signed char);
15760 vector unsigned short vec_min (vector bool short,
15761 vector unsigned short);
15762 vector unsigned short vec_min (vector unsigned short,
15763 vector bool short);
15764 vector unsigned short vec_min (vector unsigned short,
15765 vector unsigned short);
15766 vector signed short vec_min (vector bool short, vector signed short);
15767 vector signed short vec_min (vector signed short, vector bool short);
15768 vector signed short vec_min (vector signed short, vector signed short);
15769 vector unsigned int vec_min (vector bool int, vector unsigned int);
15770 vector unsigned int vec_min (vector unsigned int, vector bool int);
15771 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
15772 vector signed int vec_min (vector bool int, vector signed int);
15773 vector signed int vec_min (vector signed int, vector bool int);
15774 vector signed int vec_min (vector signed int, vector signed int);
15775 vector float vec_min (vector float, vector float);
15777 vector float vec_vminfp (vector float, vector float);
15779 vector signed int vec_vminsw (vector bool int, vector signed int);
15780 vector signed int vec_vminsw (vector signed int, vector bool int);
15781 vector signed int vec_vminsw (vector signed int, vector signed int);
15783 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
15784 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
15785 vector unsigned int vec_vminuw (vector unsigned int,
15786 vector unsigned int);
15788 vector signed short vec_vminsh (vector bool short, vector signed short);
15789 vector signed short vec_vminsh (vector signed short, vector bool short);
15790 vector signed short vec_vminsh (vector signed short,
15791 vector signed short);
15793 vector unsigned short vec_vminuh (vector bool short,
15794 vector unsigned short);
15795 vector unsigned short vec_vminuh (vector unsigned short,
15796 vector bool short);
15797 vector unsigned short vec_vminuh (vector unsigned short,
15798 vector unsigned short);
15800 vector signed char vec_vminsb (vector bool char, vector signed char);
15801 vector signed char vec_vminsb (vector signed char, vector bool char);
15802 vector signed char vec_vminsb (vector signed char, vector signed char);
15804 vector unsigned char vec_vminub (vector bool char,
15805 vector unsigned char);
15806 vector unsigned char vec_vminub (vector unsigned char,
15808 vector unsigned char vec_vminub (vector unsigned char,
15809 vector unsigned char);
15811 vector signed short vec_mladd (vector signed short,
15812 vector signed short,
15813 vector signed short);
15814 vector signed short vec_mladd (vector signed short,
15815 vector unsigned short,
15816 vector unsigned short);
15817 vector signed short vec_mladd (vector unsigned short,
15818 vector signed short,
15819 vector signed short);
15820 vector unsigned short vec_mladd (vector unsigned short,
15821 vector unsigned short,
15822 vector unsigned short);
15824 vector signed short vec_mradds (vector signed short,
15825 vector signed short,
15826 vector signed short);
15828 vector unsigned int vec_msum (vector unsigned char,
15829 vector unsigned char,
15830 vector unsigned int);
15831 vector signed int vec_msum (vector signed char,
15832 vector unsigned char,
15833 vector signed int);
15834 vector unsigned int vec_msum (vector unsigned short,
15835 vector unsigned short,
15836 vector unsigned int);
15837 vector signed int vec_msum (vector signed short,
15838 vector signed short,
15839 vector signed int);
15841 vector signed int vec_vmsumshm (vector signed short,
15842 vector signed short,
15843 vector signed int);
15845 vector unsigned int vec_vmsumuhm (vector unsigned short,
15846 vector unsigned short,
15847 vector unsigned int);
15849 vector signed int vec_vmsummbm (vector signed char,
15850 vector unsigned char,
15851 vector signed int);
15853 vector unsigned int vec_vmsumubm (vector unsigned char,
15854 vector unsigned char,
15855 vector unsigned int);
15857 vector unsigned int vec_msums (vector unsigned short,
15858 vector unsigned short,
15859 vector unsigned int);
15860 vector signed int vec_msums (vector signed short,
15861 vector signed short,
15862 vector signed int);
15864 vector signed int vec_vmsumshs (vector signed short,
15865 vector signed short,
15866 vector signed int);
15868 vector unsigned int vec_vmsumuhs (vector unsigned short,
15869 vector unsigned short,
15870 vector unsigned int);
15872 void vec_mtvscr (vector signed int);
15873 void vec_mtvscr (vector unsigned int);
15874 void vec_mtvscr (vector bool int);
15875 void vec_mtvscr (vector signed short);
15876 void vec_mtvscr (vector unsigned short);
15877 void vec_mtvscr (vector bool short);
15878 void vec_mtvscr (vector pixel);
15879 void vec_mtvscr (vector signed char);
15880 void vec_mtvscr (vector unsigned char);
15881 void vec_mtvscr (vector bool char);
15883 vector unsigned short vec_mule (vector unsigned char,
15884 vector unsigned char);
15885 vector signed short vec_mule (vector signed char,
15886 vector signed char);
15887 vector unsigned int vec_mule (vector unsigned short,
15888 vector unsigned short);
15889 vector signed int vec_mule (vector signed short, vector signed short);
15891 vector signed int vec_vmulesh (vector signed short,
15892 vector signed short);
15894 vector unsigned int vec_vmuleuh (vector unsigned short,
15895 vector unsigned short);
15897 vector signed short vec_vmulesb (vector signed char,
15898 vector signed char);
15900 vector unsigned short vec_vmuleub (vector unsigned char,
15901 vector unsigned char);
15903 vector unsigned short vec_mulo (vector unsigned char,
15904 vector unsigned char);
15905 vector signed short vec_mulo (vector signed char, vector signed char);
15906 vector unsigned int vec_mulo (vector unsigned short,
15907 vector unsigned short);
15908 vector signed int vec_mulo (vector signed short, vector signed short);
15910 vector signed int vec_vmulosh (vector signed short,
15911 vector signed short);
15913 vector unsigned int vec_vmulouh (vector unsigned short,
15914 vector unsigned short);
15916 vector signed short vec_vmulosb (vector signed char,
15917 vector signed char);
15919 vector unsigned short vec_vmuloub (vector unsigned char,
15920 vector unsigned char);
15922 vector float vec_nmsub (vector float, vector float, vector float);
15924 vector float vec_nor (vector float, vector float);
15925 vector signed int vec_nor (vector signed int, vector signed int);
15926 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
15927 vector bool int vec_nor (vector bool int, vector bool int);
15928 vector signed short vec_nor (vector signed short, vector signed short);
15929 vector unsigned short vec_nor (vector unsigned short,
15930 vector unsigned short);
15931 vector bool short vec_nor (vector bool short, vector bool short);
15932 vector signed char vec_nor (vector signed char, vector signed char);
15933 vector unsigned char vec_nor (vector unsigned char,
15934 vector unsigned char);
15935 vector bool char vec_nor (vector bool char, vector bool char);
15937 vector float vec_or (vector float, vector float);
15938 vector float vec_or (vector float, vector bool int);
15939 vector float vec_or (vector bool int, vector float);
15940 vector bool int vec_or (vector bool int, vector bool int);
15941 vector signed int vec_or (vector bool int, vector signed int);
15942 vector signed int vec_or (vector signed int, vector bool int);
15943 vector signed int vec_or (vector signed int, vector signed int);
15944 vector unsigned int vec_or (vector bool int, vector unsigned int);
15945 vector unsigned int vec_or (vector unsigned int, vector bool int);
15946 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
15947 vector bool short vec_or (vector bool short, vector bool short);
15948 vector signed short vec_or (vector bool short, vector signed short);
15949 vector signed short vec_or (vector signed short, vector bool short);
15950 vector signed short vec_or (vector signed short, vector signed short);
15951 vector unsigned short vec_or (vector bool short, vector unsigned short);
15952 vector unsigned short vec_or (vector unsigned short, vector bool short);
15953 vector unsigned short vec_or (vector unsigned short,
15954 vector unsigned short);
15955 vector signed char vec_or (vector bool char, vector signed char);
15956 vector bool char vec_or (vector bool char, vector bool char);
15957 vector signed char vec_or (vector signed char, vector bool char);
15958 vector signed char vec_or (vector signed char, vector signed char);
15959 vector unsigned char vec_or (vector bool char, vector unsigned char);
15960 vector unsigned char vec_or (vector unsigned char, vector bool char);
15961 vector unsigned char vec_or (vector unsigned char,
15962 vector unsigned char);
15964 vector signed char vec_pack (vector signed short, vector signed short);
15965 vector unsigned char vec_pack (vector unsigned short,
15966 vector unsigned short);
15967 vector bool char vec_pack (vector bool short, vector bool short);
15968 vector signed short vec_pack (vector signed int, vector signed int);
15969 vector unsigned short vec_pack (vector unsigned int,
15970 vector unsigned int);
15971 vector bool short vec_pack (vector bool int, vector bool int);
15973 vector bool short vec_vpkuwum (vector bool int, vector bool int);
15974 vector signed short vec_vpkuwum (vector signed int, vector signed int);
15975 vector unsigned short vec_vpkuwum (vector unsigned int,
15976 vector unsigned int);
15978 vector bool char vec_vpkuhum (vector bool short, vector bool short);
15979 vector signed char vec_vpkuhum (vector signed short,
15980 vector signed short);
15981 vector unsigned char vec_vpkuhum (vector unsigned short,
15982 vector unsigned short);
15984 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
15986 vector unsigned char vec_packs (vector unsigned short,
15987 vector unsigned short);
15988 vector signed char vec_packs (vector signed short, vector signed short);
15989 vector unsigned short vec_packs (vector unsigned int,
15990 vector unsigned int);
15991 vector signed short vec_packs (vector signed int, vector signed int);
15993 vector signed short vec_vpkswss (vector signed int, vector signed int);
15995 vector unsigned short vec_vpkuwus (vector unsigned int,
15996 vector unsigned int);
15998 vector signed char vec_vpkshss (vector signed short,
15999 vector signed short);
16001 vector unsigned char vec_vpkuhus (vector unsigned short,
16002 vector unsigned short);
16004 vector unsigned char vec_packsu (vector unsigned short,
16005 vector unsigned short);
16006 vector unsigned char vec_packsu (vector signed short,
16007 vector signed short);
16008 vector unsigned short vec_packsu (vector unsigned int,
16009 vector unsigned int);
16010 vector unsigned short vec_packsu (vector signed int, vector signed int);
16012 vector unsigned short vec_vpkswus (vector signed int,
16013 vector signed int);
16015 vector unsigned char vec_vpkshus (vector signed short,
16016 vector signed short);
16018 vector float vec_perm (vector float,
16020 vector unsigned char);
16021 vector signed int vec_perm (vector signed int,
16023 vector unsigned char);
16024 vector unsigned int vec_perm (vector unsigned int,
16025 vector unsigned int,
16026 vector unsigned char);
16027 vector bool int vec_perm (vector bool int,
16029 vector unsigned char);
16030 vector signed short vec_perm (vector signed short,
16031 vector signed short,
16032 vector unsigned char);
16033 vector unsigned short vec_perm (vector unsigned short,
16034 vector unsigned short,
16035 vector unsigned char);
16036 vector bool short vec_perm (vector bool short,
16038 vector unsigned char);
16039 vector pixel vec_perm (vector pixel,
16041 vector unsigned char);
16042 vector signed char vec_perm (vector signed char,
16043 vector signed char,
16044 vector unsigned char);
16045 vector unsigned char vec_perm (vector unsigned char,
16046 vector unsigned char,
16047 vector unsigned char);
16048 vector bool char vec_perm (vector bool char,
16050 vector unsigned char);
16052 vector float vec_re (vector float);
16054 vector signed char vec_rl (vector signed char,
16055 vector unsigned char);
16056 vector unsigned char vec_rl (vector unsigned char,
16057 vector unsigned char);
16058 vector signed short vec_rl (vector signed short, vector unsigned short);
16059 vector unsigned short vec_rl (vector unsigned short,
16060 vector unsigned short);
16061 vector signed int vec_rl (vector signed int, vector unsigned int);
16062 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
16064 vector signed int vec_vrlw (vector signed int, vector unsigned int);
16065 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
16067 vector signed short vec_vrlh (vector signed short,
16068 vector unsigned short);
16069 vector unsigned short vec_vrlh (vector unsigned short,
16070 vector unsigned short);
16072 vector signed char vec_vrlb (vector signed char, vector unsigned char);
16073 vector unsigned char vec_vrlb (vector unsigned char,
16074 vector unsigned char);
16076 vector float vec_round (vector float);
16078 vector float vec_recip (vector float, vector float);
16080 vector float vec_rsqrt (vector float);
16082 vector float vec_rsqrte (vector float);
16084 vector float vec_sel (vector float, vector float, vector bool int);
16085 vector float vec_sel (vector float, vector float, vector unsigned int);
16086 vector signed int vec_sel (vector signed int,
16089 vector signed int vec_sel (vector signed int,
16091 vector unsigned int);
16092 vector unsigned int vec_sel (vector unsigned int,
16093 vector unsigned int,
16095 vector unsigned int vec_sel (vector unsigned int,
16096 vector unsigned int,
16097 vector unsigned int);
16098 vector bool int vec_sel (vector bool int,
16101 vector bool int vec_sel (vector bool int,
16103 vector unsigned int);
16104 vector signed short vec_sel (vector signed short,
16105 vector signed short,
16106 vector bool short);
16107 vector signed short vec_sel (vector signed short,
16108 vector signed short,
16109 vector unsigned short);
16110 vector unsigned short vec_sel (vector unsigned short,
16111 vector unsigned short,
16112 vector bool short);
16113 vector unsigned short vec_sel (vector unsigned short,
16114 vector unsigned short,
16115 vector unsigned short);
16116 vector bool short vec_sel (vector bool short,
16118 vector bool short);
16119 vector bool short vec_sel (vector bool short,
16121 vector unsigned short);
16122 vector signed char vec_sel (vector signed char,
16123 vector signed char,
16125 vector signed char vec_sel (vector signed char,
16126 vector signed char,
16127 vector unsigned char);
16128 vector unsigned char vec_sel (vector unsigned char,
16129 vector unsigned char,
16131 vector unsigned char vec_sel (vector unsigned char,
16132 vector unsigned char,
16133 vector unsigned char);
16134 vector bool char vec_sel (vector bool char,
16137 vector bool char vec_sel (vector bool char,
16139 vector unsigned char);
16141 vector signed char vec_sl (vector signed char,
16142 vector unsigned char);
16143 vector unsigned char vec_sl (vector unsigned char,
16144 vector unsigned char);
16145 vector signed short vec_sl (vector signed short, vector unsigned short);
16146 vector unsigned short vec_sl (vector unsigned short,
16147 vector unsigned short);
16148 vector signed int vec_sl (vector signed int, vector unsigned int);
16149 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
16151 vector signed int vec_vslw (vector signed int, vector unsigned int);
16152 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
16154 vector signed short vec_vslh (vector signed short,
16155 vector unsigned short);
16156 vector unsigned short vec_vslh (vector unsigned short,
16157 vector unsigned short);
16159 vector signed char vec_vslb (vector signed char, vector unsigned char);
16160 vector unsigned char vec_vslb (vector unsigned char,
16161 vector unsigned char);
16163 vector float vec_sld (vector float, vector float, const int);
16164 vector signed int vec_sld (vector signed int,
16167 vector unsigned int vec_sld (vector unsigned int,
16168 vector unsigned int,
16170 vector bool int vec_sld (vector bool int,
16173 vector signed short vec_sld (vector signed short,
16174 vector signed short,
16176 vector unsigned short vec_sld (vector unsigned short,
16177 vector unsigned short,
16179 vector bool short vec_sld (vector bool short,
16182 vector pixel vec_sld (vector pixel,
16185 vector signed char vec_sld (vector signed char,
16186 vector signed char,
16188 vector unsigned char vec_sld (vector unsigned char,
16189 vector unsigned char,
16191 vector bool char vec_sld (vector bool char,
16195 vector signed int vec_sll (vector signed int,
16196 vector unsigned int);
16197 vector signed int vec_sll (vector signed int,
16198 vector unsigned short);
16199 vector signed int vec_sll (vector signed int,
16200 vector unsigned char);
16201 vector unsigned int vec_sll (vector unsigned int,
16202 vector unsigned int);
16203 vector unsigned int vec_sll (vector unsigned int,
16204 vector unsigned short);
16205 vector unsigned int vec_sll (vector unsigned int,
16206 vector unsigned char);
16207 vector bool int vec_sll (vector bool int,
16208 vector unsigned int);
16209 vector bool int vec_sll (vector bool int,
16210 vector unsigned short);
16211 vector bool int vec_sll (vector bool int,
16212 vector unsigned char);
16213 vector signed short vec_sll (vector signed short,
16214 vector unsigned int);
16215 vector signed short vec_sll (vector signed short,
16216 vector unsigned short);
16217 vector signed short vec_sll (vector signed short,
16218 vector unsigned char);
16219 vector unsigned short vec_sll (vector unsigned short,
16220 vector unsigned int);
16221 vector unsigned short vec_sll (vector unsigned short,
16222 vector unsigned short);
16223 vector unsigned short vec_sll (vector unsigned short,
16224 vector unsigned char);
16225 vector bool short vec_sll (vector bool short, vector unsigned int);
16226 vector bool short vec_sll (vector bool short, vector unsigned short);
16227 vector bool short vec_sll (vector bool short, vector unsigned char);
16228 vector pixel vec_sll (vector pixel, vector unsigned int);
16229 vector pixel vec_sll (vector pixel, vector unsigned short);
16230 vector pixel vec_sll (vector pixel, vector unsigned char);
16231 vector signed char vec_sll (vector signed char, vector unsigned int);
16232 vector signed char vec_sll (vector signed char, vector unsigned short);
16233 vector signed char vec_sll (vector signed char, vector unsigned char);
16234 vector unsigned char vec_sll (vector unsigned char,
16235 vector unsigned int);
16236 vector unsigned char vec_sll (vector unsigned char,
16237 vector unsigned short);
16238 vector unsigned char vec_sll (vector unsigned char,
16239 vector unsigned char);
16240 vector bool char vec_sll (vector bool char, vector unsigned int);
16241 vector bool char vec_sll (vector bool char, vector unsigned short);
16242 vector bool char vec_sll (vector bool char, vector unsigned char);
16244 vector float vec_slo (vector float, vector signed char);
16245 vector float vec_slo (vector float, vector unsigned char);
16246 vector signed int vec_slo (vector signed int, vector signed char);
16247 vector signed int vec_slo (vector signed int, vector unsigned char);
16248 vector unsigned int vec_slo (vector unsigned int, vector signed char);
16249 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
16250 vector signed short vec_slo (vector signed short, vector signed char);
16251 vector signed short vec_slo (vector signed short, vector unsigned char);
16252 vector unsigned short vec_slo (vector unsigned short,
16253 vector signed char);
16254 vector unsigned short vec_slo (vector unsigned short,
16255 vector unsigned char);
16256 vector pixel vec_slo (vector pixel, vector signed char);
16257 vector pixel vec_slo (vector pixel, vector unsigned char);
16258 vector signed char vec_slo (vector signed char, vector signed char);
16259 vector signed char vec_slo (vector signed char, vector unsigned char);
16260 vector unsigned char vec_slo (vector unsigned char, vector signed char);
16261 vector unsigned char vec_slo (vector unsigned char,
16262 vector unsigned char);
16264 vector signed char vec_splat (vector signed char, const int);
16265 vector unsigned char vec_splat (vector unsigned char, const int);
16266 vector bool char vec_splat (vector bool char, const int);
16267 vector signed short vec_splat (vector signed short, const int);
16268 vector unsigned short vec_splat (vector unsigned short, const int);
16269 vector bool short vec_splat (vector bool short, const int);
16270 vector pixel vec_splat (vector pixel, const int);
16271 vector float vec_splat (vector float, const int);
16272 vector signed int vec_splat (vector signed int, const int);
16273 vector unsigned int vec_splat (vector unsigned int, const int);
16274 vector bool int vec_splat (vector bool int, const int);
16275 vector signed long vec_splat (vector signed long, const int);
16276 vector unsigned long vec_splat (vector unsigned long, const int);
16278 vector signed char vec_splats (signed char);
16279 vector unsigned char vec_splats (unsigned char);
16280 vector signed short vec_splats (signed short);
16281 vector unsigned short vec_splats (unsigned short);
16282 vector signed int vec_splats (signed int);
16283 vector unsigned int vec_splats (unsigned int);
16284 vector float vec_splats (float);
16286 vector float vec_vspltw (vector float, const int);
16287 vector signed int vec_vspltw (vector signed int, const int);
16288 vector unsigned int vec_vspltw (vector unsigned int, const int);
16289 vector bool int vec_vspltw (vector bool int, const int);
16291 vector bool short vec_vsplth (vector bool short, const int);
16292 vector signed short vec_vsplth (vector signed short, const int);
16293 vector unsigned short vec_vsplth (vector unsigned short, const int);
16294 vector pixel vec_vsplth (vector pixel, const int);
16296 vector signed char vec_vspltb (vector signed char, const int);
16297 vector unsigned char vec_vspltb (vector unsigned char, const int);
16298 vector bool char vec_vspltb (vector bool char, const int);
16300 vector signed char vec_splat_s8 (const int);
16302 vector signed short vec_splat_s16 (const int);
16304 vector signed int vec_splat_s32 (const int);
16306 vector unsigned char vec_splat_u8 (const int);
16308 vector unsigned short vec_splat_u16 (const int);
16310 vector unsigned int vec_splat_u32 (const int);
16312 vector signed char vec_sr (vector signed char, vector unsigned char);
16313 vector unsigned char vec_sr (vector unsigned char,
16314 vector unsigned char);
16315 vector signed short vec_sr (vector signed short,
16316 vector unsigned short);
16317 vector unsigned short vec_sr (vector unsigned short,
16318 vector unsigned short);
16319 vector signed int vec_sr (vector signed int, vector unsigned int);
16320 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
16322 vector signed int vec_vsrw (vector signed int, vector unsigned int);
16323 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
16325 vector signed short vec_vsrh (vector signed short,
16326 vector unsigned short);
16327 vector unsigned short vec_vsrh (vector unsigned short,
16328 vector unsigned short);
16330 vector signed char vec_vsrb (vector signed char, vector unsigned char);
16331 vector unsigned char vec_vsrb (vector unsigned char,
16332 vector unsigned char);
16334 vector signed char vec_sra (vector signed char, vector unsigned char);
16335 vector unsigned char vec_sra (vector unsigned char,
16336 vector unsigned char);
16337 vector signed short vec_sra (vector signed short,
16338 vector unsigned short);
16339 vector unsigned short vec_sra (vector unsigned short,
16340 vector unsigned short);
16341 vector signed int vec_sra (vector signed int, vector unsigned int);
16342 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
16344 vector signed int vec_vsraw (vector signed int, vector unsigned int);
16345 vector unsigned int vec_vsraw (vector unsigned int,
16346 vector unsigned int);
16348 vector signed short vec_vsrah (vector signed short,
16349 vector unsigned short);
16350 vector unsigned short vec_vsrah (vector unsigned short,
16351 vector unsigned short);
16353 vector signed char vec_vsrab (vector signed char, vector unsigned char);
16354 vector unsigned char vec_vsrab (vector unsigned char,
16355 vector unsigned char);
16357 vector signed int vec_srl (vector signed int, vector unsigned int);
16358 vector signed int vec_srl (vector signed int, vector unsigned short);
16359 vector signed int vec_srl (vector signed int, vector unsigned char);
16360 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
16361 vector unsigned int vec_srl (vector unsigned int,
16362 vector unsigned short);
16363 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
16364 vector bool int vec_srl (vector bool int, vector unsigned int);
16365 vector bool int vec_srl (vector bool int, vector unsigned short);
16366 vector bool int vec_srl (vector bool int, vector unsigned char);
16367 vector signed short vec_srl (vector signed short, vector unsigned int);
16368 vector signed short vec_srl (vector signed short,
16369 vector unsigned short);
16370 vector signed short vec_srl (vector signed short, vector unsigned char);
16371 vector unsigned short vec_srl (vector unsigned short,
16372 vector unsigned int);
16373 vector unsigned short vec_srl (vector unsigned short,
16374 vector unsigned short);
16375 vector unsigned short vec_srl (vector unsigned short,
16376 vector unsigned char);
16377 vector bool short vec_srl (vector bool short, vector unsigned int);
16378 vector bool short vec_srl (vector bool short, vector unsigned short);
16379 vector bool short vec_srl (vector bool short, vector unsigned char);
16380 vector pixel vec_srl (vector pixel, vector unsigned int);
16381 vector pixel vec_srl (vector pixel, vector unsigned short);
16382 vector pixel vec_srl (vector pixel, vector unsigned char);
16383 vector signed char vec_srl (vector signed char, vector unsigned int);
16384 vector signed char vec_srl (vector signed char, vector unsigned short);
16385 vector signed char vec_srl (vector signed char, vector unsigned char);
16386 vector unsigned char vec_srl (vector unsigned char,
16387 vector unsigned int);
16388 vector unsigned char vec_srl (vector unsigned char,
16389 vector unsigned short);
16390 vector unsigned char vec_srl (vector unsigned char,
16391 vector unsigned char);
16392 vector bool char vec_srl (vector bool char, vector unsigned int);
16393 vector bool char vec_srl (vector bool char, vector unsigned short);
16394 vector bool char vec_srl (vector bool char, vector unsigned char);
16396 vector float vec_sro (vector float, vector signed char);
16397 vector float vec_sro (vector float, vector unsigned char);
16398 vector signed int vec_sro (vector signed int, vector signed char);
16399 vector signed int vec_sro (vector signed int, vector unsigned char);
16400 vector unsigned int vec_sro (vector unsigned int, vector signed char);
16401 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
16402 vector signed short vec_sro (vector signed short, vector signed char);
16403 vector signed short vec_sro (vector signed short, vector unsigned char);
16404 vector unsigned short vec_sro (vector unsigned short,
16405 vector signed char);
16406 vector unsigned short vec_sro (vector unsigned short,
16407 vector unsigned char);
16408 vector pixel vec_sro (vector pixel, vector signed char);
16409 vector pixel vec_sro (vector pixel, vector unsigned char);
16410 vector signed char vec_sro (vector signed char, vector signed char);
16411 vector signed char vec_sro (vector signed char, vector unsigned char);
16412 vector unsigned char vec_sro (vector unsigned char, vector signed char);
16413 vector unsigned char vec_sro (vector unsigned char,
16414 vector unsigned char);
16416 void vec_st (vector float, int, vector float *);
16417 void vec_st (vector float, int, float *);
16418 void vec_st (vector signed int, int, vector signed int *);
16419 void vec_st (vector signed int, int, int *);
16420 void vec_st (vector unsigned int, int, vector unsigned int *);
16421 void vec_st (vector unsigned int, int, unsigned int *);
16422 void vec_st (vector bool int, int, vector bool int *);
16423 void vec_st (vector bool int, int, unsigned int *);
16424 void vec_st (vector bool int, int, int *);
16425 void vec_st (vector signed short, int, vector signed short *);
16426 void vec_st (vector signed short, int, short *);
16427 void vec_st (vector unsigned short, int, vector unsigned short *);
16428 void vec_st (vector unsigned short, int, unsigned short *);
16429 void vec_st (vector bool short, int, vector bool short *);
16430 void vec_st (vector bool short, int, unsigned short *);
16431 void vec_st (vector pixel, int, vector pixel *);
16432 void vec_st (vector pixel, int, unsigned short *);
16433 void vec_st (vector pixel, int, short *);
16434 void vec_st (vector bool short, int, short *);
16435 void vec_st (vector signed char, int, vector signed char *);
16436 void vec_st (vector signed char, int, signed char *);
16437 void vec_st (vector unsigned char, int, vector unsigned char *);
16438 void vec_st (vector unsigned char, int, unsigned char *);
16439 void vec_st (vector bool char, int, vector bool char *);
16440 void vec_st (vector bool char, int, unsigned char *);
16441 void vec_st (vector bool char, int, signed char *);
16443 void vec_ste (vector signed char, int, signed char *);
16444 void vec_ste (vector unsigned char, int, unsigned char *);
16445 void vec_ste (vector bool char, int, signed char *);
16446 void vec_ste (vector bool char, int, unsigned char *);
16447 void vec_ste (vector signed short, int, short *);
16448 void vec_ste (vector unsigned short, int, unsigned short *);
16449 void vec_ste (vector bool short, int, short *);
16450 void vec_ste (vector bool short, int, unsigned short *);
16451 void vec_ste (vector pixel, int, short *);
16452 void vec_ste (vector pixel, int, unsigned short *);
16453 void vec_ste (vector float, int, float *);
16454 void vec_ste (vector signed int, int, int *);
16455 void vec_ste (vector unsigned int, int, unsigned int *);
16456 void vec_ste (vector bool int, int, int *);
16457 void vec_ste (vector bool int, int, unsigned int *);
16459 void vec_stvewx (vector float, int, float *);
16460 void vec_stvewx (vector signed int, int, int *);
16461 void vec_stvewx (vector unsigned int, int, unsigned int *);
16462 void vec_stvewx (vector bool int, int, int *);
16463 void vec_stvewx (vector bool int, int, unsigned int *);
16465 void vec_stvehx (vector signed short, int, short *);
16466 void vec_stvehx (vector unsigned short, int, unsigned short *);
16467 void vec_stvehx (vector bool short, int, short *);
16468 void vec_stvehx (vector bool short, int, unsigned short *);
16469 void vec_stvehx (vector pixel, int, short *);
16470 void vec_stvehx (vector pixel, int, unsigned short *);
16472 void vec_stvebx (vector signed char, int, signed char *);
16473 void vec_stvebx (vector unsigned char, int, unsigned char *);
16474 void vec_stvebx (vector bool char, int, signed char *);
16475 void vec_stvebx (vector bool char, int, unsigned char *);
16477 void vec_stl (vector float, int, vector float *);
16478 void vec_stl (vector float, int, float *);
16479 void vec_stl (vector signed int, int, vector signed int *);
16480 void vec_stl (vector signed int, int, int *);
16481 void vec_stl (vector unsigned int, int, vector unsigned int *);
16482 void vec_stl (vector unsigned int, int, unsigned int *);
16483 void vec_stl (vector bool int, int, vector bool int *);
16484 void vec_stl (vector bool int, int, unsigned int *);
16485 void vec_stl (vector bool int, int, int *);
16486 void vec_stl (vector signed short, int, vector signed short *);
16487 void vec_stl (vector signed short, int, short *);
16488 void vec_stl (vector unsigned short, int, vector unsigned short *);
16489 void vec_stl (vector unsigned short, int, unsigned short *);
16490 void vec_stl (vector bool short, int, vector bool short *);
16491 void vec_stl (vector bool short, int, unsigned short *);
16492 void vec_stl (vector bool short, int, short *);
16493 void vec_stl (vector pixel, int, vector pixel *);
16494 void vec_stl (vector pixel, int, unsigned short *);
16495 void vec_stl (vector pixel, int, short *);
16496 void vec_stl (vector signed char, int, vector signed char *);
16497 void vec_stl (vector signed char, int, signed char *);
16498 void vec_stl (vector unsigned char, int, vector unsigned char *);
16499 void vec_stl (vector unsigned char, int, unsigned char *);
16500 void vec_stl (vector bool char, int, vector bool char *);
16501 void vec_stl (vector bool char, int, unsigned char *);
16502 void vec_stl (vector bool char, int, signed char *);
16504 vector signed char vec_sub (vector bool char, vector signed char);
16505 vector signed char vec_sub (vector signed char, vector bool char);
16506 vector signed char vec_sub (vector signed char, vector signed char);
16507 vector unsigned char vec_sub (vector bool char, vector unsigned char);
16508 vector unsigned char vec_sub (vector unsigned char, vector bool char);
16509 vector unsigned char vec_sub (vector unsigned char,
16510 vector unsigned char);
16511 vector signed short vec_sub (vector bool short, vector signed short);
16512 vector signed short vec_sub (vector signed short, vector bool short);
16513 vector signed short vec_sub (vector signed short, vector signed short);
16514 vector unsigned short vec_sub (vector bool short,
16515 vector unsigned short);
16516 vector unsigned short vec_sub (vector unsigned short,
16517 vector bool short);
16518 vector unsigned short vec_sub (vector unsigned short,
16519 vector unsigned short);
16520 vector signed int vec_sub (vector bool int, vector signed int);
16521 vector signed int vec_sub (vector signed int, vector bool int);
16522 vector signed int vec_sub (vector signed int, vector signed int);
16523 vector unsigned int vec_sub (vector bool int, vector unsigned int);
16524 vector unsigned int vec_sub (vector unsigned int, vector bool int);
16525 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
16526 vector float vec_sub (vector float, vector float);
16528 vector float vec_vsubfp (vector float, vector float);
16530 vector signed int vec_vsubuwm (vector bool int, vector signed int);
16531 vector signed int vec_vsubuwm (vector signed int, vector bool int);
16532 vector signed int vec_vsubuwm (vector signed int, vector signed int);
16533 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
16534 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
16535 vector unsigned int vec_vsubuwm (vector unsigned int,
16536 vector unsigned int);
16538 vector signed short vec_vsubuhm (vector bool short,
16539 vector signed short);
16540 vector signed short vec_vsubuhm (vector signed short,
16541 vector bool short);
16542 vector signed short vec_vsubuhm (vector signed short,
16543 vector signed short);
16544 vector unsigned short vec_vsubuhm (vector bool short,
16545 vector unsigned short);
16546 vector unsigned short vec_vsubuhm (vector unsigned short,
16547 vector bool short);
16548 vector unsigned short vec_vsubuhm (vector unsigned short,
16549 vector unsigned short);
16551 vector signed char vec_vsububm (vector bool char, vector signed char);
16552 vector signed char vec_vsububm (vector signed char, vector bool char);
16553 vector signed char vec_vsububm (vector signed char, vector signed char);
16554 vector unsigned char vec_vsububm (vector bool char,
16555 vector unsigned char);
16556 vector unsigned char vec_vsububm (vector unsigned char,
16558 vector unsigned char vec_vsububm (vector unsigned char,
16559 vector unsigned char);
16561 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
16563 vector unsigned char vec_subs (vector bool char, vector unsigned char);
16564 vector unsigned char vec_subs (vector unsigned char, vector bool char);
16565 vector unsigned char vec_subs (vector unsigned char,
16566 vector unsigned char);
16567 vector signed char vec_subs (vector bool char, vector signed char);
16568 vector signed char vec_subs (vector signed char, vector bool char);
16569 vector signed char vec_subs (vector signed char, vector signed char);
16570 vector unsigned short vec_subs (vector bool short,
16571 vector unsigned short);
16572 vector unsigned short vec_subs (vector unsigned short,
16573 vector bool short);
16574 vector unsigned short vec_subs (vector unsigned short,
16575 vector unsigned short);
16576 vector signed short vec_subs (vector bool short, vector signed short);
16577 vector signed short vec_subs (vector signed short, vector bool short);
16578 vector signed short vec_subs (vector signed short, vector signed short);
16579 vector unsigned int vec_subs (vector bool int, vector unsigned int);
16580 vector unsigned int vec_subs (vector unsigned int, vector bool int);
16581 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
16582 vector signed int vec_subs (vector bool int, vector signed int);
16583 vector signed int vec_subs (vector signed int, vector bool int);
16584 vector signed int vec_subs (vector signed int, vector signed int);
16586 vector signed int vec_vsubsws (vector bool int, vector signed int);
16587 vector signed int vec_vsubsws (vector signed int, vector bool int);
16588 vector signed int vec_vsubsws (vector signed int, vector signed int);
16590 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
16591 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
16592 vector unsigned int vec_vsubuws (vector unsigned int,
16593 vector unsigned int);
16595 vector signed short vec_vsubshs (vector bool short,
16596 vector signed short);
16597 vector signed short vec_vsubshs (vector signed short,
16598 vector bool short);
16599 vector signed short vec_vsubshs (vector signed short,
16600 vector signed short);
16602 vector unsigned short vec_vsubuhs (vector bool short,
16603 vector unsigned short);
16604 vector unsigned short vec_vsubuhs (vector unsigned short,
16605 vector bool short);
16606 vector unsigned short vec_vsubuhs (vector unsigned short,
16607 vector unsigned short);
16609 vector signed char vec_vsubsbs (vector bool char, vector signed char);
16610 vector signed char vec_vsubsbs (vector signed char, vector bool char);
16611 vector signed char vec_vsubsbs (vector signed char, vector signed char);
16613 vector unsigned char vec_vsububs (vector bool char,
16614 vector unsigned char);
16615 vector unsigned char vec_vsububs (vector unsigned char,
16617 vector unsigned char vec_vsububs (vector unsigned char,
16618 vector unsigned char);
16620 vector unsigned int vec_sum4s (vector unsigned char,
16621 vector unsigned int);
16622 vector signed int vec_sum4s (vector signed char, vector signed int);
16623 vector signed int vec_sum4s (vector signed short, vector signed int);
16625 vector signed int vec_vsum4shs (vector signed short, vector signed int);
16627 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
16629 vector unsigned int vec_vsum4ubs (vector unsigned char,
16630 vector unsigned int);
16632 vector signed int vec_sum2s (vector signed int, vector signed int);
16634 vector signed int vec_sums (vector signed int, vector signed int);
16636 vector float vec_trunc (vector float);
16638 vector signed short vec_unpackh (vector signed char);
16639 vector bool short vec_unpackh (vector bool char);
16640 vector signed int vec_unpackh (vector signed short);
16641 vector bool int vec_unpackh (vector bool short);
16642 vector unsigned int vec_unpackh (vector pixel);
16644 vector bool int vec_vupkhsh (vector bool short);
16645 vector signed int vec_vupkhsh (vector signed short);
16647 vector unsigned int vec_vupkhpx (vector pixel);
16649 vector bool short vec_vupkhsb (vector bool char);
16650 vector signed short vec_vupkhsb (vector signed char);
16652 vector signed short vec_unpackl (vector signed char);
16653 vector bool short vec_unpackl (vector bool char);
16654 vector unsigned int vec_unpackl (vector pixel);
16655 vector signed int vec_unpackl (vector signed short);
16656 vector bool int vec_unpackl (vector bool short);
16658 vector unsigned int vec_vupklpx (vector pixel);
16660 vector bool int vec_vupklsh (vector bool short);
16661 vector signed int vec_vupklsh (vector signed short);
16663 vector bool short vec_vupklsb (vector bool char);
16664 vector signed short vec_vupklsb (vector signed char);
16666 vector float vec_xor (vector float, vector float);
16667 vector float vec_xor (vector float, vector bool int);
16668 vector float vec_xor (vector bool int, vector float);
16669 vector bool int vec_xor (vector bool int, vector bool int);
16670 vector signed int vec_xor (vector bool int, vector signed int);
16671 vector signed int vec_xor (vector signed int, vector bool int);
16672 vector signed int vec_xor (vector signed int, vector signed int);
16673 vector unsigned int vec_xor (vector bool int, vector unsigned int);
16674 vector unsigned int vec_xor (vector unsigned int, vector bool int);
16675 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
16676 vector bool short vec_xor (vector bool short, vector bool short);
16677 vector signed short vec_xor (vector bool short, vector signed short);
16678 vector signed short vec_xor (vector signed short, vector bool short);
16679 vector signed short vec_xor (vector signed short, vector signed short);
16680 vector unsigned short vec_xor (vector bool short,
16681 vector unsigned short);
16682 vector unsigned short vec_xor (vector unsigned short,
16683 vector bool short);
16684 vector unsigned short vec_xor (vector unsigned short,
16685 vector unsigned short);
16686 vector signed char vec_xor (vector bool char, vector signed char);
16687 vector bool char vec_xor (vector bool char, vector bool char);
16688 vector signed char vec_xor (vector signed char, vector bool char);
16689 vector signed char vec_xor (vector signed char, vector signed char);
16690 vector unsigned char vec_xor (vector bool char, vector unsigned char);
16691 vector unsigned char vec_xor (vector unsigned char, vector bool char);
16692 vector unsigned char vec_xor (vector unsigned char,
16693 vector unsigned char);
16695 int vec_all_eq (vector signed char, vector bool char);
16696 int vec_all_eq (vector signed char, vector signed char);
16697 int vec_all_eq (vector unsigned char, vector bool char);
16698 int vec_all_eq (vector unsigned char, vector unsigned char);
16699 int vec_all_eq (vector bool char, vector bool char);
16700 int vec_all_eq (vector bool char, vector unsigned char);
16701 int vec_all_eq (vector bool char, vector signed char);
16702 int vec_all_eq (vector signed short, vector bool short);
16703 int vec_all_eq (vector signed short, vector signed short);
16704 int vec_all_eq (vector unsigned short, vector bool short);
16705 int vec_all_eq (vector unsigned short, vector unsigned short);
16706 int vec_all_eq (vector bool short, vector bool short);
16707 int vec_all_eq (vector bool short, vector unsigned short);
16708 int vec_all_eq (vector bool short, vector signed short);
16709 int vec_all_eq (vector pixel, vector pixel);
16710 int vec_all_eq (vector signed int, vector bool int);
16711 int vec_all_eq (vector signed int, vector signed int);
16712 int vec_all_eq (vector unsigned int, vector bool int);
16713 int vec_all_eq (vector unsigned int, vector unsigned int);
16714 int vec_all_eq (vector bool int, vector bool int);
16715 int vec_all_eq (vector bool int, vector unsigned int);
16716 int vec_all_eq (vector bool int, vector signed int);
16717 int vec_all_eq (vector float, vector float);
16719 int vec_all_ge (vector bool char, vector unsigned char);
16720 int vec_all_ge (vector unsigned char, vector bool char);
16721 int vec_all_ge (vector unsigned char, vector unsigned char);
16722 int vec_all_ge (vector bool char, vector signed char);
16723 int vec_all_ge (vector signed char, vector bool char);
16724 int vec_all_ge (vector signed char, vector signed char);
16725 int vec_all_ge (vector bool short, vector unsigned short);
16726 int vec_all_ge (vector unsigned short, vector bool short);
16727 int vec_all_ge (vector unsigned short, vector unsigned short);
16728 int vec_all_ge (vector signed short, vector signed short);
16729 int vec_all_ge (vector bool short, vector signed short);
16730 int vec_all_ge (vector signed short, vector bool short);
16731 int vec_all_ge (vector bool int, vector unsigned int);
16732 int vec_all_ge (vector unsigned int, vector bool int);
16733 int vec_all_ge (vector unsigned int, vector unsigned int);
16734 int vec_all_ge (vector bool int, vector signed int);
16735 int vec_all_ge (vector signed int, vector bool int);
16736 int vec_all_ge (vector signed int, vector signed int);
16737 int vec_all_ge (vector float, vector float);
16739 int vec_all_gt (vector bool char, vector unsigned char);
16740 int vec_all_gt (vector unsigned char, vector bool char);
16741 int vec_all_gt (vector unsigned char, vector unsigned char);
16742 int vec_all_gt (vector bool char, vector signed char);
16743 int vec_all_gt (vector signed char, vector bool char);
16744 int vec_all_gt (vector signed char, vector signed char);
16745 int vec_all_gt (vector bool short, vector unsigned short);
16746 int vec_all_gt (vector unsigned short, vector bool short);
16747 int vec_all_gt (vector unsigned short, vector unsigned short);
16748 int vec_all_gt (vector bool short, vector signed short);
16749 int vec_all_gt (vector signed short, vector bool short);
16750 int vec_all_gt (vector signed short, vector signed short);
16751 int vec_all_gt (vector bool int, vector unsigned int);
16752 int vec_all_gt (vector unsigned int, vector bool int);
16753 int vec_all_gt (vector unsigned int, vector unsigned int);
16754 int vec_all_gt (vector bool int, vector signed int);
16755 int vec_all_gt (vector signed int, vector bool int);
16756 int vec_all_gt (vector signed int, vector signed int);
16757 int vec_all_gt (vector float, vector float);
16759 int vec_all_in (vector float, vector float);
16761 int vec_all_le (vector bool char, vector unsigned char);
16762 int vec_all_le (vector unsigned char, vector bool char);
16763 int vec_all_le (vector unsigned char, vector unsigned char);
16764 int vec_all_le (vector bool char, vector signed char);
16765 int vec_all_le (vector signed char, vector bool char);
16766 int vec_all_le (vector signed char, vector signed char);
16767 int vec_all_le (vector bool short, vector unsigned short);
16768 int vec_all_le (vector unsigned short, vector bool short);
16769 int vec_all_le (vector unsigned short, vector unsigned short);
16770 int vec_all_le (vector bool short, vector signed short);
16771 int vec_all_le (vector signed short, vector bool short);
16772 int vec_all_le (vector signed short, vector signed short);
16773 int vec_all_le (vector bool int, vector unsigned int);
16774 int vec_all_le (vector unsigned int, vector bool int);
16775 int vec_all_le (vector unsigned int, vector unsigned int);
16776 int vec_all_le (vector bool int, vector signed int);
16777 int vec_all_le (vector signed int, vector bool int);
16778 int vec_all_le (vector signed int, vector signed int);
16779 int vec_all_le (vector float, vector float);
16781 int vec_all_lt (vector bool char, vector unsigned char);
16782 int vec_all_lt (vector unsigned char, vector bool char);
16783 int vec_all_lt (vector unsigned char, vector unsigned char);
16784 int vec_all_lt (vector bool char, vector signed char);
16785 int vec_all_lt (vector signed char, vector bool char);
16786 int vec_all_lt (vector signed char, vector signed char);
16787 int vec_all_lt (vector bool short, vector unsigned short);
16788 int vec_all_lt (vector unsigned short, vector bool short);
16789 int vec_all_lt (vector unsigned short, vector unsigned short);
16790 int vec_all_lt (vector bool short, vector signed short);
16791 int vec_all_lt (vector signed short, vector bool short);
16792 int vec_all_lt (vector signed short, vector signed short);
16793 int vec_all_lt (vector bool int, vector unsigned int);
16794 int vec_all_lt (vector unsigned int, vector bool int);
16795 int vec_all_lt (vector unsigned int, vector unsigned int);
16796 int vec_all_lt (vector bool int, vector signed int);
16797 int vec_all_lt (vector signed int, vector bool int);
16798 int vec_all_lt (vector signed int, vector signed int);
16799 int vec_all_lt (vector float, vector float);
16801 int vec_all_nan (vector float);
16803 int vec_all_ne (vector signed char, vector bool char);
16804 int vec_all_ne (vector signed char, vector signed char);
16805 int vec_all_ne (vector unsigned char, vector bool char);
16806 int vec_all_ne (vector unsigned char, vector unsigned char);
16807 int vec_all_ne (vector bool char, vector bool char);
16808 int vec_all_ne (vector bool char, vector unsigned char);
16809 int vec_all_ne (vector bool char, vector signed char);
16810 int vec_all_ne (vector signed short, vector bool short);
16811 int vec_all_ne (vector signed short, vector signed short);
16812 int vec_all_ne (vector unsigned short, vector bool short);
16813 int vec_all_ne (vector unsigned short, vector unsigned short);
16814 int vec_all_ne (vector bool short, vector bool short);
16815 int vec_all_ne (vector bool short, vector unsigned short);
16816 int vec_all_ne (vector bool short, vector signed short);
16817 int vec_all_ne (vector pixel, vector pixel);
16818 int vec_all_ne (vector signed int, vector bool int);
16819 int vec_all_ne (vector signed int, vector signed int);
16820 int vec_all_ne (vector unsigned int, vector bool int);
16821 int vec_all_ne (vector unsigned int, vector unsigned int);
16822 int vec_all_ne (vector bool int, vector bool int);
16823 int vec_all_ne (vector bool int, vector unsigned int);
16824 int vec_all_ne (vector bool int, vector signed int);
16825 int vec_all_ne (vector float, vector float);
16827 int vec_all_nge (vector float, vector float);
16829 int vec_all_ngt (vector float, vector float);
16831 int vec_all_nle (vector float, vector float);
16833 int vec_all_nlt (vector float, vector float);
16835 int vec_all_numeric (vector float);
16837 int vec_any_eq (vector signed char, vector bool char);
16838 int vec_any_eq (vector signed char, vector signed char);
16839 int vec_any_eq (vector unsigned char, vector bool char);
16840 int vec_any_eq (vector unsigned char, vector unsigned char);
16841 int vec_any_eq (vector bool char, vector bool char);
16842 int vec_any_eq (vector bool char, vector unsigned char);
16843 int vec_any_eq (vector bool char, vector signed char);
16844 int vec_any_eq (vector signed short, vector bool short);
16845 int vec_any_eq (vector signed short, vector signed short);
16846 int vec_any_eq (vector unsigned short, vector bool short);
16847 int vec_any_eq (vector unsigned short, vector unsigned short);
16848 int vec_any_eq (vector bool short, vector bool short);
16849 int vec_any_eq (vector bool short, vector unsigned short);
16850 int vec_any_eq (vector bool short, vector signed short);
16851 int vec_any_eq (vector pixel, vector pixel);
16852 int vec_any_eq (vector signed int, vector bool int);
16853 int vec_any_eq (vector signed int, vector signed int);
16854 int vec_any_eq (vector unsigned int, vector bool int);
16855 int vec_any_eq (vector unsigned int, vector unsigned int);
16856 int vec_any_eq (vector bool int, vector bool int);
16857 int vec_any_eq (vector bool int, vector unsigned int);
16858 int vec_any_eq (vector bool int, vector signed int);
16859 int vec_any_eq (vector float, vector float);
16861 int vec_any_ge (vector signed char, vector bool char);
16862 int vec_any_ge (vector unsigned char, vector bool char);
16863 int vec_any_ge (vector unsigned char, vector unsigned char);
16864 int vec_any_ge (vector signed char, vector signed char);
16865 int vec_any_ge (vector bool char, vector unsigned char);
16866 int vec_any_ge (vector bool char, vector signed char);
16867 int vec_any_ge (vector unsigned short, vector bool short);
16868 int vec_any_ge (vector unsigned short, vector unsigned short);
16869 int vec_any_ge (vector signed short, vector signed short);
16870 int vec_any_ge (vector signed short, vector bool short);
16871 int vec_any_ge (vector bool short, vector unsigned short);
16872 int vec_any_ge (vector bool short, vector signed short);
16873 int vec_any_ge (vector signed int, vector bool int);
16874 int vec_any_ge (vector unsigned int, vector bool int);
16875 int vec_any_ge (vector unsigned int, vector unsigned int);
16876 int vec_any_ge (vector signed int, vector signed int);
16877 int vec_any_ge (vector bool int, vector unsigned int);
16878 int vec_any_ge (vector bool int, vector signed int);
16879 int vec_any_ge (vector float, vector float);
16881 int vec_any_gt (vector bool char, vector unsigned char);
16882 int vec_any_gt (vector unsigned char, vector bool char);
16883 int vec_any_gt (vector unsigned char, vector unsigned char);
16884 int vec_any_gt (vector bool char, vector signed char);
16885 int vec_any_gt (vector signed char, vector bool char);
16886 int vec_any_gt (vector signed char, vector signed char);
16887 int vec_any_gt (vector bool short, vector unsigned short);
16888 int vec_any_gt (vector unsigned short, vector bool short);
16889 int vec_any_gt (vector unsigned short, vector unsigned short);
16890 int vec_any_gt (vector bool short, vector signed short);
16891 int vec_any_gt (vector signed short, vector bool short);
16892 int vec_any_gt (vector signed short, vector signed short);
16893 int vec_any_gt (vector bool int, vector unsigned int);
16894 int vec_any_gt (vector unsigned int, vector bool int);
16895 int vec_any_gt (vector unsigned int, vector unsigned int);
16896 int vec_any_gt (vector bool int, vector signed int);
16897 int vec_any_gt (vector signed int, vector bool int);
16898 int vec_any_gt (vector signed int, vector signed int);
16899 int vec_any_gt (vector float, vector float);
16901 int vec_any_le (vector bool char, vector unsigned char);
16902 int vec_any_le (vector unsigned char, vector bool char);
16903 int vec_any_le (vector unsigned char, vector unsigned char);
16904 int vec_any_le (vector bool char, vector signed char);
16905 int vec_any_le (vector signed char, vector bool char);
16906 int vec_any_le (vector signed char, vector signed char);
16907 int vec_any_le (vector bool short, vector unsigned short);
16908 int vec_any_le (vector unsigned short, vector bool short);
16909 int vec_any_le (vector unsigned short, vector unsigned short);
16910 int vec_any_le (vector bool short, vector signed short);
16911 int vec_any_le (vector signed short, vector bool short);
16912 int vec_any_le (vector signed short, vector signed short);
16913 int vec_any_le (vector bool int, vector unsigned int);
16914 int vec_any_le (vector unsigned int, vector bool int);
16915 int vec_any_le (vector unsigned int, vector unsigned int);
16916 int vec_any_le (vector bool int, vector signed int);
16917 int vec_any_le (vector signed int, vector bool int);
16918 int vec_any_le (vector signed int, vector signed int);
16919 int vec_any_le (vector float, vector float);
16921 int vec_any_lt (vector bool char, vector unsigned char);
16922 int vec_any_lt (vector unsigned char, vector bool char);
16923 int vec_any_lt (vector unsigned char, vector unsigned char);
16924 int vec_any_lt (vector bool char, vector signed char);
16925 int vec_any_lt (vector signed char, vector bool char);
16926 int vec_any_lt (vector signed char, vector signed char);
16927 int vec_any_lt (vector bool short, vector unsigned short);
16928 int vec_any_lt (vector unsigned short, vector bool short);
16929 int vec_any_lt (vector unsigned short, vector unsigned short);
16930 int vec_any_lt (vector bool short, vector signed short);
16931 int vec_any_lt (vector signed short, vector bool short);
16932 int vec_any_lt (vector signed short, vector signed short);
16933 int vec_any_lt (vector bool int, vector unsigned int);
16934 int vec_any_lt (vector unsigned int, vector bool int);
16935 int vec_any_lt (vector unsigned int, vector unsigned int);
16936 int vec_any_lt (vector bool int, vector signed int);
16937 int vec_any_lt (vector signed int, vector bool int);
16938 int vec_any_lt (vector signed int, vector signed int);
16939 int vec_any_lt (vector float, vector float);
16941 int vec_any_nan (vector float);
16943 int vec_any_ne (vector signed char, vector bool char);
16944 int vec_any_ne (vector signed char, vector signed char);
16945 int vec_any_ne (vector unsigned char, vector bool char);
16946 int vec_any_ne (vector unsigned char, vector unsigned char);
16947 int vec_any_ne (vector bool char, vector bool char);
16948 int vec_any_ne (vector bool char, vector unsigned char);
16949 int vec_any_ne (vector bool char, vector signed char);
16950 int vec_any_ne (vector signed short, vector bool short);
16951 int vec_any_ne (vector signed short, vector signed short);
16952 int vec_any_ne (vector unsigned short, vector bool short);
16953 int vec_any_ne (vector unsigned short, vector unsigned short);
16954 int vec_any_ne (vector bool short, vector bool short);
16955 int vec_any_ne (vector bool short, vector unsigned short);
16956 int vec_any_ne (vector bool short, vector signed short);
16957 int vec_any_ne (vector pixel, vector pixel);
16958 int vec_any_ne (vector signed int, vector bool int);
16959 int vec_any_ne (vector signed int, vector signed int);
16960 int vec_any_ne (vector unsigned int, vector bool int);
16961 int vec_any_ne (vector unsigned int, vector unsigned int);
16962 int vec_any_ne (vector bool int, vector bool int);
16963 int vec_any_ne (vector bool int, vector unsigned int);
16964 int vec_any_ne (vector bool int, vector signed int);
16965 int vec_any_ne (vector float, vector float);
16967 int vec_any_nge (vector float, vector float);
16969 int vec_any_ngt (vector float, vector float);
16971 int vec_any_nle (vector float, vector float);
16973 int vec_any_nlt (vector float, vector float);
16975 int vec_any_numeric (vector float);
16977 int vec_any_out (vector float, vector float);
16980 If the vector/scalar (VSX) instruction set is available, the following
16981 additional functions are available:
16984 vector double vec_abs (vector double);
16985 vector double vec_add (vector double, vector double);
16986 vector double vec_and (vector double, vector double);
16987 vector double vec_and (vector double, vector bool long);
16988 vector double vec_and (vector bool long, vector double);
16989 vector long vec_and (vector long, vector long);
16990 vector long vec_and (vector long, vector bool long);
16991 vector long vec_and (vector bool long, vector long);
16992 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
16993 vector unsigned long vec_and (vector unsigned long, vector bool long);
16994 vector unsigned long vec_and (vector bool long, vector unsigned long);
16995 vector double vec_andc (vector double, vector double);
16996 vector double vec_andc (vector double, vector bool long);
16997 vector double vec_andc (vector bool long, vector double);
16998 vector long vec_andc (vector long, vector long);
16999 vector long vec_andc (vector long, vector bool long);
17000 vector long vec_andc (vector bool long, vector long);
17001 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
17002 vector unsigned long vec_andc (vector unsigned long, vector bool long);
17003 vector unsigned long vec_andc (vector bool long, vector unsigned long);
17004 vector double vec_ceil (vector double);
17005 vector bool long vec_cmpeq (vector double, vector double);
17006 vector bool long vec_cmpge (vector double, vector double);
17007 vector bool long vec_cmpgt (vector double, vector double);
17008 vector bool long vec_cmple (vector double, vector double);
17009 vector bool long vec_cmplt (vector double, vector double);
17010 vector double vec_cpsgn (vector double, vector double);
17011 vector float vec_div (vector float, vector float);
17012 vector double vec_div (vector double, vector double);
17013 vector long vec_div (vector long, vector long);
17014 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
17015 vector double vec_floor (vector double);
17016 vector double vec_ld (int, const vector double *);
17017 vector double vec_ld (int, const double *);
17018 vector double vec_ldl (int, const vector double *);
17019 vector double vec_ldl (int, const double *);
17020 vector unsigned char vec_lvsl (int, const volatile double *);
17021 vector unsigned char vec_lvsr (int, const volatile double *);
17022 vector double vec_madd (vector double, vector double, vector double);
17023 vector double vec_max (vector double, vector double);
17024 vector signed long vec_mergeh (vector signed long, vector signed long);
17025 vector signed long vec_mergeh (vector signed long, vector bool long);
17026 vector signed long vec_mergeh (vector bool long, vector signed long);
17027 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
17028 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
17029 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
17030 vector signed long vec_mergel (vector signed long, vector signed long);
17031 vector signed long vec_mergel (vector signed long, vector bool long);
17032 vector signed long vec_mergel (vector bool long, vector signed long);
17033 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
17034 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
17035 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
17036 vector double vec_min (vector double, vector double);
17037 vector float vec_msub (vector float, vector float, vector float);
17038 vector double vec_msub (vector double, vector double, vector double);
17039 vector float vec_mul (vector float, vector float);
17040 vector double vec_mul (vector double, vector double);
17041 vector long vec_mul (vector long, vector long);
17042 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
17043 vector float vec_nearbyint (vector float);
17044 vector double vec_nearbyint (vector double);
17045 vector float vec_nmadd (vector float, vector float, vector float);
17046 vector double vec_nmadd (vector double, vector double, vector double);
17047 vector double vec_nmsub (vector double, vector double, vector double);
17048 vector double vec_nor (vector double, vector double);
17049 vector long vec_nor (vector long, vector long);
17050 vector long vec_nor (vector long, vector bool long);
17051 vector long vec_nor (vector bool long, vector long);
17052 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
17053 vector unsigned long vec_nor (vector unsigned long, vector bool long);
17054 vector unsigned long vec_nor (vector bool long, vector unsigned long);
17055 vector double vec_or (vector double, vector double);
17056 vector double vec_or (vector double, vector bool long);
17057 vector double vec_or (vector bool long, vector double);
17058 vector long vec_or (vector long, vector long);
17059 vector long vec_or (vector long, vector bool long);
17060 vector long vec_or (vector bool long, vector long);
17061 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
17062 vector unsigned long vec_or (vector unsigned long, vector bool long);
17063 vector unsigned long vec_or (vector bool long, vector unsigned long);
17064 vector double vec_perm (vector double, vector double, vector unsigned char);
17065 vector long vec_perm (vector long, vector long, vector unsigned char);
17066 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
17067 vector unsigned char);
17068 vector double vec_rint (vector double);
17069 vector double vec_recip (vector double, vector double);
17070 vector double vec_rsqrt (vector double);
17071 vector double vec_rsqrte (vector double);
17072 vector double vec_sel (vector double, vector double, vector bool long);
17073 vector double vec_sel (vector double, vector double, vector unsigned long);
17074 vector long vec_sel (vector long, vector long, vector long);
17075 vector long vec_sel (vector long, vector long, vector unsigned long);
17076 vector long vec_sel (vector long, vector long, vector bool long);
17077 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17079 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17080 vector unsigned long);
17081 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17083 vector double vec_splats (double);
17084 vector signed long vec_splats (signed long);
17085 vector unsigned long vec_splats (unsigned long);
17086 vector float vec_sqrt (vector float);
17087 vector double vec_sqrt (vector double);
17088 void vec_st (vector double, int, vector double *);
17089 void vec_st (vector double, int, double *);
17090 vector double vec_sub (vector double, vector double);
17091 vector double vec_trunc (vector double);
17092 vector double vec_xl (int, vector double *);
17093 vector double vec_xl (int, double *);
17094 vector long long vec_xl (int, vector long long *);
17095 vector long long vec_xl (int, long long *);
17096 vector unsigned long long vec_xl (int, vector unsigned long long *);
17097 vector unsigned long long vec_xl (int, unsigned long long *);
17098 vector float vec_xl (int, vector float *);
17099 vector float vec_xl (int, float *);
17100 vector int vec_xl (int, vector int *);
17101 vector int vec_xl (int, int *);
17102 vector unsigned int vec_xl (int, vector unsigned int *);
17103 vector unsigned int vec_xl (int, unsigned int *);
17104 vector double vec_xor (vector double, vector double);
17105 vector double vec_xor (vector double, vector bool long);
17106 vector double vec_xor (vector bool long, vector double);
17107 vector long vec_xor (vector long, vector long);
17108 vector long vec_xor (vector long, vector bool long);
17109 vector long vec_xor (vector bool long, vector long);
17110 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
17111 vector unsigned long vec_xor (vector unsigned long, vector bool long);
17112 vector unsigned long vec_xor (vector bool long, vector unsigned long);
17113 void vec_xst (vector double, int, vector double *);
17114 void vec_xst (vector double, int, double *);
17115 void vec_xst (vector long long, int, vector long long *);
17116 void vec_xst (vector long long, int, long long *);
17117 void vec_xst (vector unsigned long long, int, vector unsigned long long *);
17118 void vec_xst (vector unsigned long long, int, unsigned long long *);
17119 void vec_xst (vector float, int, vector float *);
17120 void vec_xst (vector float, int, float *);
17121 void vec_xst (vector int, int, vector int *);
17122 void vec_xst (vector int, int, int *);
17123 void vec_xst (vector unsigned int, int, vector unsigned int *);
17124 void vec_xst (vector unsigned int, int, unsigned int *);
17125 int vec_all_eq (vector double, vector double);
17126 int vec_all_ge (vector double, vector double);
17127 int vec_all_gt (vector double, vector double);
17128 int vec_all_le (vector double, vector double);
17129 int vec_all_lt (vector double, vector double);
17130 int vec_all_nan (vector double);
17131 int vec_all_ne (vector double, vector double);
17132 int vec_all_nge (vector double, vector double);
17133 int vec_all_ngt (vector double, vector double);
17134 int vec_all_nle (vector double, vector double);
17135 int vec_all_nlt (vector double, vector double);
17136 int vec_all_numeric (vector double);
17137 int vec_any_eq (vector double, vector double);
17138 int vec_any_ge (vector double, vector double);
17139 int vec_any_gt (vector double, vector double);
17140 int vec_any_le (vector double, vector double);
17141 int vec_any_lt (vector double, vector double);
17142 int vec_any_nan (vector double);
17143 int vec_any_ne (vector double, vector double);
17144 int vec_any_nge (vector double, vector double);
17145 int vec_any_ngt (vector double, vector double);
17146 int vec_any_nle (vector double, vector double);
17147 int vec_any_nlt (vector double, vector double);
17148 int vec_any_numeric (vector double);
17150 vector double vec_vsx_ld (int, const vector double *);
17151 vector double vec_vsx_ld (int, const double *);
17152 vector float vec_vsx_ld (int, const vector float *);
17153 vector float vec_vsx_ld (int, const float *);
17154 vector bool int vec_vsx_ld (int, const vector bool int *);
17155 vector signed int vec_vsx_ld (int, const vector signed int *);
17156 vector signed int vec_vsx_ld (int, const int *);
17157 vector signed int vec_vsx_ld (int, const long *);
17158 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
17159 vector unsigned int vec_vsx_ld (int, const unsigned int *);
17160 vector unsigned int vec_vsx_ld (int, const unsigned long *);
17161 vector bool short vec_vsx_ld (int, const vector bool short *);
17162 vector pixel vec_vsx_ld (int, const vector pixel *);
17163 vector signed short vec_vsx_ld (int, const vector signed short *);
17164 vector signed short vec_vsx_ld (int, const short *);
17165 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
17166 vector unsigned short vec_vsx_ld (int, const unsigned short *);
17167 vector bool char vec_vsx_ld (int, const vector bool char *);
17168 vector signed char vec_vsx_ld (int, const vector signed char *);
17169 vector signed char vec_vsx_ld (int, const signed char *);
17170 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
17171 vector unsigned char vec_vsx_ld (int, const unsigned char *);
17173 void vec_vsx_st (vector double, int, vector double *);
17174 void vec_vsx_st (vector double, int, double *);
17175 void vec_vsx_st (vector float, int, vector float *);
17176 void vec_vsx_st (vector float, int, float *);
17177 void vec_vsx_st (vector signed int, int, vector signed int *);
17178 void vec_vsx_st (vector signed int, int, int *);
17179 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
17180 void vec_vsx_st (vector unsigned int, int, unsigned int *);
17181 void vec_vsx_st (vector bool int, int, vector bool int *);
17182 void vec_vsx_st (vector bool int, int, unsigned int *);
17183 void vec_vsx_st (vector bool int, int, int *);
17184 void vec_vsx_st (vector signed short, int, vector signed short *);
17185 void vec_vsx_st (vector signed short, int, short *);
17186 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
17187 void vec_vsx_st (vector unsigned short, int, unsigned short *);
17188 void vec_vsx_st (vector bool short, int, vector bool short *);
17189 void vec_vsx_st (vector bool short, int, unsigned short *);
17190 void vec_vsx_st (vector pixel, int, vector pixel *);
17191 void vec_vsx_st (vector pixel, int, unsigned short *);
17192 void vec_vsx_st (vector pixel, int, short *);
17193 void vec_vsx_st (vector bool short, int, short *);
17194 void vec_vsx_st (vector signed char, int, vector signed char *);
17195 void vec_vsx_st (vector signed char, int, signed char *);
17196 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
17197 void vec_vsx_st (vector unsigned char, int, unsigned char *);
17198 void vec_vsx_st (vector bool char, int, vector bool char *);
17199 void vec_vsx_st (vector bool char, int, unsigned char *);
17200 void vec_vsx_st (vector bool char, int, signed char *);
17202 vector double vec_xxpermdi (vector double, vector double, int);
17203 vector float vec_xxpermdi (vector float, vector float, int);
17204 vector long long vec_xxpermdi (vector long long, vector long long, int);
17205 vector unsigned long long vec_xxpermdi (vector unsigned long long,
17206 vector unsigned long long, int);
17207 vector int vec_xxpermdi (vector int, vector int, int);
17208 vector unsigned int vec_xxpermdi (vector unsigned int,
17209 vector unsigned int, int);
17210 vector short vec_xxpermdi (vector short, vector short, int);
17211 vector unsigned short vec_xxpermdi (vector unsigned short,
17212 vector unsigned short, int);
17213 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
17214 vector unsigned char vec_xxpermdi (vector unsigned char,
17215 vector unsigned char, int);
17217 vector double vec_xxsldi (vector double, vector double, int);
17218 vector float vec_xxsldi (vector float, vector float, int);
17219 vector long long vec_xxsldi (vector long long, vector long long, int);
17220 vector unsigned long long vec_xxsldi (vector unsigned long long,
17221 vector unsigned long long, int);
17222 vector int vec_xxsldi (vector int, vector int, int);
17223 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
17224 vector short vec_xxsldi (vector short, vector short, int);
17225 vector unsigned short vec_xxsldi (vector unsigned short,
17226 vector unsigned short, int);
17227 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
17228 vector unsigned char vec_xxsldi (vector unsigned char,
17229 vector unsigned char, int);
17232 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
17233 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
17234 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
17235 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
17236 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
17238 If the ISA 2.07 additions to the vector/scalar (power8-vector)
17239 instruction set are available, the following additional functions are
17240 available for both 32-bit and 64-bit targets. For 64-bit targets, you
17241 can use @var{vector long} instead of @var{vector long long},
17242 @var{vector bool long} instead of @var{vector bool long long}, and
17243 @var{vector unsigned long} instead of @var{vector unsigned long long}.
17246 vector long long vec_abs (vector long long);
17248 vector long long vec_add (vector long long, vector long long);
17249 vector unsigned long long vec_add (vector unsigned long long,
17250 vector unsigned long long);
17252 int vec_all_eq (vector long long, vector long long);
17253 int vec_all_eq (vector unsigned long long, vector unsigned long long);
17254 int vec_all_ge (vector long long, vector long long);
17255 int vec_all_ge (vector unsigned long long, vector unsigned long long);
17256 int vec_all_gt (vector long long, vector long long);
17257 int vec_all_gt (vector unsigned long long, vector unsigned long long);
17258 int vec_all_le (vector long long, vector long long);
17259 int vec_all_le (vector unsigned long long, vector unsigned long long);
17260 int vec_all_lt (vector long long, vector long long);
17261 int vec_all_lt (vector unsigned long long, vector unsigned long long);
17262 int vec_all_ne (vector long long, vector long long);
17263 int vec_all_ne (vector unsigned long long, vector unsigned long long);
17265 int vec_any_eq (vector long long, vector long long);
17266 int vec_any_eq (vector unsigned long long, vector unsigned long long);
17267 int vec_any_ge (vector long long, vector long long);
17268 int vec_any_ge (vector unsigned long long, vector unsigned long long);
17269 int vec_any_gt (vector long long, vector long long);
17270 int vec_any_gt (vector unsigned long long, vector unsigned long long);
17271 int vec_any_le (vector long long, vector long long);
17272 int vec_any_le (vector unsigned long long, vector unsigned long long);
17273 int vec_any_lt (vector long long, vector long long);
17274 int vec_any_lt (vector unsigned long long, vector unsigned long long);
17275 int vec_any_ne (vector long long, vector long long);
17276 int vec_any_ne (vector unsigned long long, vector unsigned long long);
17278 vector long long vec_eqv (vector long long, vector long long);
17279 vector long long vec_eqv (vector bool long long, vector long long);
17280 vector long long vec_eqv (vector long long, vector bool long long);
17281 vector unsigned long long vec_eqv (vector unsigned long long,
17282 vector unsigned long long);
17283 vector unsigned long long vec_eqv (vector bool long long,
17284 vector unsigned long long);
17285 vector unsigned long long vec_eqv (vector unsigned long long,
17286 vector bool long long);
17287 vector int vec_eqv (vector int, vector int);
17288 vector int vec_eqv (vector bool int, vector int);
17289 vector int vec_eqv (vector int, vector bool int);
17290 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
17291 vector unsigned int vec_eqv (vector bool unsigned int,
17292 vector unsigned int);
17293 vector unsigned int vec_eqv (vector unsigned int,
17294 vector bool unsigned int);
17295 vector short vec_eqv (vector short, vector short);
17296 vector short vec_eqv (vector bool short, vector short);
17297 vector short vec_eqv (vector short, vector bool short);
17298 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
17299 vector unsigned short vec_eqv (vector bool unsigned short,
17300 vector unsigned short);
17301 vector unsigned short vec_eqv (vector unsigned short,
17302 vector bool unsigned short);
17303 vector signed char vec_eqv (vector signed char, vector signed char);
17304 vector signed char vec_eqv (vector bool signed char, vector signed char);
17305 vector signed char vec_eqv (vector signed char, vector bool signed char);
17306 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
17307 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
17308 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
17310 vector long long vec_max (vector long long, vector long long);
17311 vector unsigned long long vec_max (vector unsigned long long,
17312 vector unsigned long long);
17314 vector signed int vec_mergee (vector signed int, vector signed int);
17315 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
17316 vector bool int vec_mergee (vector bool int, vector bool int);
17318 vector signed int vec_mergeo (vector signed int, vector signed int);
17319 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
17320 vector bool int vec_mergeo (vector bool int, vector bool int);
17322 vector long long vec_min (vector long long, vector long long);
17323 vector unsigned long long vec_min (vector unsigned long long,
17324 vector unsigned long long);
17326 vector long long vec_nand (vector long long, vector long long);
17327 vector long long vec_nand (vector bool long long, vector long long);
17328 vector long long vec_nand (vector long long, vector bool long long);
17329 vector unsigned long long vec_nand (vector unsigned long long,
17330 vector unsigned long long);
17331 vector unsigned long long vec_nand (vector bool long long,
17332 vector unsigned long long);
17333 vector unsigned long long vec_nand (vector unsigned long long,
17334 vector bool long long);
17335 vector int vec_nand (vector int, vector int);
17336 vector int vec_nand (vector bool int, vector int);
17337 vector int vec_nand (vector int, vector bool int);
17338 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
17339 vector unsigned int vec_nand (vector bool unsigned int,
17340 vector unsigned int);
17341 vector unsigned int vec_nand (vector unsigned int,
17342 vector bool unsigned int);
17343 vector short vec_nand (vector short, vector short);
17344 vector short vec_nand (vector bool short, vector short);
17345 vector short vec_nand (vector short, vector bool short);
17346 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
17347 vector unsigned short vec_nand (vector bool unsigned short,
17348 vector unsigned short);
17349 vector unsigned short vec_nand (vector unsigned short,
17350 vector bool unsigned short);
17351 vector signed char vec_nand (vector signed char, vector signed char);
17352 vector signed char vec_nand (vector bool signed char, vector signed char);
17353 vector signed char vec_nand (vector signed char, vector bool signed char);
17354 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
17355 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
17356 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
17358 vector long long vec_orc (vector long long, vector long long);
17359 vector long long vec_orc (vector bool long long, vector long long);
17360 vector long long vec_orc (vector long long, vector bool long long);
17361 vector unsigned long long vec_orc (vector unsigned long long,
17362 vector unsigned long long);
17363 vector unsigned long long vec_orc (vector bool long long,
17364 vector unsigned long long);
17365 vector unsigned long long vec_orc (vector unsigned long long,
17366 vector bool long long);
17367 vector int vec_orc (vector int, vector int);
17368 vector int vec_orc (vector bool int, vector int);
17369 vector int vec_orc (vector int, vector bool int);
17370 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
17371 vector unsigned int vec_orc (vector bool unsigned int,
17372 vector unsigned int);
17373 vector unsigned int vec_orc (vector unsigned int,
17374 vector bool unsigned int);
17375 vector short vec_orc (vector short, vector short);
17376 vector short vec_orc (vector bool short, vector short);
17377 vector short vec_orc (vector short, vector bool short);
17378 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
17379 vector unsigned short vec_orc (vector bool unsigned short,
17380 vector unsigned short);
17381 vector unsigned short vec_orc (vector unsigned short,
17382 vector bool unsigned short);
17383 vector signed char vec_orc (vector signed char, vector signed char);
17384 vector signed char vec_orc (vector bool signed char, vector signed char);
17385 vector signed char vec_orc (vector signed char, vector bool signed char);
17386 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
17387 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
17388 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
17390 vector int vec_pack (vector long long, vector long long);
17391 vector unsigned int vec_pack (vector unsigned long long,
17392 vector unsigned long long);
17393 vector bool int vec_pack (vector bool long long, vector bool long long);
17395 vector int vec_packs (vector long long, vector long long);
17396 vector unsigned int vec_packs (vector unsigned long long,
17397 vector unsigned long long);
17399 vector unsigned int vec_packsu (vector long long, vector long long);
17400 vector unsigned int vec_packsu (vector unsigned long long,
17401 vector unsigned long long);
17403 vector long long vec_rl (vector long long,
17404 vector unsigned long long);
17405 vector long long vec_rl (vector unsigned long long,
17406 vector unsigned long long);
17408 vector long long vec_sl (vector long long, vector unsigned long long);
17409 vector long long vec_sl (vector unsigned long long,
17410 vector unsigned long long);
17412 vector long long vec_sr (vector long long, vector unsigned long long);
17413 vector unsigned long long char vec_sr (vector unsigned long long,
17414 vector unsigned long long);
17416 vector long long vec_sra (vector long long, vector unsigned long long);
17417 vector unsigned long long vec_sra (vector unsigned long long,
17418 vector unsigned long long);
17420 vector long long vec_sub (vector long long, vector long long);
17421 vector unsigned long long vec_sub (vector unsigned long long,
17422 vector unsigned long long);
17424 vector long long vec_unpackh (vector int);
17425 vector unsigned long long vec_unpackh (vector unsigned int);
17427 vector long long vec_unpackl (vector int);
17428 vector unsigned long long vec_unpackl (vector unsigned int);
17430 vector long long vec_vaddudm (vector long long, vector long long);
17431 vector long long vec_vaddudm (vector bool long long, vector long long);
17432 vector long long vec_vaddudm (vector long long, vector bool long long);
17433 vector unsigned long long vec_vaddudm (vector unsigned long long,
17434 vector unsigned long long);
17435 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
17436 vector unsigned long long);
17437 vector unsigned long long vec_vaddudm (vector unsigned long long,
17438 vector bool unsigned long long);
17440 vector long long vec_vbpermq (vector signed char, vector signed char);
17441 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
17443 vector long long vec_cntlz (vector long long);
17444 vector unsigned long long vec_cntlz (vector unsigned long long);
17445 vector int vec_cntlz (vector int);
17446 vector unsigned int vec_cntlz (vector int);
17447 vector short vec_cntlz (vector short);
17448 vector unsigned short vec_cntlz (vector unsigned short);
17449 vector signed char vec_cntlz (vector signed char);
17450 vector unsigned char vec_cntlz (vector unsigned char);
17452 vector long long vec_vclz (vector long long);
17453 vector unsigned long long vec_vclz (vector unsigned long long);
17454 vector int vec_vclz (vector int);
17455 vector unsigned int vec_vclz (vector int);
17456 vector short vec_vclz (vector short);
17457 vector unsigned short vec_vclz (vector unsigned short);
17458 vector signed char vec_vclz (vector signed char);
17459 vector unsigned char vec_vclz (vector unsigned char);
17461 vector signed char vec_vclzb (vector signed char);
17462 vector unsigned char vec_vclzb (vector unsigned char);
17464 vector long long vec_vclzd (vector long long);
17465 vector unsigned long long vec_vclzd (vector unsigned long long);
17467 vector short vec_vclzh (vector short);
17468 vector unsigned short vec_vclzh (vector unsigned short);
17470 vector int vec_vclzw (vector int);
17471 vector unsigned int vec_vclzw (vector int);
17473 vector signed char vec_vgbbd (vector signed char);
17474 vector unsigned char vec_vgbbd (vector unsigned char);
17476 vector long long vec_vmaxsd (vector long long, vector long long);
17478 vector unsigned long long vec_vmaxud (vector unsigned long long,
17479 unsigned vector long long);
17481 vector long long vec_vminsd (vector long long, vector long long);
17483 vector unsigned long long vec_vminud (vector long long,
17486 vector int vec_vpksdss (vector long long, vector long long);
17487 vector unsigned int vec_vpksdss (vector long long, vector long long);
17489 vector unsigned int vec_vpkudus (vector unsigned long long,
17490 vector unsigned long long);
17492 vector int vec_vpkudum (vector long long, vector long long);
17493 vector unsigned int vec_vpkudum (vector unsigned long long,
17494 vector unsigned long long);
17495 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
17497 vector long long vec_vpopcnt (vector long long);
17498 vector unsigned long long vec_vpopcnt (vector unsigned long long);
17499 vector int vec_vpopcnt (vector int);
17500 vector unsigned int vec_vpopcnt (vector int);
17501 vector short vec_vpopcnt (vector short);
17502 vector unsigned short vec_vpopcnt (vector unsigned short);
17503 vector signed char vec_vpopcnt (vector signed char);
17504 vector unsigned char vec_vpopcnt (vector unsigned char);
17506 vector signed char vec_vpopcntb (vector signed char);
17507 vector unsigned char vec_vpopcntb (vector unsigned char);
17509 vector long long vec_vpopcntd (vector long long);
17510 vector unsigned long long vec_vpopcntd (vector unsigned long long);
17512 vector short vec_vpopcnth (vector short);
17513 vector unsigned short vec_vpopcnth (vector unsigned short);
17515 vector int vec_vpopcntw (vector int);
17516 vector unsigned int vec_vpopcntw (vector int);
17518 vector long long vec_vrld (vector long long, vector unsigned long long);
17519 vector unsigned long long vec_vrld (vector unsigned long long,
17520 vector unsigned long long);
17522 vector long long vec_vsld (vector long long, vector unsigned long long);
17523 vector long long vec_vsld (vector unsigned long long,
17524 vector unsigned long long);
17526 vector long long vec_vsrad (vector long long, vector unsigned long long);
17527 vector unsigned long long vec_vsrad (vector unsigned long long,
17528 vector unsigned long long);
17530 vector long long vec_vsrd (vector long long, vector unsigned long long);
17531 vector unsigned long long char vec_vsrd (vector unsigned long long,
17532 vector unsigned long long);
17534 vector long long vec_vsubudm (vector long long, vector long long);
17535 vector long long vec_vsubudm (vector bool long long, vector long long);
17536 vector long long vec_vsubudm (vector long long, vector bool long long);
17537 vector unsigned long long vec_vsubudm (vector unsigned long long,
17538 vector unsigned long long);
17539 vector unsigned long long vec_vsubudm (vector bool long long,
17540 vector unsigned long long);
17541 vector unsigned long long vec_vsubudm (vector unsigned long long,
17542 vector bool long long);
17544 vector long long vec_vupkhsw (vector int);
17545 vector unsigned long long vec_vupkhsw (vector unsigned int);
17547 vector long long vec_vupklsw (vector int);
17548 vector unsigned long long vec_vupklsw (vector int);
17551 If the ISA 2.07 additions to the vector/scalar (power8-vector)
17552 instruction set are available, the following additional functions are
17553 available for 64-bit targets. New vector types
17554 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
17555 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
17558 The normal vector extract, and set operations work on
17559 @var{vector __int128_t} and @var{vector __uint128_t} types,
17560 but the index value must be 0.
17563 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
17564 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
17566 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
17567 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
17569 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
17570 vector __int128_t);
17571 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
17572 vector __uint128_t);
17574 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
17575 vector __int128_t);
17576 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
17577 vector __uint128_t);
17579 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
17580 vector __int128_t);
17581 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
17582 vector __uint128_t);
17584 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
17585 vector __int128_t);
17586 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
17587 vector __uint128_t);
17589 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
17590 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
17592 __int128_t vec_vsubuqm (__int128_t, __int128_t);
17593 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
17595 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
17596 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
17597 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
17598 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
17599 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
17600 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
17601 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
17602 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
17603 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
17604 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
17607 If the ISA 3.0 instruction set additions (@option{-mcpu=power9})
17611 vector long long vec_vctz (vector long long);
17612 vector unsigned long long vec_vctz (vector unsigned long long);
17613 vector int vec_vctz (vector int);
17614 vector unsigned int vec_vctz (vector int);
17615 vector short vec_vctz (vector short);
17616 vector unsigned short vec_vctz (vector unsigned short);
17617 vector signed char vec_vctz (vector signed char);
17618 vector unsigned char vec_vctz (vector unsigned char);
17620 vector signed char vec_vctzb (vector signed char);
17621 vector unsigned char vec_vctzb (vector unsigned char);
17623 vector long long vec_vctzd (vector long long);
17624 vector unsigned long long vec_vctzd (vector unsigned long long);
17626 vector short vec_vctzh (vector short);
17627 vector unsigned short vec_vctzh (vector unsigned short);
17629 vector int vec_vctzw (vector int);
17630 vector unsigned int vec_vctzw (vector int);
17632 vector int vec_vprtyb (vector int);
17633 vector unsigned int vec_vprtyb (vector unsigned int);
17634 vector long long vec_vprtyb (vector long long);
17635 vector unsigned long long vec_vprtyb (vector unsigned long long);
17637 vector int vec_vprtybw (vector int);
17638 vector unsigned int vec_vprtybw (vector unsigned int);
17640 vector long long vec_vprtybd (vector long long);
17641 vector unsigned long long vec_vprtybd (vector unsigned long long);
17644 On 64-bit targets, if the ISA 3.0 additions (@option{-mcpu=power9})
17648 vector long vec_vprtyb (vector long);
17649 vector unsigned long vec_vprtyb (vector unsigned long);
17650 vector __int128_t vec_vprtyb (vector __int128_t);
17651 vector __uint128_t vec_vprtyb (vector __uint128_t);
17653 vector long vec_vprtybd (vector long);
17654 vector unsigned long vec_vprtybd (vector unsigned long);
17656 vector __int128_t vec_vprtybq (vector __int128_t);
17657 vector __uint128_t vec_vprtybd (vector __uint128_t);
17660 The following built-in vector functions are available for the PowerPC family
17661 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
17663 __vector unsigned char
17664 vec_slv (__vector unsigned char src, __vector unsigned char shift_distance);
17665 __vector unsigned char
17666 vec_srv (__vector unsigned char src, __vector unsigned char shift_distance);
17669 The @code{vec_slv} and @code{vec_srv} functions operate on
17670 all of the bytes of their @code{src} and @code{shift_distance}
17671 arguments in parallel. The behavior of the @code{vec_slv} is as if
17672 there existed a temporary array of 17 unsigned characters
17673 @code{slv_array} within which elements 0 through 15 are the same as
17674 the entries in the @code{src} array and element 16 equals 0. The
17675 result returned from the @code{vec_slv} function is a
17676 @code{__vector} of 16 unsigned characters within which element
17677 @code{i} is computed using the C expression
17678 @code{0xff & (*((unsigned short *)(slv_array + i)) << (0x07 &
17679 shift_distance[i]))},
17680 with this resulting value coerced to the @code{unsigned char} type.
17681 The behavior of the @code{vec_srv} is as if
17682 there existed a temporary array of 17 unsigned characters
17683 @code{srv_array} within which element 0 equals zero and
17684 elements 1 through 16 equal the elements 0 through 15 of
17685 the @code{src} array. The
17686 result returned from the @code{vec_srv} function is a
17687 @code{__vector} of 16 unsigned characters within which element
17688 @code{i} is computed using the C expression
17689 @code{0xff & (*((unsigned short *)(srv_array + i)) >>
17690 (0x07 & shift_distance[i]))},
17691 with this resulting value coerced to the @code{unsigned char} type.
17693 The following built-in functions are available for the PowerPC family
17694 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
17696 __vector unsigned char
17697 vec_absd (__vector unsigned char arg1, __vector unsigned char arg2);
17698 __vector unsigned short
17699 vec_absd (__vector unsigned short arg1, __vector unsigned short arg2);
17700 __vector unsigned int
17701 vec_absd (__vector unsigned int arg1, __vector unsigned int arg2);
17703 __vector unsigned char
17704 vec_absdb (__vector unsigned char arg1, __vector unsigned char arg2);
17705 __vector unsigned short
17706 vec_absdh (__vector unsigned short arg1, __vector unsigned short arg2);
17707 __vector unsigned int
17708 vec_absdw (__vector unsigned int arg1, __vector unsigned int arg2);
17711 The @code{vec_absd}, @code{vec_absdb}, @code{vec_absdh}, and
17712 @code{vec_absdw} built-in functions each computes the absolute
17713 differences of the pairs of vector elements supplied in its two vector
17714 arguments, placing the absolute differences into the corresponding
17715 elements of the vector result.
17717 The following built-in functions are available for the PowerPC family
17718 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
17721 vec_extract_exp (__vector float source);
17722 __vector long long int
17723 vec_extract_exp (__vector double source);
17726 vec_extract_sig (__vector float source);
17727 __vector long long int
17728 vec_extract_sig (__vector double source);
17731 vec_insert_exp (__vector unsigned int significands, __vector unsigned int exponents);
17733 vec_insert_exp (__vector unsigned long long int significands,
17734 __vector unsigned long long int exponents);
17736 __vector int vec_test_data_class (__vector float source, unsigned int condition);
17737 __vector long long int vec_test_data_class (__vector double source, unsigned int condition);
17740 The @code{vec_extract_sig} and @code{vec_extract_exp} built-in
17741 functions return vectors representing the significands and exponents
17742 of their @code{source} arguments respectively. The
17743 @code{vec_insert_exp} built-in functions return a vector of single- or
17744 double-precision floating
17745 point values constructed by assembling the values of their
17746 @code{significands} and @code{exponents} arguments into the
17747 corresponding elements of the returned vector. The sign of each
17748 element of the result is copied from the most significant bit of the
17749 corresponding entry within the @code{significands} argument. The
17750 significand and exponent components of each element of the result are
17751 composed of the least significant bits of the corresponding
17752 @code{significands} element and the least significant bits of the
17753 corresponding @code{exponents} element.
17755 The @code{vec_test_data_class} built-in function returns a vector
17756 representing the results of testing the @code{source} vector for the
17757 condition selected by the @code{condition} argument. The
17758 @code{condition} argument must be an unsigned integer with value not
17760 @code{condition} argument is encoded as a bitmask with each bit
17761 enabling the testing of a different condition, as characterized by the
17765 0x20 Test for +Infinity
17766 0x10 Test for -Infinity
17767 0x08 Test for +Zero
17768 0x04 Test for -Zero
17769 0x02 Test for +Denormal
17770 0x01 Test for -Denormal
17773 If any of the enabled test conditions is true, the corresponding entry
17774 in the result vector is -1. Otherwise (all of the enabled test
17775 conditions are false), the corresponding entry of the result vector is 0.
17777 If the cryptographic instructions are enabled (@option{-mcrypto} or
17778 @option{-mcpu=power8}), the following builtins are enabled.
17781 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
17783 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
17784 vector unsigned long long);
17786 vector unsigned long long __builtin_crypto_vcipherlast
17787 (vector unsigned long long,
17788 vector unsigned long long);
17790 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
17791 vector unsigned long long);
17793 vector unsigned long long __builtin_crypto_vncipherlast
17794 (vector unsigned long long,
17795 vector unsigned long long);
17797 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
17798 vector unsigned char,
17799 vector unsigned char);
17801 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
17802 vector unsigned short,
17803 vector unsigned short);
17805 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
17806 vector unsigned int,
17807 vector unsigned int);
17809 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
17810 vector unsigned long long,
17811 vector unsigned long long);
17813 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
17814 vector unsigned char);
17816 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
17817 vector unsigned short);
17819 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
17820 vector unsigned int);
17822 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
17823 vector unsigned long long);
17825 vector unsigned long long __builtin_crypto_vshasigmad
17826 (vector unsigned long long, int, int);
17828 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
17832 The second argument to the @var{__builtin_crypto_vshasigmad} and
17833 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
17834 integer that is 0 or 1. The third argument to these builtin functions
17835 must be a constant integer in the range of 0 to 15.
17837 If the ISA 3.0 instruction set additions
17838 are enabled (@option{-mcpu=power9}), the following additional
17839 functions are available for both 32-bit and 64-bit targets.
17841 vector short vec_xl (int, vector short *);
17842 vector short vec_xl (int, short *);
17843 vector unsigned short vec_xl (int, vector unsigned short *);
17844 vector unsigned short vec_xl (int, unsigned short *);
17845 vector char vec_xl (int, vector char *);
17846 vector char vec_xl (int, char *);
17847 vector unsigned char vec_xl (int, vector unsigned char *);
17848 vector unsigned char vec_xl (int, unsigned char *);
17850 void vec_xst (vector short, int, vector short *);
17851 void vec_xst (vector short, int, short *);
17852 void vec_xst (vector unsigned short, int, vector unsigned short *);
17853 void vec_xst (vector unsigned short, int, unsigned short *);
17854 void vec_xst (vector char, int, vector char *);
17855 void vec_xst (vector char, int, char *);
17856 void vec_xst (vector unsigned char, int, vector unsigned char *);
17857 void vec_xst (vector unsigned char, int, unsigned char *);
17859 @node PowerPC Hardware Transactional Memory Built-in Functions
17860 @subsection PowerPC Hardware Transactional Memory Built-in Functions
17861 GCC provides two interfaces for accessing the Hardware Transactional
17862 Memory (HTM) instructions available on some of the PowerPC family
17863 of processors (eg, POWER8). The two interfaces come in a low level
17864 interface, consisting of built-in functions specific to PowerPC and a
17865 higher level interface consisting of inline functions that are common
17866 between PowerPC and S/390.
17868 @subsubsection PowerPC HTM Low Level Built-in Functions
17870 The following low level built-in functions are available with
17871 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
17872 They all generate the machine instruction that is part of the name.
17874 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
17875 the full 4-bit condition register value set by their associated hardware
17876 instruction. The header file @code{htmintrin.h} defines some macros that can
17877 be used to decipher the return value. The @code{__builtin_tbegin} builtin
17878 returns a simple true or false value depending on whether a transaction was
17879 successfully started or not. The arguments of the builtins match exactly the
17880 type and order of the associated hardware instruction's operands, except for
17881 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
17882 Refer to the ISA manual for a description of each instruction's operands.
17885 unsigned int __builtin_tbegin (unsigned int)
17886 unsigned int __builtin_tend (unsigned int)
17888 unsigned int __builtin_tabort (unsigned int)
17889 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
17890 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
17891 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
17892 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
17894 unsigned int __builtin_tcheck (void)
17895 unsigned int __builtin_treclaim (unsigned int)
17896 unsigned int __builtin_trechkpt (void)
17897 unsigned int __builtin_tsr (unsigned int)
17900 In addition to the above HTM built-ins, we have added built-ins for
17901 some common extended mnemonics of the HTM instructions:
17904 unsigned int __builtin_tendall (void)
17905 unsigned int __builtin_tresume (void)
17906 unsigned int __builtin_tsuspend (void)
17909 Note that the semantics of the above HTM builtins are required to mimic
17910 the locking semantics used for critical sections. Builtins that are used
17911 to create a new transaction or restart a suspended transaction must have
17912 lock acquisition like semantics while those builtins that end or suspend a
17913 transaction must have lock release like semantics. Specifically, this must
17914 mimic lock semantics as specified by C++11, for example: Lock acquisition is
17915 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
17916 that returns 0, and lock release is as-if an execution of
17917 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
17918 implicit implementation-defined lock used for all transactions. The HTM
17919 instructions associated with with the builtins inherently provide the
17920 correct acquisition and release hardware barriers required. However,
17921 the compiler must also be prohibited from moving loads and stores across
17922 the builtins in a way that would violate their semantics. This has been
17923 accomplished by adding memory barriers to the associated HTM instructions
17924 (which is a conservative approach to provide acquire and release semantics).
17925 Earlier versions of the compiler did not treat the HTM instructions as
17926 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
17927 be used to determine whether the current compiler treats HTM instructions
17928 as memory barriers or not. This allows the user to explicitly add memory
17929 barriers to their code when using an older version of the compiler.
17931 The following set of built-in functions are available to gain access
17932 to the HTM specific special purpose registers.
17935 unsigned long __builtin_get_texasr (void)
17936 unsigned long __builtin_get_texasru (void)
17937 unsigned long __builtin_get_tfhar (void)
17938 unsigned long __builtin_get_tfiar (void)
17940 void __builtin_set_texasr (unsigned long);
17941 void __builtin_set_texasru (unsigned long);
17942 void __builtin_set_tfhar (unsigned long);
17943 void __builtin_set_tfiar (unsigned long);
17946 Example usage of these low level built-in functions may look like:
17949 #include <htmintrin.h>
17951 int num_retries = 10;
17955 if (__builtin_tbegin (0))
17957 /* Transaction State Initiated. */
17958 if (is_locked (lock))
17959 __builtin_tabort (0);
17960 ... transaction code...
17961 __builtin_tend (0);
17966 /* Transaction State Failed. Use locks if the transaction
17967 failure is "persistent" or we've tried too many times. */
17968 if (num_retries-- <= 0
17969 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
17971 acquire_lock (lock);
17972 ... non transactional fallback path...
17973 release_lock (lock);
17980 One final built-in function has been added that returns the value of
17981 the 2-bit Transaction State field of the Machine Status Register (MSR)
17982 as stored in @code{CR0}.
17985 unsigned long __builtin_ttest (void)
17988 This built-in can be used to determine the current transaction state
17989 using the following code example:
17992 #include <htmintrin.h>
17994 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
17996 if (tx_state == _HTM_TRANSACTIONAL)
17998 /* Code to use in transactional state. */
18000 else if (tx_state == _HTM_NONTRANSACTIONAL)
18002 /* Code to use in non-transactional state. */
18004 else if (tx_state == _HTM_SUSPENDED)
18006 /* Code to use in transaction suspended state. */
18010 @subsubsection PowerPC HTM High Level Inline Functions
18012 The following high level HTM interface is made available by including
18013 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
18014 where CPU is `power8' or later. This interface is common between PowerPC
18015 and S/390, allowing users to write one HTM source implementation that
18016 can be compiled and executed on either system.
18019 long __TM_simple_begin (void)
18020 long __TM_begin (void* const TM_buff)
18021 long __TM_end (void)
18022 void __TM_abort (void)
18023 void __TM_named_abort (unsigned char const code)
18024 void __TM_resume (void)
18025 void __TM_suspend (void)
18027 long __TM_is_user_abort (void* const TM_buff)
18028 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
18029 long __TM_is_illegal (void* const TM_buff)
18030 long __TM_is_footprint_exceeded (void* const TM_buff)
18031 long __TM_nesting_depth (void* const TM_buff)
18032 long __TM_is_nested_too_deep(void* const TM_buff)
18033 long __TM_is_conflict(void* const TM_buff)
18034 long __TM_is_failure_persistent(void* const TM_buff)
18035 long __TM_failure_address(void* const TM_buff)
18036 long long __TM_failure_code(void* const TM_buff)
18039 Using these common set of HTM inline functions, we can create
18040 a more portable version of the HTM example in the previous
18041 section that will work on either PowerPC or S/390:
18044 #include <htmxlintrin.h>
18046 int num_retries = 10;
18047 TM_buff_type TM_buff;
18051 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
18053 /* Transaction State Initiated. */
18054 if (is_locked (lock))
18056 ... transaction code...
18062 /* Transaction State Failed. Use locks if the transaction
18063 failure is "persistent" or we've tried too many times. */
18064 if (num_retries-- <= 0
18065 || __TM_is_failure_persistent (TM_buff))
18067 acquire_lock (lock);
18068 ... non transactional fallback path...
18069 release_lock (lock);
18076 @node RX Built-in Functions
18077 @subsection RX Built-in Functions
18078 GCC supports some of the RX instructions which cannot be expressed in
18079 the C programming language via the use of built-in functions. The
18080 following functions are supported:
18082 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
18083 Generates the @code{brk} machine instruction.
18086 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
18087 Generates the @code{clrpsw} machine instruction to clear the specified
18088 bit in the processor status word.
18091 @deftypefn {Built-in Function} void __builtin_rx_int (int)
18092 Generates the @code{int} machine instruction to generate an interrupt
18093 with the specified value.
18096 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
18097 Generates the @code{machi} machine instruction to add the result of
18098 multiplying the top 16 bits of the two arguments into the
18102 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
18103 Generates the @code{maclo} machine instruction to add the result of
18104 multiplying the bottom 16 bits of the two arguments into the
18108 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
18109 Generates the @code{mulhi} machine instruction to place the result of
18110 multiplying the top 16 bits of the two arguments into the
18114 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
18115 Generates the @code{mullo} machine instruction to place the result of
18116 multiplying the bottom 16 bits of the two arguments into the
18120 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
18121 Generates the @code{mvfachi} machine instruction to read the top
18122 32 bits of the accumulator.
18125 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
18126 Generates the @code{mvfacmi} machine instruction to read the middle
18127 32 bits of the accumulator.
18130 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
18131 Generates the @code{mvfc} machine instruction which reads the control
18132 register specified in its argument and returns its value.
18135 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
18136 Generates the @code{mvtachi} machine instruction to set the top
18137 32 bits of the accumulator.
18140 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
18141 Generates the @code{mvtaclo} machine instruction to set the bottom
18142 32 bits of the accumulator.
18145 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
18146 Generates the @code{mvtc} machine instruction which sets control
18147 register number @code{reg} to @code{val}.
18150 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
18151 Generates the @code{mvtipl} machine instruction set the interrupt
18155 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
18156 Generates the @code{racw} machine instruction to round the accumulator
18157 according to the specified mode.
18160 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
18161 Generates the @code{revw} machine instruction which swaps the bytes in
18162 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
18163 and also bits 16--23 occupy bits 24--31 and vice versa.
18166 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
18167 Generates the @code{rmpa} machine instruction which initiates a
18168 repeated multiply and accumulate sequence.
18171 @deftypefn {Built-in Function} void __builtin_rx_round (float)
18172 Generates the @code{round} machine instruction which returns the
18173 floating-point argument rounded according to the current rounding mode
18174 set in the floating-point status word register.
18177 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
18178 Generates the @code{sat} machine instruction which returns the
18179 saturated value of the argument.
18182 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
18183 Generates the @code{setpsw} machine instruction to set the specified
18184 bit in the processor status word.
18187 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
18188 Generates the @code{wait} machine instruction.
18191 @node S/390 System z Built-in Functions
18192 @subsection S/390 System z Built-in Functions
18193 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
18194 Generates the @code{tbegin} machine instruction starting a
18195 non-constrained hardware transaction. If the parameter is non-NULL the
18196 memory area is used to store the transaction diagnostic buffer and
18197 will be passed as first operand to @code{tbegin}. This buffer can be
18198 defined using the @code{struct __htm_tdb} C struct defined in
18199 @code{htmintrin.h} and must reside on a double-word boundary. The
18200 second tbegin operand is set to @code{0xff0c}. This enables
18201 save/restore of all GPRs and disables aborts for FPR and AR
18202 manipulations inside the transaction body. The condition code set by
18203 the tbegin instruction is returned as integer value. The tbegin
18204 instruction by definition overwrites the content of all FPRs. The
18205 compiler will generate code which saves and restores the FPRs. For
18206 soft-float code it is recommended to used the @code{*_nofloat}
18207 variant. In order to prevent a TDB from being written it is required
18208 to pass a constant zero value as parameter. Passing a zero value
18209 through a variable is not sufficient. Although modifications of
18210 access registers inside the transaction will not trigger an
18211 transaction abort it is not supported to actually modify them. Access
18212 registers do not get saved when entering a transaction. They will have
18213 undefined state when reaching the abort code.
18216 Macros for the possible return codes of tbegin are defined in the
18217 @code{htmintrin.h} header file:
18220 @item _HTM_TBEGIN_STARTED
18221 @code{tbegin} has been executed as part of normal processing. The
18222 transaction body is supposed to be executed.
18223 @item _HTM_TBEGIN_INDETERMINATE
18224 The transaction was aborted due to an indeterminate condition which
18225 might be persistent.
18226 @item _HTM_TBEGIN_TRANSIENT
18227 The transaction aborted due to a transient failure. The transaction
18228 should be re-executed in that case.
18229 @item _HTM_TBEGIN_PERSISTENT
18230 The transaction aborted due to a persistent failure. Re-execution
18231 under same circumstances will not be productive.
18234 @defmac _HTM_FIRST_USER_ABORT_CODE
18235 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
18236 specifies the first abort code which can be used for
18237 @code{__builtin_tabort}. Values below this threshold are reserved for
18241 @deftp {Data type} {struct __htm_tdb}
18242 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
18243 the structure of the transaction diagnostic block as specified in the
18244 Principles of Operation manual chapter 5-91.
18247 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
18248 Same as @code{__builtin_tbegin} but without FPR saves and restores.
18249 Using this variant in code making use of FPRs will leave the FPRs in
18250 undefined state when entering the transaction abort handler code.
18253 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
18254 In addition to @code{__builtin_tbegin} a loop for transient failures
18255 is generated. If tbegin returns a condition code of 2 the transaction
18256 will be retried as often as specified in the second argument. The
18257 perform processor assist instruction is used to tell the CPU about the
18258 number of fails so far.
18261 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
18262 Same as @code{__builtin_tbegin_retry} but without FPR saves and
18263 restores. Using this variant in code making use of FPRs will leave
18264 the FPRs in undefined state when entering the transaction abort
18268 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
18269 Generates the @code{tbeginc} machine instruction starting a constrained
18270 hardware transaction. The second operand is set to @code{0xff08}.
18273 @deftypefn {Built-in Function} int __builtin_tend (void)
18274 Generates the @code{tend} machine instruction finishing a transaction
18275 and making the changes visible to other threads. The condition code
18276 generated by tend is returned as integer value.
18279 @deftypefn {Built-in Function} void __builtin_tabort (int)
18280 Generates the @code{tabort} machine instruction with the specified
18281 abort code. Abort codes from 0 through 255 are reserved and will
18282 result in an error message.
18285 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
18286 Generates the @code{ppa rX,rY,1} machine instruction. Where the
18287 integer parameter is loaded into rX and a value of zero is loaded into
18288 rY. The integer parameter specifies the number of times the
18289 transaction repeatedly aborted.
18292 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
18293 Generates the @code{etnd} machine instruction. The current nesting
18294 depth is returned as integer value. For a nesting depth of 0 the code
18295 is not executed as part of an transaction.
18298 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
18300 Generates the @code{ntstg} machine instruction. The second argument
18301 is written to the first arguments location. The store operation will
18302 not be rolled-back in case of an transaction abort.
18305 @node SH Built-in Functions
18306 @subsection SH Built-in Functions
18307 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
18308 families of processors:
18310 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
18311 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
18312 used by system code that manages threads and execution contexts. The compiler
18313 normally does not generate code that modifies the contents of @samp{GBR} and
18314 thus the value is preserved across function calls. Changing the @samp{GBR}
18315 value in user code must be done with caution, since the compiler might use
18316 @samp{GBR} in order to access thread local variables.
18320 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
18321 Returns the value that is currently set in the @samp{GBR} register.
18322 Memory loads and stores that use the thread pointer as a base address are
18323 turned into @samp{GBR} based displacement loads and stores, if possible.
18331 int get_tcb_value (void)
18333 // Generate @samp{mov.l @@(8,gbr),r0} instruction
18334 return ((my_tcb*)__builtin_thread_pointer ())->c;
18340 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
18341 Returns the value that is currently set in the @samp{FPSCR} register.
18344 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
18345 Sets the @samp{FPSCR} register to the specified value @var{val}, while
18346 preserving the current values of the FR, SZ and PR bits.
18349 @node SPARC VIS Built-in Functions
18350 @subsection SPARC VIS Built-in Functions
18352 GCC supports SIMD operations on the SPARC using both the generic vector
18353 extensions (@pxref{Vector Extensions}) as well as built-in functions for
18354 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
18355 switch, the VIS extension is exposed as the following built-in functions:
18358 typedef int v1si __attribute__ ((vector_size (4)));
18359 typedef int v2si __attribute__ ((vector_size (8)));
18360 typedef short v4hi __attribute__ ((vector_size (8)));
18361 typedef short v2hi __attribute__ ((vector_size (4)));
18362 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
18363 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
18365 void __builtin_vis_write_gsr (int64_t);
18366 int64_t __builtin_vis_read_gsr (void);
18368 void * __builtin_vis_alignaddr (void *, long);
18369 void * __builtin_vis_alignaddrl (void *, long);
18370 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
18371 v2si __builtin_vis_faligndatav2si (v2si, v2si);
18372 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
18373 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
18375 v4hi __builtin_vis_fexpand (v4qi);
18377 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
18378 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
18379 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
18380 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
18381 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
18382 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
18383 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
18385 v4qi __builtin_vis_fpack16 (v4hi);
18386 v8qi __builtin_vis_fpack32 (v2si, v8qi);
18387 v2hi __builtin_vis_fpackfix (v2si);
18388 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
18390 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
18392 long __builtin_vis_edge8 (void *, void *);
18393 long __builtin_vis_edge8l (void *, void *);
18394 long __builtin_vis_edge16 (void *, void *);
18395 long __builtin_vis_edge16l (void *, void *);
18396 long __builtin_vis_edge32 (void *, void *);
18397 long __builtin_vis_edge32l (void *, void *);
18399 long __builtin_vis_fcmple16 (v4hi, v4hi);
18400 long __builtin_vis_fcmple32 (v2si, v2si);
18401 long __builtin_vis_fcmpne16 (v4hi, v4hi);
18402 long __builtin_vis_fcmpne32 (v2si, v2si);
18403 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
18404 long __builtin_vis_fcmpgt32 (v2si, v2si);
18405 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
18406 long __builtin_vis_fcmpeq32 (v2si, v2si);
18408 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
18409 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
18410 v2si __builtin_vis_fpadd32 (v2si, v2si);
18411 v1si __builtin_vis_fpadd32s (v1si, v1si);
18412 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
18413 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
18414 v2si __builtin_vis_fpsub32 (v2si, v2si);
18415 v1si __builtin_vis_fpsub32s (v1si, v1si);
18417 long __builtin_vis_array8 (long, long);
18418 long __builtin_vis_array16 (long, long);
18419 long __builtin_vis_array32 (long, long);
18422 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
18423 functions also become available:
18426 long __builtin_vis_bmask (long, long);
18427 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
18428 v2si __builtin_vis_bshufflev2si (v2si, v2si);
18429 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
18430 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
18432 long __builtin_vis_edge8n (void *, void *);
18433 long __builtin_vis_edge8ln (void *, void *);
18434 long __builtin_vis_edge16n (void *, void *);
18435 long __builtin_vis_edge16ln (void *, void *);
18436 long __builtin_vis_edge32n (void *, void *);
18437 long __builtin_vis_edge32ln (void *, void *);
18440 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
18441 functions also become available:
18444 void __builtin_vis_cmask8 (long);
18445 void __builtin_vis_cmask16 (long);
18446 void __builtin_vis_cmask32 (long);
18448 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
18450 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
18451 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
18452 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
18453 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
18454 v2si __builtin_vis_fsll16 (v2si, v2si);
18455 v2si __builtin_vis_fslas16 (v2si, v2si);
18456 v2si __builtin_vis_fsrl16 (v2si, v2si);
18457 v2si __builtin_vis_fsra16 (v2si, v2si);
18459 long __builtin_vis_pdistn (v8qi, v8qi);
18461 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
18463 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
18464 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
18466 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
18467 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
18468 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
18469 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
18470 v2si __builtin_vis_fpadds32 (v2si, v2si);
18471 v1si __builtin_vis_fpadds32s (v1si, v1si);
18472 v2si __builtin_vis_fpsubs32 (v2si, v2si);
18473 v1si __builtin_vis_fpsubs32s (v1si, v1si);
18475 long __builtin_vis_fucmple8 (v8qi, v8qi);
18476 long __builtin_vis_fucmpne8 (v8qi, v8qi);
18477 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
18478 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
18480 float __builtin_vis_fhadds (float, float);
18481 double __builtin_vis_fhaddd (double, double);
18482 float __builtin_vis_fhsubs (float, float);
18483 double __builtin_vis_fhsubd (double, double);
18484 float __builtin_vis_fnhadds (float, float);
18485 double __builtin_vis_fnhaddd (double, double);
18487 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
18488 int64_t __builtin_vis_xmulx (int64_t, int64_t);
18489 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
18492 When you use the @option{-mvis4} switch, the VIS version 4.0 built-in
18493 functions also become available:
18496 v8qi __builtin_vis_fpadd8 (v8qi, v8qi);
18497 v8qi __builtin_vis_fpadds8 (v8qi, v8qi);
18498 v8qi __builtin_vis_fpaddus8 (v8qi, v8qi);
18499 v4hi __builtin_vis_fpaddus16 (v4hi, v4hi);
18501 v8qi __builtin_vis_fpsub8 (v8qi, v8qi);
18502 v8qi __builtin_vis_fpsubs8 (v8qi, v8qi);
18503 v8qi __builtin_vis_fpsubus8 (v8qi, v8qi);
18504 v4hi __builtin_vis_fpsubus16 (v4hi, v4hi);
18506 long __builtin_vis_fpcmple8 (v8qi, v8qi);
18507 long __builtin_vis_fpcmpgt8 (v8qi, v8qi);
18508 long __builtin_vis_fpcmpule16 (v4hi, v4hi);
18509 long __builtin_vis_fpcmpugt16 (v4hi, v4hi);
18510 long __builtin_vis_fpcmpule32 (v2si, v2si);
18511 long __builtin_vis_fpcmpugt32 (v2si, v2si);
18513 v8qi __builtin_vis_fpmax8 (v8qi, v8qi);
18514 v4hi __builtin_vis_fpmax16 (v4hi, v4hi);
18515 v2si __builtin_vis_fpmax32 (v2si, v2si);
18517 v8qi __builtin_vis_fpmaxu8 (v8qi, v8qi);
18518 v4hi __builtin_vis_fpmaxu16 (v4hi, v4hi);
18519 v2si __builtin_vis_fpmaxu32 (v2si, v2si);
18522 v8qi __builtin_vis_fpmin8 (v8qi, v8qi);
18523 v4hi __builtin_vis_fpmin16 (v4hi, v4hi);
18524 v2si __builtin_vis_fpmin32 (v2si, v2si);
18526 v8qi __builtin_vis_fpminu8 (v8qi, v8qi);
18527 v4hi __builtin_vis_fpminu16 (v4hi, v4hi);
18528 v2si __builtin_vis_fpminu32 (v2si, v2si);
18531 @node SPU Built-in Functions
18532 @subsection SPU Built-in Functions
18534 GCC provides extensions for the SPU processor as described in the
18535 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
18536 found at @uref{http://cell.scei.co.jp/} or
18537 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
18538 implementation differs in several ways.
18543 The optional extension of specifying vector constants in parentheses is
18547 A vector initializer requires no cast if the vector constant is of the
18548 same type as the variable it is initializing.
18551 If @code{signed} or @code{unsigned} is omitted, the signedness of the
18552 vector type is the default signedness of the base type. The default
18553 varies depending on the operating system, so a portable program should
18554 always specify the signedness.
18557 By default, the keyword @code{__vector} is added. The macro
18558 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
18562 GCC allows using a @code{typedef} name as the type specifier for a
18566 For C, overloaded functions are implemented with macros so the following
18570 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
18574 Since @code{spu_add} is a macro, the vector constant in the example
18575 is treated as four separate arguments. Wrap the entire argument in
18576 parentheses for this to work.
18579 The extended version of @code{__builtin_expect} is not supported.
18583 @emph{Note:} Only the interface described in the aforementioned
18584 specification is supported. Internally, GCC uses built-in functions to
18585 implement the required functionality, but these are not supported and
18586 are subject to change without notice.
18588 @node TI C6X Built-in Functions
18589 @subsection TI C6X Built-in Functions
18591 GCC provides intrinsics to access certain instructions of the TI C6X
18592 processors. These intrinsics, listed below, are available after
18593 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
18594 to C6X instructions.
18598 int _sadd (int, int)
18599 int _ssub (int, int)
18600 int _sadd2 (int, int)
18601 int _ssub2 (int, int)
18602 long long _mpy2 (int, int)
18603 long long _smpy2 (int, int)
18604 int _add4 (int, int)
18605 int _sub4 (int, int)
18606 int _saddu4 (int, int)
18608 int _smpy (int, int)
18609 int _smpyh (int, int)
18610 int _smpyhl (int, int)
18611 int _smpylh (int, int)
18613 int _sshl (int, int)
18614 int _subc (int, int)
18616 int _avg2 (int, int)
18617 int _avgu4 (int, int)
18619 int _clrr (int, int)
18620 int _extr (int, int)
18621 int _extru (int, int)
18627 @node TILE-Gx Built-in Functions
18628 @subsection TILE-Gx Built-in Functions
18630 GCC provides intrinsics to access every instruction of the TILE-Gx
18631 processor. The intrinsics are of the form:
18635 unsigned long long __insn_@var{op} (...)
18639 Where @var{op} is the name of the instruction. Refer to the ISA manual
18640 for the complete list of instructions.
18642 GCC also provides intrinsics to directly access the network registers.
18643 The intrinsics are:
18647 unsigned long long __tile_idn0_receive (void)
18648 unsigned long long __tile_idn1_receive (void)
18649 unsigned long long __tile_udn0_receive (void)
18650 unsigned long long __tile_udn1_receive (void)
18651 unsigned long long __tile_udn2_receive (void)
18652 unsigned long long __tile_udn3_receive (void)
18653 void __tile_idn_send (unsigned long long)
18654 void __tile_udn_send (unsigned long long)
18658 The intrinsic @code{void __tile_network_barrier (void)} is used to
18659 guarantee that no network operations before it are reordered with
18662 @node TILEPro Built-in Functions
18663 @subsection TILEPro Built-in Functions
18665 GCC provides intrinsics to access every instruction of the TILEPro
18666 processor. The intrinsics are of the form:
18670 unsigned __insn_@var{op} (...)
18675 where @var{op} is the name of the instruction. Refer to the ISA manual
18676 for the complete list of instructions.
18678 GCC also provides intrinsics to directly access the network registers.
18679 The intrinsics are:
18683 unsigned __tile_idn0_receive (void)
18684 unsigned __tile_idn1_receive (void)
18685 unsigned __tile_sn_receive (void)
18686 unsigned __tile_udn0_receive (void)
18687 unsigned __tile_udn1_receive (void)
18688 unsigned __tile_udn2_receive (void)
18689 unsigned __tile_udn3_receive (void)
18690 void __tile_idn_send (unsigned)
18691 void __tile_sn_send (unsigned)
18692 void __tile_udn_send (unsigned)
18696 The intrinsic @code{void __tile_network_barrier (void)} is used to
18697 guarantee that no network operations before it are reordered with
18700 @node x86 Built-in Functions
18701 @subsection x86 Built-in Functions
18703 These built-in functions are available for the x86-32 and x86-64 family
18704 of computers, depending on the command-line switches used.
18706 If you specify command-line switches such as @option{-msse},
18707 the compiler could use the extended instruction sets even if the built-ins
18708 are not used explicitly in the program. For this reason, applications
18709 that perform run-time CPU detection must compile separate files for each
18710 supported architecture, using the appropriate flags. In particular,
18711 the file containing the CPU detection code should be compiled without
18714 The following machine modes are available for use with MMX built-in functions
18715 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
18716 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
18717 vector of eight 8-bit integers. Some of the built-in functions operate on
18718 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
18720 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
18721 of two 32-bit floating-point values.
18723 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
18724 floating-point values. Some instructions use a vector of four 32-bit
18725 integers, these use @code{V4SI}. Finally, some instructions operate on an
18726 entire vector register, interpreting it as a 128-bit integer, these use mode
18729 The x86-32 and x86-64 family of processors use additional built-in
18730 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
18731 floating point and @code{TC} 128-bit complex floating-point values.
18733 The following floating-point built-in functions are always available. All
18734 of them implement the function that is part of the name.
18737 __float128 __builtin_fabsq (__float128)
18738 __float128 __builtin_copysignq (__float128, __float128)
18741 The following built-in functions are always available.
18744 @item __float128 __builtin_infq (void)
18745 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
18746 @findex __builtin_infq
18748 @item __float128 __builtin_huge_valq (void)
18749 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
18750 @findex __builtin_huge_valq
18752 @item __float128 __builtin_nanq (void)
18753 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
18754 @findex __builtin_nanq
18756 @item __float128 __builtin_nansq (void)
18757 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
18758 @findex __builtin_nansq
18761 The following built-in function is always available.
18764 @item void __builtin_ia32_pause (void)
18765 Generates the @code{pause} machine instruction with a compiler memory
18769 The following built-in functions are always available and can be used to
18770 check the target platform type.
18772 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
18773 This function runs the CPU detection code to check the type of CPU and the
18774 features supported. This built-in function needs to be invoked along with the built-in functions
18775 to check CPU type and features, @code{__builtin_cpu_is} and
18776 @code{__builtin_cpu_supports}, only when used in a function that is
18777 executed before any constructors are called. The CPU detection code is
18778 automatically executed in a very high priority constructor.
18780 For example, this function has to be used in @code{ifunc} resolvers that
18781 check for CPU type using the built-in functions @code{__builtin_cpu_is}
18782 and @code{__builtin_cpu_supports}, or in constructors on targets that
18783 don't support constructor priority.
18786 static void (*resolve_memcpy (void)) (void)
18788 // ifunc resolvers fire before constructors, explicitly call the init
18790 __builtin_cpu_init ();
18791 if (__builtin_cpu_supports ("ssse3"))
18792 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
18794 return default_memcpy;
18797 void *memcpy (void *, const void *, size_t)
18798 __attribute__ ((ifunc ("resolve_memcpy")));
18803 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
18804 This function returns a positive integer if the run-time CPU
18805 is of type @var{cpuname}
18806 and returns @code{0} otherwise. The following CPU names can be detected:
18822 Intel Core i7 Nehalem CPU.
18825 Intel Core i7 Westmere CPU.
18828 Intel Core i7 Sandy Bridge CPU.
18834 AMD Family 10h CPU.
18837 AMD Family 10h Barcelona CPU.
18840 AMD Family 10h Shanghai CPU.
18843 AMD Family 10h Istanbul CPU.
18846 AMD Family 14h CPU.
18849 AMD Family 15h CPU.
18852 AMD Family 15h Bulldozer version 1.
18855 AMD Family 15h Bulldozer version 2.
18858 AMD Family 15h Bulldozer version 3.
18861 AMD Family 15h Bulldozer version 4.
18864 AMD Family 16h CPU.
18867 AMD Family 17h CPU.
18870 Here is an example:
18872 if (__builtin_cpu_is ("corei7"))
18874 do_corei7 (); // Core i7 specific implementation.
18878 do_generic (); // Generic implementation.
18883 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
18884 This function returns a positive integer if the run-time CPU
18885 supports @var{feature}
18886 and returns @code{0} otherwise. The following features can be detected:
18894 POPCNT instruction.
18902 SSSE3 instructions.
18904 SSE4.1 instructions.
18906 SSE4.2 instructions.
18912 AVX512F instructions.
18915 Here is an example:
18917 if (__builtin_cpu_supports ("popcnt"))
18919 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
18923 count = generic_countbits (n); //generic implementation.
18929 The following built-in functions are made available by @option{-mmmx}.
18930 All of them generate the machine instruction that is part of the name.
18933 v8qi __builtin_ia32_paddb (v8qi, v8qi)
18934 v4hi __builtin_ia32_paddw (v4hi, v4hi)
18935 v2si __builtin_ia32_paddd (v2si, v2si)
18936 v8qi __builtin_ia32_psubb (v8qi, v8qi)
18937 v4hi __builtin_ia32_psubw (v4hi, v4hi)
18938 v2si __builtin_ia32_psubd (v2si, v2si)
18939 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
18940 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
18941 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
18942 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
18943 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
18944 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
18945 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
18946 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
18947 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
18948 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
18949 di __builtin_ia32_pand (di, di)
18950 di __builtin_ia32_pandn (di,di)
18951 di __builtin_ia32_por (di, di)
18952 di __builtin_ia32_pxor (di, di)
18953 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
18954 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
18955 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
18956 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
18957 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
18958 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
18959 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
18960 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
18961 v2si __builtin_ia32_punpckhdq (v2si, v2si)
18962 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
18963 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
18964 v2si __builtin_ia32_punpckldq (v2si, v2si)
18965 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
18966 v4hi __builtin_ia32_packssdw (v2si, v2si)
18967 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
18969 v4hi __builtin_ia32_psllw (v4hi, v4hi)
18970 v2si __builtin_ia32_pslld (v2si, v2si)
18971 v1di __builtin_ia32_psllq (v1di, v1di)
18972 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
18973 v2si __builtin_ia32_psrld (v2si, v2si)
18974 v1di __builtin_ia32_psrlq (v1di, v1di)
18975 v4hi __builtin_ia32_psraw (v4hi, v4hi)
18976 v2si __builtin_ia32_psrad (v2si, v2si)
18977 v4hi __builtin_ia32_psllwi (v4hi, int)
18978 v2si __builtin_ia32_pslldi (v2si, int)
18979 v1di __builtin_ia32_psllqi (v1di, int)
18980 v4hi __builtin_ia32_psrlwi (v4hi, int)
18981 v2si __builtin_ia32_psrldi (v2si, int)
18982 v1di __builtin_ia32_psrlqi (v1di, int)
18983 v4hi __builtin_ia32_psrawi (v4hi, int)
18984 v2si __builtin_ia32_psradi (v2si, int)
18988 The following built-in functions are made available either with
18989 @option{-msse}, or with a combination of @option{-m3dnow} and
18990 @option{-march=athlon}. All of them generate the machine
18991 instruction that is part of the name.
18994 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
18995 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
18996 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
18997 v1di __builtin_ia32_psadbw (v8qi, v8qi)
18998 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
18999 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
19000 v8qi __builtin_ia32_pminub (v8qi, v8qi)
19001 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
19002 int __builtin_ia32_pmovmskb (v8qi)
19003 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
19004 void __builtin_ia32_movntq (di *, di)
19005 void __builtin_ia32_sfence (void)
19008 The following built-in functions are available when @option{-msse} is used.
19009 All of them generate the machine instruction that is part of the name.
19012 int __builtin_ia32_comieq (v4sf, v4sf)
19013 int __builtin_ia32_comineq (v4sf, v4sf)
19014 int __builtin_ia32_comilt (v4sf, v4sf)
19015 int __builtin_ia32_comile (v4sf, v4sf)
19016 int __builtin_ia32_comigt (v4sf, v4sf)
19017 int __builtin_ia32_comige (v4sf, v4sf)
19018 int __builtin_ia32_ucomieq (v4sf, v4sf)
19019 int __builtin_ia32_ucomineq (v4sf, v4sf)
19020 int __builtin_ia32_ucomilt (v4sf, v4sf)
19021 int __builtin_ia32_ucomile (v4sf, v4sf)
19022 int __builtin_ia32_ucomigt (v4sf, v4sf)
19023 int __builtin_ia32_ucomige (v4sf, v4sf)
19024 v4sf __builtin_ia32_addps (v4sf, v4sf)
19025 v4sf __builtin_ia32_subps (v4sf, v4sf)
19026 v4sf __builtin_ia32_mulps (v4sf, v4sf)
19027 v4sf __builtin_ia32_divps (v4sf, v4sf)
19028 v4sf __builtin_ia32_addss (v4sf, v4sf)
19029 v4sf __builtin_ia32_subss (v4sf, v4sf)
19030 v4sf __builtin_ia32_mulss (v4sf, v4sf)
19031 v4sf __builtin_ia32_divss (v4sf, v4sf)
19032 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
19033 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
19034 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
19035 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
19036 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
19037 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
19038 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
19039 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
19040 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
19041 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
19042 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
19043 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
19044 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
19045 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
19046 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
19047 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
19048 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
19049 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
19050 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
19051 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
19052 v4sf __builtin_ia32_maxps (v4sf, v4sf)
19053 v4sf __builtin_ia32_maxss (v4sf, v4sf)
19054 v4sf __builtin_ia32_minps (v4sf, v4sf)
19055 v4sf __builtin_ia32_minss (v4sf, v4sf)
19056 v4sf __builtin_ia32_andps (v4sf, v4sf)
19057 v4sf __builtin_ia32_andnps (v4sf, v4sf)
19058 v4sf __builtin_ia32_orps (v4sf, v4sf)
19059 v4sf __builtin_ia32_xorps (v4sf, v4sf)
19060 v4sf __builtin_ia32_movss (v4sf, v4sf)
19061 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
19062 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
19063 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
19064 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
19065 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
19066 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
19067 v2si __builtin_ia32_cvtps2pi (v4sf)
19068 int __builtin_ia32_cvtss2si (v4sf)
19069 v2si __builtin_ia32_cvttps2pi (v4sf)
19070 int __builtin_ia32_cvttss2si (v4sf)
19071 v4sf __builtin_ia32_rcpps (v4sf)
19072 v4sf __builtin_ia32_rsqrtps (v4sf)
19073 v4sf __builtin_ia32_sqrtps (v4sf)
19074 v4sf __builtin_ia32_rcpss (v4sf)
19075 v4sf __builtin_ia32_rsqrtss (v4sf)
19076 v4sf __builtin_ia32_sqrtss (v4sf)
19077 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
19078 void __builtin_ia32_movntps (float *, v4sf)
19079 int __builtin_ia32_movmskps (v4sf)
19082 The following built-in functions are available when @option{-msse} is used.
19085 @item v4sf __builtin_ia32_loadups (float *)
19086 Generates the @code{movups} machine instruction as a load from memory.
19087 @item void __builtin_ia32_storeups (float *, v4sf)
19088 Generates the @code{movups} machine instruction as a store to memory.
19089 @item v4sf __builtin_ia32_loadss (float *)
19090 Generates the @code{movss} machine instruction as a load from memory.
19091 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
19092 Generates the @code{movhps} machine instruction as a load from memory.
19093 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
19094 Generates the @code{movlps} machine instruction as a load from memory
19095 @item void __builtin_ia32_storehps (v2sf *, v4sf)
19096 Generates the @code{movhps} machine instruction as a store to memory.
19097 @item void __builtin_ia32_storelps (v2sf *, v4sf)
19098 Generates the @code{movlps} machine instruction as a store to memory.
19101 The following built-in functions are available when @option{-msse2} is used.
19102 All of them generate the machine instruction that is part of the name.
19105 int __builtin_ia32_comisdeq (v2df, v2df)
19106 int __builtin_ia32_comisdlt (v2df, v2df)
19107 int __builtin_ia32_comisdle (v2df, v2df)
19108 int __builtin_ia32_comisdgt (v2df, v2df)
19109 int __builtin_ia32_comisdge (v2df, v2df)
19110 int __builtin_ia32_comisdneq (v2df, v2df)
19111 int __builtin_ia32_ucomisdeq (v2df, v2df)
19112 int __builtin_ia32_ucomisdlt (v2df, v2df)
19113 int __builtin_ia32_ucomisdle (v2df, v2df)
19114 int __builtin_ia32_ucomisdgt (v2df, v2df)
19115 int __builtin_ia32_ucomisdge (v2df, v2df)
19116 int __builtin_ia32_ucomisdneq (v2df, v2df)
19117 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
19118 v2df __builtin_ia32_cmpltpd (v2df, v2df)
19119 v2df __builtin_ia32_cmplepd (v2df, v2df)
19120 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
19121 v2df __builtin_ia32_cmpgepd (v2df, v2df)
19122 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
19123 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
19124 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
19125 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
19126 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
19127 v2df __builtin_ia32_cmpngepd (v2df, v2df)
19128 v2df __builtin_ia32_cmpordpd (v2df, v2df)
19129 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
19130 v2df __builtin_ia32_cmpltsd (v2df, v2df)
19131 v2df __builtin_ia32_cmplesd (v2df, v2df)
19132 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
19133 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
19134 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
19135 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
19136 v2df __builtin_ia32_cmpordsd (v2df, v2df)
19137 v2di __builtin_ia32_paddq (v2di, v2di)
19138 v2di __builtin_ia32_psubq (v2di, v2di)
19139 v2df __builtin_ia32_addpd (v2df, v2df)
19140 v2df __builtin_ia32_subpd (v2df, v2df)
19141 v2df __builtin_ia32_mulpd (v2df, v2df)
19142 v2df __builtin_ia32_divpd (v2df, v2df)
19143 v2df __builtin_ia32_addsd (v2df, v2df)
19144 v2df __builtin_ia32_subsd (v2df, v2df)
19145 v2df __builtin_ia32_mulsd (v2df, v2df)
19146 v2df __builtin_ia32_divsd (v2df, v2df)
19147 v2df __builtin_ia32_minpd (v2df, v2df)
19148 v2df __builtin_ia32_maxpd (v2df, v2df)
19149 v2df __builtin_ia32_minsd (v2df, v2df)
19150 v2df __builtin_ia32_maxsd (v2df, v2df)
19151 v2df __builtin_ia32_andpd (v2df, v2df)
19152 v2df __builtin_ia32_andnpd (v2df, v2df)
19153 v2df __builtin_ia32_orpd (v2df, v2df)
19154 v2df __builtin_ia32_xorpd (v2df, v2df)
19155 v2df __builtin_ia32_movsd (v2df, v2df)
19156 v2df __builtin_ia32_unpckhpd (v2df, v2df)
19157 v2df __builtin_ia32_unpcklpd (v2df, v2df)
19158 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
19159 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
19160 v4si __builtin_ia32_paddd128 (v4si, v4si)
19161 v2di __builtin_ia32_paddq128 (v2di, v2di)
19162 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
19163 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
19164 v4si __builtin_ia32_psubd128 (v4si, v4si)
19165 v2di __builtin_ia32_psubq128 (v2di, v2di)
19166 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
19167 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
19168 v2di __builtin_ia32_pand128 (v2di, v2di)
19169 v2di __builtin_ia32_pandn128 (v2di, v2di)
19170 v2di __builtin_ia32_por128 (v2di, v2di)
19171 v2di __builtin_ia32_pxor128 (v2di, v2di)
19172 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
19173 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
19174 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
19175 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
19176 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
19177 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
19178 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
19179 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
19180 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
19181 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
19182 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
19183 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
19184 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
19185 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
19186 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
19187 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
19188 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
19189 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
19190 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
19191 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
19192 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
19193 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
19194 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
19195 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
19196 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
19197 v2df __builtin_ia32_loadupd (double *)
19198 void __builtin_ia32_storeupd (double *, v2df)
19199 v2df __builtin_ia32_loadhpd (v2df, double const *)
19200 v2df __builtin_ia32_loadlpd (v2df, double const *)
19201 int __builtin_ia32_movmskpd (v2df)
19202 int __builtin_ia32_pmovmskb128 (v16qi)
19203 void __builtin_ia32_movnti (int *, int)
19204 void __builtin_ia32_movnti64 (long long int *, long long int)
19205 void __builtin_ia32_movntpd (double *, v2df)
19206 void __builtin_ia32_movntdq (v2df *, v2df)
19207 v4si __builtin_ia32_pshufd (v4si, int)
19208 v8hi __builtin_ia32_pshuflw (v8hi, int)
19209 v8hi __builtin_ia32_pshufhw (v8hi, int)
19210 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
19211 v2df __builtin_ia32_sqrtpd (v2df)
19212 v2df __builtin_ia32_sqrtsd (v2df)
19213 v2df __builtin_ia32_shufpd (v2df, v2df, int)
19214 v2df __builtin_ia32_cvtdq2pd (v4si)
19215 v4sf __builtin_ia32_cvtdq2ps (v4si)
19216 v4si __builtin_ia32_cvtpd2dq (v2df)
19217 v2si __builtin_ia32_cvtpd2pi (v2df)
19218 v4sf __builtin_ia32_cvtpd2ps (v2df)
19219 v4si __builtin_ia32_cvttpd2dq (v2df)
19220 v2si __builtin_ia32_cvttpd2pi (v2df)
19221 v2df __builtin_ia32_cvtpi2pd (v2si)
19222 int __builtin_ia32_cvtsd2si (v2df)
19223 int __builtin_ia32_cvttsd2si (v2df)
19224 long long __builtin_ia32_cvtsd2si64 (v2df)
19225 long long __builtin_ia32_cvttsd2si64 (v2df)
19226 v4si __builtin_ia32_cvtps2dq (v4sf)
19227 v2df __builtin_ia32_cvtps2pd (v4sf)
19228 v4si __builtin_ia32_cvttps2dq (v4sf)
19229 v2df __builtin_ia32_cvtsi2sd (v2df, int)
19230 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
19231 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
19232 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
19233 void __builtin_ia32_clflush (const void *)
19234 void __builtin_ia32_lfence (void)
19235 void __builtin_ia32_mfence (void)
19236 v16qi __builtin_ia32_loaddqu (const char *)
19237 void __builtin_ia32_storedqu (char *, v16qi)
19238 v1di __builtin_ia32_pmuludq (v2si, v2si)
19239 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
19240 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
19241 v4si __builtin_ia32_pslld128 (v4si, v4si)
19242 v2di __builtin_ia32_psllq128 (v2di, v2di)
19243 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
19244 v4si __builtin_ia32_psrld128 (v4si, v4si)
19245 v2di __builtin_ia32_psrlq128 (v2di, v2di)
19246 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
19247 v4si __builtin_ia32_psrad128 (v4si, v4si)
19248 v2di __builtin_ia32_pslldqi128 (v2di, int)
19249 v8hi __builtin_ia32_psllwi128 (v8hi, int)
19250 v4si __builtin_ia32_pslldi128 (v4si, int)
19251 v2di __builtin_ia32_psllqi128 (v2di, int)
19252 v2di __builtin_ia32_psrldqi128 (v2di, int)
19253 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
19254 v4si __builtin_ia32_psrldi128 (v4si, int)
19255 v2di __builtin_ia32_psrlqi128 (v2di, int)
19256 v8hi __builtin_ia32_psrawi128 (v8hi, int)
19257 v4si __builtin_ia32_psradi128 (v4si, int)
19258 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
19259 v2di __builtin_ia32_movq128 (v2di)
19262 The following built-in functions are available when @option{-msse3} is used.
19263 All of them generate the machine instruction that is part of the name.
19266 v2df __builtin_ia32_addsubpd (v2df, v2df)
19267 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
19268 v2df __builtin_ia32_haddpd (v2df, v2df)
19269 v4sf __builtin_ia32_haddps (v4sf, v4sf)
19270 v2df __builtin_ia32_hsubpd (v2df, v2df)
19271 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
19272 v16qi __builtin_ia32_lddqu (char const *)
19273 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
19274 v4sf __builtin_ia32_movshdup (v4sf)
19275 v4sf __builtin_ia32_movsldup (v4sf)
19276 void __builtin_ia32_mwait (unsigned int, unsigned int)
19279 The following built-in functions are available when @option{-mssse3} is used.
19280 All of them generate the machine instruction that is part of the name.
19283 v2si __builtin_ia32_phaddd (v2si, v2si)
19284 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
19285 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
19286 v2si __builtin_ia32_phsubd (v2si, v2si)
19287 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
19288 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
19289 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
19290 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
19291 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
19292 v8qi __builtin_ia32_psignb (v8qi, v8qi)
19293 v2si __builtin_ia32_psignd (v2si, v2si)
19294 v4hi __builtin_ia32_psignw (v4hi, v4hi)
19295 v1di __builtin_ia32_palignr (v1di, v1di, int)
19296 v8qi __builtin_ia32_pabsb (v8qi)
19297 v2si __builtin_ia32_pabsd (v2si)
19298 v4hi __builtin_ia32_pabsw (v4hi)
19301 The following built-in functions are available when @option{-mssse3} is used.
19302 All of them generate the machine instruction that is part of the name.
19305 v4si __builtin_ia32_phaddd128 (v4si, v4si)
19306 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
19307 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
19308 v4si __builtin_ia32_phsubd128 (v4si, v4si)
19309 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
19310 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
19311 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
19312 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
19313 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
19314 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
19315 v4si __builtin_ia32_psignd128 (v4si, v4si)
19316 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
19317 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
19318 v16qi __builtin_ia32_pabsb128 (v16qi)
19319 v4si __builtin_ia32_pabsd128 (v4si)
19320 v8hi __builtin_ia32_pabsw128 (v8hi)
19323 The following built-in functions are available when @option{-msse4.1} is
19324 used. All of them generate the machine instruction that is part of the
19328 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
19329 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
19330 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
19331 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
19332 v2df __builtin_ia32_dppd (v2df, v2df, const int)
19333 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
19334 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
19335 v2di __builtin_ia32_movntdqa (v2di *);
19336 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
19337 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
19338 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
19339 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
19340 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
19341 v8hi __builtin_ia32_phminposuw128 (v8hi)
19342 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
19343 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
19344 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
19345 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
19346 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
19347 v4si __builtin_ia32_pminsd128 (v4si, v4si)
19348 v4si __builtin_ia32_pminud128 (v4si, v4si)
19349 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
19350 v4si __builtin_ia32_pmovsxbd128 (v16qi)
19351 v2di __builtin_ia32_pmovsxbq128 (v16qi)
19352 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
19353 v2di __builtin_ia32_pmovsxdq128 (v4si)
19354 v4si __builtin_ia32_pmovsxwd128 (v8hi)
19355 v2di __builtin_ia32_pmovsxwq128 (v8hi)
19356 v4si __builtin_ia32_pmovzxbd128 (v16qi)
19357 v2di __builtin_ia32_pmovzxbq128 (v16qi)
19358 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
19359 v2di __builtin_ia32_pmovzxdq128 (v4si)
19360 v4si __builtin_ia32_pmovzxwd128 (v8hi)
19361 v2di __builtin_ia32_pmovzxwq128 (v8hi)
19362 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
19363 v4si __builtin_ia32_pmulld128 (v4si, v4si)
19364 int __builtin_ia32_ptestc128 (v2di, v2di)
19365 int __builtin_ia32_ptestnzc128 (v2di, v2di)
19366 int __builtin_ia32_ptestz128 (v2di, v2di)
19367 v2df __builtin_ia32_roundpd (v2df, const int)
19368 v4sf __builtin_ia32_roundps (v4sf, const int)
19369 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
19370 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
19373 The following built-in functions are available when @option{-msse4.1} is
19377 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
19378 Generates the @code{insertps} machine instruction.
19379 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
19380 Generates the @code{pextrb} machine instruction.
19381 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
19382 Generates the @code{pinsrb} machine instruction.
19383 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
19384 Generates the @code{pinsrd} machine instruction.
19385 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
19386 Generates the @code{pinsrq} machine instruction in 64bit mode.
19389 The following built-in functions are changed to generate new SSE4.1
19390 instructions when @option{-msse4.1} is used.
19393 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
19394 Generates the @code{extractps} machine instruction.
19395 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
19396 Generates the @code{pextrd} machine instruction.
19397 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
19398 Generates the @code{pextrq} machine instruction in 64bit mode.
19401 The following built-in functions are available when @option{-msse4.2} is
19402 used. All of them generate the machine instruction that is part of the
19406 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
19407 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
19408 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
19409 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
19410 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
19411 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
19412 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
19413 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
19414 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
19415 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
19416 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
19417 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
19418 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
19419 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
19420 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
19423 The following built-in functions are available when @option{-msse4.2} is
19427 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
19428 Generates the @code{crc32b} machine instruction.
19429 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
19430 Generates the @code{crc32w} machine instruction.
19431 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
19432 Generates the @code{crc32l} machine instruction.
19433 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
19434 Generates the @code{crc32q} machine instruction.
19437 The following built-in functions are changed to generate new SSE4.2
19438 instructions when @option{-msse4.2} is used.
19441 @item int __builtin_popcount (unsigned int)
19442 Generates the @code{popcntl} machine instruction.
19443 @item int __builtin_popcountl (unsigned long)
19444 Generates the @code{popcntl} or @code{popcntq} machine instruction,
19445 depending on the size of @code{unsigned long}.
19446 @item int __builtin_popcountll (unsigned long long)
19447 Generates the @code{popcntq} machine instruction.
19450 The following built-in functions are available when @option{-mavx} is
19451 used. All of them generate the machine instruction that is part of the
19455 v4df __builtin_ia32_addpd256 (v4df,v4df)
19456 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
19457 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
19458 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
19459 v4df __builtin_ia32_andnpd256 (v4df,v4df)
19460 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
19461 v4df __builtin_ia32_andpd256 (v4df,v4df)
19462 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
19463 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
19464 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
19465 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
19466 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
19467 v2df __builtin_ia32_cmppd (v2df,v2df,int)
19468 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
19469 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
19470 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
19471 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
19472 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
19473 v4df __builtin_ia32_cvtdq2pd256 (v4si)
19474 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
19475 v4si __builtin_ia32_cvtpd2dq256 (v4df)
19476 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
19477 v8si __builtin_ia32_cvtps2dq256 (v8sf)
19478 v4df __builtin_ia32_cvtps2pd256 (v4sf)
19479 v4si __builtin_ia32_cvttpd2dq256 (v4df)
19480 v8si __builtin_ia32_cvttps2dq256 (v8sf)
19481 v4df __builtin_ia32_divpd256 (v4df,v4df)
19482 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
19483 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
19484 v4df __builtin_ia32_haddpd256 (v4df,v4df)
19485 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
19486 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
19487 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
19488 v32qi __builtin_ia32_lddqu256 (pcchar)
19489 v32qi __builtin_ia32_loaddqu256 (pcchar)
19490 v4df __builtin_ia32_loadupd256 (pcdouble)
19491 v8sf __builtin_ia32_loadups256 (pcfloat)
19492 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
19493 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
19494 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
19495 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
19496 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
19497 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
19498 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
19499 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
19500 v4df __builtin_ia32_maxpd256 (v4df,v4df)
19501 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
19502 v4df __builtin_ia32_minpd256 (v4df,v4df)
19503 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
19504 v4df __builtin_ia32_movddup256 (v4df)
19505 int __builtin_ia32_movmskpd256 (v4df)
19506 int __builtin_ia32_movmskps256 (v8sf)
19507 v8sf __builtin_ia32_movshdup256 (v8sf)
19508 v8sf __builtin_ia32_movsldup256 (v8sf)
19509 v4df __builtin_ia32_mulpd256 (v4df,v4df)
19510 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
19511 v4df __builtin_ia32_orpd256 (v4df,v4df)
19512 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
19513 v2df __builtin_ia32_pd_pd256 (v4df)
19514 v4df __builtin_ia32_pd256_pd (v2df)
19515 v4sf __builtin_ia32_ps_ps256 (v8sf)
19516 v8sf __builtin_ia32_ps256_ps (v4sf)
19517 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
19518 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
19519 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
19520 v8sf __builtin_ia32_rcpps256 (v8sf)
19521 v4df __builtin_ia32_roundpd256 (v4df,int)
19522 v8sf __builtin_ia32_roundps256 (v8sf,int)
19523 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
19524 v8sf __builtin_ia32_rsqrtps256 (v8sf)
19525 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
19526 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
19527 v4si __builtin_ia32_si_si256 (v8si)
19528 v8si __builtin_ia32_si256_si (v4si)
19529 v4df __builtin_ia32_sqrtpd256 (v4df)
19530 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
19531 v8sf __builtin_ia32_sqrtps256 (v8sf)
19532 void __builtin_ia32_storedqu256 (pchar,v32qi)
19533 void __builtin_ia32_storeupd256 (pdouble,v4df)
19534 void __builtin_ia32_storeups256 (pfloat,v8sf)
19535 v4df __builtin_ia32_subpd256 (v4df,v4df)
19536 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
19537 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
19538 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
19539 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
19540 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
19541 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
19542 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
19543 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
19544 v4sf __builtin_ia32_vbroadcastss (pcfloat)
19545 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
19546 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
19547 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
19548 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
19549 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
19550 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
19551 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
19552 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
19553 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
19554 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
19555 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
19556 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
19557 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
19558 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
19559 v2df __builtin_ia32_vpermilpd (v2df,int)
19560 v4df __builtin_ia32_vpermilpd256 (v4df,int)
19561 v4sf __builtin_ia32_vpermilps (v4sf,int)
19562 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
19563 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
19564 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
19565 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
19566 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
19567 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
19568 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
19569 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
19570 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
19571 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
19572 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
19573 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
19574 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
19575 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
19576 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
19577 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
19578 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
19579 void __builtin_ia32_vzeroall (void)
19580 void __builtin_ia32_vzeroupper (void)
19581 v4df __builtin_ia32_xorpd256 (v4df,v4df)
19582 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
19585 The following built-in functions are available when @option{-mavx2} is
19586 used. All of them generate the machine instruction that is part of the
19590 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
19591 v32qi __builtin_ia32_pabsb256 (v32qi)
19592 v16hi __builtin_ia32_pabsw256 (v16hi)
19593 v8si __builtin_ia32_pabsd256 (v8si)
19594 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
19595 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
19596 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
19597 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
19598 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
19599 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
19600 v8si __builtin_ia32_paddd256 (v8si,v8si)
19601 v4di __builtin_ia32_paddq256 (v4di,v4di)
19602 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
19603 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
19604 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
19605 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
19606 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
19607 v4di __builtin_ia32_andsi256 (v4di,v4di)
19608 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
19609 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
19610 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
19611 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
19612 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
19613 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
19614 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
19615 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
19616 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
19617 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
19618 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
19619 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
19620 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
19621 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
19622 v8si __builtin_ia32_phaddd256 (v8si,v8si)
19623 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
19624 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
19625 v8si __builtin_ia32_phsubd256 (v8si,v8si)
19626 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
19627 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
19628 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
19629 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
19630 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
19631 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
19632 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
19633 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
19634 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
19635 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
19636 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
19637 v8si __builtin_ia32_pminsd256 (v8si,v8si)
19638 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
19639 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
19640 v8si __builtin_ia32_pminud256 (v8si,v8si)
19641 int __builtin_ia32_pmovmskb256 (v32qi)
19642 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
19643 v8si __builtin_ia32_pmovsxbd256 (v16qi)
19644 v4di __builtin_ia32_pmovsxbq256 (v16qi)
19645 v8si __builtin_ia32_pmovsxwd256 (v8hi)
19646 v4di __builtin_ia32_pmovsxwq256 (v8hi)
19647 v4di __builtin_ia32_pmovsxdq256 (v4si)
19648 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
19649 v8si __builtin_ia32_pmovzxbd256 (v16qi)
19650 v4di __builtin_ia32_pmovzxbq256 (v16qi)
19651 v8si __builtin_ia32_pmovzxwd256 (v8hi)
19652 v4di __builtin_ia32_pmovzxwq256 (v8hi)
19653 v4di __builtin_ia32_pmovzxdq256 (v4si)
19654 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
19655 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
19656 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
19657 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
19658 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
19659 v8si __builtin_ia32_pmulld256 (v8si,v8si)
19660 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
19661 v4di __builtin_ia32_por256 (v4di,v4di)
19662 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
19663 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
19664 v8si __builtin_ia32_pshufd256 (v8si,int)
19665 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
19666 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
19667 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
19668 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
19669 v8si __builtin_ia32_psignd256 (v8si,v8si)
19670 v4di __builtin_ia32_pslldqi256 (v4di,int)
19671 v16hi __builtin_ia32_psllwi256 (16hi,int)
19672 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
19673 v8si __builtin_ia32_pslldi256 (v8si,int)
19674 v8si __builtin_ia32_pslld256(v8si,v4si)
19675 v4di __builtin_ia32_psllqi256 (v4di,int)
19676 v4di __builtin_ia32_psllq256(v4di,v2di)
19677 v16hi __builtin_ia32_psrawi256 (v16hi,int)
19678 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
19679 v8si __builtin_ia32_psradi256 (v8si,int)
19680 v8si __builtin_ia32_psrad256 (v8si,v4si)
19681 v4di __builtin_ia32_psrldqi256 (v4di, int)
19682 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
19683 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
19684 v8si __builtin_ia32_psrldi256 (v8si,int)
19685 v8si __builtin_ia32_psrld256 (v8si,v4si)
19686 v4di __builtin_ia32_psrlqi256 (v4di,int)
19687 v4di __builtin_ia32_psrlq256(v4di,v2di)
19688 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
19689 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
19690 v8si __builtin_ia32_psubd256 (v8si,v8si)
19691 v4di __builtin_ia32_psubq256 (v4di,v4di)
19692 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
19693 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
19694 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
19695 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
19696 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
19697 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
19698 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
19699 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
19700 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
19701 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
19702 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
19703 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
19704 v4di __builtin_ia32_pxor256 (v4di,v4di)
19705 v4di __builtin_ia32_movntdqa256 (pv4di)
19706 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
19707 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
19708 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
19709 v4di __builtin_ia32_vbroadcastsi256 (v2di)
19710 v4si __builtin_ia32_pblendd128 (v4si,v4si)
19711 v8si __builtin_ia32_pblendd256 (v8si,v8si)
19712 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
19713 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
19714 v8si __builtin_ia32_pbroadcastd256 (v4si)
19715 v4di __builtin_ia32_pbroadcastq256 (v2di)
19716 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
19717 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
19718 v4si __builtin_ia32_pbroadcastd128 (v4si)
19719 v2di __builtin_ia32_pbroadcastq128 (v2di)
19720 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
19721 v4df __builtin_ia32_permdf256 (v4df,int)
19722 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
19723 v4di __builtin_ia32_permdi256 (v4di,int)
19724 v4di __builtin_ia32_permti256 (v4di,v4di,int)
19725 v4di __builtin_ia32_extract128i256 (v4di,int)
19726 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
19727 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
19728 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
19729 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
19730 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
19731 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
19732 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
19733 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
19734 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
19735 v8si __builtin_ia32_psllv8si (v8si,v8si)
19736 v4si __builtin_ia32_psllv4si (v4si,v4si)
19737 v4di __builtin_ia32_psllv4di (v4di,v4di)
19738 v2di __builtin_ia32_psllv2di (v2di,v2di)
19739 v8si __builtin_ia32_psrav8si (v8si,v8si)
19740 v4si __builtin_ia32_psrav4si (v4si,v4si)
19741 v8si __builtin_ia32_psrlv8si (v8si,v8si)
19742 v4si __builtin_ia32_psrlv4si (v4si,v4si)
19743 v4di __builtin_ia32_psrlv4di (v4di,v4di)
19744 v2di __builtin_ia32_psrlv2di (v2di,v2di)
19745 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
19746 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
19747 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
19748 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
19749 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
19750 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
19751 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
19752 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
19753 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
19754 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
19755 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
19756 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
19757 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
19758 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
19759 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
19760 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
19763 The following built-in functions are available when @option{-maes} is
19764 used. All of them generate the machine instruction that is part of the
19768 v2di __builtin_ia32_aesenc128 (v2di, v2di)
19769 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
19770 v2di __builtin_ia32_aesdec128 (v2di, v2di)
19771 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
19772 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
19773 v2di __builtin_ia32_aesimc128 (v2di)
19776 The following built-in function is available when @option{-mpclmul} is
19780 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
19781 Generates the @code{pclmulqdq} machine instruction.
19784 The following built-in function is available when @option{-mfsgsbase} is
19785 used. All of them generate the machine instruction that is part of the
19789 unsigned int __builtin_ia32_rdfsbase32 (void)
19790 unsigned long long __builtin_ia32_rdfsbase64 (void)
19791 unsigned int __builtin_ia32_rdgsbase32 (void)
19792 unsigned long long __builtin_ia32_rdgsbase64 (void)
19793 void _writefsbase_u32 (unsigned int)
19794 void _writefsbase_u64 (unsigned long long)
19795 void _writegsbase_u32 (unsigned int)
19796 void _writegsbase_u64 (unsigned long long)
19799 The following built-in function is available when @option{-mrdrnd} is
19800 used. All of them generate the machine instruction that is part of the
19804 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
19805 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
19806 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
19809 The following built-in functions are available when @option{-msse4a} is used.
19810 All of them generate the machine instruction that is part of the name.
19813 void __builtin_ia32_movntsd (double *, v2df)
19814 void __builtin_ia32_movntss (float *, v4sf)
19815 v2di __builtin_ia32_extrq (v2di, v16qi)
19816 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
19817 v2di __builtin_ia32_insertq (v2di, v2di)
19818 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
19821 The following built-in functions are available when @option{-mxop} is used.
19823 v2df __builtin_ia32_vfrczpd (v2df)
19824 v4sf __builtin_ia32_vfrczps (v4sf)
19825 v2df __builtin_ia32_vfrczsd (v2df)
19826 v4sf __builtin_ia32_vfrczss (v4sf)
19827 v4df __builtin_ia32_vfrczpd256 (v4df)
19828 v8sf __builtin_ia32_vfrczps256 (v8sf)
19829 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
19830 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
19831 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
19832 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
19833 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
19834 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
19835 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
19836 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
19837 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
19838 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
19839 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
19840 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
19841 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
19842 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
19843 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
19844 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
19845 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
19846 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
19847 v4si __builtin_ia32_vpcomequd (v4si, v4si)
19848 v2di __builtin_ia32_vpcomequq (v2di, v2di)
19849 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
19850 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
19851 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
19852 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
19853 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
19854 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
19855 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
19856 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
19857 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
19858 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
19859 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
19860 v4si __builtin_ia32_vpcomged (v4si, v4si)
19861 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
19862 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
19863 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
19864 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
19865 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
19866 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
19867 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
19868 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
19869 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
19870 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
19871 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
19872 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
19873 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
19874 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
19875 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
19876 v4si __builtin_ia32_vpcomled (v4si, v4si)
19877 v2di __builtin_ia32_vpcomleq (v2di, v2di)
19878 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
19879 v4si __builtin_ia32_vpcomleud (v4si, v4si)
19880 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
19881 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
19882 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
19883 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
19884 v4si __builtin_ia32_vpcomltd (v4si, v4si)
19885 v2di __builtin_ia32_vpcomltq (v2di, v2di)
19886 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
19887 v4si __builtin_ia32_vpcomltud (v4si, v4si)
19888 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
19889 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
19890 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
19891 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
19892 v4si __builtin_ia32_vpcomned (v4si, v4si)
19893 v2di __builtin_ia32_vpcomneq (v2di, v2di)
19894 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
19895 v4si __builtin_ia32_vpcomneud (v4si, v4si)
19896 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
19897 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
19898 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
19899 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
19900 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
19901 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
19902 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
19903 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
19904 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
19905 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
19906 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
19907 v4si __builtin_ia32_vphaddbd (v16qi)
19908 v2di __builtin_ia32_vphaddbq (v16qi)
19909 v8hi __builtin_ia32_vphaddbw (v16qi)
19910 v2di __builtin_ia32_vphadddq (v4si)
19911 v4si __builtin_ia32_vphaddubd (v16qi)
19912 v2di __builtin_ia32_vphaddubq (v16qi)
19913 v8hi __builtin_ia32_vphaddubw (v16qi)
19914 v2di __builtin_ia32_vphaddudq (v4si)
19915 v4si __builtin_ia32_vphadduwd (v8hi)
19916 v2di __builtin_ia32_vphadduwq (v8hi)
19917 v4si __builtin_ia32_vphaddwd (v8hi)
19918 v2di __builtin_ia32_vphaddwq (v8hi)
19919 v8hi __builtin_ia32_vphsubbw (v16qi)
19920 v2di __builtin_ia32_vphsubdq (v4si)
19921 v4si __builtin_ia32_vphsubwd (v8hi)
19922 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
19923 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
19924 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
19925 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
19926 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
19927 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
19928 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
19929 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
19930 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
19931 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
19932 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
19933 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
19934 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
19935 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
19936 v4si __builtin_ia32_vprotd (v4si, v4si)
19937 v2di __builtin_ia32_vprotq (v2di, v2di)
19938 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
19939 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
19940 v4si __builtin_ia32_vpshad (v4si, v4si)
19941 v2di __builtin_ia32_vpshaq (v2di, v2di)
19942 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
19943 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
19944 v4si __builtin_ia32_vpshld (v4si, v4si)
19945 v2di __builtin_ia32_vpshlq (v2di, v2di)
19946 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
19949 The following built-in functions are available when @option{-mfma4} is used.
19950 All of them generate the machine instruction that is part of the name.
19953 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
19954 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
19955 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
19956 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
19957 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
19958 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
19959 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
19960 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
19961 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
19962 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
19963 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
19964 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
19965 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
19966 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
19967 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
19968 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
19969 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
19970 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
19971 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
19972 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
19973 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
19974 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
19975 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
19976 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
19977 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
19978 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
19979 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
19980 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
19981 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
19982 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
19983 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
19984 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
19988 The following built-in functions are available when @option{-mlwp} is used.
19991 void __builtin_ia32_llwpcb16 (void *);
19992 void __builtin_ia32_llwpcb32 (void *);
19993 void __builtin_ia32_llwpcb64 (void *);
19994 void * __builtin_ia32_llwpcb16 (void);
19995 void * __builtin_ia32_llwpcb32 (void);
19996 void * __builtin_ia32_llwpcb64 (void);
19997 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
19998 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
19999 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
20000 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
20001 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
20002 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
20005 The following built-in functions are available when @option{-mbmi} is used.
20006 All of them generate the machine instruction that is part of the name.
20008 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
20009 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
20012 The following built-in functions are available when @option{-mbmi2} is used.
20013 All of them generate the machine instruction that is part of the name.
20015 unsigned int _bzhi_u32 (unsigned int, unsigned int)
20016 unsigned int _pdep_u32 (unsigned int, unsigned int)
20017 unsigned int _pext_u32 (unsigned int, unsigned int)
20018 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
20019 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
20020 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
20023 The following built-in functions are available when @option{-mlzcnt} is used.
20024 All of them generate the machine instruction that is part of the name.
20026 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
20027 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
20028 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
20031 The following built-in functions are available when @option{-mfxsr} is used.
20032 All of them generate the machine instruction that is part of the name.
20034 void __builtin_ia32_fxsave (void *)
20035 void __builtin_ia32_fxrstor (void *)
20036 void __builtin_ia32_fxsave64 (void *)
20037 void __builtin_ia32_fxrstor64 (void *)
20040 The following built-in functions are available when @option{-mxsave} is used.
20041 All of them generate the machine instruction that is part of the name.
20043 void __builtin_ia32_xsave (void *, long long)
20044 void __builtin_ia32_xrstor (void *, long long)
20045 void __builtin_ia32_xsave64 (void *, long long)
20046 void __builtin_ia32_xrstor64 (void *, long long)
20049 The following built-in functions are available when @option{-mxsaveopt} is used.
20050 All of them generate the machine instruction that is part of the name.
20052 void __builtin_ia32_xsaveopt (void *, long long)
20053 void __builtin_ia32_xsaveopt64 (void *, long long)
20056 The following built-in functions are available when @option{-mtbm} is used.
20057 Both of them generate the immediate form of the bextr machine instruction.
20059 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
20060 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
20064 The following built-in functions are available when @option{-m3dnow} is used.
20065 All of them generate the machine instruction that is part of the name.
20068 void __builtin_ia32_femms (void)
20069 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
20070 v2si __builtin_ia32_pf2id (v2sf)
20071 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
20072 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
20073 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
20074 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
20075 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
20076 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
20077 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
20078 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
20079 v2sf __builtin_ia32_pfrcp (v2sf)
20080 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
20081 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
20082 v2sf __builtin_ia32_pfrsqrt (v2sf)
20083 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
20084 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
20085 v2sf __builtin_ia32_pi2fd (v2si)
20086 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
20089 The following built-in functions are available when both @option{-m3dnow}
20090 and @option{-march=athlon} are used. All of them generate the machine
20091 instruction that is part of the name.
20094 v2si __builtin_ia32_pf2iw (v2sf)
20095 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
20096 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
20097 v2sf __builtin_ia32_pi2fw (v2si)
20098 v2sf __builtin_ia32_pswapdsf (v2sf)
20099 v2si __builtin_ia32_pswapdsi (v2si)
20102 The following built-in functions are available when @option{-mrtm} is used
20103 They are used for restricted transactional memory. These are the internal
20104 low level functions. Normally the functions in
20105 @ref{x86 transactional memory intrinsics} should be used instead.
20108 int __builtin_ia32_xbegin ()
20109 void __builtin_ia32_xend ()
20110 void __builtin_ia32_xabort (status)
20111 int __builtin_ia32_xtest ()
20114 The following built-in functions are available when @option{-mmwaitx} is used.
20115 All of them generate the machine instruction that is part of the name.
20117 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
20118 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
20121 The following built-in functions are available when @option{-mclzero} is used.
20122 All of them generate the machine instruction that is part of the name.
20124 void __builtin_i32_clzero (void *)
20127 The following built-in functions are available when @option{-mpku} is used.
20128 They generate reads and writes to PKRU.
20130 void __builtin_ia32_wrpkru (unsigned int)
20131 unsigned int __builtin_ia32_rdpkru ()
20134 @node x86 transactional memory intrinsics
20135 @subsection x86 Transactional Memory Intrinsics
20137 These hardware transactional memory intrinsics for x86 allow you to use
20138 memory transactions with RTM (Restricted Transactional Memory).
20139 This support is enabled with the @option{-mrtm} option.
20140 For using HLE (Hardware Lock Elision) see
20141 @ref{x86 specific memory model extensions for transactional memory} instead.
20143 A memory transaction commits all changes to memory in an atomic way,
20144 as visible to other threads. If the transaction fails it is rolled back
20145 and all side effects discarded.
20147 Generally there is no guarantee that a memory transaction ever succeeds
20148 and suitable fallback code always needs to be supplied.
20150 @deftypefn {RTM Function} {unsigned} _xbegin ()
20151 Start a RTM (Restricted Transactional Memory) transaction.
20152 Returns @code{_XBEGIN_STARTED} when the transaction
20153 started successfully (note this is not 0, so the constant has to be
20154 explicitly tested).
20156 If the transaction aborts, all side-effects
20157 are undone and an abort code encoded as a bit mask is returned.
20158 The following macros are defined:
20161 @item _XABORT_EXPLICIT
20162 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
20163 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
20164 @item _XABORT_RETRY
20165 Transaction retry is possible.
20166 @item _XABORT_CONFLICT
20167 Transaction abort due to a memory conflict with another thread.
20168 @item _XABORT_CAPACITY
20169 Transaction abort due to the transaction using too much memory.
20170 @item _XABORT_DEBUG
20171 Transaction abort due to a debug trap.
20172 @item _XABORT_NESTED
20173 Transaction abort in an inner nested transaction.
20176 There is no guarantee
20177 any transaction ever succeeds, so there always needs to be a valid
20181 @deftypefn {RTM Function} {void} _xend ()
20182 Commit the current transaction. When no transaction is active this faults.
20183 All memory side-effects of the transaction become visible
20184 to other threads in an atomic manner.
20187 @deftypefn {RTM Function} {int} _xtest ()
20188 Return a nonzero value if a transaction is currently active, otherwise 0.
20191 @deftypefn {RTM Function} {void} _xabort (status)
20192 Abort the current transaction. When no transaction is active this is a no-op.
20193 The @var{status} is an 8-bit constant; its value is encoded in the return
20194 value from @code{_xbegin}.
20197 Here is an example showing handling for @code{_XABORT_RETRY}
20198 and a fallback path for other failures:
20201 #include <immintrin.h>
20203 int n_tries, max_tries;
20204 unsigned status = _XABORT_EXPLICIT;
20207 for (n_tries = 0; n_tries < max_tries; n_tries++)
20209 status = _xbegin ();
20210 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
20213 if (status == _XBEGIN_STARTED)
20215 ... transaction code...
20220 ... non-transactional fallback path...
20225 Note that, in most cases, the transactional and non-transactional code
20226 must synchronize together to ensure consistency.
20228 @node Target Format Checks
20229 @section Format Checks Specific to Particular Target Machines
20231 For some target machines, GCC supports additional options to the
20233 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
20236 * Solaris Format Checks::
20237 * Darwin Format Checks::
20240 @node Solaris Format Checks
20241 @subsection Solaris Format Checks
20243 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
20244 check. @code{cmn_err} accepts a subset of the standard @code{printf}
20245 conversions, and the two-argument @code{%b} conversion for displaying
20246 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
20248 @node Darwin Format Checks
20249 @subsection Darwin Format Checks
20251 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
20252 attribute context. Declarations made with such attribution are parsed for correct syntax
20253 and format argument types. However, parsing of the format string itself is currently undefined
20254 and is not carried out by this version of the compiler.
20256 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
20257 also be used as format arguments. Note that the relevant headers are only likely to be
20258 available on Darwin (OSX) installations. On such installations, the XCode and system
20259 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
20260 associated functions.
20263 @section Pragmas Accepted by GCC
20265 @cindex @code{#pragma}
20267 GCC supports several types of pragmas, primarily in order to compile
20268 code originally written for other compilers. Note that in general
20269 we do not recommend the use of pragmas; @xref{Function Attributes},
20270 for further explanation.
20273 * AArch64 Pragmas::
20277 * RS/6000 and PowerPC Pragmas::
20280 * Solaris Pragmas::
20281 * Symbol-Renaming Pragmas::
20282 * Structure-Layout Pragmas::
20284 * Diagnostic Pragmas::
20285 * Visibility Pragmas::
20286 * Push/Pop Macro Pragmas::
20287 * Function Specific Option Pragmas::
20288 * Loop-Specific Pragmas::
20291 @node AArch64 Pragmas
20292 @subsection AArch64 Pragmas
20294 The pragmas defined by the AArch64 target correspond to the AArch64
20295 target function attributes. They can be specified as below:
20297 #pragma GCC target("string")
20300 where @code{@var{string}} can be any string accepted as an AArch64 target
20301 attribute. @xref{AArch64 Function Attributes}, for more details
20302 on the permissible values of @code{string}.
20305 @subsection ARM Pragmas
20307 The ARM target defines pragmas for controlling the default addition of
20308 @code{long_call} and @code{short_call} attributes to functions.
20309 @xref{Function Attributes}, for information about the effects of these
20314 @cindex pragma, long_calls
20315 Set all subsequent functions to have the @code{long_call} attribute.
20317 @item no_long_calls
20318 @cindex pragma, no_long_calls
20319 Set all subsequent functions to have the @code{short_call} attribute.
20321 @item long_calls_off
20322 @cindex pragma, long_calls_off
20323 Do not affect the @code{long_call} or @code{short_call} attributes of
20324 subsequent functions.
20328 @subsection M32C Pragmas
20331 @item GCC memregs @var{number}
20332 @cindex pragma, memregs
20333 Overrides the command-line option @code{-memregs=} for the current
20334 file. Use with care! This pragma must be before any function in the
20335 file, and mixing different memregs values in different objects may
20336 make them incompatible. This pragma is useful when a
20337 performance-critical function uses a memreg for temporary values,
20338 as it may allow you to reduce the number of memregs used.
20340 @item ADDRESS @var{name} @var{address}
20341 @cindex pragma, address
20342 For any declared symbols matching @var{name}, this does three things
20343 to that symbol: it forces the symbol to be located at the given
20344 address (a number), it forces the symbol to be volatile, and it
20345 changes the symbol's scope to be static. This pragma exists for
20346 compatibility with other compilers, but note that the common
20347 @code{1234H} numeric syntax is not supported (use @code{0x1234}
20351 #pragma ADDRESS port3 0x103
20358 @subsection MeP Pragmas
20362 @item custom io_volatile (on|off)
20363 @cindex pragma, custom io_volatile
20364 Overrides the command-line option @code{-mio-volatile} for the current
20365 file. Note that for compatibility with future GCC releases, this
20366 option should only be used once before any @code{io} variables in each
20369 @item GCC coprocessor available @var{registers}
20370 @cindex pragma, coprocessor available
20371 Specifies which coprocessor registers are available to the register
20372 allocator. @var{registers} may be a single register, register range
20373 separated by ellipses, or comma-separated list of those. Example:
20376 #pragma GCC coprocessor available $c0...$c10, $c28
20379 @item GCC coprocessor call_saved @var{registers}
20380 @cindex pragma, coprocessor call_saved
20381 Specifies which coprocessor registers are to be saved and restored by
20382 any function using them. @var{registers} may be a single register,
20383 register range separated by ellipses, or comma-separated list of
20387 #pragma GCC coprocessor call_saved $c4...$c6, $c31
20390 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
20391 @cindex pragma, coprocessor subclass
20392 Creates and defines a register class. These register classes can be
20393 used by inline @code{asm} constructs. @var{registers} may be a single
20394 register, register range separated by ellipses, or comma-separated
20395 list of those. Example:
20398 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
20400 asm ("cpfoo %0" : "=B" (x));
20403 @item GCC disinterrupt @var{name} , @var{name} @dots{}
20404 @cindex pragma, disinterrupt
20405 For the named functions, the compiler adds code to disable interrupts
20406 for the duration of those functions. If any functions so named
20407 are not encountered in the source, a warning is emitted that the pragma is
20408 not used. Examples:
20411 #pragma disinterrupt foo
20412 #pragma disinterrupt bar, grill
20413 int foo () @{ @dots{} @}
20416 @item GCC call @var{name} , @var{name} @dots{}
20417 @cindex pragma, call
20418 For the named functions, the compiler always uses a register-indirect
20419 call model when calling the named functions. Examples:
20428 @node RS/6000 and PowerPC Pragmas
20429 @subsection RS/6000 and PowerPC Pragmas
20431 The RS/6000 and PowerPC targets define one pragma for controlling
20432 whether or not the @code{longcall} attribute is added to function
20433 declarations by default. This pragma overrides the @option{-mlongcall}
20434 option, but not the @code{longcall} and @code{shortcall} attributes.
20435 @xref{RS/6000 and PowerPC Options}, for more information about when long
20436 calls are and are not necessary.
20440 @cindex pragma, longcall
20441 Apply the @code{longcall} attribute to all subsequent function
20445 Do not apply the @code{longcall} attribute to subsequent function
20449 @c Describe h8300 pragmas here.
20450 @c Describe sh pragmas here.
20451 @c Describe v850 pragmas here.
20453 @node S/390 Pragmas
20454 @subsection S/390 Pragmas
20456 The pragmas defined by the S/390 target correspond to the S/390
20457 target function attributes and some the additional options:
20464 Note that options of the pragma, unlike options of the target
20465 attribute, do change the value of preprocessor macros like
20466 @code{__VEC__}. They can be specified as below:
20469 #pragma GCC target("string[,string]...")
20470 #pragma GCC target("string"[,"string"]...)
20473 @node Darwin Pragmas
20474 @subsection Darwin Pragmas
20476 The following pragmas are available for all architectures running the
20477 Darwin operating system. These are useful for compatibility with other
20481 @item mark @var{tokens}@dots{}
20482 @cindex pragma, mark
20483 This pragma is accepted, but has no effect.
20485 @item options align=@var{alignment}
20486 @cindex pragma, options align
20487 This pragma sets the alignment of fields in structures. The values of
20488 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
20489 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
20490 properly; to restore the previous setting, use @code{reset} for the
20493 @item segment @var{tokens}@dots{}
20494 @cindex pragma, segment
20495 This pragma is accepted, but has no effect.
20497 @item unused (@var{var} [, @var{var}]@dots{})
20498 @cindex pragma, unused
20499 This pragma declares variables to be possibly unused. GCC does not
20500 produce warnings for the listed variables. The effect is similar to
20501 that of the @code{unused} attribute, except that this pragma may appear
20502 anywhere within the variables' scopes.
20505 @node Solaris Pragmas
20506 @subsection Solaris Pragmas
20508 The Solaris target supports @code{#pragma redefine_extname}
20509 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
20510 @code{#pragma} directives for compatibility with the system compiler.
20513 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
20514 @cindex pragma, align
20516 Increase the minimum alignment of each @var{variable} to @var{alignment}.
20517 This is the same as GCC's @code{aligned} attribute @pxref{Variable
20518 Attributes}). Macro expansion occurs on the arguments to this pragma
20519 when compiling C and Objective-C@. It does not currently occur when
20520 compiling C++, but this is a bug which may be fixed in a future
20523 @item fini (@var{function} [, @var{function}]...)
20524 @cindex pragma, fini
20526 This pragma causes each listed @var{function} to be called after
20527 main, or during shared module unloading, by adding a call to the
20528 @code{.fini} section.
20530 @item init (@var{function} [, @var{function}]...)
20531 @cindex pragma, init
20533 This pragma causes each listed @var{function} to be called during
20534 initialization (before @code{main}) or during shared module loading, by
20535 adding a call to the @code{.init} section.
20539 @node Symbol-Renaming Pragmas
20540 @subsection Symbol-Renaming Pragmas
20542 GCC supports a @code{#pragma} directive that changes the name used in
20543 assembly for a given declaration. While this pragma is supported on all
20544 platforms, it is intended primarily to provide compatibility with the
20545 Solaris system headers. This effect can also be achieved using the asm
20546 labels extension (@pxref{Asm Labels}).
20549 @item redefine_extname @var{oldname} @var{newname}
20550 @cindex pragma, redefine_extname
20552 This pragma gives the C function @var{oldname} the assembly symbol
20553 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
20554 is defined if this pragma is available (currently on all platforms).
20557 This pragma and the asm labels extension interact in a complicated
20558 manner. Here are some corner cases you may want to be aware of:
20561 @item This pragma silently applies only to declarations with external
20562 linkage. Asm labels do not have this restriction.
20564 @item In C++, this pragma silently applies only to declarations with
20565 ``C'' linkage. Again, asm labels do not have this restriction.
20567 @item If either of the ways of changing the assembly name of a
20568 declaration are applied to a declaration whose assembly name has
20569 already been determined (either by a previous use of one of these
20570 features, or because the compiler needed the assembly name in order to
20571 generate code), and the new name is different, a warning issues and
20572 the name does not change.
20574 @item The @var{oldname} used by @code{#pragma redefine_extname} is
20575 always the C-language name.
20578 @node Structure-Layout Pragmas
20579 @subsection Structure-Layout Pragmas
20581 For compatibility with Microsoft Windows compilers, GCC supports a
20582 set of @code{#pragma} directives that change the maximum alignment of
20583 members of structures (other than zero-width bit-fields), unions, and
20584 classes subsequently defined. The @var{n} value below always is required
20585 to be a small power of two and specifies the new alignment in bytes.
20588 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
20589 @item @code{#pragma pack()} sets the alignment to the one that was in
20590 effect when compilation started (see also command-line option
20591 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
20592 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
20593 setting on an internal stack and then optionally sets the new alignment.
20594 @item @code{#pragma pack(pop)} restores the alignment setting to the one
20595 saved at the top of the internal stack (and removes that stack entry).
20596 Note that @code{#pragma pack([@var{n}])} does not influence this internal
20597 stack; thus it is possible to have @code{#pragma pack(push)} followed by
20598 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
20599 @code{#pragma pack(pop)}.
20602 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
20603 directive which lays out structures and unions subsequently defined as the
20604 documented @code{__attribute__ ((ms_struct))}.
20607 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
20608 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
20609 @item @code{#pragma ms_struct reset} goes back to the default layout.
20612 Most targets also support the @code{#pragma scalar_storage_order} directive
20613 which lays out structures and unions subsequently defined as the documented
20614 @code{__attribute__ ((scalar_storage_order))}.
20617 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
20618 of the scalar fields to big-endian.
20619 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
20620 of the scalar fields to little-endian.
20621 @item @code{#pragma scalar_storage_order default} goes back to the endianness
20622 that was in effect when compilation started (see also command-line option
20623 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
20627 @subsection Weak Pragmas
20629 For compatibility with SVR4, GCC supports a set of @code{#pragma}
20630 directives for declaring symbols to be weak, and defining weak
20634 @item #pragma weak @var{symbol}
20635 @cindex pragma, weak
20636 This pragma declares @var{symbol} to be weak, as if the declaration
20637 had the attribute of the same name. The pragma may appear before
20638 or after the declaration of @var{symbol}. It is not an error for
20639 @var{symbol} to never be defined at all.
20641 @item #pragma weak @var{symbol1} = @var{symbol2}
20642 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
20643 It is an error if @var{symbol2} is not defined in the current
20647 @node Diagnostic Pragmas
20648 @subsection Diagnostic Pragmas
20650 GCC allows the user to selectively enable or disable certain types of
20651 diagnostics, and change the kind of the diagnostic. For example, a
20652 project's policy might require that all sources compile with
20653 @option{-Werror} but certain files might have exceptions allowing
20654 specific types of warnings. Or, a project might selectively enable
20655 diagnostics and treat them as errors depending on which preprocessor
20656 macros are defined.
20659 @item #pragma GCC diagnostic @var{kind} @var{option}
20660 @cindex pragma, diagnostic
20662 Modifies the disposition of a diagnostic. Note that not all
20663 diagnostics are modifiable; at the moment only warnings (normally
20664 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
20665 Use @option{-fdiagnostics-show-option} to determine which diagnostics
20666 are controllable and which option controls them.
20668 @var{kind} is @samp{error} to treat this diagnostic as an error,
20669 @samp{warning} to treat it like a warning (even if @option{-Werror} is
20670 in effect), or @samp{ignored} if the diagnostic is to be ignored.
20671 @var{option} is a double quoted string that matches the command-line
20675 #pragma GCC diagnostic warning "-Wformat"
20676 #pragma GCC diagnostic error "-Wformat"
20677 #pragma GCC diagnostic ignored "-Wformat"
20680 Note that these pragmas override any command-line options. GCC keeps
20681 track of the location of each pragma, and issues diagnostics according
20682 to the state as of that point in the source file. Thus, pragmas occurring
20683 after a line do not affect diagnostics caused by that line.
20685 @item #pragma GCC diagnostic push
20686 @itemx #pragma GCC diagnostic pop
20688 Causes GCC to remember the state of the diagnostics as of each
20689 @code{push}, and restore to that point at each @code{pop}. If a
20690 @code{pop} has no matching @code{push}, the command-line options are
20694 #pragma GCC diagnostic error "-Wuninitialized"
20695 foo(a); /* error is given for this one */
20696 #pragma GCC diagnostic push
20697 #pragma GCC diagnostic ignored "-Wuninitialized"
20698 foo(b); /* no diagnostic for this one */
20699 #pragma GCC diagnostic pop
20700 foo(c); /* error is given for this one */
20701 #pragma GCC diagnostic pop
20702 foo(d); /* depends on command-line options */
20707 GCC also offers a simple mechanism for printing messages during
20711 @item #pragma message @var{string}
20712 @cindex pragma, diagnostic
20714 Prints @var{string} as a compiler message on compilation. The message
20715 is informational only, and is neither a compilation warning nor an error.
20718 #pragma message "Compiling " __FILE__ "..."
20721 @var{string} may be parenthesized, and is printed with location
20722 information. For example,
20725 #define DO_PRAGMA(x) _Pragma (#x)
20726 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
20728 TODO(Remember to fix this)
20732 prints @samp{/tmp/file.c:4: note: #pragma message:
20733 TODO - Remember to fix this}.
20737 @node Visibility Pragmas
20738 @subsection Visibility Pragmas
20741 @item #pragma GCC visibility push(@var{visibility})
20742 @itemx #pragma GCC visibility pop
20743 @cindex pragma, visibility
20745 This pragma allows the user to set the visibility for multiple
20746 declarations without having to give each a visibility attribute
20747 (@pxref{Function Attributes}).
20749 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
20750 declarations. Class members and template specializations are not
20751 affected; if you want to override the visibility for a particular
20752 member or instantiation, you must use an attribute.
20757 @node Push/Pop Macro Pragmas
20758 @subsection Push/Pop Macro Pragmas
20760 For compatibility with Microsoft Windows compilers, GCC supports
20761 @samp{#pragma push_macro(@var{"macro_name"})}
20762 and @samp{#pragma pop_macro(@var{"macro_name"})}.
20765 @item #pragma push_macro(@var{"macro_name"})
20766 @cindex pragma, push_macro
20767 This pragma saves the value of the macro named as @var{macro_name} to
20768 the top of the stack for this macro.
20770 @item #pragma pop_macro(@var{"macro_name"})
20771 @cindex pragma, pop_macro
20772 This pragma sets the value of the macro named as @var{macro_name} to
20773 the value on top of the stack for this macro. If the stack for
20774 @var{macro_name} is empty, the value of the macro remains unchanged.
20781 #pragma push_macro("X")
20784 #pragma pop_macro("X")
20789 In this example, the definition of X as 1 is saved by @code{#pragma
20790 push_macro} and restored by @code{#pragma pop_macro}.
20792 @node Function Specific Option Pragmas
20793 @subsection Function Specific Option Pragmas
20796 @item #pragma GCC target (@var{"string"}...)
20797 @cindex pragma GCC target
20799 This pragma allows you to set target specific options for functions
20800 defined later in the source file. One or more strings can be
20801 specified. Each function that is defined after this point is as
20802 if @code{attribute((target("STRING")))} was specified for that
20803 function. The parenthesis around the options is optional.
20804 @xref{Function Attributes}, for more information about the
20805 @code{target} attribute and the attribute syntax.
20807 The @code{#pragma GCC target} pragma is presently implemented for
20808 x86, PowerPC, and Nios II targets only.
20812 @item #pragma GCC optimize (@var{"string"}...)
20813 @cindex pragma GCC optimize
20815 This pragma allows you to set global optimization options for functions
20816 defined later in the source file. One or more strings can be
20817 specified. Each function that is defined after this point is as
20818 if @code{attribute((optimize("STRING")))} was specified for that
20819 function. The parenthesis around the options is optional.
20820 @xref{Function Attributes}, for more information about the
20821 @code{optimize} attribute and the attribute syntax.
20825 @item #pragma GCC push_options
20826 @itemx #pragma GCC pop_options
20827 @cindex pragma GCC push_options
20828 @cindex pragma GCC pop_options
20830 These pragmas maintain a stack of the current target and optimization
20831 options. It is intended for include files where you temporarily want
20832 to switch to using a different @samp{#pragma GCC target} or
20833 @samp{#pragma GCC optimize} and then to pop back to the previous
20838 @item #pragma GCC reset_options
20839 @cindex pragma GCC reset_options
20841 This pragma clears the current @code{#pragma GCC target} and
20842 @code{#pragma GCC optimize} to use the default switches as specified
20843 on the command line.
20846 @node Loop-Specific Pragmas
20847 @subsection Loop-Specific Pragmas
20850 @item #pragma GCC ivdep
20851 @cindex pragma GCC ivdep
20854 With this pragma, the programmer asserts that there are no loop-carried
20855 dependencies which would prevent consecutive iterations of
20856 the following loop from executing concurrently with SIMD
20857 (single instruction multiple data) instructions.
20859 For example, the compiler can only unconditionally vectorize the following
20860 loop with the pragma:
20863 void foo (int n, int *a, int *b, int *c)
20867 for (i = 0; i < n; ++i)
20868 a[i] = b[i] + c[i];
20873 In this example, using the @code{restrict} qualifier had the same
20874 effect. In the following example, that would not be possible. Assume
20875 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
20876 that it can unconditionally vectorize the following loop:
20879 void ignore_vec_dep (int *a, int k, int c, int m)
20882 for (int i = 0; i < m; i++)
20883 a[i] = a[i + k] * c;
20888 @node Unnamed Fields
20889 @section Unnamed Structure and Union Fields
20890 @cindex @code{struct}
20891 @cindex @code{union}
20893 As permitted by ISO C11 and for compatibility with other compilers,
20894 GCC allows you to define
20895 a structure or union that contains, as fields, structures and unions
20896 without names. For example:
20910 In this example, you are able to access members of the unnamed
20911 union with code like @samp{foo.b}. Note that only unnamed structs and
20912 unions are allowed, you may not have, for example, an unnamed
20915 You must never create such structures that cause ambiguous field definitions.
20916 For example, in this structure:
20928 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
20929 The compiler gives errors for such constructs.
20931 @opindex fms-extensions
20932 Unless @option{-fms-extensions} is used, the unnamed field must be a
20933 structure or union definition without a tag (for example, @samp{struct
20934 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
20935 also be a definition with a tag such as @samp{struct foo @{ int a;
20936 @};}, a reference to a previously defined structure or union such as
20937 @samp{struct foo;}, or a reference to a @code{typedef} name for a
20938 previously defined structure or union type.
20940 @opindex fplan9-extensions
20941 The option @option{-fplan9-extensions} enables
20942 @option{-fms-extensions} as well as two other extensions. First, a
20943 pointer to a structure is automatically converted to a pointer to an
20944 anonymous field for assignments and function calls. For example:
20947 struct s1 @{ int a; @};
20948 struct s2 @{ struct s1; @};
20949 extern void f1 (struct s1 *);
20950 void f2 (struct s2 *p) @{ f1 (p); @}
20954 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
20955 converted into a pointer to the anonymous field.
20957 Second, when the type of an anonymous field is a @code{typedef} for a
20958 @code{struct} or @code{union}, code may refer to the field using the
20959 name of the @code{typedef}.
20962 typedef struct @{ int a; @} s1;
20963 struct s2 @{ s1; @};
20964 s1 f1 (struct s2 *p) @{ return p->s1; @}
20967 These usages are only permitted when they are not ambiguous.
20970 @section Thread-Local Storage
20971 @cindex Thread-Local Storage
20972 @cindex @acronym{TLS}
20973 @cindex @code{__thread}
20975 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
20976 are allocated such that there is one instance of the variable per extant
20977 thread. The runtime model GCC uses to implement this originates
20978 in the IA-64 processor-specific ABI, but has since been migrated
20979 to other processors as well. It requires significant support from
20980 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
20981 system libraries (@file{libc.so} and @file{libpthread.so}), so it
20982 is not available everywhere.
20984 At the user level, the extension is visible with a new storage
20985 class keyword: @code{__thread}. For example:
20989 extern __thread struct state s;
20990 static __thread char *p;
20993 The @code{__thread} specifier may be used alone, with the @code{extern}
20994 or @code{static} specifiers, but with no other storage class specifier.
20995 When used with @code{extern} or @code{static}, @code{__thread} must appear
20996 immediately after the other storage class specifier.
20998 The @code{__thread} specifier may be applied to any global, file-scoped
20999 static, function-scoped static, or static data member of a class. It may
21000 not be applied to block-scoped automatic or non-static data member.
21002 When the address-of operator is applied to a thread-local variable, it is
21003 evaluated at run time and returns the address of the current thread's
21004 instance of that variable. An address so obtained may be used by any
21005 thread. When a thread terminates, any pointers to thread-local variables
21006 in that thread become invalid.
21008 No static initialization may refer to the address of a thread-local variable.
21010 In C++, if an initializer is present for a thread-local variable, it must
21011 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
21014 See @uref{http://www.akkadia.org/drepper/tls.pdf,
21015 ELF Handling For Thread-Local Storage} for a detailed explanation of
21016 the four thread-local storage addressing models, and how the runtime
21017 is expected to function.
21020 * C99 Thread-Local Edits::
21021 * C++98 Thread-Local Edits::
21024 @node C99 Thread-Local Edits
21025 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
21027 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
21028 that document the exact semantics of the language extension.
21032 @cite{5.1.2 Execution environments}
21034 Add new text after paragraph 1
21037 Within either execution environment, a @dfn{thread} is a flow of
21038 control within a program. It is implementation defined whether
21039 or not there may be more than one thread associated with a program.
21040 It is implementation defined how threads beyond the first are
21041 created, the name and type of the function called at thread
21042 startup, and how threads may be terminated. However, objects
21043 with thread storage duration shall be initialized before thread
21048 @cite{6.2.4 Storage durations of objects}
21050 Add new text before paragraph 3
21053 An object whose identifier is declared with the storage-class
21054 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
21055 Its lifetime is the entire execution of the thread, and its
21056 stored value is initialized only once, prior to thread startup.
21060 @cite{6.4.1 Keywords}
21062 Add @code{__thread}.
21065 @cite{6.7.1 Storage-class specifiers}
21067 Add @code{__thread} to the list of storage class specifiers in
21070 Change paragraph 2 to
21073 With the exception of @code{__thread}, at most one storage-class
21074 specifier may be given [@dots{}]. The @code{__thread} specifier may
21075 be used alone, or immediately following @code{extern} or
21079 Add new text after paragraph 6
21082 The declaration of an identifier for a variable that has
21083 block scope that specifies @code{__thread} shall also
21084 specify either @code{extern} or @code{static}.
21086 The @code{__thread} specifier shall be used only with
21091 @node C++98 Thread-Local Edits
21092 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
21094 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
21095 that document the exact semantics of the language extension.
21099 @b{[intro.execution]}
21101 New text after paragraph 4
21104 A @dfn{thread} is a flow of control within the abstract machine.
21105 It is implementation defined whether or not there may be more than
21109 New text after paragraph 7
21112 It is unspecified whether additional action must be taken to
21113 ensure when and whether side effects are visible to other threads.
21119 Add @code{__thread}.
21122 @b{[basic.start.main]}
21124 Add after paragraph 5
21127 The thread that begins execution at the @code{main} function is called
21128 the @dfn{main thread}. It is implementation defined how functions
21129 beginning threads other than the main thread are designated or typed.
21130 A function so designated, as well as the @code{main} function, is called
21131 a @dfn{thread startup function}. It is implementation defined what
21132 happens if a thread startup function returns. It is implementation
21133 defined what happens to other threads when any thread calls @code{exit}.
21137 @b{[basic.start.init]}
21139 Add after paragraph 4
21142 The storage for an object of thread storage duration shall be
21143 statically initialized before the first statement of the thread startup
21144 function. An object of thread storage duration shall not require
21145 dynamic initialization.
21149 @b{[basic.start.term]}
21151 Add after paragraph 3
21154 The type of an object with thread storage duration shall not have a
21155 non-trivial destructor, nor shall it be an array type whose elements
21156 (directly or indirectly) have non-trivial destructors.
21162 Add ``thread storage duration'' to the list in paragraph 1.
21167 Thread, static, and automatic storage durations are associated with
21168 objects introduced by declarations [@dots{}].
21171 Add @code{__thread} to the list of specifiers in paragraph 3.
21174 @b{[basic.stc.thread]}
21176 New section before @b{[basic.stc.static]}
21179 The keyword @code{__thread} applied to a non-local object gives the
21180 object thread storage duration.
21182 A local variable or class data member declared both @code{static}
21183 and @code{__thread} gives the variable or member thread storage
21188 @b{[basic.stc.static]}
21193 All objects that have neither thread storage duration, dynamic
21194 storage duration nor are local [@dots{}].
21200 Add @code{__thread} to the list in paragraph 1.
21205 With the exception of @code{__thread}, at most one
21206 @var{storage-class-specifier} shall appear in a given
21207 @var{decl-specifier-seq}. The @code{__thread} specifier may
21208 be used alone, or immediately following the @code{extern} or
21209 @code{static} specifiers. [@dots{}]
21212 Add after paragraph 5
21215 The @code{__thread} specifier can be applied only to the names of objects
21216 and to anonymous unions.
21222 Add after paragraph 6
21225 Non-@code{static} members shall not be @code{__thread}.
21229 @node Binary constants
21230 @section Binary Constants using the @samp{0b} Prefix
21231 @cindex Binary constants using the @samp{0b} prefix
21233 Integer constants can be written as binary constants, consisting of a
21234 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
21235 @samp{0B}. This is particularly useful in environments that operate a
21236 lot on the bit level (like microcontrollers).
21238 The following statements are identical:
21247 The type of these constants follows the same rules as for octal or
21248 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
21251 @node C++ Extensions
21252 @chapter Extensions to the C++ Language
21253 @cindex extensions, C++ language
21254 @cindex C++ language extensions
21256 The GNU compiler provides these extensions to the C++ language (and you
21257 can also use most of the C language extensions in your C++ programs). If you
21258 want to write code that checks whether these features are available, you can
21259 test for the GNU compiler the same way as for C programs: check for a
21260 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
21261 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
21262 Predefined Macros,cpp,The GNU C Preprocessor}).
21265 * C++ Volatiles:: What constitutes an access to a volatile object.
21266 * Restricted Pointers:: C99 restricted pointers and references.
21267 * Vague Linkage:: Where G++ puts inlines, vtables and such.
21268 * C++ Interface:: You can use a single C++ header file for both
21269 declarations and definitions.
21270 * Template Instantiation:: Methods for ensuring that exactly one copy of
21271 each needed template instantiation is emitted.
21272 * Bound member functions:: You can extract a function pointer to the
21273 method denoted by a @samp{->*} or @samp{.*} expression.
21274 * C++ Attributes:: Variable, function, and type attributes for C++ only.
21275 * Function Multiversioning:: Declaring multiple function versions.
21276 * Namespace Association:: Strong using-directives for namespace association.
21277 * Type Traits:: Compiler support for type traits.
21278 * C++ Concepts:: Improved support for generic programming.
21279 * Java Exceptions:: Tweaking exception handling to work with Java.
21280 * Deprecated Features:: Things will disappear from G++.
21281 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
21284 @node C++ Volatiles
21285 @section When is a Volatile C++ Object Accessed?
21286 @cindex accessing volatiles
21287 @cindex volatile read
21288 @cindex volatile write
21289 @cindex volatile access
21291 The C++ standard differs from the C standard in its treatment of
21292 volatile objects. It fails to specify what constitutes a volatile
21293 access, except to say that C++ should behave in a similar manner to C
21294 with respect to volatiles, where possible. However, the different
21295 lvalueness of expressions between C and C++ complicate the behavior.
21296 G++ behaves the same as GCC for volatile access, @xref{C
21297 Extensions,,Volatiles}, for a description of GCC's behavior.
21299 The C and C++ language specifications differ when an object is
21300 accessed in a void context:
21303 volatile int *src = @var{somevalue};
21307 The C++ standard specifies that such expressions do not undergo lvalue
21308 to rvalue conversion, and that the type of the dereferenced object may
21309 be incomplete. The C++ standard does not specify explicitly that it
21310 is lvalue to rvalue conversion that is responsible for causing an
21311 access. There is reason to believe that it is, because otherwise
21312 certain simple expressions become undefined. However, because it
21313 would surprise most programmers, G++ treats dereferencing a pointer to
21314 volatile object of complete type as GCC would do for an equivalent
21315 type in C@. When the object has incomplete type, G++ issues a
21316 warning; if you wish to force an error, you must force a conversion to
21317 rvalue with, for instance, a static cast.
21319 When using a reference to volatile, G++ does not treat equivalent
21320 expressions as accesses to volatiles, but instead issues a warning that
21321 no volatile is accessed. The rationale for this is that otherwise it
21322 becomes difficult to determine where volatile access occur, and not
21323 possible to ignore the return value from functions returning volatile
21324 references. Again, if you wish to force a read, cast the reference to
21327 G++ implements the same behavior as GCC does when assigning to a
21328 volatile object---there is no reread of the assigned-to object, the
21329 assigned rvalue is reused. Note that in C++ assignment expressions
21330 are lvalues, and if used as an lvalue, the volatile object is
21331 referred to. For instance, @var{vref} refers to @var{vobj}, as
21332 expected, in the following example:
21336 volatile int &vref = vobj = @var{something};
21339 @node Restricted Pointers
21340 @section Restricting Pointer Aliasing
21341 @cindex restricted pointers
21342 @cindex restricted references
21343 @cindex restricted this pointer
21345 As with the C front end, G++ understands the C99 feature of restricted pointers,
21346 specified with the @code{__restrict__}, or @code{__restrict} type
21347 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
21348 language flag, @code{restrict} is not a keyword in C++.
21350 In addition to allowing restricted pointers, you can specify restricted
21351 references, which indicate that the reference is not aliased in the local
21355 void fn (int *__restrict__ rptr, int &__restrict__ rref)
21362 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
21363 @var{rref} refers to a (different) unaliased integer.
21365 You may also specify whether a member function's @var{this} pointer is
21366 unaliased by using @code{__restrict__} as a member function qualifier.
21369 void T::fn () __restrict__
21376 Within the body of @code{T::fn}, @var{this} has the effective
21377 definition @code{T *__restrict__ const this}. Notice that the
21378 interpretation of a @code{__restrict__} member function qualifier is
21379 different to that of @code{const} or @code{volatile} qualifier, in that it
21380 is applied to the pointer rather than the object. This is consistent with
21381 other compilers that implement restricted pointers.
21383 As with all outermost parameter qualifiers, @code{__restrict__} is
21384 ignored in function definition matching. This means you only need to
21385 specify @code{__restrict__} in a function definition, rather than
21386 in a function prototype as well.
21388 @node Vague Linkage
21389 @section Vague Linkage
21390 @cindex vague linkage
21392 There are several constructs in C++ that require space in the object
21393 file but are not clearly tied to a single translation unit. We say that
21394 these constructs have ``vague linkage''. Typically such constructs are
21395 emitted wherever they are needed, though sometimes we can be more
21399 @item Inline Functions
21400 Inline functions are typically defined in a header file which can be
21401 included in many different compilations. Hopefully they can usually be
21402 inlined, but sometimes an out-of-line copy is necessary, if the address
21403 of the function is taken or if inlining fails. In general, we emit an
21404 out-of-line copy in all translation units where one is needed. As an
21405 exception, we only emit inline virtual functions with the vtable, since
21406 it always requires a copy.
21408 Local static variables and string constants used in an inline function
21409 are also considered to have vague linkage, since they must be shared
21410 between all inlined and out-of-line instances of the function.
21414 C++ virtual functions are implemented in most compilers using a lookup
21415 table, known as a vtable. The vtable contains pointers to the virtual
21416 functions provided by a class, and each object of the class contains a
21417 pointer to its vtable (or vtables, in some multiple-inheritance
21418 situations). If the class declares any non-inline, non-pure virtual
21419 functions, the first one is chosen as the ``key method'' for the class,
21420 and the vtable is only emitted in the translation unit where the key
21423 @emph{Note:} If the chosen key method is later defined as inline, the
21424 vtable is still emitted in every translation unit that defines it.
21425 Make sure that any inline virtuals are declared inline in the class
21426 body, even if they are not defined there.
21428 @item @code{type_info} objects
21429 @cindex @code{type_info}
21431 C++ requires information about types to be written out in order to
21432 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
21433 For polymorphic classes (classes with virtual functions), the @samp{type_info}
21434 object is written out along with the vtable so that @samp{dynamic_cast}
21435 can determine the dynamic type of a class object at run time. For all
21436 other types, we write out the @samp{type_info} object when it is used: when
21437 applying @samp{typeid} to an expression, throwing an object, or
21438 referring to a type in a catch clause or exception specification.
21440 @item Template Instantiations
21441 Most everything in this section also applies to template instantiations,
21442 but there are other options as well.
21443 @xref{Template Instantiation,,Where's the Template?}.
21447 When used with GNU ld version 2.8 or later on an ELF system such as
21448 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
21449 these constructs will be discarded at link time. This is known as
21452 On targets that don't support COMDAT, but do support weak symbols, GCC
21453 uses them. This way one copy overrides all the others, but
21454 the unused copies still take up space in the executable.
21456 For targets that do not support either COMDAT or weak symbols,
21457 most entities with vague linkage are emitted as local symbols to
21458 avoid duplicate definition errors from the linker. This does not happen
21459 for local statics in inlines, however, as having multiple copies
21460 almost certainly breaks things.
21462 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
21463 another way to control placement of these constructs.
21465 @node C++ Interface
21466 @section C++ Interface and Implementation Pragmas
21468 @cindex interface and implementation headers, C++
21469 @cindex C++ interface and implementation headers
21470 @cindex pragmas, interface and implementation
21472 @code{#pragma interface} and @code{#pragma implementation} provide the
21473 user with a way of explicitly directing the compiler to emit entities
21474 with vague linkage (and debugging information) in a particular
21477 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
21478 by COMDAT support and the ``key method'' heuristic
21479 mentioned in @ref{Vague Linkage}. Using them can actually cause your
21480 program to grow due to unnecessary out-of-line copies of inline
21484 @item #pragma interface
21485 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
21486 @kindex #pragma interface
21487 Use this directive in @emph{header files} that define object classes, to save
21488 space in most of the object files that use those classes. Normally,
21489 local copies of certain information (backup copies of inline member
21490 functions, debugging information, and the internal tables that implement
21491 virtual functions) must be kept in each object file that includes class
21492 definitions. You can use this pragma to avoid such duplication. When a
21493 header file containing @samp{#pragma interface} is included in a
21494 compilation, this auxiliary information is not generated (unless
21495 the main input source file itself uses @samp{#pragma implementation}).
21496 Instead, the object files contain references to be resolved at link
21499 The second form of this directive is useful for the case where you have
21500 multiple headers with the same name in different directories. If you
21501 use this form, you must specify the same string to @samp{#pragma
21504 @item #pragma implementation
21505 @itemx #pragma implementation "@var{objects}.h"
21506 @kindex #pragma implementation
21507 Use this pragma in a @emph{main input file}, when you want full output from
21508 included header files to be generated (and made globally visible). The
21509 included header file, in turn, should use @samp{#pragma interface}.
21510 Backup copies of inline member functions, debugging information, and the
21511 internal tables used to implement virtual functions are all generated in
21512 implementation files.
21514 @cindex implied @code{#pragma implementation}
21515 @cindex @code{#pragma implementation}, implied
21516 @cindex naming convention, implementation headers
21517 If you use @samp{#pragma implementation} with no argument, it applies to
21518 an include file with the same basename@footnote{A file's @dfn{basename}
21519 is the name stripped of all leading path information and of trailing
21520 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
21521 file. For example, in @file{allclass.cc}, giving just
21522 @samp{#pragma implementation}
21523 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
21525 Use the string argument if you want a single implementation file to
21526 include code from multiple header files. (You must also use
21527 @samp{#include} to include the header file; @samp{#pragma
21528 implementation} only specifies how to use the file---it doesn't actually
21531 There is no way to split up the contents of a single header file into
21532 multiple implementation files.
21535 @cindex inlining and C++ pragmas
21536 @cindex C++ pragmas, effect on inlining
21537 @cindex pragmas in C++, effect on inlining
21538 @samp{#pragma implementation} and @samp{#pragma interface} also have an
21539 effect on function inlining.
21541 If you define a class in a header file marked with @samp{#pragma
21542 interface}, the effect on an inline function defined in that class is
21543 similar to an explicit @code{extern} declaration---the compiler emits
21544 no code at all to define an independent version of the function. Its
21545 definition is used only for inlining with its callers.
21547 @opindex fno-implement-inlines
21548 Conversely, when you include the same header file in a main source file
21549 that declares it as @samp{#pragma implementation}, the compiler emits
21550 code for the function itself; this defines a version of the function
21551 that can be found via pointers (or by callers compiled without
21552 inlining). If all calls to the function can be inlined, you can avoid
21553 emitting the function by compiling with @option{-fno-implement-inlines}.
21554 If any calls are not inlined, you will get linker errors.
21556 @node Template Instantiation
21557 @section Where's the Template?
21558 @cindex template instantiation
21560 C++ templates were the first language feature to require more
21561 intelligence from the environment than was traditionally found on a UNIX
21562 system. Somehow the compiler and linker have to make sure that each
21563 template instance occurs exactly once in the executable if it is needed,
21564 and not at all otherwise. There are two basic approaches to this
21565 problem, which are referred to as the Borland model and the Cfront model.
21568 @item Borland model
21569 Borland C++ solved the template instantiation problem by adding the code
21570 equivalent of common blocks to their linker; the compiler emits template
21571 instances in each translation unit that uses them, and the linker
21572 collapses them together. The advantage of this model is that the linker
21573 only has to consider the object files themselves; there is no external
21574 complexity to worry about. The disadvantage is that compilation time
21575 is increased because the template code is being compiled repeatedly.
21576 Code written for this model tends to include definitions of all
21577 templates in the header file, since they must be seen to be
21581 The AT&T C++ translator, Cfront, solved the template instantiation
21582 problem by creating the notion of a template repository, an
21583 automatically maintained place where template instances are stored. A
21584 more modern version of the repository works as follows: As individual
21585 object files are built, the compiler places any template definitions and
21586 instantiations encountered in the repository. At link time, the link
21587 wrapper adds in the objects in the repository and compiles any needed
21588 instances that were not previously emitted. The advantages of this
21589 model are more optimal compilation speed and the ability to use the
21590 system linker; to implement the Borland model a compiler vendor also
21591 needs to replace the linker. The disadvantages are vastly increased
21592 complexity, and thus potential for error; for some code this can be
21593 just as transparent, but in practice it can been very difficult to build
21594 multiple programs in one directory and one program in multiple
21595 directories. Code written for this model tends to separate definitions
21596 of non-inline member templates into a separate file, which should be
21597 compiled separately.
21600 G++ implements the Borland model on targets where the linker supports it,
21601 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
21602 Otherwise G++ implements neither automatic model.
21604 You have the following options for dealing with template instantiations:
21608 Do nothing. Code written for the Borland model works fine, but
21609 each translation unit contains instances of each of the templates it
21610 uses. The duplicate instances will be discarded by the linker, but in
21611 a large program, this can lead to an unacceptable amount of code
21612 duplication in object files or shared libraries.
21614 Duplicate instances of a template can be avoided by defining an explicit
21615 instantiation in one object file, and preventing the compiler from doing
21616 implicit instantiations in any other object files by using an explicit
21617 instantiation declaration, using the @code{extern template} syntax:
21620 extern template int max (int, int);
21623 This syntax is defined in the C++ 2011 standard, but has been supported by
21624 G++ and other compilers since well before 2011.
21626 Explicit instantiations can be used for the largest or most frequently
21627 duplicated instances, without having to know exactly which other instances
21628 are used in the rest of the program. You can scatter the explicit
21629 instantiations throughout your program, perhaps putting them in the
21630 translation units where the instances are used or the translation units
21631 that define the templates themselves; you can put all of the explicit
21632 instantiations you need into one big file; or you can create small files
21639 template class Foo<int>;
21640 template ostream& operator <<
21641 (ostream&, const Foo<int>&);
21645 for each of the instances you need, and create a template instantiation
21646 library from those.
21648 This is the simplest option, but also offers flexibility and
21649 fine-grained control when necessary. It is also the most portable
21650 alternative and programs using this approach will work with most modern
21655 Compile your template-using code with @option{-frepo}. The compiler
21656 generates files with the extension @samp{.rpo} listing all of the
21657 template instantiations used in the corresponding object files that
21658 could be instantiated there; the link wrapper, @samp{collect2},
21659 then updates the @samp{.rpo} files to tell the compiler where to place
21660 those instantiations and rebuild any affected object files. The
21661 link-time overhead is negligible after the first pass, as the compiler
21662 continues to place the instantiations in the same files.
21664 This can be a suitable option for application code written for the Borland
21665 model, as it usually just works. Code written for the Cfront model
21666 needs to be modified so that the template definitions are available at
21667 one or more points of instantiation; usually this is as simple as adding
21668 @code{#include <tmethods.cc>} to the end of each template header.
21670 For library code, if you want the library to provide all of the template
21671 instantiations it needs, just try to link all of its object files
21672 together; the link will fail, but cause the instantiations to be
21673 generated as a side effect. Be warned, however, that this may cause
21674 conflicts if multiple libraries try to provide the same instantiations.
21675 For greater control, use explicit instantiation as described in the next
21679 @opindex fno-implicit-templates
21680 Compile your code with @option{-fno-implicit-templates} to disable the
21681 implicit generation of template instances, and explicitly instantiate
21682 all the ones you use. This approach requires more knowledge of exactly
21683 which instances you need than do the others, but it's less
21684 mysterious and allows greater control if you want to ensure that only
21685 the intended instances are used.
21687 If you are using Cfront-model code, you can probably get away with not
21688 using @option{-fno-implicit-templates} when compiling files that don't
21689 @samp{#include} the member template definitions.
21691 If you use one big file to do the instantiations, you may want to
21692 compile it without @option{-fno-implicit-templates} so you get all of the
21693 instances required by your explicit instantiations (but not by any
21694 other files) without having to specify them as well.
21696 In addition to forward declaration of explicit instantiations
21697 (with @code{extern}), G++ has extended the template instantiation
21698 syntax to support instantiation of the compiler support data for a
21699 template class (i.e.@: the vtable) without instantiating any of its
21700 members (with @code{inline}), and instantiation of only the static data
21701 members of a template class, without the support data or member
21702 functions (with @code{static}):
21705 inline template class Foo<int>;
21706 static template class Foo<int>;
21710 @node Bound member functions
21711 @section Extracting the Function Pointer from a Bound Pointer to Member Function
21713 @cindex pointer to member function
21714 @cindex bound pointer to member function
21716 In C++, pointer to member functions (PMFs) are implemented using a wide
21717 pointer of sorts to handle all the possible call mechanisms; the PMF
21718 needs to store information about how to adjust the @samp{this} pointer,
21719 and if the function pointed to is virtual, where to find the vtable, and
21720 where in the vtable to look for the member function. If you are using
21721 PMFs in an inner loop, you should really reconsider that decision. If
21722 that is not an option, you can extract the pointer to the function that
21723 would be called for a given object/PMF pair and call it directly inside
21724 the inner loop, to save a bit of time.
21726 Note that you still pay the penalty for the call through a
21727 function pointer; on most modern architectures, such a call defeats the
21728 branch prediction features of the CPU@. This is also true of normal
21729 virtual function calls.
21731 The syntax for this extension is
21735 extern int (A::*fp)();
21736 typedef int (*fptr)(A *);
21738 fptr p = (fptr)(a.*fp);
21741 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
21742 no object is needed to obtain the address of the function. They can be
21743 converted to function pointers directly:
21746 fptr p1 = (fptr)(&A::foo);
21749 @opindex Wno-pmf-conversions
21750 You must specify @option{-Wno-pmf-conversions} to use this extension.
21752 @node C++ Attributes
21753 @section C++-Specific Variable, Function, and Type Attributes
21755 Some attributes only make sense for C++ programs.
21758 @item abi_tag ("@var{tag}", ...)
21759 @cindex @code{abi_tag} function attribute
21760 @cindex @code{abi_tag} variable attribute
21761 @cindex @code{abi_tag} type attribute
21762 The @code{abi_tag} attribute can be applied to a function, variable, or class
21763 declaration. It modifies the mangled name of the entity to
21764 incorporate the tag name, in order to distinguish the function or
21765 class from an earlier version with a different ABI; perhaps the class
21766 has changed size, or the function has a different return type that is
21767 not encoded in the mangled name.
21769 The attribute can also be applied to an inline namespace, but does not
21770 affect the mangled name of the namespace; in this case it is only used
21771 for @option{-Wabi-tag} warnings and automatic tagging of functions and
21772 variables. Tagging inline namespaces is generally preferable to
21773 tagging individual declarations, but the latter is sometimes
21774 necessary, such as when only certain members of a class need to be
21777 The argument can be a list of strings of arbitrary length. The
21778 strings are sorted on output, so the order of the list is
21781 A redeclaration of an entity must not add new ABI tags,
21782 since doing so would change the mangled name.
21784 The ABI tags apply to a name, so all instantiations and
21785 specializations of a template have the same tags. The attribute will
21786 be ignored if applied to an explicit specialization or instantiation.
21788 The @option{-Wabi-tag} flag enables a warning about a class which does
21789 not have all the ABI tags used by its subobjects and virtual functions; for users with code
21790 that needs to coexist with an earlier ABI, using this option can help
21791 to find all affected types that need to be tagged.
21793 When a type involving an ABI tag is used as the type of a variable or
21794 return type of a function where that tag is not already present in the
21795 signature of the function, the tag is automatically applied to the
21796 variable or function. @option{-Wabi-tag} also warns about this
21797 situation; this warning can be avoided by explicitly tagging the
21798 variable or function or moving it into a tagged inline namespace.
21800 @item init_priority (@var{priority})
21801 @cindex @code{init_priority} variable attribute
21803 In Standard C++, objects defined at namespace scope are guaranteed to be
21804 initialized in an order in strict accordance with that of their definitions
21805 @emph{in a given translation unit}. No guarantee is made for initializations
21806 across translation units. However, GNU C++ allows users to control the
21807 order of initialization of objects defined at namespace scope with the
21808 @code{init_priority} attribute by specifying a relative @var{priority},
21809 a constant integral expression currently bounded between 101 and 65535
21810 inclusive. Lower numbers indicate a higher priority.
21812 In the following example, @code{A} would normally be created before
21813 @code{B}, but the @code{init_priority} attribute reverses that order:
21816 Some_Class A __attribute__ ((init_priority (2000)));
21817 Some_Class B __attribute__ ((init_priority (543)));
21821 Note that the particular values of @var{priority} do not matter; only their
21824 @item java_interface
21825 @cindex @code{java_interface} type attribute
21827 This type attribute informs C++ that the class is a Java interface. It may
21828 only be applied to classes declared within an @code{extern "Java"} block.
21829 Calls to methods declared in this interface are dispatched using GCJ's
21830 interface table mechanism, instead of regular virtual table dispatch.
21833 @cindex @code{warn_unused} type attribute
21835 For C++ types with non-trivial constructors and/or destructors it is
21836 impossible for the compiler to determine whether a variable of this
21837 type is truly unused if it is not referenced. This type attribute
21838 informs the compiler that variables of this type should be warned
21839 about if they appear to be unused, just like variables of fundamental
21842 This attribute is appropriate for types which just represent a value,
21843 such as @code{std::string}; it is not appropriate for types which
21844 control a resource, such as @code{std::lock_guard}.
21846 This attribute is also accepted in C, but it is unnecessary because C
21847 does not have constructors or destructors.
21851 See also @ref{Namespace Association}.
21853 @node Function Multiversioning
21854 @section Function Multiversioning
21855 @cindex function versions
21857 With the GNU C++ front end, for x86 targets, you may specify multiple
21858 versions of a function, where each function is specialized for a
21859 specific target feature. At runtime, the appropriate version of the
21860 function is automatically executed depending on the characteristics of
21861 the execution platform. Here is an example.
21864 __attribute__ ((target ("default")))
21867 // The default version of foo.
21871 __attribute__ ((target ("sse4.2")))
21874 // foo version for SSE4.2
21878 __attribute__ ((target ("arch=atom")))
21881 // foo version for the Intel ATOM processor
21885 __attribute__ ((target ("arch=amdfam10")))
21888 // foo version for the AMD Family 0x10 processors.
21895 assert ((*p) () == foo ());
21900 In the above example, four versions of function foo are created. The
21901 first version of foo with the target attribute "default" is the default
21902 version. This version gets executed when no other target specific
21903 version qualifies for execution on a particular platform. A new version
21904 of foo is created by using the same function signature but with a
21905 different target string. Function foo is called or a pointer to it is
21906 taken just like a regular function. GCC takes care of doing the
21907 dispatching to call the right version at runtime. Refer to the
21908 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
21909 Function Multiversioning} for more details.
21911 @node Namespace Association
21912 @section Namespace Association
21914 @strong{Caution:} The semantics of this extension are equivalent
21915 to C++ 2011 inline namespaces. Users should use inline namespaces
21916 instead as this extension will be removed in future versions of G++.
21918 A using-directive with @code{__attribute ((strong))} is stronger
21919 than a normal using-directive in two ways:
21923 Templates from the used namespace can be specialized and explicitly
21924 instantiated as though they were members of the using namespace.
21927 The using namespace is considered an associated namespace of all
21928 templates in the used namespace for purposes of argument-dependent
21932 The used namespace must be nested within the using namespace so that
21933 normal unqualified lookup works properly.
21935 This is useful for composing a namespace transparently from
21936 implementation namespaces. For example:
21941 template <class T> struct A @{ @};
21943 using namespace debug __attribute ((__strong__));
21944 template <> struct A<int> @{ @}; // @r{OK to specialize}
21946 template <class T> void f (A<T>);
21951 f (std::A<float>()); // @r{lookup finds} std::f
21957 @section Type Traits
21959 The C++ front end implements syntactic extensions that allow
21960 compile-time determination of
21961 various characteristics of a type (or of a
21965 @item __has_nothrow_assign (type)
21966 If @code{type} is const qualified or is a reference type then the trait is
21967 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
21968 is true, else if @code{type} is a cv class or union type with copy assignment
21969 operators that are known not to throw an exception then the trait is true,
21970 else it is false. Requires: @code{type} shall be a complete type,
21971 (possibly cv-qualified) @code{void}, or an array of unknown bound.
21973 @item __has_nothrow_copy (type)
21974 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
21975 @code{type} is a cv class or union type with copy constructors that
21976 are known not to throw an exception then the trait is true, else it is false.
21977 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
21978 @code{void}, or an array of unknown bound.
21980 @item __has_nothrow_constructor (type)
21981 If @code{__has_trivial_constructor (type)} is true then the trait is
21982 true, else if @code{type} is a cv class or union type (or array
21983 thereof) with a default constructor that is known not to throw an
21984 exception then the trait is true, else it is false. Requires:
21985 @code{type} shall be a complete type, (possibly cv-qualified)
21986 @code{void}, or an array of unknown bound.
21988 @item __has_trivial_assign (type)
21989 If @code{type} is const qualified or is a reference type then the trait is
21990 false. Otherwise if @code{__is_pod (type)} is true then the trait is
21991 true, else if @code{type} is a cv class or union type with a trivial
21992 copy assignment ([class.copy]) then the trait is true, else it is
21993 false. Requires: @code{type} shall be a complete type, (possibly
21994 cv-qualified) @code{void}, or an array of unknown bound.
21996 @item __has_trivial_copy (type)
21997 If @code{__is_pod (type)} is true or @code{type} is a reference type
21998 then the trait is true, else if @code{type} is a cv class or union type
21999 with a trivial copy constructor ([class.copy]) then the trait
22000 is true, else it is false. Requires: @code{type} shall be a complete
22001 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22003 @item __has_trivial_constructor (type)
22004 If @code{__is_pod (type)} is true then the trait is true, else if
22005 @code{type} is a cv class or union type (or array thereof) with a
22006 trivial default constructor ([class.ctor]) then the trait is true,
22007 else it is false. Requires: @code{type} shall be a complete
22008 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22010 @item __has_trivial_destructor (type)
22011 If @code{__is_pod (type)} is true or @code{type} is a reference type then
22012 the trait is true, else if @code{type} is a cv class or union type (or
22013 array thereof) with a trivial destructor ([class.dtor]) then the trait
22014 is true, else it is false. Requires: @code{type} shall be a complete
22015 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22017 @item __has_virtual_destructor (type)
22018 If @code{type} is a class type with a virtual destructor
22019 ([class.dtor]) then the trait is true, else it is false. Requires:
22020 @code{type} shall be a complete type, (possibly cv-qualified)
22021 @code{void}, or an array of unknown bound.
22023 @item __is_abstract (type)
22024 If @code{type} is an abstract class ([class.abstract]) then the trait
22025 is true, else it is false. Requires: @code{type} shall be a complete
22026 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22028 @item __is_base_of (base_type, derived_type)
22029 If @code{base_type} is a base class of @code{derived_type}
22030 ([class.derived]) then the trait is true, otherwise it is false.
22031 Top-level cv qualifications of @code{base_type} and
22032 @code{derived_type} are ignored. For the purposes of this trait, a
22033 class type is considered is own base. Requires: if @code{__is_class
22034 (base_type)} and @code{__is_class (derived_type)} are true and
22035 @code{base_type} and @code{derived_type} are not the same type
22036 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
22037 type. A diagnostic is produced if this requirement is not met.
22039 @item __is_class (type)
22040 If @code{type} is a cv class type, and not a union type
22041 ([basic.compound]) the trait is true, else it is false.
22043 @item __is_empty (type)
22044 If @code{__is_class (type)} is false then the trait is false.
22045 Otherwise @code{type} is considered empty if and only if: @code{type}
22046 has no non-static data members, or all non-static data members, if
22047 any, are bit-fields of length 0, and @code{type} has no virtual
22048 members, and @code{type} has no virtual base classes, and @code{type}
22049 has no base classes @code{base_type} for which
22050 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
22051 be a complete type, (possibly cv-qualified) @code{void}, or an array
22054 @item __is_enum (type)
22055 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
22056 true, else it is false.
22058 @item __is_literal_type (type)
22059 If @code{type} is a literal type ([basic.types]) the trait is
22060 true, else it is false. Requires: @code{type} shall be a complete type,
22061 (possibly cv-qualified) @code{void}, or an array of unknown bound.
22063 @item __is_pod (type)
22064 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
22065 else it is false. Requires: @code{type} shall be a complete type,
22066 (possibly cv-qualified) @code{void}, or an array of unknown bound.
22068 @item __is_polymorphic (type)
22069 If @code{type} is a polymorphic class ([class.virtual]) then the trait
22070 is true, else it is false. Requires: @code{type} shall be a complete
22071 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22073 @item __is_standard_layout (type)
22074 If @code{type} is a standard-layout type ([basic.types]) the trait is
22075 true, else it is false. Requires: @code{type} shall be a complete
22076 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22078 @item __is_trivial (type)
22079 If @code{type} is a trivial type ([basic.types]) the trait is
22080 true, else it is false. Requires: @code{type} shall be a complete
22081 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
22083 @item __is_union (type)
22084 If @code{type} is a cv union type ([basic.compound]) the trait is
22085 true, else it is false.
22087 @item __underlying_type (type)
22088 The underlying type of @code{type}. Requires: @code{type} shall be
22089 an enumeration type ([dcl.enum]).
22095 @section C++ Concepts
22097 C++ concepts provide much-improved support for generic programming. In
22098 particular, they allow the specification of constraints on template arguments.
22099 The constraints are used to extend the usual overloading and partial
22100 specialization capabilities of the language, allowing generic data structures
22101 and algorithms to be ``refined'' based on their properties rather than their
22104 The following keywords are reserved for concepts.
22108 States an expression as an assumption, and if possible, verifies that the
22109 assumption is valid. For example, @code{assume(n > 0)}.
22112 Introduces an axiom definition. Axioms introduce requirements on values.
22115 Introduces a universally quantified object in an axiom. For example,
22116 @code{forall (int n) n + 0 == n}).
22119 Introduces a concept definition. Concepts are sets of syntactic and semantic
22120 requirements on types and their values.
22123 Introduces constraints on template arguments or requirements for a member
22124 function of a class template.
22128 The front end also exposes a number of internal mechanism that can be used
22129 to simplify the writing of type traits. Note that some of these traits are
22130 likely to be removed in the future.
22133 @item __is_same (type1, type2)
22134 A binary type trait: true whenever the type arguments are the same.
22139 @node Java Exceptions
22140 @section Java Exceptions
22142 The Java language uses a slightly different exception handling model
22143 from C++. Normally, GNU C++ automatically detects when you are
22144 writing C++ code that uses Java exceptions, and handle them
22145 appropriately. However, if C++ code only needs to execute destructors
22146 when Java exceptions are thrown through it, GCC guesses incorrectly.
22147 Sample problematic code is:
22150 struct S @{ ~S(); @};
22151 extern void bar(); // @r{is written in Java, and may throw exceptions}
22160 The usual effect of an incorrect guess is a link failure, complaining of
22161 a missing routine called @samp{__gxx_personality_v0}.
22163 You can inform the compiler that Java exceptions are to be used in a
22164 translation unit, irrespective of what it might think, by writing
22165 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
22166 @samp{#pragma} must appear before any functions that throw or catch
22167 exceptions, or run destructors when exceptions are thrown through them.
22169 You cannot mix Java and C++ exceptions in the same translation unit. It
22170 is believed to be safe to throw a C++ exception from one file through
22171 another file compiled for the Java exception model, or vice versa, but
22172 there may be bugs in this area.
22174 @node Deprecated Features
22175 @section Deprecated Features
22177 In the past, the GNU C++ compiler was extended to experiment with new
22178 features, at a time when the C++ language was still evolving. Now that
22179 the C++ standard is complete, some of those features are superseded by
22180 superior alternatives. Using the old features might cause a warning in
22181 some cases that the feature will be dropped in the future. In other
22182 cases, the feature might be gone already.
22184 While the list below is not exhaustive, it documents some of the options
22185 that are now deprecated:
22188 @item -fexternal-templates
22189 @itemx -falt-external-templates
22190 These are two of the many ways for G++ to implement template
22191 instantiation. @xref{Template Instantiation}. The C++ standard clearly
22192 defines how template definitions have to be organized across
22193 implementation units. G++ has an implicit instantiation mechanism that
22194 should work just fine for standard-conforming code.
22196 @item -fstrict-prototype
22197 @itemx -fno-strict-prototype
22198 Previously it was possible to use an empty prototype parameter list to
22199 indicate an unspecified number of parameters (like C), rather than no
22200 parameters, as C++ demands. This feature has been removed, except where
22201 it is required for backwards compatibility. @xref{Backwards Compatibility}.
22204 G++ allows a virtual function returning @samp{void *} to be overridden
22205 by one returning a different pointer type. This extension to the
22206 covariant return type rules is now deprecated and will be removed from a
22209 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
22210 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
22211 and are now removed from G++. Code using these operators should be
22212 modified to use @code{std::min} and @code{std::max} instead.
22214 The named return value extension has been deprecated, and is now
22217 The use of initializer lists with new expressions has been deprecated,
22218 and is now removed from G++.
22220 Floating and complex non-type template parameters have been deprecated,
22221 and are now removed from G++.
22223 The implicit typename extension has been deprecated and is now
22226 The use of default arguments in function pointers, function typedefs
22227 and other places where they are not permitted by the standard is
22228 deprecated and will be removed from a future version of G++.
22230 G++ allows floating-point literals to appear in integral constant expressions,
22231 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
22232 This extension is deprecated and will be removed from a future version.
22234 G++ allows static data members of const floating-point type to be declared
22235 with an initializer in a class definition. The standard only allows
22236 initializers for static members of const integral types and const
22237 enumeration types so this extension has been deprecated and will be removed
22238 from a future version.
22240 @node Backwards Compatibility
22241 @section Backwards Compatibility
22242 @cindex Backwards Compatibility
22243 @cindex ARM [Annotated C++ Reference Manual]
22245 Now that there is a definitive ISO standard C++, G++ has a specification
22246 to adhere to. The C++ language evolved over time, and features that
22247 used to be acceptable in previous drafts of the standard, such as the ARM
22248 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
22249 compilation of C++ written to such drafts, G++ contains some backwards
22250 compatibilities. @emph{All such backwards compatibility features are
22251 liable to disappear in future versions of G++.} They should be considered
22252 deprecated. @xref{Deprecated Features}.
22256 If a variable is declared at for scope, it used to remain in scope until
22257 the end of the scope that contained the for statement (rather than just
22258 within the for scope). G++ retains this, but issues a warning, if such a
22259 variable is accessed outside the for scope.
22261 @item Implicit C language
22262 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
22263 scope to set the language. On such systems, all header files are
22264 implicitly scoped inside a C language scope. Also, an empty prototype
22265 @code{()} is treated as an unspecified number of arguments, rather
22266 than no arguments, as C++ demands.
22269 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
22270 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr