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 ISO C99 supports compound literals. A compound literal looks like
1872 a cast containing an initializer. Its value is an object of the
1873 type specified in the cast, containing the elements specified in
1874 the initializer; it is an lvalue. As an extension, GCC supports
1875 compound literals in C90 mode and in C++, though the semantics are
1876 somewhat different in C++.
1878 Usually, the specified type is a structure. Assume that
1879 @code{struct foo} and @code{structure} are declared as shown:
1882 struct foo @{int a; char b[2];@} structure;
1886 Here is an example of constructing a @code{struct foo} with a compound literal:
1889 structure = ((struct foo) @{x + y, 'a', 0@});
1893 This is equivalent to writing the following:
1897 struct foo temp = @{x + y, 'a', 0@};
1902 You can also construct an array, though this is dangerous in C++, as
1903 explained below. If all the elements of the compound literal are
1904 (made up of) simple constant expressions, suitable for use in
1905 initializers of objects of static storage duration, then the compound
1906 literal can be coerced to a pointer to its first element and used in
1907 such an initializer, as shown here:
1910 char **foo = (char *[]) @{ "x", "y", "z" @};
1913 Compound literals for scalar types and union types are
1914 also allowed, but then the compound literal is equivalent
1917 As a GNU extension, GCC allows initialization of objects with static storage
1918 duration by compound literals (which is not possible in ISO C99, because
1919 the initializer is not a constant).
1920 It is handled as if the object is initialized only with the bracket
1921 enclosed list if the types of the compound literal and the object match.
1922 The initializer list of the compound literal must be constant.
1923 If the object being initialized has array type of unknown size, the size is
1924 determined by compound literal size.
1927 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1928 static int y[] = (int []) @{1, 2, 3@};
1929 static int z[] = (int [3]) @{1@};
1933 The above lines are equivalent to the following:
1935 static struct foo x = @{1, 'a', 'b'@};
1936 static int y[] = @{1, 2, 3@};
1937 static int z[] = @{1, 0, 0@};
1940 In C, a compound literal designates an unnamed object with static or
1941 automatic storage duration. In C++, a compound literal designates a
1942 temporary object, which only lives until the end of its
1943 full-expression. As a result, well-defined C code that takes the
1944 address of a subobject of a compound literal can be undefined in C++,
1945 so the C++ compiler rejects the conversion of a temporary array to a pointer.
1946 For instance, if the array compound literal example above appeared
1947 inside a function, any subsequent use of @samp{foo} in C++ has
1948 undefined behavior because the lifetime of the array ends after the
1949 declaration of @samp{foo}.
1951 As an optimization, the C++ compiler sometimes gives array compound
1952 literals longer lifetimes: when the array either appears outside a
1953 function or has const-qualified type. If @samp{foo} and its
1954 initializer had elements of @samp{char *const} type rather than
1955 @samp{char *}, or if @samp{foo} were a global variable, the array
1956 would have static storage duration. But it is probably safest just to
1957 avoid the use of array compound literals in code compiled as C++.
1959 @node Designated Inits
1960 @section Designated Initializers
1961 @cindex initializers with labeled elements
1962 @cindex labeled elements in initializers
1963 @cindex case labels in initializers
1964 @cindex designated initializers
1966 Standard C90 requires the elements of an initializer to appear in a fixed
1967 order, the same as the order of the elements in the array or structure
1970 In ISO C99 you can give the elements in any order, specifying the array
1971 indices or structure field names they apply to, and GNU C allows this as
1972 an extension in C90 mode as well. This extension is not
1973 implemented in GNU C++.
1975 To specify an array index, write
1976 @samp{[@var{index}] =} before the element value. For example,
1979 int a[6] = @{ [4] = 29, [2] = 15 @};
1986 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1990 The index values must be constant expressions, even if the array being
1991 initialized is automatic.
1993 An alternative syntax for this that has been obsolete since GCC 2.5 but
1994 GCC still accepts is to write @samp{[@var{index}]} before the element
1995 value, with no @samp{=}.
1997 To initialize a range of elements to the same value, write
1998 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1999 extension. For example,
2002 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2006 If the value in it has side-effects, the side-effects happen only once,
2007 not for each initialized field by the range initializer.
2010 Note that the length of the array is the highest value specified
2013 In a structure initializer, specify the name of a field to initialize
2014 with @samp{.@var{fieldname} =} before the element value. For example,
2015 given the following structure,
2018 struct point @{ int x, y; @};
2022 the following initialization
2025 struct point p = @{ .y = yvalue, .x = xvalue @};
2032 struct point p = @{ xvalue, yvalue @};
2035 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2036 @samp{@var{fieldname}:}, as shown here:
2039 struct point p = @{ y: yvalue, x: xvalue @};
2042 Omitted field members are implicitly initialized the same as objects
2043 that have static storage duration.
2046 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2047 @dfn{designator}. You can also use a designator (or the obsolete colon
2048 syntax) when initializing a union, to specify which element of the union
2049 should be used. For example,
2052 union foo @{ int i; double d; @};
2054 union foo f = @{ .d = 4 @};
2058 converts 4 to a @code{double} to store it in the union using
2059 the second element. By contrast, casting 4 to type @code{union foo}
2060 stores it into the union as the integer @code{i}, since it is
2061 an integer. (@xref{Cast to Union}.)
2063 You can combine this technique of naming elements with ordinary C
2064 initialization of successive elements. Each initializer element that
2065 does not have a designator applies to the next consecutive element of the
2066 array or structure. For example,
2069 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2076 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2079 Labeling the elements of an array initializer is especially useful
2080 when the indices are characters or belong to an @code{enum} type.
2085 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2086 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2089 @cindex designator lists
2090 You can also write a series of @samp{.@var{fieldname}} and
2091 @samp{[@var{index}]} designators before an @samp{=} to specify a
2092 nested subobject to initialize; the list is taken relative to the
2093 subobject corresponding to the closest surrounding brace pair. For
2094 example, with the @samp{struct point} declaration above:
2097 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2101 If the same field is initialized multiple times, it has the value from
2102 the last initialization. If any such overridden initialization has
2103 side-effect, it is unspecified whether the side-effect happens or not.
2104 Currently, GCC discards them and issues a warning.
2107 @section Case Ranges
2109 @cindex ranges in case statements
2111 You can specify a range of consecutive values in a single @code{case} label,
2115 case @var{low} ... @var{high}:
2119 This has the same effect as the proper number of individual @code{case}
2120 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2122 This feature is especially useful for ranges of ASCII character codes:
2128 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2129 it may be parsed wrong when you use it with integer values. For example,
2144 @section Cast to a Union Type
2145 @cindex cast to a union
2146 @cindex union, casting to a
2148 A cast to union type is similar to other casts, except that the type
2149 specified is a union type. You can specify the type either with
2150 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2151 a constructor, not a cast, and hence does not yield an lvalue like
2152 normal casts. (@xref{Compound Literals}.)
2154 The types that may be cast to the union type are those of the members
2155 of the union. Thus, given the following union and variables:
2158 union foo @{ int i; double d; @};
2164 both @code{x} and @code{y} can be cast to type @code{union foo}.
2166 Using the cast as the right-hand side of an assignment to a variable of
2167 union type is equivalent to storing in a member of the union:
2172 u = (union foo) x @equiv{} u.i = x
2173 u = (union foo) y @equiv{} u.d = y
2176 You can also use the union cast as a function argument:
2179 void hack (union foo);
2181 hack ((union foo) x);
2184 @node Mixed Declarations
2185 @section Mixed Declarations and Code
2186 @cindex mixed declarations and code
2187 @cindex declarations, mixed with code
2188 @cindex code, mixed with declarations
2190 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2191 within compound statements. As an extension, GNU C also allows this in
2192 C90 mode. For example, you could do:
2201 Each identifier is visible from where it is declared until the end of
2202 the enclosing block.
2204 @node Function Attributes
2205 @section Declaring Attributes of Functions
2206 @cindex function attributes
2207 @cindex declaring attributes of functions
2208 @cindex @code{volatile} applied to function
2209 @cindex @code{const} applied to function
2211 In GNU C, you can use function attributes to declare certain things
2212 about functions called in your program which help the compiler
2213 optimize calls and check your code more carefully. For example, you
2214 can use attributes to declare that a function never returns
2215 (@code{noreturn}), returns a value depending only on its arguments
2216 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2218 You can also use attributes to control memory placement, code
2219 generation options or call/return conventions within the function
2220 being annotated. Many of these attributes are target-specific. For
2221 example, many targets support attributes for defining interrupt
2222 handler functions, which typically must follow special register usage
2223 and return conventions.
2225 Function attributes are introduced by the @code{__attribute__} keyword
2226 on a declaration, followed by an attribute specification inside double
2227 parentheses. You can specify multiple attributes in a declaration by
2228 separating them by commas within the double parentheses or by
2229 immediately following an attribute declaration with another attribute
2230 declaration. @xref{Attribute Syntax}, for the exact rules on
2231 attribute syntax and placement.
2233 GCC also supports attributes on
2234 variable declarations (@pxref{Variable Attributes}),
2235 labels (@pxref{Label Attributes}),
2236 enumerators (@pxref{Enumerator Attributes}),
2237 and types (@pxref{Type Attributes}).
2239 There is some overlap between the purposes of attributes and pragmas
2240 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2241 found convenient to use @code{__attribute__} to achieve a natural
2242 attachment of attributes to their corresponding declarations, whereas
2243 @code{#pragma} is of use for compatibility with other compilers
2244 or constructs that do not naturally form part of the grammar.
2246 In addition to the attributes documented here,
2247 GCC plugins may provide their own attributes.
2250 * Common Function Attributes::
2251 * AArch64 Function Attributes::
2252 * ARC Function Attributes::
2253 * ARM Function Attributes::
2254 * AVR Function Attributes::
2255 * Blackfin Function Attributes::
2256 * CR16 Function Attributes::
2257 * Epiphany Function Attributes::
2258 * H8/300 Function Attributes::
2259 * IA-64 Function Attributes::
2260 * M32C Function Attributes::
2261 * M32R/D Function Attributes::
2262 * m68k Function Attributes::
2263 * MCORE Function Attributes::
2264 * MeP Function Attributes::
2265 * MicroBlaze Function Attributes::
2266 * Microsoft Windows Function Attributes::
2267 * MIPS Function Attributes::
2268 * MSP430 Function Attributes::
2269 * NDS32 Function Attributes::
2270 * Nios II Function Attributes::
2271 * Nvidia PTX Function Attributes::
2272 * PowerPC Function Attributes::
2273 * RL78 Function Attributes::
2274 * RX Function Attributes::
2275 * S/390 Function Attributes::
2276 * SH Function Attributes::
2277 * SPU Function Attributes::
2278 * Symbian OS Function Attributes::
2279 * V850 Function Attributes::
2280 * Visium Function Attributes::
2281 * x86 Function Attributes::
2282 * Xstormy16 Function Attributes::
2285 @node Common Function Attributes
2286 @subsection Common Function Attributes
2288 The following attributes are supported on most targets.
2291 @c Keep this table alphabetized by attribute name. Treat _ as space.
2293 @item alias ("@var{target}")
2294 @cindex @code{alias} function attribute
2295 The @code{alias} attribute causes the declaration to be emitted as an
2296 alias for another symbol, which must be specified. For instance,
2299 void __f () @{ /* @r{Do something.} */; @}
2300 void f () __attribute__ ((weak, alias ("__f")));
2304 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2305 mangled name for the target must be used. It is an error if @samp{__f}
2306 is not defined in the same translation unit.
2308 This attribute requires assembler and object file support,
2309 and may not be available on all targets.
2311 @item aligned (@var{alignment})
2312 @cindex @code{aligned} function attribute
2313 This attribute specifies a minimum alignment for the function,
2316 You cannot use this attribute to decrease the alignment of a function,
2317 only to increase it. However, when you explicitly specify a function
2318 alignment this overrides the effect of the
2319 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2322 Note that the effectiveness of @code{aligned} attributes may be
2323 limited by inherent limitations in your linker. On many systems, the
2324 linker is only able to arrange for functions to be aligned up to a
2325 certain maximum alignment. (For some linkers, the maximum supported
2326 alignment may be very very small.) See your linker documentation for
2327 further information.
2329 The @code{aligned} attribute can also be used for variables and fields
2330 (@pxref{Variable Attributes}.)
2333 @cindex @code{alloc_align} function attribute
2334 The @code{alloc_align} attribute is used to tell the compiler that the
2335 function return value points to memory, where the returned pointer minimum
2336 alignment is given by one of the functions parameters. GCC uses this
2337 information to improve pointer alignment analysis.
2339 The function parameter denoting the allocated alignment is specified by
2340 one integer argument, whose number is the argument of the attribute.
2341 Argument numbering starts at one.
2346 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2350 declares that @code{my_memalign} returns memory with minimum alignment
2351 given by parameter 1.
2354 @cindex @code{alloc_size} function attribute
2355 The @code{alloc_size} attribute is used to tell the compiler that the
2356 function return value points to memory, where the size is given by
2357 one or two of the functions parameters. GCC uses this
2358 information to improve the correctness of @code{__builtin_object_size}.
2360 The function parameter(s) denoting the allocated size are specified by
2361 one or two integer arguments supplied to the attribute. The allocated size
2362 is either the value of the single function argument specified or the product
2363 of the two function arguments specified. Argument numbering starts at
2369 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2370 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2374 declares that @code{my_calloc} returns memory of the size given by
2375 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2376 of the size given by parameter 2.
2379 @cindex @code{always_inline} function attribute
2380 Generally, functions are not inlined unless optimization is specified.
2381 For functions declared inline, this attribute inlines the function
2382 independent of any restrictions that otherwise apply to inlining.
2383 Failure to inline such a function is diagnosed as an error.
2384 Note that if such a function is called indirectly the compiler may
2385 or may not inline it depending on optimization level and a failure
2386 to inline an indirect call may or may not be diagnosed.
2389 @cindex @code{artificial} function attribute
2390 This attribute is useful for small inline wrappers that if possible
2391 should appear during debugging as a unit. Depending on the debug
2392 info format it either means marking the function as artificial
2393 or using the caller location for all instructions within the inlined
2396 @item assume_aligned
2397 @cindex @code{assume_aligned} function attribute
2398 The @code{assume_aligned} attribute is used to tell the compiler that the
2399 function return value points to memory, where the returned pointer minimum
2400 alignment is given by the first argument.
2401 If the attribute has two arguments, the second argument is misalignment offset.
2406 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2407 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2411 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2412 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2415 @item bnd_instrument
2416 @cindex @code{bnd_instrument} function attribute
2417 The @code{bnd_instrument} attribute on functions is used to inform the
2418 compiler that the function should be instrumented when compiled
2419 with the @option{-fchkp-instrument-marked-only} option.
2422 @cindex @code{bnd_legacy} function attribute
2423 @cindex Pointer Bounds Checker attributes
2424 The @code{bnd_legacy} attribute on functions is used to inform the
2425 compiler that the function should not be instrumented when compiled
2426 with the @option{-fcheck-pointer-bounds} option.
2429 @cindex @code{cold} function attribute
2430 The @code{cold} attribute on functions is used to inform the compiler that
2431 the function is unlikely to be executed. The function is optimized for
2432 size rather than speed and on many targets it is placed into a special
2433 subsection of the text section so all cold functions appear close together,
2434 improving code locality of non-cold parts of program. The paths leading
2435 to calls of cold functions within code are marked as unlikely by the branch
2436 prediction mechanism. It is thus useful to mark functions used to handle
2437 unlikely conditions, such as @code{perror}, as cold to improve optimization
2438 of hot functions that do call marked functions in rare occasions.
2440 When profile feedback is available, via @option{-fprofile-use}, cold functions
2441 are automatically detected and this attribute is ignored.
2444 @cindex @code{const} function attribute
2445 @cindex functions that have no side effects
2446 Many functions do not examine any values except their arguments, and
2447 have no effects except the return value. Basically this is just slightly
2448 more strict class than the @code{pure} attribute below, since function is not
2449 allowed to read global memory.
2451 @cindex pointer arguments
2452 Note that a function that has pointer arguments and examines the data
2453 pointed to must @emph{not} be declared @code{const}. Likewise, a
2454 function that calls a non-@code{const} function usually must not be
2455 @code{const}. It does not make sense for a @code{const} function to
2460 @itemx constructor (@var{priority})
2461 @itemx destructor (@var{priority})
2462 @cindex @code{constructor} function attribute
2463 @cindex @code{destructor} function attribute
2464 The @code{constructor} attribute causes the function to be called
2465 automatically before execution enters @code{main ()}. Similarly, the
2466 @code{destructor} attribute causes the function to be called
2467 automatically after @code{main ()} completes or @code{exit ()} is
2468 called. Functions with these attributes are useful for
2469 initializing data that is used implicitly during the execution of
2472 You may provide an optional integer priority to control the order in
2473 which constructor and destructor functions are run. A constructor
2474 with a smaller priority number runs before a constructor with a larger
2475 priority number; the opposite relationship holds for destructors. So,
2476 if you have a constructor that allocates a resource and a destructor
2477 that deallocates the same resource, both functions typically have the
2478 same priority. The priorities for constructor and destructor
2479 functions are the same as those specified for namespace-scope C++
2480 objects (@pxref{C++ Attributes}).
2482 These attributes are not currently implemented for Objective-C@.
2485 @itemx deprecated (@var{msg})
2486 @cindex @code{deprecated} function attribute
2487 The @code{deprecated} attribute results in a warning if the function
2488 is used anywhere in the source file. This is useful when identifying
2489 functions that are expected to be removed in a future version of a
2490 program. The warning also includes the location of the declaration
2491 of the deprecated function, to enable users to easily find further
2492 information about why the function is deprecated, or what they should
2493 do instead. Note that the warnings only occurs for uses:
2496 int old_fn () __attribute__ ((deprecated));
2498 int (*fn_ptr)() = old_fn;
2502 results in a warning on line 3 but not line 2. The optional @var{msg}
2503 argument, which must be a string, is printed in the warning if
2506 The @code{deprecated} attribute can also be used for variables and
2507 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2509 @item error ("@var{message}")
2510 @itemx warning ("@var{message}")
2511 @cindex @code{error} function attribute
2512 @cindex @code{warning} function attribute
2513 If the @code{error} or @code{warning} attribute
2514 is used on a function declaration and a call to such a function
2515 is not eliminated through dead code elimination or other optimizations,
2516 an error or warning (respectively) that includes @var{message} is diagnosed.
2518 for compile-time checking, especially together with @code{__builtin_constant_p}
2519 and inline functions where checking the inline function arguments is not
2520 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2522 While it is possible to leave the function undefined and thus invoke
2523 a link failure (to define the function with
2524 a message in @code{.gnu.warning*} section),
2525 when using these attributes the problem is diagnosed
2526 earlier and with exact location of the call even in presence of inline
2527 functions or when not emitting debugging information.
2529 @item externally_visible
2530 @cindex @code{externally_visible} function attribute
2531 This attribute, attached to a global variable or function, nullifies
2532 the effect of the @option{-fwhole-program} command-line option, so the
2533 object remains visible outside the current compilation unit.
2535 If @option{-fwhole-program} is used together with @option{-flto} and
2536 @command{gold} is used as the linker plugin,
2537 @code{externally_visible} attributes are automatically added to functions
2538 (not variable yet due to a current @command{gold} issue)
2539 that are accessed outside of LTO objects according to resolution file
2540 produced by @command{gold}.
2541 For other linkers that cannot generate resolution file,
2542 explicit @code{externally_visible} attributes are still necessary.
2545 @cindex @code{flatten} function attribute
2546 Generally, inlining into a function is limited. For a function marked with
2547 this attribute, every call inside this function is inlined, if possible.
2548 Whether the function itself is considered for inlining depends on its size and
2549 the current inlining parameters.
2551 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2552 @cindex @code{format} function attribute
2553 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2555 The @code{format} attribute specifies that a function takes @code{printf},
2556 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2557 should be type-checked against a format string. For example, the
2562 my_printf (void *my_object, const char *my_format, ...)
2563 __attribute__ ((format (printf, 2, 3)));
2567 causes the compiler to check the arguments in calls to @code{my_printf}
2568 for consistency with the @code{printf} style format string argument
2571 The parameter @var{archetype} determines how the format string is
2572 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2573 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2574 @code{strfmon}. (You can also use @code{__printf__},
2575 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2576 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2577 @code{ms_strftime} are also present.
2578 @var{archetype} values such as @code{printf} refer to the formats accepted
2579 by the system's C runtime library,
2580 while values prefixed with @samp{gnu_} always refer
2581 to the formats accepted by the GNU C Library. On Microsoft Windows
2582 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2583 @file{msvcrt.dll} library.
2584 The parameter @var{string-index}
2585 specifies which argument is the format string argument (starting
2586 from 1), while @var{first-to-check} is the number of the first
2587 argument to check against the format string. For functions
2588 where the arguments are not available to be checked (such as
2589 @code{vprintf}), specify the third parameter as zero. In this case the
2590 compiler only checks the format string for consistency. For
2591 @code{strftime} formats, the third parameter is required to be zero.
2592 Since non-static C++ methods have an implicit @code{this} argument, the
2593 arguments of such methods should be counted from two, not one, when
2594 giving values for @var{string-index} and @var{first-to-check}.
2596 In the example above, the format string (@code{my_format}) is the second
2597 argument of the function @code{my_print}, and the arguments to check
2598 start with the third argument, so the correct parameters for the format
2599 attribute are 2 and 3.
2601 @opindex ffreestanding
2602 @opindex fno-builtin
2603 The @code{format} attribute allows you to identify your own functions
2604 that take format strings as arguments, so that GCC can check the
2605 calls to these functions for errors. The compiler always (unless
2606 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2607 for the standard library functions @code{printf}, @code{fprintf},
2608 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2609 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2610 warnings are requested (using @option{-Wformat}), so there is no need to
2611 modify the header file @file{stdio.h}. In C99 mode, the functions
2612 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2613 @code{vsscanf} are also checked. Except in strictly conforming C
2614 standard modes, the X/Open function @code{strfmon} is also checked as
2615 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2616 @xref{C Dialect Options,,Options Controlling C Dialect}.
2618 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2619 recognized in the same context. Declarations including these format attributes
2620 are parsed for correct syntax, however the result of checking of such format
2621 strings is not yet defined, and is not carried out by this version of the
2624 The target may also provide additional types of format checks.
2625 @xref{Target Format Checks,,Format Checks Specific to Particular
2628 @item format_arg (@var{string-index})
2629 @cindex @code{format_arg} function attribute
2630 @opindex Wformat-nonliteral
2631 The @code{format_arg} attribute specifies that a function takes a format
2632 string for a @code{printf}, @code{scanf}, @code{strftime} or
2633 @code{strfmon} style function and modifies it (for example, to translate
2634 it into another language), so the result can be passed to a
2635 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2636 function (with the remaining arguments to the format function the same
2637 as they would have been for the unmodified string). For example, the
2642 my_dgettext (char *my_domain, const char *my_format)
2643 __attribute__ ((format_arg (2)));
2647 causes the compiler to check the arguments in calls to a @code{printf},
2648 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2649 format string argument is a call to the @code{my_dgettext} function, for
2650 consistency with the format string argument @code{my_format}. If the
2651 @code{format_arg} attribute had not been specified, all the compiler
2652 could tell in such calls to format functions would be that the format
2653 string argument is not constant; this would generate a warning when
2654 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2655 without the attribute.
2657 The parameter @var{string-index} specifies which argument is the format
2658 string argument (starting from one). Since non-static C++ methods have
2659 an implicit @code{this} argument, the arguments of such methods should
2660 be counted from two.
2662 The @code{format_arg} attribute allows you to identify your own
2663 functions that modify format strings, so that GCC can check the
2664 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2665 type function whose operands are a call to one of your own function.
2666 The compiler always treats @code{gettext}, @code{dgettext}, and
2667 @code{dcgettext} in this manner except when strict ISO C support is
2668 requested by @option{-ansi} or an appropriate @option{-std} option, or
2669 @option{-ffreestanding} or @option{-fno-builtin}
2670 is used. @xref{C Dialect Options,,Options
2671 Controlling C Dialect}.
2673 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2674 @code{NSString} reference for compatibility with the @code{format} attribute
2677 The target may also allow additional types in @code{format-arg} attributes.
2678 @xref{Target Format Checks,,Format Checks Specific to Particular
2682 @cindex @code{gnu_inline} function attribute
2683 This attribute should be used with a function that is also declared
2684 with the @code{inline} keyword. It directs GCC to treat the function
2685 as if it were defined in gnu90 mode even when compiling in C99 or
2688 If the function is declared @code{extern}, then this definition of the
2689 function is used only for inlining. In no case is the function
2690 compiled as a standalone function, not even if you take its address
2691 explicitly. Such an address becomes an external reference, as if you
2692 had only declared the function, and had not defined it. This has
2693 almost the effect of a macro. The way to use this is to put a
2694 function definition in a header file with this attribute, and put
2695 another copy of the function, without @code{extern}, in a library
2696 file. The definition in the header file causes most calls to the
2697 function to be inlined. If any uses of the function remain, they
2698 refer to the single copy in the library. Note that the two
2699 definitions of the functions need not be precisely the same, although
2700 if they do not have the same effect your program may behave oddly.
2702 In C, if the function is neither @code{extern} nor @code{static}, then
2703 the function is compiled as a standalone function, as well as being
2704 inlined where possible.
2706 This is how GCC traditionally handled functions declared
2707 @code{inline}. Since ISO C99 specifies a different semantics for
2708 @code{inline}, this function attribute is provided as a transition
2709 measure and as a useful feature in its own right. This attribute is
2710 available in GCC 4.1.3 and later. It is available if either of the
2711 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2712 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2713 Function is As Fast As a Macro}.
2715 In C++, this attribute does not depend on @code{extern} in any way,
2716 but it still requires the @code{inline} keyword to enable its special
2720 @cindex @code{hot} function attribute
2721 The @code{hot} attribute on a function is used to inform the compiler that
2722 the function is a hot spot of the compiled program. The function is
2723 optimized more aggressively and on many targets it is placed into a special
2724 subsection of the text section so all hot functions appear close together,
2727 When profile feedback is available, via @option{-fprofile-use}, hot functions
2728 are automatically detected and this attribute is ignored.
2730 @item ifunc ("@var{resolver}")
2731 @cindex @code{ifunc} function attribute
2732 @cindex indirect functions
2733 @cindex functions that are dynamically resolved
2734 The @code{ifunc} attribute is used to mark a function as an indirect
2735 function using the STT_GNU_IFUNC symbol type extension to the ELF
2736 standard. This allows the resolution of the symbol value to be
2737 determined dynamically at load time, and an optimized version of the
2738 routine can be selected for the particular processor or other system
2739 characteristics determined then. To use this attribute, first define
2740 the implementation functions available, and a resolver function that
2741 returns a pointer to the selected implementation function. The
2742 implementation functions' declarations must match the API of the
2743 function being implemented, the resolver's declaration is be a
2744 function returning pointer to void function returning void:
2747 void *my_memcpy (void *dst, const void *src, size_t len)
2752 static void (*resolve_memcpy (void)) (void)
2754 return my_memcpy; // we'll just always select this routine
2759 The exported header file declaring the function the user calls would
2763 extern void *memcpy (void *, const void *, size_t);
2767 allowing the user to call this as a regular function, unaware of the
2768 implementation. Finally, the indirect function needs to be defined in
2769 the same translation unit as the resolver function:
2772 void *memcpy (void *, const void *, size_t)
2773 __attribute__ ((ifunc ("resolve_memcpy")));
2776 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2777 and GNU C Library version 2.11.1 are required to use this feature.
2780 @itemx interrupt_handler
2781 Many GCC back ends support attributes to indicate that a function is
2782 an interrupt handler, which tells the compiler to generate function
2783 entry and exit sequences that differ from those from regular
2784 functions. The exact syntax and behavior are target-specific;
2785 refer to the following subsections for details.
2788 @cindex @code{leaf} function attribute
2789 Calls to external functions with this attribute must return to the
2790 current compilation unit only by return or by exception handling. In
2791 particular, a leaf function is not allowed to invoke callback functions
2792 passed to it from the current compilation unit, directly call functions
2793 exported by the unit, or @code{longjmp} into the unit. Leaf functions
2794 might still call functions from other compilation units and thus they
2795 are not necessarily leaf in the sense that they contain no function
2798 The attribute is intended for library functions to improve dataflow
2799 analysis. The compiler takes the hint that any data not escaping the
2800 current compilation unit cannot be used or modified by the leaf
2801 function. For example, the @code{sin} function is a leaf function, but
2802 @code{qsort} is not.
2804 Note that leaf functions might indirectly run a signal handler defined
2805 in the current compilation unit that uses static variables. Similarly,
2806 when lazy symbol resolution is in effect, leaf functions might invoke
2807 indirect functions whose resolver function or implementation function is
2808 defined in the current compilation unit and uses static variables. There
2809 is no standard-compliant way to write such a signal handler, resolver
2810 function, or implementation function, and the best that you can do is to
2811 remove the @code{leaf} attribute or mark all such static variables
2812 @code{volatile}. Lastly, for ELF-based systems that support symbol
2813 interposition, care should be taken that functions defined in the
2814 current compilation unit do not unexpectedly interpose other symbols
2815 based on the defined standards mode and defined feature test macros;
2816 otherwise an inadvertent callback would be added.
2818 The attribute has no effect on functions defined within the current
2819 compilation unit. This is to allow easy merging of multiple compilation
2820 units into one, for example, by using the link-time optimization. For
2821 this reason the attribute is not allowed on types to annotate indirect
2825 @cindex @code{malloc} function attribute
2826 @cindex functions that behave like malloc
2827 This tells the compiler that a function is @code{malloc}-like, i.e.,
2828 that the pointer @var{P} returned by the function cannot alias any
2829 other pointer valid when the function returns, and moreover no
2830 pointers to valid objects occur in any storage addressed by @var{P}.
2832 Using this attribute can improve optimization. Functions like
2833 @code{malloc} and @code{calloc} have this property because they return
2834 a pointer to uninitialized or zeroed-out storage. However, functions
2835 like @code{realloc} do not have this property, as they can return a
2836 pointer to storage containing pointers.
2839 @cindex @code{no_icf} function attribute
2840 This function attribute prevents a functions from being merged with another
2841 semantically equivalent function.
2843 @item no_instrument_function
2844 @cindex @code{no_instrument_function} function attribute
2845 @opindex finstrument-functions
2846 If @option{-finstrument-functions} is given, profiling function calls are
2847 generated at entry and exit of most user-compiled functions.
2848 Functions with this attribute are not so instrumented.
2851 @cindex @code{no_reorder} function attribute
2852 Do not reorder functions or variables marked @code{no_reorder}
2853 against each other or top level assembler statements the executable.
2854 The actual order in the program will depend on the linker command
2855 line. Static variables marked like this are also not removed.
2856 This has a similar effect
2857 as the @option{-fno-toplevel-reorder} option, but only applies to the
2860 @item no_sanitize_address
2861 @itemx no_address_safety_analysis
2862 @cindex @code{no_sanitize_address} function attribute
2863 The @code{no_sanitize_address} attribute on functions is used
2864 to inform the compiler that it should not instrument memory accesses
2865 in the function when compiling with the @option{-fsanitize=address} option.
2866 The @code{no_address_safety_analysis} is a deprecated alias of the
2867 @code{no_sanitize_address} attribute, new code should use
2868 @code{no_sanitize_address}.
2870 @item no_sanitize_thread
2871 @cindex @code{no_sanitize_thread} function attribute
2872 The @code{no_sanitize_thread} attribute on functions is used
2873 to inform the compiler that it should not instrument memory accesses
2874 in the function when compiling with the @option{-fsanitize=thread} option.
2876 @item no_sanitize_undefined
2877 @cindex @code{no_sanitize_undefined} function attribute
2878 The @code{no_sanitize_undefined} attribute on functions is used
2879 to inform the compiler that it should not check for undefined behavior
2880 in the function when compiling with the @option{-fsanitize=undefined} option.
2882 @item no_split_stack
2883 @cindex @code{no_split_stack} function attribute
2884 @opindex fsplit-stack
2885 If @option{-fsplit-stack} is given, functions have a small
2886 prologue which decides whether to split the stack. Functions with the
2887 @code{no_split_stack} attribute do not have that prologue, and thus
2888 may run with only a small amount of stack space available.
2890 @item no_stack_limit
2891 @cindex @code{no_stack_limit} function attribute
2892 This attribute locally overrides the @option{-fstack-limit-register}
2893 and @option{-fstack-limit-symbol} command-line options; it has the effect
2894 of disabling stack limit checking in the function it applies to.
2897 @cindex @code{noclone} function attribute
2898 This function attribute prevents a function from being considered for
2899 cloning---a mechanism that produces specialized copies of functions
2900 and which is (currently) performed by interprocedural constant
2904 @cindex @code{noinline} function attribute
2905 This function attribute prevents a function from being considered for
2907 @c Don't enumerate the optimizations by name here; we try to be
2908 @c future-compatible with this mechanism.
2909 If the function does not have side-effects, there are optimizations
2910 other than inlining that cause function calls to be optimized away,
2911 although the function call is live. To keep such calls from being
2918 (@pxref{Extended Asm}) in the called function, to serve as a special
2921 @item nonnull (@var{arg-index}, @dots{})
2922 @cindex @code{nonnull} function attribute
2923 @cindex functions with non-null pointer arguments
2924 The @code{nonnull} attribute specifies that some function parameters should
2925 be non-null pointers. For instance, the declaration:
2929 my_memcpy (void *dest, const void *src, size_t len)
2930 __attribute__((nonnull (1, 2)));
2934 causes the compiler to check that, in calls to @code{my_memcpy},
2935 arguments @var{dest} and @var{src} are non-null. If the compiler
2936 determines that a null pointer is passed in an argument slot marked
2937 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2938 is issued. The compiler may also choose to make optimizations based
2939 on the knowledge that certain function arguments will never be null.
2941 If no argument index list is given to the @code{nonnull} attribute,
2942 all pointer arguments are marked as non-null. To illustrate, the
2943 following declaration is equivalent to the previous example:
2947 my_memcpy (void *dest, const void *src, size_t len)
2948 __attribute__((nonnull));
2952 @cindex @code{noplt} function attribute
2953 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
2954 Calls to functions marked with this attribute in position-independent code
2959 /* Externally defined function foo. */
2960 int foo () __attribute__ ((noplt));
2963 main (/* @r{@dots{}} */)
2972 The @code{noplt} attribute on function @code{foo}
2973 tells the compiler to assume that
2974 the function @code{foo} is externally defined and that the call to
2975 @code{foo} must avoid the PLT
2976 in position-independent code.
2978 In position-dependent code, a few targets also convert calls to
2979 functions that are marked to not use the PLT to use the GOT instead.
2982 @cindex @code{noreturn} function attribute
2983 @cindex functions that never return
2984 A few standard library functions, such as @code{abort} and @code{exit},
2985 cannot return. GCC knows this automatically. Some programs define
2986 their own functions that never return. You can declare them
2987 @code{noreturn} to tell the compiler this fact. For example,
2991 void fatal () __attribute__ ((noreturn));
2994 fatal (/* @r{@dots{}} */)
2996 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3002 The @code{noreturn} keyword tells the compiler to assume that
3003 @code{fatal} cannot return. It can then optimize without regard to what
3004 would happen if @code{fatal} ever did return. This makes slightly
3005 better code. More importantly, it helps avoid spurious warnings of
3006 uninitialized variables.
3008 The @code{noreturn} keyword does not affect the exceptional path when that
3009 applies: a @code{noreturn}-marked function may still return to the caller
3010 by throwing an exception or calling @code{longjmp}.
3012 Do not assume that registers saved by the calling function are
3013 restored before calling the @code{noreturn} function.
3015 It does not make sense for a @code{noreturn} function to have a return
3016 type other than @code{void}.
3019 @cindex @code{nothrow} function attribute
3020 The @code{nothrow} attribute is used to inform the compiler that a
3021 function cannot throw an exception. For example, most functions in
3022 the standard C library can be guaranteed not to throw an exception
3023 with the notable exceptions of @code{qsort} and @code{bsearch} that
3024 take function pointer arguments.
3027 @cindex @code{optimize} function attribute
3028 The @code{optimize} attribute is used to specify that a function is to
3029 be compiled with different optimization options than specified on the
3030 command line. Arguments can either be numbers or strings. Numbers
3031 are assumed to be an optimization level. Strings that begin with
3032 @code{O} are assumed to be an optimization option, while other options
3033 are assumed to be used with a @code{-f} prefix. You can also use the
3034 @samp{#pragma GCC optimize} pragma to set the optimization options
3035 that affect more than one function.
3036 @xref{Function Specific Option Pragmas}, for details about the
3037 @samp{#pragma GCC optimize} pragma.
3039 This attribute should be used for debugging purposes only. It is not
3040 suitable in production code.
3043 @cindex @code{pure} function attribute
3044 @cindex functions that have no side effects
3045 Many functions have no effects except the return value and their
3046 return value depends only on the parameters and/or global variables.
3047 Such a function can be subject
3048 to common subexpression elimination and loop optimization just as an
3049 arithmetic operator would be. These functions should be declared
3050 with the attribute @code{pure}. For example,
3053 int square (int) __attribute__ ((pure));
3057 says that the hypothetical function @code{square} is safe to call
3058 fewer times than the program says.
3060 Some common examples of pure functions are @code{strlen} or @code{memcmp}.
3061 Interesting non-pure functions are functions with infinite loops or those
3062 depending on volatile memory or other system resource, that may change between
3063 two consecutive calls (such as @code{feof} in a multithreading environment).
3065 @item returns_nonnull
3066 @cindex @code{returns_nonnull} function attribute
3067 The @code{returns_nonnull} attribute specifies that the function
3068 return value should be a non-null pointer. For instance, the declaration:
3072 mymalloc (size_t len) __attribute__((returns_nonnull));
3076 lets the compiler optimize callers based on the knowledge
3077 that the return value will never be null.
3080 @cindex @code{returns_twice} function attribute
3081 @cindex functions that return more than once
3082 The @code{returns_twice} attribute tells the compiler that a function may
3083 return more than one time. The compiler ensures that all registers
3084 are dead before calling such a function and emits a warning about
3085 the variables that may be clobbered after the second return from the
3086 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3087 The @code{longjmp}-like counterpart of such function, if any, might need
3088 to be marked with the @code{noreturn} attribute.
3090 @item section ("@var{section-name}")
3091 @cindex @code{section} function attribute
3092 @cindex functions in arbitrary sections
3093 Normally, the compiler places the code it generates in the @code{text} section.
3094 Sometimes, however, you need additional sections, or you need certain
3095 particular functions to appear in special sections. The @code{section}
3096 attribute specifies that a function lives in a particular section.
3097 For example, the declaration:
3100 extern void foobar (void) __attribute__ ((section ("bar")));
3104 puts the function @code{foobar} in the @code{bar} section.
3106 Some file formats do not support arbitrary sections so the @code{section}
3107 attribute is not available on all platforms.
3108 If you need to map the entire contents of a module to a particular
3109 section, consider using the facilities of the linker instead.
3112 @cindex @code{sentinel} function attribute
3113 This function attribute ensures that a parameter in a function call is
3114 an explicit @code{NULL}. The attribute is only valid on variadic
3115 functions. By default, the sentinel is located at position zero, the
3116 last parameter of the function call. If an optional integer position
3117 argument P is supplied to the attribute, the sentinel must be located at
3118 position P counting backwards from the end of the argument list.
3121 __attribute__ ((sentinel))
3123 __attribute__ ((sentinel(0)))
3126 The attribute is automatically set with a position of 0 for the built-in
3127 functions @code{execl} and @code{execlp}. The built-in function
3128 @code{execle} has the attribute set with a position of 1.
3130 A valid @code{NULL} in this context is defined as zero with any pointer
3131 type. If your system defines the @code{NULL} macro with an integer type
3132 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3133 with a copy that redefines NULL appropriately.
3135 The warnings for missing or incorrect sentinels are enabled with
3139 @itemx simd("@var{mask}")
3140 @cindex @code{simd} function attribute
3141 This attribute enables creation of one or more function versions that
3142 can process multiple arguments using SIMD instructions from a
3143 single invocation. Specifying this attribute allows compiler to
3144 assume that such versions are available at link time (provided
3145 in the same or another translation unit). Generated versions are
3146 target-dependent and described in the corresponding Vector ABI document. For
3147 x86_64 target this document can be found
3148 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3150 The optional argument @var{mask} may have the value
3151 @code{notinbranch} or @code{inbranch},
3152 and instructs the compiler to generate non-masked or masked
3153 clones correspondingly. By default, all clones are generated.
3155 The attribute should not be used together with Cilk Plus @code{vector}
3156 attribute on the same function.
3158 If the attribute is specified and @code{#pragma omp declare simd} is
3159 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3160 switch is specified, then the attribute is ignored.
3163 @cindex @code{stack_protect} function attribute
3164 This attribute adds stack protection code to the function if
3165 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3166 or @option{-fstack-protector-explicit} are set.
3168 @item target (@var{options})
3169 @cindex @code{target} function attribute
3170 Multiple target back ends implement the @code{target} attribute
3171 to specify that a function is to
3172 be compiled with different target options than specified on the
3173 command line. This can be used for instance to have functions
3174 compiled with a different ISA (instruction set architecture) than the
3175 default. You can also use the @samp{#pragma GCC target} pragma to set
3176 more than one function to be compiled with specific target options.
3177 @xref{Function Specific Option Pragmas}, for details about the
3178 @samp{#pragma GCC target} pragma.
3180 For instance, on an x86, you could declare one function with the
3181 @code{target("sse4.1,arch=core2")} attribute and another with
3182 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3183 compiling the first function with @option{-msse4.1} and
3184 @option{-march=core2} options, and the second function with
3185 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3186 to make sure that a function is only invoked on a machine that
3187 supports the particular ISA it is compiled for (for example by using
3188 @code{cpuid} on x86 to determine what feature bits and architecture
3192 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3193 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3196 You can either use multiple
3197 strings separated by commas to specify multiple options,
3198 or separate the options with a comma (@samp{,}) within a single string.
3200 The options supported are specific to each target; refer to @ref{x86
3201 Function Attributes}, @ref{PowerPC Function Attributes},
3202 @ref{ARM Function Attributes},and @ref{Nios II Function Attributes},
3205 @item target_clones (@var{options})
3206 @cindex @code{target_clones} function attribute
3207 The @code{target_clones} attribute is used to specify that a function
3208 be cloned into multiple versions compiled with different target options
3209 than specified on the command line. The supported options and restrictions
3210 are the same as for @code{target} attribute.
3212 For instance, on an x86, you could compile a function with
3213 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3214 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3215 It also creates a resolver function (see the @code{ifunc} attribute
3216 above) that dynamically selects a clone suitable for current architecture.
3219 @cindex @code{unused} function attribute
3220 This attribute, attached to a function, means that the function is meant
3221 to be possibly unused. GCC does not produce a warning for this
3225 @cindex @code{used} function attribute
3226 This attribute, attached to a function, means that code must be emitted
3227 for the function even if it appears that the function is not referenced.
3228 This is useful, for example, when the function is referenced only in
3231 When applied to a member function of a C++ class template, the
3232 attribute also means that the function is instantiated if the
3233 class itself is instantiated.
3235 @item visibility ("@var{visibility_type}")
3236 @cindex @code{visibility} function attribute
3237 This attribute affects the linkage of the declaration to which it is attached.
3238 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3239 (@pxref{Common Type Attributes}) as well as functions.
3241 There are four supported @var{visibility_type} values: default,
3242 hidden, protected or internal visibility.
3245 void __attribute__ ((visibility ("protected")))
3246 f () @{ /* @r{Do something.} */; @}
3247 int i __attribute__ ((visibility ("hidden")));
3250 The possible values of @var{visibility_type} correspond to the
3251 visibility settings in the ELF gABI.
3254 @c keep this list of visibilities in alphabetical order.
3257 Default visibility is the normal case for the object file format.
3258 This value is available for the visibility attribute to override other
3259 options that may change the assumed visibility of entities.
3261 On ELF, default visibility means that the declaration is visible to other
3262 modules and, in shared libraries, means that the declared entity may be
3265 On Darwin, default visibility means that the declaration is visible to
3268 Default visibility corresponds to ``external linkage'' in the language.
3271 Hidden visibility indicates that the entity declared has a new
3272 form of linkage, which we call ``hidden linkage''. Two
3273 declarations of an object with hidden linkage refer to the same object
3274 if they are in the same shared object.
3277 Internal visibility is like hidden visibility, but with additional
3278 processor specific semantics. Unless otherwise specified by the
3279 psABI, GCC defines internal visibility to mean that a function is
3280 @emph{never} called from another module. Compare this with hidden
3281 functions which, while they cannot be referenced directly by other
3282 modules, can be referenced indirectly via function pointers. By
3283 indicating that a function cannot be called from outside the module,
3284 GCC may for instance omit the load of a PIC register since it is known
3285 that the calling function loaded the correct value.
3288 Protected visibility is like default visibility except that it
3289 indicates that references within the defining module bind to the
3290 definition in that module. That is, the declared entity cannot be
3291 overridden by another module.
3295 All visibilities are supported on many, but not all, ELF targets
3296 (supported when the assembler supports the @samp{.visibility}
3297 pseudo-op). Default visibility is supported everywhere. Hidden
3298 visibility is supported on Darwin targets.
3300 The visibility attribute should be applied only to declarations that
3301 would otherwise have external linkage. The attribute should be applied
3302 consistently, so that the same entity should not be declared with
3303 different settings of the attribute.
3305 In C++, the visibility attribute applies to types as well as functions
3306 and objects, because in C++ types have linkage. A class must not have
3307 greater visibility than its non-static data member types and bases,
3308 and class members default to the visibility of their class. Also, a
3309 declaration without explicit visibility is limited to the visibility
3312 In C++, you can mark member functions and static member variables of a
3313 class with the visibility attribute. This is useful if you know a
3314 particular method or static member variable should only be used from
3315 one shared object; then you can mark it hidden while the rest of the
3316 class has default visibility. Care must be taken to avoid breaking
3317 the One Definition Rule; for example, it is usually not useful to mark
3318 an inline method as hidden without marking the whole class as hidden.
3320 A C++ namespace declaration can also have the visibility attribute.
3323 namespace nspace1 __attribute__ ((visibility ("protected")))
3324 @{ /* @r{Do something.} */; @}
3327 This attribute applies only to the particular namespace body, not to
3328 other definitions of the same namespace; it is equivalent to using
3329 @samp{#pragma GCC visibility} before and after the namespace
3330 definition (@pxref{Visibility Pragmas}).
3332 In C++, if a template argument has limited visibility, this
3333 restriction is implicitly propagated to the template instantiation.
3334 Otherwise, template instantiations and specializations default to the
3335 visibility of their template.
3337 If both the template and enclosing class have explicit visibility, the
3338 visibility from the template is used.
3340 @item warn_unused_result
3341 @cindex @code{warn_unused_result} function attribute
3342 The @code{warn_unused_result} attribute causes a warning to be emitted
3343 if a caller of the function with this attribute does not use its
3344 return value. This is useful for functions where not checking
3345 the result is either a security problem or always a bug, such as
3349 int fn () __attribute__ ((warn_unused_result));
3352 if (fn () < 0) return -1;
3359 results in warning on line 5.
3362 @cindex @code{weak} function attribute
3363 The @code{weak} attribute causes the declaration to be emitted as a weak
3364 symbol rather than a global. This is primarily useful in defining
3365 library functions that can be overridden in user code, though it can
3366 also be used with non-function declarations. Weak symbols are supported
3367 for ELF targets, and also for a.out targets when using the GNU assembler
3371 @itemx weakref ("@var{target}")
3372 @cindex @code{weakref} function attribute
3373 The @code{weakref} attribute marks a declaration as a weak reference.
3374 Without arguments, it should be accompanied by an @code{alias} attribute
3375 naming the target symbol. Optionally, the @var{target} may be given as
3376 an argument to @code{weakref} itself. In either case, @code{weakref}
3377 implicitly marks the declaration as @code{weak}. Without a
3378 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3379 @code{weakref} is equivalent to @code{weak}.
3382 static int x() __attribute__ ((weakref ("y")));
3383 /* is equivalent to... */
3384 static int x() __attribute__ ((weak, weakref, alias ("y")));
3386 static int x() __attribute__ ((weakref));
3387 static int x() __attribute__ ((alias ("y")));
3390 A weak reference is an alias that does not by itself require a
3391 definition to be given for the target symbol. If the target symbol is
3392 only referenced through weak references, then it becomes a @code{weak}
3393 undefined symbol. If it is directly referenced, however, then such
3394 strong references prevail, and a definition is required for the
3395 symbol, not necessarily in the same translation unit.
3397 The effect is equivalent to moving all references to the alias to a
3398 separate translation unit, renaming the alias to the aliased symbol,
3399 declaring it as weak, compiling the two separate translation units and
3400 performing a reloadable link on them.
3402 At present, a declaration to which @code{weakref} is attached can
3403 only be @code{static}.
3408 @c This is the end of the target-independent attribute table
3410 @node AArch64 Function Attributes
3411 @subsection AArch64 Function Attributes
3413 The following target-specific function attributes are available for the
3414 AArch64 target. For the most part, these options mirror the behavior of
3415 similar command-line options (@pxref{AArch64 Options}), but on a
3419 @item general-regs-only
3420 @cindex @code{general-regs-only} function attribute, AArch64
3421 Indicates that no floating-point or Advanced SIMD registers should be
3422 used when generating code for this function. If the function explicitly
3423 uses floating-point code, then the compiler gives an error. This is
3424 the same behavior as that of the command-line option
3425 @option{-mgeneral-regs-only}.
3427 @item fix-cortex-a53-835769
3428 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3429 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3430 applied to this function. To explicitly disable the workaround for this
3431 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3432 This corresponds to the behavior of the command line options
3433 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3436 @cindex @code{cmodel=} function attribute, AArch64
3437 Indicates that code should be generated for a particular code model for
3438 this function. The behavior and permissible arguments are the same as
3439 for the command line option @option{-mcmodel=}.
3442 @cindex @code{strict-align} function attribute, AArch64
3443 Indicates that the compiler should not assume that unaligned memory references
3444 are handled by the system. The behavior is the same as for the command-line
3445 option @option{-mstrict-align}.
3447 @item omit-leaf-frame-pointer
3448 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3449 Indicates that the frame pointer should be omitted for a leaf function call.
3450 To keep the frame pointer, the inverse attribute
3451 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3452 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3453 and @option{-mno-omit-leaf-frame-pointer}.
3456 @cindex @code{tls-dialect=} function attribute, AArch64
3457 Specifies the TLS dialect to use for this function. The behavior and
3458 permissible arguments are the same as for the command-line option
3459 @option{-mtls-dialect=}.
3462 @cindex @code{arch=} function attribute, AArch64
3463 Specifies the architecture version and architectural extensions to use
3464 for this function. The behavior and permissible arguments are the same as
3465 for the @option{-march=} command-line option.
3468 @cindex @code{tune=} function attribute, AArch64
3469 Specifies the core for which to tune the performance of this function.
3470 The behavior and permissible arguments are the same as for the @option{-mtune=}
3471 command-line option.
3474 @cindex @code{cpu=} function attribute, AArch64
3475 Specifies the core for which to tune the performance of this function and also
3476 whose architectural features to use. The behavior and valid arguments are the
3477 same as for the @option{-mcpu=} command-line option.
3481 The above target attributes can be specified as follows:
3484 __attribute__((target("@var{attr-string}")))
3492 where @code{@var{attr-string}} is one of the attribute strings specified above.
3494 Additionally, the architectural extension string may be specified on its
3495 own. This can be used to turn on and off particular architectural extensions
3496 without having to specify a particular architecture version or core. Example:
3499 __attribute__((target("+crc+nocrypto")))
3507 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3508 extension and disables the @code{crypto} extension for the function @code{foo}
3509 without modifying an existing @option{-march=} or @option{-mcpu} option.
3511 Multiple target function attributes can be specified by separating them with
3512 a comma. For example:
3514 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3522 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3523 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3525 @subsubsection Inlining rules
3526 Specifying target attributes on individual functions or performing link-time
3527 optimization across translation units compiled with different target options
3528 can affect function inlining rules:
3530 In particular, a caller function can inline a callee function only if the
3531 architectural features available to the callee are a subset of the features
3532 available to the caller.
3533 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3534 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3535 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3536 because the all the architectural features that function @code{bar} requires
3537 are available to function @code{foo}. Conversely, function @code{bar} cannot
3538 inline function @code{foo}.
3540 Additionally inlining a function compiled with @option{-mstrict-align} into a
3541 function compiled without @code{-mstrict-align} is not allowed.
3542 However, inlining a function compiled without @option{-mstrict-align} into a
3543 function compiled with @option{-mstrict-align} is allowed.
3545 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3546 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3547 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3548 architectural feature rules specified above.
3550 @node ARC Function Attributes
3551 @subsection ARC Function Attributes
3553 These function attributes are supported by the ARC back end:
3557 @cindex @code{interrupt} function attribute, ARC
3558 Use this attribute to indicate
3559 that the specified function is an interrupt handler. The compiler generates
3560 function entry and exit sequences suitable for use in an interrupt handler
3561 when this attribute is present.
3563 On the ARC, you must specify the kind of interrupt to be handled
3564 in a parameter to the interrupt attribute like this:
3567 void f () __attribute__ ((interrupt ("ilink1")));
3570 Permissible values for this parameter are: @w{@code{ilink1}} and
3576 @cindex @code{long_call} function attribute, ARC
3577 @cindex @code{medium_call} function attribute, ARC
3578 @cindex @code{short_call} function attribute, ARC
3579 @cindex indirect calls, ARC
3580 These attributes specify how a particular function is called.
3581 These attributes override the
3582 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3583 command-line switches and @code{#pragma long_calls} settings.
3585 For ARC, a function marked with the @code{long_call} attribute is
3586 always called using register-indirect jump-and-link instructions,
3587 thereby enabling the called function to be placed anywhere within the
3588 32-bit address space. A function marked with the @code{medium_call}
3589 attribute will always be close enough to be called with an unconditional
3590 branch-and-link instruction, which has a 25-bit offset from
3591 the call site. A function marked with the @code{short_call}
3592 attribute will always be close enough to be called with a conditional
3593 branch-and-link instruction, which has a 21-bit offset from
3597 @node ARM Function Attributes
3598 @subsection ARM Function Attributes
3600 These function attributes are supported for ARM targets:
3604 @cindex @code{interrupt} function attribute, ARM
3605 Use this attribute to indicate
3606 that the specified function is an interrupt handler. The compiler generates
3607 function entry and exit sequences suitable for use in an interrupt handler
3608 when this attribute is present.
3610 You can specify the kind of interrupt to be handled by
3611 adding an optional parameter to the interrupt attribute like this:
3614 void f () __attribute__ ((interrupt ("IRQ")));
3618 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3619 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3621 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3622 may be called with a word-aligned stack pointer.
3625 @cindex @code{isr} function attribute, ARM
3626 Use this attribute on ARM to write Interrupt Service Routines. This is an
3627 alias to the @code{interrupt} attribute above.
3631 @cindex @code{long_call} function attribute, ARM
3632 @cindex @code{short_call} function attribute, ARM
3633 @cindex indirect calls, ARM
3634 These attributes specify how a particular function is called.
3635 These attributes override the
3636 @option{-mlong-calls} (@pxref{ARM Options})
3637 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3638 @code{long_call} attribute indicates that the function might be far
3639 away from the call site and require a different (more expensive)
3640 calling sequence. The @code{short_call} attribute always places
3641 the offset to the function from the call site into the @samp{BL}
3642 instruction directly.
3645 @cindex @code{naked} function attribute, ARM
3646 This attribute allows the compiler to construct the
3647 requisite function declaration, while allowing the body of the
3648 function to be assembly code. The specified function will not have
3649 prologue/epilogue sequences generated by the compiler. Only basic
3650 @code{asm} statements can safely be included in naked functions
3651 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3652 basic @code{asm} and C code may appear to work, they cannot be
3653 depended upon to work reliably and are not supported.
3656 @cindex @code{pcs} function attribute, ARM
3658 The @code{pcs} attribute can be used to control the calling convention
3659 used for a function on ARM. The attribute takes an argument that specifies
3660 the calling convention to use.
3662 When compiling using the AAPCS ABI (or a variant of it) then valid
3663 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3664 order to use a variant other than @code{"aapcs"} then the compiler must
3665 be permitted to use the appropriate co-processor registers (i.e., the
3666 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3670 /* Argument passed in r0, and result returned in r0+r1. */
3671 double f2d (float) __attribute__((pcs("aapcs")));
3674 Variadic functions always use the @code{"aapcs"} calling convention and
3675 the compiler rejects attempts to specify an alternative.
3677 @item target (@var{options})
3678 @cindex @code{target} function attribute
3679 As discussed in @ref{Common Function Attributes}, this attribute
3680 allows specification of target-specific compilation options.
3682 On ARM, the following options are allowed:
3686 @cindex @code{target("thumb")} function attribute, ARM
3687 Force code generation in the Thumb (T16/T32) ISA, depending on the
3691 @cindex @code{target("arm")} function attribute, ARM
3692 Force code generation in the ARM (A32) ISA.
3694 Functions from different modes can be inlined in the caller's mode.
3697 @cindex @code{target("fpu=")} function attribute, ARM
3698 Specifies the fpu for which to tune the performance of this function.
3699 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3700 command-line option.
3706 @node AVR Function Attributes
3707 @subsection AVR Function Attributes
3709 These function attributes are supported by the AVR back end:
3713 @cindex @code{interrupt} function attribute, AVR
3714 Use this attribute to indicate
3715 that the specified function is an interrupt handler. The compiler generates
3716 function entry and exit sequences suitable for use in an interrupt handler
3717 when this attribute is present.
3719 On the AVR, the hardware globally disables interrupts when an
3720 interrupt is executed. The first instruction of an interrupt handler
3721 declared with this attribute is a @code{SEI} instruction to
3722 re-enable interrupts. See also the @code{signal} function attribute
3723 that does not insert a @code{SEI} instruction. If both @code{signal} and
3724 @code{interrupt} are specified for the same function, @code{signal}
3725 is silently ignored.
3728 @cindex @code{naked} function attribute, AVR
3729 This attribute allows the compiler to construct the
3730 requisite function declaration, while allowing the body of the
3731 function to be assembly code. The specified function will not have
3732 prologue/epilogue sequences generated by the compiler. Only basic
3733 @code{asm} statements can safely be included in naked functions
3734 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3735 basic @code{asm} and C code may appear to work, they cannot be
3736 depended upon to work reliably and are not supported.
3740 @cindex @code{OS_main} function attribute, AVR
3741 @cindex @code{OS_task} function attribute, AVR
3742 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3743 do not save/restore any call-saved register in their prologue/epilogue.
3745 The @code{OS_main} attribute can be used when there @emph{is
3746 guarantee} that interrupts are disabled at the time when the function
3747 is entered. This saves resources when the stack pointer has to be
3748 changed to set up a frame for local variables.
3750 The @code{OS_task} attribute can be used when there is @emph{no
3751 guarantee} that interrupts are disabled at that time when the function
3752 is entered like for, e@.g@. task functions in a multi-threading operating
3753 system. In that case, changing the stack pointer register is
3754 guarded by save/clear/restore of the global interrupt enable flag.
3756 The differences to the @code{naked} function attribute are:
3758 @item @code{naked} functions do not have a return instruction whereas
3759 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
3760 @code{RETI} return instruction.
3761 @item @code{naked} functions do not set up a frame for local variables
3762 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3767 @cindex @code{signal} function attribute, AVR
3768 Use this attribute on the AVR to indicate that the specified
3769 function is an interrupt handler. The compiler generates function
3770 entry and exit sequences suitable for use in an interrupt handler when this
3771 attribute is present.
3773 See also the @code{interrupt} function attribute.
3775 The AVR hardware globally disables interrupts when an interrupt is executed.
3776 Interrupt handler functions defined with the @code{signal} attribute
3777 do not re-enable interrupts. It is save to enable interrupts in a
3778 @code{signal} handler. This ``save'' only applies to the code
3779 generated by the compiler and not to the IRQ layout of the
3780 application which is responsibility of the application.
3782 If both @code{signal} and @code{interrupt} are specified for the same
3783 function, @code{signal} is silently ignored.
3786 @node Blackfin Function Attributes
3787 @subsection Blackfin Function Attributes
3789 These function attributes are supported by the Blackfin back end:
3793 @item exception_handler
3794 @cindex @code{exception_handler} function attribute
3795 @cindex exception handler functions, Blackfin
3796 Use this attribute on the Blackfin to indicate that the specified function
3797 is an exception handler. The compiler generates function entry and
3798 exit sequences suitable for use in an exception handler when this
3799 attribute is present.
3801 @item interrupt_handler
3802 @cindex @code{interrupt_handler} function attribute, Blackfin
3803 Use this attribute to
3804 indicate that the specified function is an interrupt handler. The compiler
3805 generates function entry and exit sequences suitable for use in an
3806 interrupt handler when this attribute is present.
3809 @cindex @code{kspisusp} function attribute, Blackfin
3810 @cindex User stack pointer in interrupts on the Blackfin
3811 When used together with @code{interrupt_handler}, @code{exception_handler}
3812 or @code{nmi_handler}, code is generated to load the stack pointer
3813 from the USP register in the function prologue.
3816 @cindex @code{l1_text} function attribute, Blackfin
3817 This attribute specifies a function to be placed into L1 Instruction
3818 SRAM@. The function is put into a specific section named @code{.l1.text}.
3819 With @option{-mfdpic}, function calls with a such function as the callee
3820 or caller uses inlined PLT.
3823 @cindex @code{l2} function attribute, Blackfin
3824 This attribute specifies a function to be placed into L2
3825 SRAM. The function is put into a specific section named
3826 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
3831 @cindex indirect calls, Blackfin
3832 @cindex @code{longcall} function attribute, Blackfin
3833 @cindex @code{shortcall} function attribute, Blackfin
3834 The @code{longcall} attribute
3835 indicates that the function might be far away from the call site and
3836 require a different (more expensive) calling sequence. The
3837 @code{shortcall} attribute indicates that the function is always close
3838 enough for the shorter calling sequence to be used. These attributes
3839 override the @option{-mlongcall} switch.
3842 @cindex @code{nesting} function attribute, Blackfin
3843 @cindex Allow nesting in an interrupt handler on the Blackfin processor
3844 Use this attribute together with @code{interrupt_handler},
3845 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3846 entry code should enable nested interrupts or exceptions.
3849 @cindex @code{nmi_handler} function attribute, Blackfin
3850 @cindex NMI handler functions on the Blackfin processor
3851 Use this attribute on the Blackfin to indicate that the specified function
3852 is an NMI handler. The compiler generates function entry and
3853 exit sequences suitable for use in an NMI handler when this
3854 attribute is present.
3857 @cindex @code{saveall} function attribute, Blackfin
3858 @cindex save all registers on the Blackfin
3859 Use this attribute to indicate that
3860 all registers except the stack pointer should be saved in the prologue
3861 regardless of whether they are used or not.
3864 @node CR16 Function Attributes
3865 @subsection CR16 Function Attributes
3867 These function attributes are supported by the CR16 back end:
3871 @cindex @code{interrupt} function attribute, CR16
3872 Use this attribute to indicate
3873 that the specified function is an interrupt handler. The compiler generates
3874 function entry and exit sequences suitable for use in an interrupt handler
3875 when this attribute is present.
3878 @node Epiphany Function Attributes
3879 @subsection Epiphany Function Attributes
3881 These function attributes are supported by the Epiphany back end:
3885 @cindex @code{disinterrupt} function attribute, Epiphany
3886 This attribute causes the compiler to emit
3887 instructions to disable interrupts for the duration of the given
3890 @item forwarder_section
3891 @cindex @code{forwarder_section} function attribute, Epiphany
3892 This attribute modifies the behavior of an interrupt handler.
3893 The interrupt handler may be in external memory which cannot be
3894 reached by a branch instruction, so generate a local memory trampoline
3895 to transfer control. The single parameter identifies the section where
3896 the trampoline is placed.
3899 @cindex @code{interrupt} function attribute, Epiphany
3900 Use this attribute to indicate
3901 that the specified function is an interrupt handler. The compiler generates
3902 function entry and exit sequences suitable for use in an interrupt handler
3903 when this attribute is present. It may also generate
3904 a special section with code to initialize the interrupt vector table.
3906 On Epiphany targets one or more optional parameters can be added like this:
3909 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
3912 Permissible values for these parameters are: @w{@code{reset}},
3913 @w{@code{software_exception}}, @w{@code{page_miss}},
3914 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
3915 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
3916 Multiple parameters indicate that multiple entries in the interrupt
3917 vector table should be initialized for this function, i.e.@: for each
3918 parameter @w{@var{name}}, a jump to the function is emitted in
3919 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
3920 entirely, in which case no interrupt vector table entry is provided.
3922 Note that interrupts are enabled inside the function
3923 unless the @code{disinterrupt} attribute is also specified.
3925 The following examples are all valid uses of these attributes on
3928 void __attribute__ ((interrupt)) universal_handler ();
3929 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
3930 void __attribute__ ((interrupt ("dma0, dma1")))
3931 universal_dma_handler ();
3932 void __attribute__ ((interrupt ("timer0"), disinterrupt))
3933 fast_timer_handler ();
3934 void __attribute__ ((interrupt ("dma0, dma1"),
3935 forwarder_section ("tramp")))
3936 external_dma_handler ();
3941 @cindex @code{long_call} function attribute, Epiphany
3942 @cindex @code{short_call} function attribute, Epiphany
3943 @cindex indirect calls, Epiphany
3944 These attributes specify how a particular function is called.
3945 These attributes override the
3946 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
3947 command-line switch and @code{#pragma long_calls} settings.
3951 @node H8/300 Function Attributes
3952 @subsection H8/300 Function Attributes
3954 These function attributes are available for H8/300 targets:
3957 @item function_vector
3958 @cindex @code{function_vector} function attribute, H8/300
3959 Use this attribute on the H8/300, H8/300H, and H8S to indicate
3960 that the specified function should be called through the function vector.
3961 Calling a function through the function vector reduces code size; however,
3962 the function vector has a limited size (maximum 128 entries on the H8/300
3963 and 64 entries on the H8/300H and H8S)
3964 and shares space with the interrupt vector.
3966 @item interrupt_handler
3967 @cindex @code{interrupt_handler} function attribute, H8/300
3968 Use this attribute on the H8/300, H8/300H, and H8S to
3969 indicate that the specified function is an interrupt handler. The compiler
3970 generates function entry and exit sequences suitable for use in an
3971 interrupt handler when this attribute is present.
3974 @cindex @code{saveall} function attribute, H8/300
3975 @cindex save all registers on the H8/300, H8/300H, and H8S
3976 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
3977 all registers except the stack pointer should be saved in the prologue
3978 regardless of whether they are used or not.
3981 @node IA-64 Function Attributes
3982 @subsection IA-64 Function Attributes
3984 These function attributes are supported on IA-64 targets:
3987 @item syscall_linkage
3988 @cindex @code{syscall_linkage} function attribute, IA-64
3989 This attribute is used to modify the IA-64 calling convention by marking
3990 all input registers as live at all function exits. This makes it possible
3991 to restart a system call after an interrupt without having to save/restore
3992 the input registers. This also prevents kernel data from leaking into
3996 @cindex @code{version_id} function attribute, IA-64
3997 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
3998 symbol to contain a version string, thus allowing for function level
3999 versioning. HP-UX system header files may use function level versioning
4000 for some system calls.
4003 extern int foo () __attribute__((version_id ("20040821")));
4007 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4010 @node M32C Function Attributes
4011 @subsection M32C Function Attributes
4013 These function attributes are supported by the M32C back end:
4017 @cindex @code{bank_switch} function attribute, M32C
4018 When added to an interrupt handler with the M32C port, causes the
4019 prologue and epilogue to use bank switching to preserve the registers
4020 rather than saving them on the stack.
4022 @item fast_interrupt
4023 @cindex @code{fast_interrupt} function attribute, M32C
4024 Use this attribute on the M32C port to indicate that the specified
4025 function is a fast interrupt handler. This is just like the
4026 @code{interrupt} attribute, except that @code{freit} is used to return
4027 instead of @code{reit}.
4029 @item function_vector
4030 @cindex @code{function_vector} function attribute, M16C/M32C
4031 On M16C/M32C targets, the @code{function_vector} attribute declares a
4032 special page subroutine call function. Use of this attribute reduces
4033 the code size by 2 bytes for each call generated to the
4034 subroutine. The argument to the attribute is the vector number entry
4035 from the special page vector table which contains the 16 low-order
4036 bits of the subroutine's entry address. Each vector table has special
4037 page number (18 to 255) that is used in @code{jsrs} instructions.
4038 Jump addresses of the routines are generated by adding 0x0F0000 (in
4039 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4040 2-byte addresses set in the vector table. Therefore you need to ensure
4041 that all the special page vector routines should get mapped within the
4042 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4045 In the following example 2 bytes are saved for each call to
4046 function @code{foo}.
4049 void foo (void) __attribute__((function_vector(0x18)));
4060 If functions are defined in one file and are called in another file,
4061 then be sure to write this declaration in both files.
4063 This attribute is ignored for R8C target.
4066 @cindex @code{interrupt} function attribute, M32C
4067 Use this attribute to indicate
4068 that the specified function is an interrupt handler. The compiler generates
4069 function entry and exit sequences suitable for use in an interrupt handler
4070 when this attribute is present.
4073 @node M32R/D Function Attributes
4074 @subsection M32R/D Function Attributes
4076 These function attributes are supported by the M32R/D back end:
4080 @cindex @code{interrupt} function attribute, M32R/D
4081 Use this attribute to indicate
4082 that the specified function is an interrupt handler. The compiler generates
4083 function entry and exit sequences suitable for use in an interrupt handler
4084 when this attribute is present.
4086 @item model (@var{model-name})
4087 @cindex @code{model} function attribute, M32R/D
4088 @cindex function addressability on the M32R/D
4090 On the M32R/D, use this attribute to set the addressability of an
4091 object, and of the code generated for a function. The identifier
4092 @var{model-name} is one of @code{small}, @code{medium}, or
4093 @code{large}, representing each of the code models.
4095 Small model objects live in the lower 16MB of memory (so that their
4096 addresses can be loaded with the @code{ld24} instruction), and are
4097 callable with the @code{bl} instruction.
4099 Medium model objects may live anywhere in the 32-bit address space (the
4100 compiler generates @code{seth/add3} instructions to load their addresses),
4101 and are callable with the @code{bl} instruction.
4103 Large model objects may live anywhere in the 32-bit address space (the
4104 compiler generates @code{seth/add3} instructions to load their addresses),
4105 and may not be reachable with the @code{bl} instruction (the compiler
4106 generates the much slower @code{seth/add3/jl} instruction sequence).
4109 @node m68k Function Attributes
4110 @subsection m68k Function Attributes
4112 These function attributes are supported by the m68k back end:
4116 @itemx interrupt_handler
4117 @cindex @code{interrupt} function attribute, m68k
4118 @cindex @code{interrupt_handler} function attribute, m68k
4119 Use this attribute to
4120 indicate that the specified function is an interrupt handler. The compiler
4121 generates function entry and exit sequences suitable for use in an
4122 interrupt handler when this attribute is present. Either name may be used.
4124 @item interrupt_thread
4125 @cindex @code{interrupt_thread} function attribute, fido
4126 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4127 that the specified function is an interrupt handler that is designed
4128 to run as a thread. The compiler omits generate prologue/epilogue
4129 sequences and replaces the return instruction with a @code{sleep}
4130 instruction. This attribute is available only on fido.
4133 @node MCORE Function Attributes
4134 @subsection MCORE Function Attributes
4136 These function attributes are supported by the MCORE back end:
4140 @cindex @code{naked} function attribute, MCORE
4141 This attribute allows the compiler to construct the
4142 requisite function declaration, while allowing the body of the
4143 function to be assembly code. The specified function will not have
4144 prologue/epilogue sequences generated by the compiler. Only basic
4145 @code{asm} statements can safely be included in naked functions
4146 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4147 basic @code{asm} and C code may appear to work, they cannot be
4148 depended upon to work reliably and are not supported.
4151 @node MeP Function Attributes
4152 @subsection MeP Function Attributes
4154 These function attributes are supported by the MeP back end:
4158 @cindex @code{disinterrupt} function attribute, MeP
4159 On MeP targets, this attribute causes the compiler to emit
4160 instructions to disable interrupts for the duration of the given
4164 @cindex @code{interrupt} function attribute, MeP
4165 Use this attribute to indicate
4166 that the specified function is an interrupt handler. The compiler generates
4167 function entry and exit sequences suitable for use in an interrupt handler
4168 when this attribute is present.
4171 @cindex @code{near} function attribute, MeP
4172 This attribute causes the compiler to assume the called
4173 function is close enough to use the normal calling convention,
4174 overriding the @option{-mtf} command-line option.
4177 @cindex @code{far} function attribute, MeP
4178 On MeP targets this causes the compiler to use a calling convention
4179 that assumes the called function is too far away for the built-in
4183 @cindex @code{vliw} function attribute, MeP
4184 The @code{vliw} attribute tells the compiler to emit
4185 instructions in VLIW mode instead of core mode. Note that this
4186 attribute is not allowed unless a VLIW coprocessor has been configured
4187 and enabled through command-line options.
4190 @node MicroBlaze Function Attributes
4191 @subsection MicroBlaze Function Attributes
4193 These function attributes are supported on MicroBlaze targets:
4196 @item save_volatiles
4197 @cindex @code{save_volatiles} function attribute, MicroBlaze
4198 Use this attribute to indicate that the function is
4199 an interrupt handler. All volatile registers (in addition to non-volatile
4200 registers) are saved in the function prologue. If the function is a leaf
4201 function, only volatiles used by the function are saved. A normal function
4202 return is generated instead of a return from interrupt.
4205 @cindex @code{break_handler} function attribute, MicroBlaze
4206 @cindex break handler functions
4207 Use this attribute to indicate that
4208 the specified function is a break handler. The compiler generates function
4209 entry and exit sequences suitable for use in an break handler when this
4210 attribute is present. The return from @code{break_handler} is done through
4211 the @code{rtbd} instead of @code{rtsd}.
4214 void f () __attribute__ ((break_handler));
4217 @item interrupt_handler
4218 @itemx fast_interrupt
4219 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4220 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4221 These attributes indicate that the specified function is an interrupt
4222 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4223 used in low-latency interrupt mode, and @code{interrupt_handler} for
4224 interrupts that do not use low-latency handlers. In both cases, GCC
4225 emits appropriate prologue code and generates a return from the handler
4226 using @code{rtid} instead of @code{rtsd}.
4229 @node Microsoft Windows Function Attributes
4230 @subsection Microsoft Windows Function Attributes
4232 The following attributes are available on Microsoft Windows and Symbian OS
4237 @cindex @code{dllexport} function attribute
4238 @cindex @code{__declspec(dllexport)}
4239 On Microsoft Windows targets and Symbian OS targets the
4240 @code{dllexport} attribute causes the compiler to provide a global
4241 pointer to a pointer in a DLL, so that it can be referenced with the
4242 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4243 name is formed by combining @code{_imp__} and the function or variable
4246 You can use @code{__declspec(dllexport)} as a synonym for
4247 @code{__attribute__ ((dllexport))} for compatibility with other
4250 On systems that support the @code{visibility} attribute, this
4251 attribute also implies ``default'' visibility. It is an error to
4252 explicitly specify any other visibility.
4254 GCC's default behavior is to emit all inline functions with the
4255 @code{dllexport} attribute. Since this can cause object file-size bloat,
4256 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4257 ignore the attribute for inlined functions unless the
4258 @option{-fkeep-inline-functions} flag is used instead.
4260 The attribute is ignored for undefined symbols.
4262 When applied to C++ classes, the attribute marks defined non-inlined
4263 member functions and static data members as exports. Static consts
4264 initialized in-class are not marked unless they are also defined
4267 For Microsoft Windows targets there are alternative methods for
4268 including the symbol in the DLL's export table such as using a
4269 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4270 the @option{--export-all} linker flag.
4273 @cindex @code{dllimport} function attribute
4274 @cindex @code{__declspec(dllimport)}
4275 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4276 attribute causes the compiler to reference a function or variable via
4277 a global pointer to a pointer that is set up by the DLL exporting the
4278 symbol. The attribute implies @code{extern}. On Microsoft Windows
4279 targets, the pointer name is formed by combining @code{_imp__} and the
4280 function or variable name.
4282 You can use @code{__declspec(dllimport)} as a synonym for
4283 @code{__attribute__ ((dllimport))} for compatibility with other
4286 On systems that support the @code{visibility} attribute, this
4287 attribute also implies ``default'' visibility. It is an error to
4288 explicitly specify any other visibility.
4290 Currently, the attribute is ignored for inlined functions. If the
4291 attribute is applied to a symbol @emph{definition}, an error is reported.
4292 If a symbol previously declared @code{dllimport} is later defined, the
4293 attribute is ignored in subsequent references, and a warning is emitted.
4294 The attribute is also overridden by a subsequent declaration as
4297 When applied to C++ classes, the attribute marks non-inlined
4298 member functions and static data members as imports. However, the
4299 attribute is ignored for virtual methods to allow creation of vtables
4302 On the SH Symbian OS target the @code{dllimport} attribute also has
4303 another affect---it can cause the vtable and run-time type information
4304 for a class to be exported. This happens when the class has a
4305 dllimported constructor or a non-inline, non-pure virtual function
4306 and, for either of those two conditions, the class also has an inline
4307 constructor or destructor and has a key function that is defined in
4308 the current translation unit.
4310 For Microsoft Windows targets the use of the @code{dllimport}
4311 attribute on functions is not necessary, but provides a small
4312 performance benefit by eliminating a thunk in the DLL@. The use of the
4313 @code{dllimport} attribute on imported variables can be avoided by passing the
4314 @option{--enable-auto-import} switch to the GNU linker. As with
4315 functions, using the attribute for a variable eliminates a thunk in
4318 One drawback to using this attribute is that a pointer to a
4319 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4320 address. However, a pointer to a @emph{function} with the
4321 @code{dllimport} attribute can be used as a constant initializer; in
4322 this case, the address of a stub function in the import lib is
4323 referenced. On Microsoft Windows targets, the attribute can be disabled
4324 for functions by setting the @option{-mnop-fun-dllimport} flag.
4327 @node MIPS Function Attributes
4328 @subsection MIPS Function Attributes
4330 These function attributes are supported by the MIPS back end:
4334 @cindex @code{interrupt} function attribute, MIPS
4335 Use this attribute to indicate that the specified function is an interrupt
4336 handler. The compiler generates function entry and exit sequences suitable
4337 for use in an interrupt handler when this attribute is present.
4338 An optional argument is supported for the interrupt attribute which allows
4339 the interrupt mode to be described. By default GCC assumes the external
4340 interrupt controller (EIC) mode is in use, this can be explicitly set using
4341 @code{eic}. When interrupts are non-masked then the requested Interrupt
4342 Priority Level (IPL) is copied to the current IPL which has the effect of only
4343 enabling higher priority interrupts. To use vectored interrupt mode use
4344 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4345 the behavior of the non-masked interrupt support and GCC will arrange to mask
4346 all interrupts from sw0 up to and including the specified interrupt vector.
4348 You can use the following attributes to modify the behavior
4349 of an interrupt handler:
4351 @item use_shadow_register_set
4352 @cindex @code{use_shadow_register_set} function attribute, MIPS
4353 Assume that the handler uses a shadow register set, instead of
4354 the main general-purpose registers. An optional argument @code{intstack} is
4355 supported to indicate that the shadow register set contains a valid stack
4358 @item keep_interrupts_masked
4359 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4360 Keep interrupts masked for the whole function. Without this attribute,
4361 GCC tries to reenable interrupts for as much of the function as it can.
4363 @item use_debug_exception_return
4364 @cindex @code{use_debug_exception_return} function attribute, MIPS
4365 Return using the @code{deret} instruction. Interrupt handlers that don't
4366 have this attribute return using @code{eret} instead.
4369 You can use any combination of these attributes, as shown below:
4371 void __attribute__ ((interrupt)) v0 ();
4372 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4373 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4374 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4375 void __attribute__ ((interrupt, use_shadow_register_set,
4376 keep_interrupts_masked)) v4 ();
4377 void __attribute__ ((interrupt, use_shadow_register_set,
4378 use_debug_exception_return)) v5 ();
4379 void __attribute__ ((interrupt, keep_interrupts_masked,
4380 use_debug_exception_return)) v6 ();
4381 void __attribute__ ((interrupt, use_shadow_register_set,
4382 keep_interrupts_masked,
4383 use_debug_exception_return)) v7 ();
4384 void __attribute__ ((interrupt("eic"))) v8 ();
4385 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4391 @cindex indirect calls, MIPS
4392 @cindex @code{long_call} function attribute, MIPS
4393 @cindex @code{near} function attribute, MIPS
4394 @cindex @code{far} function attribute, MIPS
4395 These attributes specify how a particular function is called on MIPS@.
4396 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4397 command-line switch. The @code{long_call} and @code{far} attributes are
4398 synonyms, and cause the compiler to always call
4399 the function by first loading its address into a register, and then using
4400 the contents of that register. The @code{near} attribute has the opposite
4401 effect; it specifies that non-PIC calls should be made using the more
4402 efficient @code{jal} instruction.
4406 @cindex @code{mips16} function attribute, MIPS
4407 @cindex @code{nomips16} function attribute, MIPS
4409 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4410 function attributes to locally select or turn off MIPS16 code generation.
4411 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4412 while MIPS16 code generation is disabled for functions with the
4413 @code{nomips16} attribute. These attributes override the
4414 @option{-mips16} and @option{-mno-mips16} options on the command line
4415 (@pxref{MIPS Options}).
4417 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4418 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4419 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4420 may interact badly with some GCC extensions such as @code{__builtin_apply}
4421 (@pxref{Constructing Calls}).
4423 @item micromips, MIPS
4424 @itemx nomicromips, MIPS
4425 @cindex @code{micromips} function attribute
4426 @cindex @code{nomicromips} function attribute
4428 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4429 function attributes to locally select or turn off microMIPS code generation.
4430 A function with the @code{micromips} attribute is emitted as microMIPS code,
4431 while microMIPS code generation is disabled for functions with the
4432 @code{nomicromips} attribute. These attributes override the
4433 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4434 (@pxref{MIPS Options}).
4436 When compiling files containing mixed microMIPS and non-microMIPS code, the
4437 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4439 not that within individual functions. Mixed microMIPS and non-microMIPS code
4440 may interact badly with some GCC extensions such as @code{__builtin_apply}
4441 (@pxref{Constructing Calls}).
4444 @cindex @code{nocompression} function attribute, MIPS
4445 On MIPS targets, you can use the @code{nocompression} function attribute
4446 to locally turn off MIPS16 and microMIPS code generation. This attribute
4447 overrides the @option{-mips16} and @option{-mmicromips} options on the
4448 command line (@pxref{MIPS Options}).
4451 @node MSP430 Function Attributes
4452 @subsection MSP430 Function Attributes
4454 These function attributes are supported by the MSP430 back end:
4458 @cindex @code{critical} function attribute, MSP430
4459 Critical functions disable interrupts upon entry and restore the
4460 previous interrupt state upon exit. Critical functions cannot also
4461 have the @code{naked} or @code{reentrant} attributes. They can have
4462 the @code{interrupt} attribute.
4465 @cindex @code{interrupt} function attribute, MSP430
4466 Use this attribute to indicate
4467 that the specified function is an interrupt handler. The compiler generates
4468 function entry and exit sequences suitable for use in an interrupt handler
4469 when this attribute is present.
4471 You can provide an argument to the interrupt
4472 attribute which specifies a name or number. If the argument is a
4473 number it indicates the slot in the interrupt vector table (0 - 31) to
4474 which this handler should be assigned. If the argument is a name it
4475 is treated as a symbolic name for the vector slot. These names should
4476 match up with appropriate entries in the linker script. By default
4477 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4478 @code{reset} for vector 31 are recognized.
4481 @cindex @code{naked} function attribute, MSP430
4482 This attribute allows the compiler to construct the
4483 requisite function declaration, while allowing the body of the
4484 function to be assembly code. The specified function will not have
4485 prologue/epilogue sequences generated by the compiler. Only basic
4486 @code{asm} statements can safely be included in naked functions
4487 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4488 basic @code{asm} and C code may appear to work, they cannot be
4489 depended upon to work reliably and are not supported.
4492 @cindex @code{reentrant} function attribute, MSP430
4493 Reentrant functions disable interrupts upon entry and enable them
4494 upon exit. Reentrant functions cannot also have the @code{naked}
4495 or @code{critical} attributes. They can have the @code{interrupt}
4499 @cindex @code{wakeup} function attribute, MSP430
4500 This attribute only applies to interrupt functions. It is silently
4501 ignored if applied to a non-interrupt function. A wakeup interrupt
4502 function will rouse the processor from any low-power state that it
4503 might be in when the function exits.
4508 @cindex @code{lower} function attribute, MSP430
4509 @cindex @code{upper} function attribute, MSP430
4510 @cindex @code{either} function attribute, MSP430
4511 On the MSP430 target these attributes can be used to specify whether
4512 the function or variable should be placed into low memory, high
4513 memory, or the placement should be left to the linker to decide. The
4514 attributes are only significant if compiling for the MSP430X
4517 The attributes work in conjunction with a linker script that has been
4518 augmented to specify where to place sections with a @code{.lower} and
4519 a @code{.upper} prefix. So, for example, as well as placing the
4520 @code{.data} section, the script also specifies the placement of a
4521 @code{.lower.data} and a @code{.upper.data} section. The intention
4522 is that @code{lower} sections are placed into a small but easier to
4523 access memory region and the upper sections are placed into a larger, but
4524 slower to access, region.
4526 The @code{either} attribute is special. It tells the linker to place
4527 the object into the corresponding @code{lower} section if there is
4528 room for it. If there is insufficient room then the object is placed
4529 into the corresponding @code{upper} section instead. Note that the
4530 placement algorithm is not very sophisticated. It does not attempt to
4531 find an optimal packing of the @code{lower} sections. It just makes
4532 one pass over the objects and does the best that it can. Using the
4533 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4534 options can help the packing, however, since they produce smaller,
4535 easier to pack regions.
4538 @node NDS32 Function Attributes
4539 @subsection NDS32 Function Attributes
4541 These function attributes are supported by the NDS32 back end:
4545 @cindex @code{exception} function attribute
4546 @cindex exception handler functions, NDS32
4547 Use this attribute on the NDS32 target to indicate that the specified function
4548 is an exception handler. The compiler will generate corresponding sections
4549 for use in an exception handler.
4552 @cindex @code{interrupt} function attribute, NDS32
4553 On NDS32 target, this attribute indicates that the specified function
4554 is an interrupt handler. The compiler generates corresponding sections
4555 for use in an interrupt handler. You can use the following attributes
4556 to modify the behavior:
4559 @cindex @code{nested} function attribute, NDS32
4560 This interrupt service routine is interruptible.
4562 @cindex @code{not_nested} function attribute, NDS32
4563 This interrupt service routine is not interruptible.
4565 @cindex @code{nested_ready} function attribute, NDS32
4566 This interrupt service routine is interruptible after @code{PSW.GIE}
4567 (global interrupt enable) is set. This allows interrupt service routine to
4568 finish some short critical code before enabling interrupts.
4570 @cindex @code{save_all} function attribute, NDS32
4571 The system will help save all registers into stack before entering
4574 @cindex @code{partial_save} function attribute, NDS32
4575 The system will help save caller registers into stack before entering
4580 @cindex @code{naked} function attribute, NDS32
4581 This attribute allows the compiler to construct the
4582 requisite function declaration, while allowing the body of the
4583 function to be assembly code. The specified function will not have
4584 prologue/epilogue sequences generated by the compiler. Only basic
4585 @code{asm} statements can safely be included in naked functions
4586 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4587 basic @code{asm} and C code may appear to work, they cannot be
4588 depended upon to work reliably and are not supported.
4591 @cindex @code{reset} function attribute, NDS32
4592 @cindex reset handler functions
4593 Use this attribute on the NDS32 target to indicate that the specified function
4594 is a reset handler. The compiler will generate corresponding sections
4595 for use in a reset handler. You can use the following attributes
4596 to provide extra exception handling:
4599 @cindex @code{nmi} function attribute, NDS32
4600 Provide a user-defined function to handle NMI exception.
4602 @cindex @code{warm} function attribute, NDS32
4603 Provide a user-defined function to handle warm reset exception.
4607 @node Nios II Function Attributes
4608 @subsection Nios II Function Attributes
4610 These function attributes are supported by the Nios II back end:
4613 @item target (@var{options})
4614 @cindex @code{target} function attribute
4615 As discussed in @ref{Common Function Attributes}, this attribute
4616 allows specification of target-specific compilation options.
4618 When compiling for Nios II, the following options are allowed:
4621 @item custom-@var{insn}=@var{N}
4622 @itemx no-custom-@var{insn}
4623 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4624 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4625 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4626 custom instruction with encoding @var{N} when generating code that uses
4627 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4628 the custom instruction @var{insn}.
4629 These target attributes correspond to the
4630 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4631 command-line options, and support the same set of @var{insn} keywords.
4632 @xref{Nios II Options}, for more information.
4634 @item custom-fpu-cfg=@var{name}
4635 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4636 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4637 command-line option, to select a predefined set of custom instructions
4639 @xref{Nios II Options}, for more information.
4643 @node Nvidia PTX Function Attributes
4644 @subsection Nvidia PTX Function Attributes
4646 These function attributes are supported by the Nvidia PTX back end:
4650 @cindex @code{kernel} attribute, Nvidia PTX
4651 This attribute indicates that the corresponding function should be compiled
4652 as a kernel function, which can be invoked from the host via the CUDA RT
4654 By default functions are only callable only from other PTX functions.
4656 Kernel functions must have @code{void} return type.
4659 @node PowerPC Function Attributes
4660 @subsection PowerPC Function Attributes
4662 These function attributes are supported by the PowerPC back end:
4667 @cindex indirect calls, PowerPC
4668 @cindex @code{longcall} function attribute, PowerPC
4669 @cindex @code{shortcall} function attribute, PowerPC
4670 The @code{longcall} attribute
4671 indicates that the function might be far away from the call site and
4672 require a different (more expensive) calling sequence. The
4673 @code{shortcall} attribute indicates that the function is always close
4674 enough for the shorter calling sequence to be used. These attributes
4675 override both the @option{-mlongcall} switch and
4676 the @code{#pragma longcall} setting.
4678 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4679 calls are necessary.
4681 @item target (@var{options})
4682 @cindex @code{target} function attribute
4683 As discussed in @ref{Common Function Attributes}, this attribute
4684 allows specification of target-specific compilation options.
4686 On the PowerPC, the following options are allowed:
4691 @cindex @code{target("altivec")} function attribute, PowerPC
4692 Generate code that uses (does not use) AltiVec instructions. In
4693 32-bit code, you cannot enable AltiVec instructions unless
4694 @option{-mabi=altivec} is used on the command line.
4698 @cindex @code{target("cmpb")} function attribute, PowerPC
4699 Generate code that uses (does not use) the compare bytes instruction
4700 implemented on the POWER6 processor and other processors that support
4701 the PowerPC V2.05 architecture.
4705 @cindex @code{target("dlmzb")} function attribute, PowerPC
4706 Generate code that uses (does not use) the string-search @samp{dlmzb}
4707 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4708 generated by default when targeting those processors.
4712 @cindex @code{target("fprnd")} function attribute, PowerPC
4713 Generate code that uses (does not use) the FP round to integer
4714 instructions implemented on the POWER5+ processor and other processors
4715 that support the PowerPC V2.03 architecture.
4719 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4720 Generate code that uses (does not use) the decimal floating-point
4721 instructions implemented on some POWER processors.
4725 @cindex @code{target("isel")} function attribute, PowerPC
4726 Generate code that uses (does not use) ISEL instruction.
4730 @cindex @code{target("mfcrf")} function attribute, PowerPC
4731 Generate code that uses (does not use) the move from condition
4732 register field instruction implemented on the POWER4 processor and
4733 other processors that support the PowerPC V2.01 architecture.
4737 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4738 Generate code that uses (does not use) the FP move to/from general
4739 purpose register instructions implemented on the POWER6X processor and
4740 other processors that support the extended PowerPC V2.05 architecture.
4744 @cindex @code{target("mulhw")} function attribute, PowerPC
4745 Generate code that uses (does not use) the half-word multiply and
4746 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
4747 These instructions are generated by default when targeting those
4752 @cindex @code{target("multiple")} function attribute, PowerPC
4753 Generate code that uses (does not use) the load multiple word
4754 instructions and the store multiple word instructions.
4758 @cindex @code{target("update")} function attribute, PowerPC
4759 Generate code that uses (does not use) the load or store instructions
4760 that update the base register to the address of the calculated memory
4765 @cindex @code{target("popcntb")} function attribute, PowerPC
4766 Generate code that uses (does not use) the popcount and double-precision
4767 FP reciprocal estimate instruction implemented on the POWER5
4768 processor and other processors that support the PowerPC V2.02
4773 @cindex @code{target("popcntd")} function attribute, PowerPC
4774 Generate code that uses (does not use) the popcount instruction
4775 implemented on the POWER7 processor and other processors that support
4776 the PowerPC V2.06 architecture.
4778 @item powerpc-gfxopt
4779 @itemx no-powerpc-gfxopt
4780 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
4781 Generate code that uses (does not use) the optional PowerPC
4782 architecture instructions in the Graphics group, including
4783 floating-point select.
4786 @itemx no-powerpc-gpopt
4787 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
4788 Generate code that uses (does not use) the optional PowerPC
4789 architecture instructions in the General Purpose group, including
4790 floating-point square root.
4792 @item recip-precision
4793 @itemx no-recip-precision
4794 @cindex @code{target("recip-precision")} function attribute, PowerPC
4795 Assume (do not assume) that the reciprocal estimate instructions
4796 provide higher-precision estimates than is mandated by the PowerPC
4801 @cindex @code{target("string")} function attribute, PowerPC
4802 Generate code that uses (does not use) the load string instructions
4803 and the store string word instructions to save multiple registers and
4804 do small block moves.
4808 @cindex @code{target("vsx")} function attribute, PowerPC
4809 Generate code that uses (does not use) vector/scalar (VSX)
4810 instructions, and also enable the use of built-in functions that allow
4811 more direct access to the VSX instruction set. In 32-bit code, you
4812 cannot enable VSX or AltiVec instructions unless
4813 @option{-mabi=altivec} is used on the command line.
4817 @cindex @code{target("friz")} function attribute, PowerPC
4818 Generate (do not generate) the @code{friz} instruction when the
4819 @option{-funsafe-math-optimizations} option is used to optimize
4820 rounding a floating-point value to 64-bit integer and back to floating
4821 point. The @code{friz} instruction does not return the same value if
4822 the floating-point number is too large to fit in an integer.
4824 @item avoid-indexed-addresses
4825 @itemx no-avoid-indexed-addresses
4826 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
4827 Generate code that tries to avoid (not avoid) the use of indexed load
4828 or store instructions.
4832 @cindex @code{target("paired")} function attribute, PowerPC
4833 Generate code that uses (does not use) the generation of PAIRED simd
4838 @cindex @code{target("longcall")} function attribute, PowerPC
4839 Generate code that assumes (does not assume) that all calls are far
4840 away so that a longer more expensive calling sequence is required.
4843 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
4844 Specify the architecture to generate code for when compiling the
4845 function. If you select the @code{target("cpu=power7")} attribute when
4846 generating 32-bit code, VSX and AltiVec instructions are not generated
4847 unless you use the @option{-mabi=altivec} option on the command line.
4849 @item tune=@var{TUNE}
4850 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
4851 Specify the architecture to tune for when compiling the function. If
4852 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
4853 you do specify the @code{target("cpu=@var{CPU}")} attribute,
4854 compilation tunes for the @var{CPU} architecture, and not the
4855 default tuning specified on the command line.
4858 On the PowerPC, the inliner does not inline a
4859 function that has different target options than the caller, unless the
4860 callee has a subset of the target options of the caller.
4863 @node RL78 Function Attributes
4864 @subsection RL78 Function Attributes
4866 These function attributes are supported by the RL78 back end:
4870 @itemx brk_interrupt
4871 @cindex @code{interrupt} function attribute, RL78
4872 @cindex @code{brk_interrupt} function attribute, RL78
4873 These attributes indicate
4874 that the specified function is an interrupt handler. The compiler generates
4875 function entry and exit sequences suitable for use in an interrupt handler
4876 when this attribute is present.
4878 Use @code{brk_interrupt} instead of @code{interrupt} for
4879 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
4880 that must end with @code{RETB} instead of @code{RETI}).
4883 @cindex @code{naked} function attribute, RL78
4884 This attribute allows the compiler to construct the
4885 requisite function declaration, while allowing the body of the
4886 function to be assembly code. The specified function will not have
4887 prologue/epilogue sequences generated by the compiler. Only basic
4888 @code{asm} statements can safely be included in naked functions
4889 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4890 basic @code{asm} and C code may appear to work, they cannot be
4891 depended upon to work reliably and are not supported.
4894 @node RX Function Attributes
4895 @subsection RX Function Attributes
4897 These function attributes are supported by the RX back end:
4900 @item fast_interrupt
4901 @cindex @code{fast_interrupt} function attribute, RX
4902 Use this attribute on the RX port to indicate that the specified
4903 function is a fast interrupt handler. This is just like the
4904 @code{interrupt} attribute, except that @code{freit} is used to return
4905 instead of @code{reit}.
4908 @cindex @code{interrupt} function attribute, RX
4909 Use this attribute to indicate
4910 that the specified function is an interrupt handler. The compiler generates
4911 function entry and exit sequences suitable for use in an interrupt handler
4912 when this attribute is present.
4914 On RX targets, you may specify one or more vector numbers as arguments
4915 to the attribute, as well as naming an alternate table name.
4916 Parameters are handled sequentially, so one handler can be assigned to
4917 multiple entries in multiple tables. One may also pass the magic
4918 string @code{"$default"} which causes the function to be used for any
4919 unfilled slots in the current table.
4921 This example shows a simple assignment of a function to one vector in
4922 the default table (note that preprocessor macros may be used for
4923 chip-specific symbolic vector names):
4925 void __attribute__ ((interrupt (5))) txd1_handler ();
4928 This example assigns a function to two slots in the default table
4929 (using preprocessor macros defined elsewhere) and makes it the default
4930 for the @code{dct} table:
4932 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
4937 @cindex @code{naked} function attribute, RX
4938 This attribute allows the compiler to construct the
4939 requisite function declaration, while allowing the body of the
4940 function to be assembly code. The specified function will not have
4941 prologue/epilogue sequences generated by the compiler. Only basic
4942 @code{asm} statements can safely be included in naked functions
4943 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4944 basic @code{asm} and C code may appear to work, they cannot be
4945 depended upon to work reliably and are not supported.
4948 @cindex @code{vector} function attribute, RX
4949 This RX attribute is similar to the @code{interrupt} attribute, including its
4950 parameters, but does not make the function an interrupt-handler type
4951 function (i.e. it retains the normal C function calling ABI). See the
4952 @code{interrupt} attribute for a description of its arguments.
4955 @node S/390 Function Attributes
4956 @subsection S/390 Function Attributes
4958 These function attributes are supported on the S/390:
4961 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
4962 @cindex @code{hotpatch} function attribute, S/390
4964 On S/390 System z targets, you can use this function attribute to
4965 make GCC generate a ``hot-patching'' function prologue. If the
4966 @option{-mhotpatch=} command-line option is used at the same time,
4967 the @code{hotpatch} attribute takes precedence. The first of the
4968 two arguments specifies the number of halfwords to be added before
4969 the function label. A second argument can be used to specify the
4970 number of halfwords to be added after the function label. For
4971 both arguments the maximum allowed value is 1000000.
4973 If both arguments are zero, hotpatching is disabled.
4975 @item target (@var{options})
4976 @cindex @code{target} function attribute
4977 As discussed in @ref{Common Function Attributes}, this attribute
4978 allows specification of target-specific compilation options.
4980 On S/390, the following options are supported:
4988 @item warn-framesize=
5000 @itemx no-packed-stack
5002 @itemx no-small-exec
5005 @item warn-dynamicstack
5006 @itemx no-warn-dynamicstack
5009 The options work exactly like the S/390 specific command line
5010 options (without the prefix @option{-m}) except that they do not
5011 change any feature macros. For example,
5014 @code{target("no-vx")}
5017 does not undefine the @code{__VEC__} macro.
5020 @node SH Function Attributes
5021 @subsection SH Function Attributes
5023 These function attributes are supported on the SH family of processors:
5026 @item function_vector
5027 @cindex @code{function_vector} function attribute, SH
5028 @cindex calling functions through the function vector on SH2A
5029 On SH2A targets, this attribute declares a function to be called using the
5030 TBR relative addressing mode. The argument to this attribute is the entry
5031 number of the same function in a vector table containing all the TBR
5032 relative addressable functions. For correct operation the TBR must be setup
5033 accordingly to point to the start of the vector table before any functions with
5034 this attribute are invoked. Usually a good place to do the initialization is
5035 the startup routine. The TBR relative vector table can have at max 256 function
5036 entries. The jumps to these functions are generated using a SH2A specific,
5037 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5038 from GNU binutils version 2.7 or later for this attribute to work correctly.
5040 In an application, for a function being called once, this attribute
5041 saves at least 8 bytes of code; and if other successive calls are being
5042 made to the same function, it saves 2 bytes of code per each of these
5045 @item interrupt_handler
5046 @cindex @code{interrupt_handler} function attribute, SH
5047 Use this attribute to
5048 indicate that the specified function is an interrupt handler. The compiler
5049 generates function entry and exit sequences suitable for use in an
5050 interrupt handler when this attribute is present.
5052 @item nosave_low_regs
5053 @cindex @code{nosave_low_regs} function attribute, SH
5054 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5055 function should not save and restore registers R0..R7. This can be used on SH3*
5056 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5060 @cindex @code{renesas} function attribute, SH
5061 On SH targets this attribute specifies that the function or struct follows the
5065 @cindex @code{resbank} function attribute, SH
5066 On the SH2A target, this attribute enables the high-speed register
5067 saving and restoration using a register bank for @code{interrupt_handler}
5068 routines. Saving to the bank is performed automatically after the CPU
5069 accepts an interrupt that uses a register bank.
5071 The nineteen 32-bit registers comprising general register R0 to R14,
5072 control register GBR, and system registers MACH, MACL, and PR and the
5073 vector table address offset are saved into a register bank. Register
5074 banks are stacked in first-in last-out (FILO) sequence. Restoration
5075 from the bank is executed by issuing a RESBANK instruction.
5078 @cindex @code{sp_switch} function attribute, SH
5079 Use this attribute on the SH to indicate an @code{interrupt_handler}
5080 function should switch to an alternate stack. It expects a string
5081 argument that names a global variable holding the address of the
5086 void f () __attribute__ ((interrupt_handler,
5087 sp_switch ("alt_stack")));
5091 @cindex @code{trap_exit} function attribute, SH
5092 Use this attribute on the SH for an @code{interrupt_handler} to return using
5093 @code{trapa} instead of @code{rte}. This attribute expects an integer
5094 argument specifying the trap number to be used.
5097 @cindex @code{trapa_handler} function attribute, SH
5098 On SH targets this function attribute is similar to @code{interrupt_handler}
5099 but it does not save and restore all registers.
5102 @node SPU Function Attributes
5103 @subsection SPU Function Attributes
5105 These function attributes are supported by the SPU back end:
5109 @cindex @code{naked} function attribute, SPU
5110 This attribute allows the compiler to construct the
5111 requisite function declaration, while allowing the body of the
5112 function to be assembly code. The specified function will not have
5113 prologue/epilogue sequences generated by the compiler. Only basic
5114 @code{asm} statements can safely be included in naked functions
5115 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5116 basic @code{asm} and C code may appear to work, they cannot be
5117 depended upon to work reliably and are not supported.
5120 @node Symbian OS Function Attributes
5121 @subsection Symbian OS Function Attributes
5123 @xref{Microsoft Windows Function Attributes}, for discussion of the
5124 @code{dllexport} and @code{dllimport} attributes.
5126 @node V850 Function Attributes
5127 @subsection V850 Function Attributes
5129 The V850 back end supports these function attributes:
5133 @itemx interrupt_handler
5134 @cindex @code{interrupt} function attribute, V850
5135 @cindex @code{interrupt_handler} function attribute, V850
5136 Use these attributes to indicate
5137 that the specified function is an interrupt handler. The compiler generates
5138 function entry and exit sequences suitable for use in an interrupt handler
5139 when either attribute is present.
5142 @node Visium Function Attributes
5143 @subsection Visium Function Attributes
5145 These function attributes are supported by the Visium back end:
5149 @cindex @code{interrupt} function attribute, Visium
5150 Use this attribute to indicate
5151 that the specified function is an interrupt handler. The compiler generates
5152 function entry and exit sequences suitable for use in an interrupt handler
5153 when this attribute is present.
5156 @node x86 Function Attributes
5157 @subsection x86 Function Attributes
5159 These function attributes are supported by the x86 back end:
5163 @cindex @code{cdecl} function attribute, x86-32
5164 @cindex functions that pop the argument stack on x86-32
5166 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5167 assume that the calling function pops off the stack space used to
5168 pass arguments. This is
5169 useful to override the effects of the @option{-mrtd} switch.
5172 @cindex @code{fastcall} function attribute, x86-32
5173 @cindex functions that pop the argument stack on x86-32
5174 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5175 pass the first argument (if of integral type) in the register ECX and
5176 the second argument (if of integral type) in the register EDX@. Subsequent
5177 and other typed arguments are passed on the stack. The called function
5178 pops the arguments off the stack. If the number of arguments is variable all
5179 arguments are pushed on the stack.
5182 @cindex @code{thiscall} function attribute, x86-32
5183 @cindex functions that pop the argument stack on x86-32
5184 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5185 pass the first argument (if of integral type) in the register ECX.
5186 Subsequent and other typed arguments are passed on the stack. The called
5187 function pops the arguments off the stack.
5188 If the number of arguments is variable all arguments are pushed on the
5190 The @code{thiscall} attribute is intended for C++ non-static member functions.
5191 As a GCC extension, this calling convention can be used for C functions
5192 and for static member methods.
5196 @cindex @code{ms_abi} function attribute, x86
5197 @cindex @code{sysv_abi} function attribute, x86
5199 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5200 to indicate which calling convention should be used for a function. The
5201 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5202 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5203 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5204 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5206 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5207 requires the @option{-maccumulate-outgoing-args} option.
5209 @item callee_pop_aggregate_return (@var{number})
5210 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5212 On x86-32 targets, you can use this attribute to control how
5213 aggregates are returned in memory. If the caller is responsible for
5214 popping the hidden pointer together with the rest of the arguments, specify
5215 @var{number} equal to zero. If callee is responsible for popping the
5216 hidden pointer, specify @var{number} equal to one.
5218 The default x86-32 ABI assumes that the callee pops the
5219 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5220 the compiler assumes that the
5221 caller pops the stack for hidden pointer.
5223 @item ms_hook_prologue
5224 @cindex @code{ms_hook_prologue} function attribute, x86
5226 On 32-bit and 64-bit x86 targets, you can use
5227 this function attribute to make GCC generate the ``hot-patching'' function
5228 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5231 @item regparm (@var{number})
5232 @cindex @code{regparm} function attribute, x86
5233 @cindex functions that are passed arguments in registers on x86-32
5234 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5235 pass arguments number one to @var{number} if they are of integral type
5236 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5237 take a variable number of arguments continue to be passed all of their
5238 arguments on the stack.
5240 Beware that on some ELF systems this attribute is unsuitable for
5241 global functions in shared libraries with lazy binding (which is the
5242 default). Lazy binding sends the first call via resolving code in
5243 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5244 per the standard calling conventions. Solaris 8 is affected by this.
5245 Systems with the GNU C Library version 2.1 or higher
5246 and FreeBSD are believed to be
5247 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5248 disabled with the linker or the loader if desired, to avoid the
5252 @cindex @code{sseregparm} function attribute, x86
5253 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5254 causes the compiler to pass up to 3 floating-point arguments in
5255 SSE registers instead of on the stack. Functions that take a
5256 variable number of arguments continue to pass all of their
5257 floating-point arguments on the stack.
5259 @item force_align_arg_pointer
5260 @cindex @code{force_align_arg_pointer} function attribute, x86
5261 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5262 applied to individual function definitions, generating an alternate
5263 prologue and epilogue that realigns the run-time stack if necessary.
5264 This supports mixing legacy codes that run with a 4-byte aligned stack
5265 with modern codes that keep a 16-byte stack for SSE compatibility.
5268 @cindex @code{stdcall} function attribute, x86-32
5269 @cindex functions that pop the argument stack on x86-32
5270 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5271 assume that the called function pops off the stack space used to
5272 pass arguments, unless it takes a variable number of arguments.
5274 @item no_caller_saved_registers
5275 @cindex @code{no_caller_saved_registers} function attribute, x86
5276 Use this attribute to indicate that the specified function has no
5277 caller-saved registers. That is, all registers are callee-saved. For
5278 example, this attribute can be used for a function called from an
5279 interrupt handler. The compiler generates proper function entry and
5280 exit sequences to save and restore any modified registers, except for
5281 the EFLAGS register. Since GCC doesn't preserve MPX, SSE, MMX nor x87
5282 states, the GCC option @option{-mgeneral-regs-only} should be used to
5283 compile functions with @code{no_caller_saved_registers} attribute.
5286 @cindex @code{interrupt} function attribute, x86
5287 Use this attribute to indicate that the specified function is an
5288 interrupt handler or an exception handler (depending on parameters passed
5289 to the function, explained further). The compiler generates function
5290 entry and exit sequences suitable for use in an interrupt handler when
5291 this attribute is present. The @code{IRET} instruction, instead of the
5292 @code{RET} instruction, is used to return from interrupt handlers. All
5293 registers, except for the EFLAGS register which is restored by the
5294 @code{IRET} instruction, are preserved by the compiler. Since GCC
5295 doesn't preserve MPX, SSE, MMX nor x87 states, the GCC option
5296 @option{-mgeneral-regs-only} should be used to compile interrupt and
5299 Any interruptible-without-stack-switch code must be compiled with
5300 @option{-mno-red-zone} since interrupt handlers can and will, because
5301 of the hardware design, touch the red zone.
5303 An interrupt handler must be declared with a mandatory pointer
5307 struct interrupt_frame;
5309 __attribute__ ((interrupt))
5311 f (struct interrupt_frame *frame)
5317 and you must define @code{struct interrupt_frame} as described in the
5320 Exception handlers differ from interrupt handlers because the system
5321 pushes an error code on the stack. An exception handler declaration is
5322 similar to that for an interrupt handler, but with a different mandatory
5323 function signature. The compiler arranges to pop the error code off the
5324 stack before the @code{IRET} instruction.
5328 typedef unsigned long long int uword_t;
5330 typedef unsigned int uword_t;
5333 struct interrupt_frame;
5335 __attribute__ ((interrupt))
5337 f (struct interrupt_frame *frame, uword_t error_code)
5343 Exception handlers should only be used for exceptions that push an error
5344 code; you should use an interrupt handler in other cases. The system
5345 will crash if the wrong kind of handler is used.
5347 @item target (@var{options})
5348 @cindex @code{target} function attribute
5349 As discussed in @ref{Common Function Attributes}, this attribute
5350 allows specification of target-specific compilation options.
5352 On the x86, the following options are allowed:
5356 @cindex @code{target("abm")} function attribute, x86
5357 Enable/disable the generation of the advanced bit instructions.
5361 @cindex @code{target("aes")} function attribute, x86
5362 Enable/disable the generation of the AES instructions.
5365 @cindex @code{target("default")} function attribute, x86
5366 @xref{Function Multiversioning}, where it is used to specify the
5367 default function version.
5371 @cindex @code{target("mmx")} function attribute, x86
5372 Enable/disable the generation of the MMX instructions.
5376 @cindex @code{target("pclmul")} function attribute, x86
5377 Enable/disable the generation of the PCLMUL instructions.
5381 @cindex @code{target("popcnt")} function attribute, x86
5382 Enable/disable the generation of the POPCNT instruction.
5386 @cindex @code{target("sse")} function attribute, x86
5387 Enable/disable the generation of the SSE instructions.
5391 @cindex @code{target("sse2")} function attribute, x86
5392 Enable/disable the generation of the SSE2 instructions.
5396 @cindex @code{target("sse3")} function attribute, x86
5397 Enable/disable the generation of the SSE3 instructions.
5401 @cindex @code{target("sse4")} function attribute, x86
5402 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5407 @cindex @code{target("sse4.1")} function attribute, x86
5408 Enable/disable the generation of the sse4.1 instructions.
5412 @cindex @code{target("sse4.2")} function attribute, x86
5413 Enable/disable the generation of the sse4.2 instructions.
5417 @cindex @code{target("sse4a")} function attribute, x86
5418 Enable/disable the generation of the SSE4A instructions.
5422 @cindex @code{target("fma4")} function attribute, x86
5423 Enable/disable the generation of the FMA4 instructions.
5427 @cindex @code{target("xop")} function attribute, x86
5428 Enable/disable the generation of the XOP instructions.
5432 @cindex @code{target("lwp")} function attribute, x86
5433 Enable/disable the generation of the LWP instructions.
5437 @cindex @code{target("ssse3")} function attribute, x86
5438 Enable/disable the generation of the SSSE3 instructions.
5442 @cindex @code{target("cld")} function attribute, x86
5443 Enable/disable the generation of the CLD before string moves.
5445 @item fancy-math-387
5446 @itemx no-fancy-math-387
5447 @cindex @code{target("fancy-math-387")} function attribute, x86
5448 Enable/disable the generation of the @code{sin}, @code{cos}, and
5449 @code{sqrt} instructions on the 387 floating-point unit.
5452 @itemx no-fused-madd
5453 @cindex @code{target("fused-madd")} function attribute, x86
5454 Enable/disable the generation of the fused multiply/add instructions.
5458 @cindex @code{target("ieee-fp")} function attribute, x86
5459 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5461 @item inline-all-stringops
5462 @itemx no-inline-all-stringops
5463 @cindex @code{target("inline-all-stringops")} function attribute, x86
5464 Enable/disable inlining of string operations.
5466 @item inline-stringops-dynamically
5467 @itemx no-inline-stringops-dynamically
5468 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5469 Enable/disable the generation of the inline code to do small string
5470 operations and calling the library routines for large operations.
5472 @item align-stringops
5473 @itemx no-align-stringops
5474 @cindex @code{target("align-stringops")} function attribute, x86
5475 Do/do not align destination of inlined string operations.
5479 @cindex @code{target("recip")} function attribute, x86
5480 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5481 instructions followed an additional Newton-Raphson step instead of
5482 doing a floating-point division.
5484 @item arch=@var{ARCH}
5485 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5486 Specify the architecture to generate code for in compiling the function.
5488 @item tune=@var{TUNE}
5489 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5490 Specify the architecture to tune for in compiling the function.
5492 @item fpmath=@var{FPMATH}
5493 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5494 Specify which floating-point unit to use. You must specify the
5495 @code{target("fpmath=sse,387")} option as
5496 @code{target("fpmath=sse+387")} because the comma would separate
5500 On the x86, the inliner does not inline a
5501 function that has different target options than the caller, unless the
5502 callee has a subset of the target options of the caller. For example
5503 a function declared with @code{target("sse3")} can inline a function
5504 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5507 @node Xstormy16 Function Attributes
5508 @subsection Xstormy16 Function Attributes
5510 These function attributes are supported by the Xstormy16 back end:
5514 @cindex @code{interrupt} function attribute, Xstormy16
5515 Use this attribute to indicate
5516 that the specified function is an interrupt handler. The compiler generates
5517 function entry and exit sequences suitable for use in an interrupt handler
5518 when this attribute is present.
5521 @node Variable Attributes
5522 @section Specifying Attributes of Variables
5523 @cindex attribute of variables
5524 @cindex variable attributes
5526 The keyword @code{__attribute__} allows you to specify special
5527 attributes of variables or structure fields. This keyword is followed
5528 by an attribute specification inside double parentheses. Some
5529 attributes are currently defined generically for variables.
5530 Other attributes are defined for variables on particular target
5531 systems. Other attributes are available for functions
5532 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5533 enumerators (@pxref{Enumerator Attributes}), and for types
5534 (@pxref{Type Attributes}).
5535 Other front ends might define more attributes
5536 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5538 @xref{Attribute Syntax}, for details of the exact syntax for using
5542 * Common Variable Attributes::
5543 * AVR Variable Attributes::
5544 * Blackfin Variable Attributes::
5545 * H8/300 Variable Attributes::
5546 * IA-64 Variable Attributes::
5547 * M32R/D Variable Attributes::
5548 * MeP Variable Attributes::
5549 * Microsoft Windows Variable Attributes::
5550 * MSP430 Variable Attributes::
5551 * PowerPC Variable Attributes::
5552 * RL78 Variable Attributes::
5553 * SPU Variable Attributes::
5554 * V850 Variable Attributes::
5555 * x86 Variable Attributes::
5556 * Xstormy16 Variable Attributes::
5559 @node Common Variable Attributes
5560 @subsection Common Variable Attributes
5562 The following attributes are supported on most targets.
5565 @cindex @code{aligned} variable attribute
5566 @item aligned (@var{alignment})
5567 This attribute specifies a minimum alignment for the variable or
5568 structure field, measured in bytes. For example, the declaration:
5571 int x __attribute__ ((aligned (16))) = 0;
5575 causes the compiler to allocate the global variable @code{x} on a
5576 16-byte boundary. On a 68040, this could be used in conjunction with
5577 an @code{asm} expression to access the @code{move16} instruction which
5578 requires 16-byte aligned operands.
5580 You can also specify the alignment of structure fields. For example, to
5581 create a double-word aligned @code{int} pair, you could write:
5584 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5588 This is an alternative to creating a union with a @code{double} member,
5589 which forces the union to be double-word aligned.
5591 As in the preceding examples, you can explicitly specify the alignment
5592 (in bytes) that you wish the compiler to use for a given variable or
5593 structure field. Alternatively, you can leave out the alignment factor
5594 and just ask the compiler to align a variable or field to the
5595 default alignment for the target architecture you are compiling for.
5596 The default alignment is sufficient for all scalar types, but may not be
5597 enough for all vector types on a target that supports vector operations.
5598 The default alignment is fixed for a particular target ABI.
5600 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5601 which is the largest alignment ever used for any data type on the
5602 target machine you are compiling for. For example, you could write:
5605 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5608 The compiler automatically sets the alignment for the declared
5609 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5610 often make copy operations more efficient, because the compiler can
5611 use whatever instructions copy the biggest chunks of memory when
5612 performing copies to or from the variables or fields that you have
5613 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5614 may change depending on command-line options.
5616 When used on a struct, or struct member, the @code{aligned} attribute can
5617 only increase the alignment; in order to decrease it, the @code{packed}
5618 attribute must be specified as well. When used as part of a typedef, the
5619 @code{aligned} attribute can both increase and decrease alignment, and
5620 specifying the @code{packed} attribute generates a warning.
5622 Note that the effectiveness of @code{aligned} attributes may be limited
5623 by inherent limitations in your linker. On many systems, the linker is
5624 only able to arrange for variables to be aligned up to a certain maximum
5625 alignment. (For some linkers, the maximum supported alignment may
5626 be very very small.) If your linker is only able to align variables
5627 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5628 in an @code{__attribute__} still only provides you with 8-byte
5629 alignment. See your linker documentation for further information.
5631 The @code{aligned} attribute can also be used for functions
5632 (@pxref{Common Function Attributes}.)
5634 @item cleanup (@var{cleanup_function})
5635 @cindex @code{cleanup} variable attribute
5636 The @code{cleanup} attribute runs a function when the variable goes
5637 out of scope. This attribute can only be applied to auto function
5638 scope variables; it may not be applied to parameters or variables
5639 with static storage duration. The function must take one parameter,
5640 a pointer to a type compatible with the variable. The return value
5641 of the function (if any) is ignored.
5643 If @option{-fexceptions} is enabled, then @var{cleanup_function}
5644 is run during the stack unwinding that happens during the
5645 processing of the exception. Note that the @code{cleanup} attribute
5646 does not allow the exception to be caught, only to perform an action.
5647 It is undefined what happens if @var{cleanup_function} does not
5652 @cindex @code{common} variable attribute
5653 @cindex @code{nocommon} variable attribute
5656 The @code{common} attribute requests GCC to place a variable in
5657 ``common'' storage. The @code{nocommon} attribute requests the
5658 opposite---to allocate space for it directly.
5660 These attributes override the default chosen by the
5661 @option{-fno-common} and @option{-fcommon} flags respectively.
5664 @itemx deprecated (@var{msg})
5665 @cindex @code{deprecated} variable attribute
5666 The @code{deprecated} attribute results in a warning if the variable
5667 is used anywhere in the source file. This is useful when identifying
5668 variables that are expected to be removed in a future version of a
5669 program. The warning also includes the location of the declaration
5670 of the deprecated variable, to enable users to easily find further
5671 information about why the variable is deprecated, or what they should
5672 do instead. Note that the warning only occurs for uses:
5675 extern int old_var __attribute__ ((deprecated));
5677 int new_fn () @{ return old_var; @}
5681 results in a warning on line 3 but not line 2. The optional @var{msg}
5682 argument, which must be a string, is printed in the warning if
5685 The @code{deprecated} attribute can also be used for functions and
5686 types (@pxref{Common Function Attributes},
5687 @pxref{Common Type Attributes}).
5689 @item mode (@var{mode})
5690 @cindex @code{mode} variable attribute
5691 This attribute specifies the data type for the declaration---whichever
5692 type corresponds to the mode @var{mode}. This in effect lets you
5693 request an integer or floating-point type according to its width.
5695 You may also specify a mode of @code{byte} or @code{__byte__} to
5696 indicate the mode corresponding to a one-byte integer, @code{word} or
5697 @code{__word__} for the mode of a one-word integer, and @code{pointer}
5698 or @code{__pointer__} for the mode used to represent pointers.
5701 @cindex @code{packed} variable attribute
5702 The @code{packed} attribute specifies that a variable or structure field
5703 should have the smallest possible alignment---one byte for a variable,
5704 and one bit for a field, unless you specify a larger value with the
5705 @code{aligned} attribute.
5707 Here is a structure in which the field @code{x} is packed, so that it
5708 immediately follows @code{a}:
5714 int x[2] __attribute__ ((packed));
5718 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
5719 @code{packed} attribute on bit-fields of type @code{char}. This has
5720 been fixed in GCC 4.4 but the change can lead to differences in the
5721 structure layout. See the documentation of
5722 @option{-Wpacked-bitfield-compat} for more information.
5724 @item section ("@var{section-name}")
5725 @cindex @code{section} variable attribute
5726 Normally, the compiler places the objects it generates in sections like
5727 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
5728 or you need certain particular variables to appear in special sections,
5729 for example to map to special hardware. The @code{section}
5730 attribute specifies that a variable (or function) lives in a particular
5731 section. For example, this small program uses several specific section names:
5734 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
5735 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
5736 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
5737 int init_data __attribute__ ((section ("INITDATA")));
5741 /* @r{Initialize stack pointer} */
5742 init_sp (stack + sizeof (stack));
5744 /* @r{Initialize initialized data} */
5745 memcpy (&init_data, &data, &edata - &data);
5747 /* @r{Turn on the serial ports} */
5754 Use the @code{section} attribute with
5755 @emph{global} variables and not @emph{local} variables,
5756 as shown in the example.
5758 You may use the @code{section} attribute with initialized or
5759 uninitialized global variables but the linker requires
5760 each object be defined once, with the exception that uninitialized
5761 variables tentatively go in the @code{common} (or @code{bss}) section
5762 and can be multiply ``defined''. Using the @code{section} attribute
5763 changes what section the variable goes into and may cause the
5764 linker to issue an error if an uninitialized variable has multiple
5765 definitions. You can force a variable to be initialized with the
5766 @option{-fno-common} flag or the @code{nocommon} attribute.
5768 Some file formats do not support arbitrary sections so the @code{section}
5769 attribute is not available on all platforms.
5770 If you need to map the entire contents of a module to a particular
5771 section, consider using the facilities of the linker instead.
5773 @item tls_model ("@var{tls_model}")
5774 @cindex @code{tls_model} variable attribute
5775 The @code{tls_model} attribute sets thread-local storage model
5776 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
5777 overriding @option{-ftls-model=} command-line switch on a per-variable
5779 The @var{tls_model} argument should be one of @code{global-dynamic},
5780 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
5782 Not all targets support this attribute.
5785 @cindex @code{unused} variable attribute
5786 This attribute, attached to a variable, means that the variable is meant
5787 to be possibly unused. GCC does not produce a warning for this
5791 @cindex @code{used} variable attribute
5792 This attribute, attached to a variable with static storage, means that
5793 the variable must be emitted even if it appears that the variable is not
5796 When applied to a static data member of a C++ class template, the
5797 attribute also means that the member is instantiated if the
5798 class itself is instantiated.
5800 @item vector_size (@var{bytes})
5801 @cindex @code{vector_size} variable attribute
5802 This attribute specifies the vector size for the variable, measured in
5803 bytes. For example, the declaration:
5806 int foo __attribute__ ((vector_size (16)));
5810 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
5811 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
5812 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
5814 This attribute is only applicable to integral and float scalars,
5815 although arrays, pointers, and function return values are allowed in
5816 conjunction with this construct.
5818 Aggregates with this attribute are invalid, even if they are of the same
5819 size as a corresponding scalar. For example, the declaration:
5822 struct S @{ int a; @};
5823 struct S __attribute__ ((vector_size (16))) foo;
5827 is invalid even if the size of the structure is the same as the size of
5830 @item visibility ("@var{visibility_type}")
5831 @cindex @code{visibility} variable attribute
5832 This attribute affects the linkage of the declaration to which it is attached.
5833 The @code{visibility} attribute is described in
5834 @ref{Common Function Attributes}.
5837 @cindex @code{weak} variable attribute
5838 The @code{weak} attribute is described in
5839 @ref{Common Function Attributes}.
5843 @node AVR Variable Attributes
5844 @subsection AVR Variable Attributes
5848 @cindex @code{progmem} variable attribute, AVR
5849 The @code{progmem} attribute is used on the AVR to place read-only
5850 data in the non-volatile program memory (flash). The @code{progmem}
5851 attribute accomplishes this by putting respective variables into a
5852 section whose name starts with @code{.progmem}.
5854 This attribute works similar to the @code{section} attribute
5855 but adds additional checking.
5858 @item @bullet{}@tie{} Ordinary AVR cores with 32 general purpose registers:
5859 @code{progmem} affects the location
5860 of the data but not how this data is accessed.
5861 In order to read data located with the @code{progmem} attribute
5862 (inline) assembler must be used.
5864 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
5865 #include <avr/pgmspace.h>
5867 /* Locate var in flash memory */
5868 const int var[2] PROGMEM = @{ 1, 2 @};
5870 int read_var (int i)
5872 /* Access var[] by accessor macro from avr/pgmspace.h */
5873 return (int) pgm_read_word (& var[i]);
5877 AVR is a Harvard architecture processor and data and read-only data
5878 normally resides in the data memory (RAM).
5880 See also the @ref{AVR Named Address Spaces} section for
5881 an alternate way to locate and access data in flash memory.
5883 @item @bullet{}@tie{}Reduced AVR Tiny cores like ATtiny40:
5884 The compiler adds @code{0x4000}
5885 to the addresses of objects and declarations in @code{progmem} and locates
5886 the objects in flash memory, namely in section @code{.progmem.data}.
5887 The offset is needed because the flash memory is visible in the RAM
5888 address space starting at address @code{0x4000}.
5890 Data in @code{progmem} can be accessed by means of ordinary C@tie{}code,
5891 no special functions or macros are needed.
5894 /* var is located in flash memory */
5895 extern const int var[2] __attribute__((progmem));
5897 int read_var (int i)
5906 @itemx io (@var{addr})
5907 @cindex @code{io} variable attribute, AVR
5908 Variables with the @code{io} attribute are used to address
5909 memory-mapped peripherals in the io address range.
5910 If an address is specified, the variable
5911 is assigned that address, and the value is interpreted as an
5912 address in the data address space.
5916 volatile int porta __attribute__((io (0x22)));
5919 The address specified in the address in the data address range.
5921 Otherwise, the variable it is not assigned an address, but the
5922 compiler will still use in/out instructions where applicable,
5923 assuming some other module assigns an address in the io address range.
5927 extern volatile int porta __attribute__((io));
5931 @itemx io_low (@var{addr})
5932 @cindex @code{io_low} variable attribute, AVR
5933 This is like the @code{io} attribute, but additionally it informs the
5934 compiler that the object lies in the lower half of the I/O area,
5935 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
5939 @itemx address (@var{addr})
5940 @cindex @code{address} variable attribute, AVR
5941 Variables with the @code{address} attribute are used to address
5942 memory-mapped peripherals that may lie outside the io address range.
5945 volatile int porta __attribute__((address (0x600)));
5950 @node Blackfin Variable Attributes
5951 @subsection Blackfin Variable Attributes
5953 Three attributes are currently defined for the Blackfin.
5959 @cindex @code{l1_data} variable attribute, Blackfin
5960 @cindex @code{l1_data_A} variable attribute, Blackfin
5961 @cindex @code{l1_data_B} variable attribute, Blackfin
5962 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
5963 Variables with @code{l1_data} attribute are put into the specific section
5964 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
5965 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
5966 attribute are put into the specific section named @code{.l1.data.B}.
5969 @cindex @code{l2} variable attribute, Blackfin
5970 Use this attribute on the Blackfin to place the variable into L2 SRAM.
5971 Variables with @code{l2} attribute are put into the specific section
5972 named @code{.l2.data}.
5975 @node H8/300 Variable Attributes
5976 @subsection H8/300 Variable Attributes
5978 These variable attributes are available for H8/300 targets:
5982 @cindex @code{eightbit_data} variable attribute, H8/300
5983 @cindex eight-bit data on the H8/300, H8/300H, and H8S
5984 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
5985 variable should be placed into the eight-bit data section.
5986 The compiler generates more efficient code for certain operations
5987 on data in the eight-bit data area. Note the eight-bit data area is limited to
5990 You must use GAS and GLD from GNU binutils version 2.7 or later for
5991 this attribute to work correctly.
5994 @cindex @code{tiny_data} variable attribute, H8/300
5995 @cindex tiny data section on the H8/300H and H8S
5996 Use this attribute on the H8/300H and H8S to indicate that the specified
5997 variable should be placed into the tiny data section.
5998 The compiler generates more efficient code for loads and stores
5999 on data in the tiny data section. Note the tiny data area is limited to
6000 slightly under 32KB of data.
6004 @node IA-64 Variable Attributes
6005 @subsection IA-64 Variable Attributes
6007 The IA-64 back end supports the following variable attribute:
6010 @item model (@var{model-name})
6011 @cindex @code{model} variable attribute, IA-64
6013 On IA-64, use this attribute to set the addressability of an object.
6014 At present, the only supported identifier for @var{model-name} is
6015 @code{small}, indicating addressability via ``small'' (22-bit)
6016 addresses (so that their addresses can be loaded with the @code{addl}
6017 instruction). Caveat: such addressing is by definition not position
6018 independent and hence this attribute must not be used for objects
6019 defined by shared libraries.
6023 @node M32R/D Variable Attributes
6024 @subsection M32R/D Variable Attributes
6026 One attribute is currently defined for the M32R/D@.
6029 @item model (@var{model-name})
6030 @cindex @code{model-name} variable attribute, M32R/D
6031 @cindex variable addressability on the M32R/D
6032 Use this attribute on the M32R/D to set the addressability of an object.
6033 The identifier @var{model-name} is one of @code{small}, @code{medium},
6034 or @code{large}, representing each of the code models.
6036 Small model objects live in the lower 16MB of memory (so that their
6037 addresses can be loaded with the @code{ld24} instruction).
6039 Medium and large model objects may live anywhere in the 32-bit address space
6040 (the compiler generates @code{seth/add3} instructions to load their
6044 @node MeP Variable Attributes
6045 @subsection MeP Variable Attributes
6047 The MeP target has a number of addressing modes and busses. The
6048 @code{near} space spans the standard memory space's first 16 megabytes
6049 (24 bits). The @code{far} space spans the entire 32-bit memory space.
6050 The @code{based} space is a 128-byte region in the memory space that
6051 is addressed relative to the @code{$tp} register. The @code{tiny}
6052 space is a 65536-byte region relative to the @code{$gp} register. In
6053 addition to these memory regions, the MeP target has a separate 16-bit
6054 control bus which is specified with @code{cb} attributes.
6059 @cindex @code{based} variable attribute, MeP
6060 Any variable with the @code{based} attribute is assigned to the
6061 @code{.based} section, and is accessed with relative to the
6062 @code{$tp} register.
6065 @cindex @code{tiny} variable attribute, MeP
6066 Likewise, the @code{tiny} attribute assigned variables to the
6067 @code{.tiny} section, relative to the @code{$gp} register.
6070 @cindex @code{near} variable attribute, MeP
6071 Variables with the @code{near} attribute are assumed to have addresses
6072 that fit in a 24-bit addressing mode. This is the default for large
6073 variables (@code{-mtiny=4} is the default) but this attribute can
6074 override @code{-mtiny=} for small variables, or override @code{-ml}.
6077 @cindex @code{far} variable attribute, MeP
6078 Variables with the @code{far} attribute are addressed using a full
6079 32-bit address. Since this covers the entire memory space, this
6080 allows modules to make no assumptions about where variables might be
6084 @cindex @code{io} variable attribute, MeP
6085 @itemx io (@var{addr})
6086 Variables with the @code{io} attribute are used to address
6087 memory-mapped peripherals. If an address is specified, the variable
6088 is assigned that address, else it is not assigned an address (it is
6089 assumed some other module assigns an address). Example:
6092 int timer_count __attribute__((io(0x123)));
6096 @itemx cb (@var{addr})
6097 @cindex @code{cb} variable attribute, MeP
6098 Variables with the @code{cb} attribute are used to access the control
6099 bus, using special instructions. @code{addr} indicates the control bus
6103 int cpu_clock __attribute__((cb(0x123)));
6108 @node Microsoft Windows Variable Attributes
6109 @subsection Microsoft Windows Variable Attributes
6111 You can use these attributes on Microsoft Windows targets.
6112 @ref{x86 Variable Attributes} for additional Windows compatibility
6113 attributes available on all x86 targets.
6118 @cindex @code{dllimport} variable attribute
6119 @cindex @code{dllexport} variable attribute
6120 The @code{dllimport} and @code{dllexport} attributes are described in
6121 @ref{Microsoft Windows Function Attributes}.
6124 @cindex @code{selectany} variable attribute
6125 The @code{selectany} attribute causes an initialized global variable to
6126 have link-once semantics. When multiple definitions of the variable are
6127 encountered by the linker, the first is selected and the remainder are
6128 discarded. Following usage by the Microsoft compiler, the linker is told
6129 @emph{not} to warn about size or content differences of the multiple
6132 Although the primary usage of this attribute is for POD types, the
6133 attribute can also be applied to global C++ objects that are initialized
6134 by a constructor. In this case, the static initialization and destruction
6135 code for the object is emitted in each translation defining the object,
6136 but the calls to the constructor and destructor are protected by a
6137 link-once guard variable.
6139 The @code{selectany} attribute is only available on Microsoft Windows
6140 targets. You can use @code{__declspec (selectany)} as a synonym for
6141 @code{__attribute__ ((selectany))} for compatibility with other
6145 @cindex @code{shared} variable attribute
6146 On Microsoft Windows, in addition to putting variable definitions in a named
6147 section, the section can also be shared among all running copies of an
6148 executable or DLL@. For example, this small program defines shared data
6149 by putting it in a named section @code{shared} and marking the section
6153 int foo __attribute__((section ("shared"), shared)) = 0;
6158 /* @r{Read and write foo. All running
6159 copies see the same value.} */
6165 You may only use the @code{shared} attribute along with @code{section}
6166 attribute with a fully-initialized global definition because of the way
6167 linkers work. See @code{section} attribute for more information.
6169 The @code{shared} attribute is only available on Microsoft Windows@.
6173 @node MSP430 Variable Attributes
6174 @subsection MSP430 Variable Attributes
6178 @cindex @code{noinit} variable attribute, MSP430
6179 Any data with the @code{noinit} attribute will not be initialised by
6180 the C runtime startup code, or the program loader. Not initialising
6181 data in this way can reduce program startup times.
6184 @cindex @code{persistent} variable attribute, MSP430
6185 Any variable with the @code{persistent} attribute will not be
6186 initialised by the C runtime startup code. Instead its value will be
6187 set once, when the application is loaded, and then never initialised
6188 again, even if the processor is reset or the program restarts.
6189 Persistent data is intended to be placed into FLASH RAM, where its
6190 value will be retained across resets. The linker script being used to
6191 create the application should ensure that persistent data is correctly
6197 @cindex @code{lower} variable attribute, MSP430
6198 @cindex @code{upper} variable attribute, MSP430
6199 @cindex @code{either} variable attribute, MSP430
6200 These attributes are the same as the MSP430 function attributes of the
6201 same name (@pxref{MSP430 Function Attributes}).
6202 These attributes can be applied to both functions and variables.
6205 @node PowerPC Variable Attributes
6206 @subsection PowerPC Variable Attributes
6208 Three attributes currently are defined for PowerPC configurations:
6209 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6211 @cindex @code{ms_struct} variable attribute, PowerPC
6212 @cindex @code{gcc_struct} variable attribute, PowerPC
6213 For full documentation of the struct attributes please see the
6214 documentation in @ref{x86 Variable Attributes}.
6216 @cindex @code{altivec} variable attribute, PowerPC
6217 For documentation of @code{altivec} attribute please see the
6218 documentation in @ref{PowerPC Type Attributes}.
6220 @node RL78 Variable Attributes
6221 @subsection RL78 Variable Attributes
6223 @cindex @code{saddr} variable attribute, RL78
6224 The RL78 back end supports the @code{saddr} variable attribute. This
6225 specifies placement of the corresponding variable in the SADDR area,
6226 which can be accessed more efficiently than the default memory region.
6228 @node SPU Variable Attributes
6229 @subsection SPU Variable Attributes
6231 @cindex @code{spu_vector} variable attribute, SPU
6232 The SPU supports the @code{spu_vector} attribute for variables. For
6233 documentation of this attribute please see the documentation in
6234 @ref{SPU Type Attributes}.
6236 @node V850 Variable Attributes
6237 @subsection V850 Variable Attributes
6239 These variable attributes are supported by the V850 back end:
6244 @cindex @code{sda} variable attribute, V850
6245 Use this attribute to explicitly place a variable in the small data area,
6246 which can hold up to 64 kilobytes.
6249 @cindex @code{tda} variable attribute, V850
6250 Use this attribute to explicitly place a variable in the tiny data area,
6251 which can hold up to 256 bytes in total.
6254 @cindex @code{zda} variable attribute, V850
6255 Use this attribute to explicitly place a variable in the first 32 kilobytes
6259 @node x86 Variable Attributes
6260 @subsection x86 Variable Attributes
6262 Two attributes are currently defined for x86 configurations:
6263 @code{ms_struct} and @code{gcc_struct}.
6268 @cindex @code{ms_struct} variable attribute, x86
6269 @cindex @code{gcc_struct} variable attribute, x86
6271 If @code{packed} is used on a structure, or if bit-fields are used,
6272 it may be that the Microsoft ABI lays out the structure differently
6273 than the way GCC normally does. Particularly when moving packed
6274 data between functions compiled with GCC and the native Microsoft compiler
6275 (either via function call or as data in a file), it may be necessary to access
6278 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6279 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6280 command-line options, respectively;
6281 see @ref{x86 Options}, for details of how structure layout is affected.
6282 @xref{x86 Type Attributes}, for information about the corresponding
6283 attributes on types.
6287 @node Xstormy16 Variable Attributes
6288 @subsection Xstormy16 Variable Attributes
6290 One attribute is currently defined for xstormy16 configurations:
6295 @cindex @code{below100} variable attribute, Xstormy16
6297 If a variable has the @code{below100} attribute (@code{BELOW100} is
6298 allowed also), GCC places the variable in the first 0x100 bytes of
6299 memory and use special opcodes to access it. Such variables are
6300 placed in either the @code{.bss_below100} section or the
6301 @code{.data_below100} section.
6305 @node Type Attributes
6306 @section Specifying Attributes of Types
6307 @cindex attribute of types
6308 @cindex type attributes
6310 The keyword @code{__attribute__} allows you to specify special
6311 attributes of types. Some type attributes apply only to @code{struct}
6312 and @code{union} types, while others can apply to any type defined
6313 via a @code{typedef} declaration. Other attributes are defined for
6314 functions (@pxref{Function Attributes}), labels (@pxref{Label
6315 Attributes}), enumerators (@pxref{Enumerator Attributes}), and for
6316 variables (@pxref{Variable Attributes}).
6318 The @code{__attribute__} keyword is followed by an attribute specification
6319 inside double parentheses.
6321 You may specify type attributes in an enum, struct or union type
6322 declaration or definition by placing them immediately after the
6323 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6324 syntax is to place them just past the closing curly brace of the
6327 You can also include type attributes in a @code{typedef} declaration.
6328 @xref{Attribute Syntax}, for details of the exact syntax for using
6332 * Common Type Attributes::
6333 * ARM Type Attributes::
6334 * MeP Type Attributes::
6335 * PowerPC Type Attributes::
6336 * SPU Type Attributes::
6337 * x86 Type Attributes::
6340 @node Common Type Attributes
6341 @subsection Common Type Attributes
6343 The following type attributes are supported on most targets.
6346 @cindex @code{aligned} type attribute
6347 @item aligned (@var{alignment})
6348 This attribute specifies a minimum alignment (in bytes) for variables
6349 of the specified type. For example, the declarations:
6352 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6353 typedef int more_aligned_int __attribute__ ((aligned (8)));
6357 force the compiler to ensure (as far as it can) that each variable whose
6358 type is @code{struct S} or @code{more_aligned_int} is allocated and
6359 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6360 variables of type @code{struct S} aligned to 8-byte boundaries allows
6361 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6362 store) instructions when copying one variable of type @code{struct S} to
6363 another, thus improving run-time efficiency.
6365 Note that the alignment of any given @code{struct} or @code{union} type
6366 is required by the ISO C standard to be at least a perfect multiple of
6367 the lowest common multiple of the alignments of all of the members of
6368 the @code{struct} or @code{union} in question. This means that you @emph{can}
6369 effectively adjust the alignment of a @code{struct} or @code{union}
6370 type by attaching an @code{aligned} attribute to any one of the members
6371 of such a type, but the notation illustrated in the example above is a
6372 more obvious, intuitive, and readable way to request the compiler to
6373 adjust the alignment of an entire @code{struct} or @code{union} type.
6375 As in the preceding example, you can explicitly specify the alignment
6376 (in bytes) that you wish the compiler to use for a given @code{struct}
6377 or @code{union} type. Alternatively, you can leave out the alignment factor
6378 and just ask the compiler to align a type to the maximum
6379 useful alignment for the target machine you are compiling for. For
6380 example, you could write:
6383 struct S @{ short f[3]; @} __attribute__ ((aligned));
6386 Whenever you leave out the alignment factor in an @code{aligned}
6387 attribute specification, the compiler automatically sets the alignment
6388 for the type to the largest alignment that is ever used for any data
6389 type on the target machine you are compiling for. Doing this can often
6390 make copy operations more efficient, because the compiler can use
6391 whatever instructions copy the biggest chunks of memory when performing
6392 copies to or from the variables that have types that you have aligned
6395 In the example above, if the size of each @code{short} is 2 bytes, then
6396 the size of the entire @code{struct S} type is 6 bytes. The smallest
6397 power of two that is greater than or equal to that is 8, so the
6398 compiler sets the alignment for the entire @code{struct S} type to 8
6401 Note that although you can ask the compiler to select a time-efficient
6402 alignment for a given type and then declare only individual stand-alone
6403 objects of that type, the compiler's ability to select a time-efficient
6404 alignment is primarily useful only when you plan to create arrays of
6405 variables having the relevant (efficiently aligned) type. If you
6406 declare or use arrays of variables of an efficiently-aligned type, then
6407 it is likely that your program also does pointer arithmetic (or
6408 subscripting, which amounts to the same thing) on pointers to the
6409 relevant type, and the code that the compiler generates for these
6410 pointer arithmetic operations is often more efficient for
6411 efficiently-aligned types than for other types.
6413 Note that the effectiveness of @code{aligned} attributes may be limited
6414 by inherent limitations in your linker. On many systems, the linker is
6415 only able to arrange for variables to be aligned up to a certain maximum
6416 alignment. (For some linkers, the maximum supported alignment may
6417 be very very small.) If your linker is only able to align variables
6418 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6419 in an @code{__attribute__} still only provides you with 8-byte
6420 alignment. See your linker documentation for further information.
6422 The @code{aligned} attribute can only increase alignment. Alignment
6423 can be decreased by specifying the @code{packed} attribute. See below.
6425 @item bnd_variable_size
6426 @cindex @code{bnd_variable_size} type attribute
6427 @cindex Pointer Bounds Checker attributes
6428 When applied to a structure field, this attribute tells Pointer
6429 Bounds Checker that the size of this field should not be computed
6430 using static type information. It may be used to mark variably-sized
6431 static array fields placed at the end of a structure.
6439 S *p = (S *)malloc (sizeof(S) + 100);
6440 p->data[10] = 0; //Bounds violation
6444 By using an attribute for the field we may avoid unwanted bound
6451 char data[1] __attribute__((bnd_variable_size));
6453 S *p = (S *)malloc (sizeof(S) + 100);
6454 p->data[10] = 0; //OK
6458 @itemx deprecated (@var{msg})
6459 @cindex @code{deprecated} type attribute
6460 The @code{deprecated} attribute results in a warning if the type
6461 is used anywhere in the source file. This is useful when identifying
6462 types that are expected to be removed in a future version of a program.
6463 If possible, the warning also includes the location of the declaration
6464 of the deprecated type, to enable users to easily find further
6465 information about why the type is deprecated, or what they should do
6466 instead. Note that the warnings only occur for uses and then only
6467 if the type is being applied to an identifier that itself is not being
6468 declared as deprecated.
6471 typedef int T1 __attribute__ ((deprecated));
6475 typedef T1 T3 __attribute__ ((deprecated));
6476 T3 z __attribute__ ((deprecated));
6480 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
6481 warning is issued for line 4 because T2 is not explicitly
6482 deprecated. Line 5 has no warning because T3 is explicitly
6483 deprecated. Similarly for line 6. The optional @var{msg}
6484 argument, which must be a string, is printed in the warning if
6487 The @code{deprecated} attribute can also be used for functions and
6488 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
6490 @item designated_init
6491 @cindex @code{designated_init} type attribute
6492 This attribute may only be applied to structure types. It indicates
6493 that any initialization of an object of this type must use designated
6494 initializers rather than positional initializers. The intent of this
6495 attribute is to allow the programmer to indicate that a structure's
6496 layout may change, and that therefore relying on positional
6497 initialization will result in future breakage.
6499 GCC emits warnings based on this attribute by default; use
6500 @option{-Wno-designated-init} to suppress them.
6503 @cindex @code{may_alias} type attribute
6504 Accesses through pointers to types with this attribute are not subject
6505 to type-based alias analysis, but are instead assumed to be able to alias
6506 any other type of objects.
6507 In the context of section 6.5 paragraph 7 of the C99 standard,
6508 an lvalue expression
6509 dereferencing such a pointer is treated like having a character type.
6510 See @option{-fstrict-aliasing} for more information on aliasing issues.
6511 This extension exists to support some vector APIs, in which pointers to
6512 one vector type are permitted to alias pointers to a different vector type.
6514 Note that an object of a type with this attribute does not have any
6520 typedef short __attribute__((__may_alias__)) short_a;
6526 short_a *b = (short_a *) &a;
6530 if (a == 0x12345678)
6538 If you replaced @code{short_a} with @code{short} in the variable
6539 declaration, the above program would abort when compiled with
6540 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
6544 @cindex @code{packed} type attribute
6545 This attribute, attached to @code{struct} or @code{union} type
6546 definition, specifies that each member (other than zero-width bit-fields)
6547 of the structure or union is placed to minimize the memory required. When
6548 attached to an @code{enum} definition, it indicates that the smallest
6549 integral type should be used.
6551 @opindex fshort-enums
6552 Specifying the @code{packed} attribute for @code{struct} and @code{union}
6553 types is equivalent to specifying the @code{packed} attribute on each
6554 of the structure or union members. Specifying the @option{-fshort-enums}
6555 flag on the command line is equivalent to specifying the @code{packed}
6556 attribute on all @code{enum} definitions.
6558 In the following example @code{struct my_packed_struct}'s members are
6559 packed closely together, but the internal layout of its @code{s} member
6560 is not packed---to do that, @code{struct my_unpacked_struct} needs to
6564 struct my_unpacked_struct
6570 struct __attribute__ ((__packed__)) my_packed_struct
6574 struct my_unpacked_struct s;
6578 You may only specify the @code{packed} attribute attribute on the definition
6579 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
6580 that does not also define the enumerated type, structure or union.
6582 @item scalar_storage_order ("@var{endianness}")
6583 @cindex @code{scalar_storage_order} type attribute
6584 When attached to a @code{union} or a @code{struct}, this attribute sets
6585 the storage order, aka endianness, of the scalar fields of the type, as
6586 well as the array fields whose component is scalar. The supported
6587 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
6588 has no effects on fields which are themselves a @code{union}, a @code{struct}
6589 or an array whose component is a @code{union} or a @code{struct}, and it is
6590 possible for these fields to have a different scalar storage order than the
6593 This attribute is supported only for targets that use a uniform default
6594 scalar storage order (fortunately, most of them), i.e. targets that store
6595 the scalars either all in big-endian or all in little-endian.
6597 Additional restrictions are enforced for types with the reverse scalar
6598 storage order with regard to the scalar storage order of the target:
6601 @item Taking the address of a scalar field of a @code{union} or a
6602 @code{struct} with reverse scalar storage order is not permitted and yields
6604 @item Taking the address of an array field, whose component is scalar, of
6605 a @code{union} or a @code{struct} with reverse scalar storage order is
6606 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
6608 @item Taking the address of a @code{union} or a @code{struct} with reverse
6609 scalar storage order is permitted.
6612 These restrictions exist because the storage order attribute is lost when
6613 the address of a scalar or the address of an array with scalar component is
6614 taken, so storing indirectly through this address generally does not work.
6615 The second case is nevertheless allowed to be able to perform a block copy
6616 from or to the array.
6618 Moreover, the use of type punning or aliasing to toggle the storage order
6619 is not supported; that is to say, a given scalar object cannot be accessed
6620 through distinct types that assign a different storage order to it.
6622 @item transparent_union
6623 @cindex @code{transparent_union} type attribute
6625 This attribute, attached to a @code{union} type definition, indicates
6626 that any function parameter having that union type causes calls to that
6627 function to be treated in a special way.
6629 First, the argument corresponding to a transparent union type can be of
6630 any type in the union; no cast is required. Also, if the union contains
6631 a pointer type, the corresponding argument can be a null pointer
6632 constant or a void pointer expression; and if the union contains a void
6633 pointer type, the corresponding argument can be any pointer expression.
6634 If the union member type is a pointer, qualifiers like @code{const} on
6635 the referenced type must be respected, just as with normal pointer
6638 Second, the argument is passed to the function using the calling
6639 conventions of the first member of the transparent union, not the calling
6640 conventions of the union itself. All members of the union must have the
6641 same machine representation; this is necessary for this argument passing
6644 Transparent unions are designed for library functions that have multiple
6645 interfaces for compatibility reasons. For example, suppose the
6646 @code{wait} function must accept either a value of type @code{int *} to
6647 comply with POSIX, or a value of type @code{union wait *} to comply with
6648 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
6649 @code{wait} would accept both kinds of arguments, but it would also
6650 accept any other pointer type and this would make argument type checking
6651 less useful. Instead, @code{<sys/wait.h>} might define the interface
6655 typedef union __attribute__ ((__transparent_union__))
6659 @} wait_status_ptr_t;
6661 pid_t wait (wait_status_ptr_t);
6665 This interface allows either @code{int *} or @code{union wait *}
6666 arguments to be passed, using the @code{int *} calling convention.
6667 The program can call @code{wait} with arguments of either type:
6670 int w1 () @{ int w; return wait (&w); @}
6671 int w2 () @{ union wait w; return wait (&w); @}
6675 With this interface, @code{wait}'s implementation might look like this:
6678 pid_t wait (wait_status_ptr_t p)
6680 return waitpid (-1, p.__ip, 0);
6685 @cindex @code{unused} type attribute
6686 When attached to a type (including a @code{union} or a @code{struct}),
6687 this attribute means that variables of that type are meant to appear
6688 possibly unused. GCC does not produce a warning for any variables of
6689 that type, even if the variable appears to do nothing. This is often
6690 the case with lock or thread classes, which are usually defined and then
6691 not referenced, but contain constructors and destructors that have
6692 nontrivial bookkeeping functions.
6695 @cindex @code{visibility} type attribute
6696 In C++, attribute visibility (@pxref{Function Attributes}) can also be
6697 applied to class, struct, union and enum types. Unlike other type
6698 attributes, the attribute must appear between the initial keyword and
6699 the name of the type; it cannot appear after the body of the type.
6701 Note that the type visibility is applied to vague linkage entities
6702 associated with the class (vtable, typeinfo node, etc.). In
6703 particular, if a class is thrown as an exception in one shared object
6704 and caught in another, the class must have default visibility.
6705 Otherwise the two shared objects are unable to use the same
6706 typeinfo node and exception handling will break.
6710 To specify multiple attributes, separate them by commas within the
6711 double parentheses: for example, @samp{__attribute__ ((aligned (16),
6714 @node ARM Type Attributes
6715 @subsection ARM Type Attributes
6717 @cindex @code{notshared} type attribute, ARM
6718 On those ARM targets that support @code{dllimport} (such as Symbian
6719 OS), you can use the @code{notshared} attribute to indicate that the
6720 virtual table and other similar data for a class should not be
6721 exported from a DLL@. For example:
6724 class __declspec(notshared) C @{
6726 __declspec(dllimport) C();
6730 __declspec(dllexport)
6735 In this code, @code{C::C} is exported from the current DLL, but the
6736 virtual table for @code{C} is not exported. (You can use
6737 @code{__attribute__} instead of @code{__declspec} if you prefer, but
6738 most Symbian OS code uses @code{__declspec}.)
6740 @node MeP Type Attributes
6741 @subsection MeP Type Attributes
6743 @cindex @code{based} type attribute, MeP
6744 @cindex @code{tiny} type attribute, MeP
6745 @cindex @code{near} type attribute, MeP
6746 @cindex @code{far} type attribute, MeP
6747 Many of the MeP variable attributes may be applied to types as well.
6748 Specifically, the @code{based}, @code{tiny}, @code{near}, and
6749 @code{far} attributes may be applied to either. The @code{io} and
6750 @code{cb} attributes may not be applied to types.
6752 @node PowerPC Type Attributes
6753 @subsection PowerPC Type Attributes
6755 Three attributes currently are defined for PowerPC configurations:
6756 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6758 @cindex @code{ms_struct} type attribute, PowerPC
6759 @cindex @code{gcc_struct} type attribute, PowerPC
6760 For full documentation of the @code{ms_struct} and @code{gcc_struct}
6761 attributes please see the documentation in @ref{x86 Type Attributes}.
6763 @cindex @code{altivec} type attribute, PowerPC
6764 The @code{altivec} attribute allows one to declare AltiVec vector data
6765 types supported by the AltiVec Programming Interface Manual. The
6766 attribute requires an argument to specify one of three vector types:
6767 @code{vector__}, @code{pixel__} (always followed by unsigned short),
6768 and @code{bool__} (always followed by unsigned).
6771 __attribute__((altivec(vector__)))
6772 __attribute__((altivec(pixel__))) unsigned short
6773 __attribute__((altivec(bool__))) unsigned
6776 These attributes mainly are intended to support the @code{__vector},
6777 @code{__pixel}, and @code{__bool} AltiVec keywords.
6779 @node SPU Type Attributes
6780 @subsection SPU Type Attributes
6782 @cindex @code{spu_vector} type attribute, SPU
6783 The SPU supports the @code{spu_vector} attribute for types. This attribute
6784 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
6785 Language Extensions Specification. It is intended to support the
6786 @code{__vector} keyword.
6788 @node x86 Type Attributes
6789 @subsection x86 Type Attributes
6791 Two attributes are currently defined for x86 configurations:
6792 @code{ms_struct} and @code{gcc_struct}.
6798 @cindex @code{ms_struct} type attribute, x86
6799 @cindex @code{gcc_struct} type attribute, x86
6801 If @code{packed} is used on a structure, or if bit-fields are used
6802 it may be that the Microsoft ABI packs them differently
6803 than GCC normally packs them. Particularly when moving packed
6804 data between functions compiled with GCC and the native Microsoft compiler
6805 (either via function call or as data in a file), it may be necessary to access
6808 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6809 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6810 command-line options, respectively;
6811 see @ref{x86 Options}, for details of how structure layout is affected.
6812 @xref{x86 Variable Attributes}, for information about the corresponding
6813 attributes on variables.
6817 @node Label Attributes
6818 @section Label Attributes
6819 @cindex Label Attributes
6821 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
6822 details of the exact syntax for using attributes. Other attributes are
6823 available for functions (@pxref{Function Attributes}), variables
6824 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
6825 and for types (@pxref{Type Attributes}).
6827 This example uses the @code{cold} label attribute to indicate the
6828 @code{ErrorHandling} branch is unlikely to be taken and that the
6829 @code{ErrorHandling} label is unused:
6833 asm goto ("some asm" : : : : NoError);
6835 /* This branch (the fall-through from the asm) is less commonly used */
6837 __attribute__((cold, unused)); /* Semi-colon is required here */
6842 printf("no error\n");
6848 @cindex @code{unused} label attribute
6849 This feature is intended for program-generated code that may contain
6850 unused labels, but which is compiled with @option{-Wall}. It is
6851 not normally appropriate to use in it human-written code, though it
6852 could be useful in cases where the code that jumps to the label is
6853 contained within an @code{#ifdef} conditional.
6856 @cindex @code{hot} label attribute
6857 The @code{hot} attribute on a label is used to inform the compiler that
6858 the path following the label is more likely than paths that are not so
6859 annotated. This attribute is used in cases where @code{__builtin_expect}
6860 cannot be used, for instance with computed goto or @code{asm goto}.
6863 @cindex @code{cold} label attribute
6864 The @code{cold} attribute on labels is used to inform the compiler that
6865 the path following the label is unlikely to be executed. This attribute
6866 is used in cases where @code{__builtin_expect} cannot be used, for instance
6867 with computed goto or @code{asm goto}.
6871 @node Enumerator Attributes
6872 @section Enumerator Attributes
6873 @cindex Enumerator Attributes
6875 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
6876 details of the exact syntax for using attributes. Other attributes are
6877 available for functions (@pxref{Function Attributes}), variables
6878 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}),
6879 and for types (@pxref{Type Attributes}).
6881 This example uses the @code{deprecated} enumerator attribute to indicate the
6882 @code{oldval} enumerator is deprecated:
6886 oldval __attribute__((deprecated)),
6899 @cindex @code{deprecated} enumerator attribute
6900 The @code{deprecated} attribute results in a warning if the enumerator
6901 is used anywhere in the source file. This is useful when identifying
6902 enumerators that are expected to be removed in a future version of a
6903 program. The warning also includes the location of the declaration
6904 of the deprecated enumerator, to enable users to easily find further
6905 information about why the enumerator is deprecated, or what they should
6906 do instead. Note that the warnings only occurs for uses.
6910 @node Attribute Syntax
6911 @section Attribute Syntax
6912 @cindex attribute syntax
6914 This section describes the syntax with which @code{__attribute__} may be
6915 used, and the constructs to which attribute specifiers bind, for the C
6916 language. Some details may vary for C++ and Objective-C@. Because of
6917 infelicities in the grammar for attributes, some forms described here
6918 may not be successfully parsed in all cases.
6920 There are some problems with the semantics of attributes in C++. For
6921 example, there are no manglings for attributes, although they may affect
6922 code generation, so problems may arise when attributed types are used in
6923 conjunction with templates or overloading. Similarly, @code{typeid}
6924 does not distinguish between types with different attributes. Support
6925 for attributes in C++ may be restricted in future to attributes on
6926 declarations only, but not on nested declarators.
6928 @xref{Function Attributes}, for details of the semantics of attributes
6929 applying to functions. @xref{Variable Attributes}, for details of the
6930 semantics of attributes applying to variables. @xref{Type Attributes},
6931 for details of the semantics of attributes applying to structure, union
6932 and enumerated types.
6933 @xref{Label Attributes}, for details of the semantics of attributes
6935 @xref{Enumerator Attributes}, for details of the semantics of attributes
6936 applying to enumerators.
6938 An @dfn{attribute specifier} is of the form
6939 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
6940 is a possibly empty comma-separated sequence of @dfn{attributes}, where
6941 each attribute is one of the following:
6945 Empty. Empty attributes are ignored.
6949 (which may be an identifier such as @code{unused}, or a reserved
6950 word such as @code{const}).
6953 An attribute name followed by a parenthesized list of
6954 parameters for the attribute.
6955 These parameters take one of the following forms:
6959 An identifier. For example, @code{mode} attributes use this form.
6962 An identifier followed by a comma and a non-empty comma-separated list
6963 of expressions. For example, @code{format} attributes use this form.
6966 A possibly empty comma-separated list of expressions. For example,
6967 @code{format_arg} attributes use this form with the list being a single
6968 integer constant expression, and @code{alias} attributes use this form
6969 with the list being a single string constant.
6973 An @dfn{attribute specifier list} is a sequence of one or more attribute
6974 specifiers, not separated by any other tokens.
6976 You may optionally specify attribute names with @samp{__}
6977 preceding and following the name.
6978 This allows you to use them in header files without
6979 being concerned about a possible macro of the same name. For example,
6980 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
6983 @subsubheading Label Attributes
6985 In GNU C, an attribute specifier list may appear after the colon following a
6986 label, other than a @code{case} or @code{default} label. GNU C++ only permits
6987 attributes on labels if the attribute specifier is immediately
6988 followed by a semicolon (i.e., the label applies to an empty
6989 statement). If the semicolon is missing, C++ label attributes are
6990 ambiguous, as it is permissible for a declaration, which could begin
6991 with an attribute list, to be labelled in C++. Declarations cannot be
6992 labelled in C90 or C99, so the ambiguity does not arise there.
6994 @subsubheading Enumerator Attributes
6996 In GNU C, an attribute specifier list may appear as part of an enumerator.
6997 The attribute goes after the enumeration constant, before @code{=}, if
6998 present. The optional attribute in the enumerator appertains to the
6999 enumeration constant. It is not possible to place the attribute after
7000 the constant expression, if present.
7002 @subsubheading Type Attributes
7004 An attribute specifier list may appear as part of a @code{struct},
7005 @code{union} or @code{enum} specifier. It may go either immediately
7006 after the @code{struct}, @code{union} or @code{enum} keyword, or after
7007 the closing brace. The former syntax is preferred.
7008 Where attribute specifiers follow the closing brace, they are considered
7009 to relate to the structure, union or enumerated type defined, not to any
7010 enclosing declaration the type specifier appears in, and the type
7011 defined is not complete until after the attribute specifiers.
7012 @c Otherwise, there would be the following problems: a shift/reduce
7013 @c conflict between attributes binding the struct/union/enum and
7014 @c binding to the list of specifiers/qualifiers; and "aligned"
7015 @c attributes could use sizeof for the structure, but the size could be
7016 @c changed later by "packed" attributes.
7019 @subsubheading All other attributes
7021 Otherwise, an attribute specifier appears as part of a declaration,
7022 counting declarations of unnamed parameters and type names, and relates
7023 to that declaration (which may be nested in another declaration, for
7024 example in the case of a parameter declaration), or to a particular declarator
7025 within a declaration. Where an
7026 attribute specifier is applied to a parameter declared as a function or
7027 an array, it should apply to the function or array rather than the
7028 pointer to which the parameter is implicitly converted, but this is not
7029 yet correctly implemented.
7031 Any list of specifiers and qualifiers at the start of a declaration may
7032 contain attribute specifiers, whether or not such a list may in that
7033 context contain storage class specifiers. (Some attributes, however,
7034 are essentially in the nature of storage class specifiers, and only make
7035 sense where storage class specifiers may be used; for example,
7036 @code{section}.) There is one necessary limitation to this syntax: the
7037 first old-style parameter declaration in a function definition cannot
7038 begin with an attribute specifier, because such an attribute applies to
7039 the function instead by syntax described below (which, however, is not
7040 yet implemented in this case). In some other cases, attribute
7041 specifiers are permitted by this grammar but not yet supported by the
7042 compiler. All attribute specifiers in this place relate to the
7043 declaration as a whole. In the obsolescent usage where a type of
7044 @code{int} is implied by the absence of type specifiers, such a list of
7045 specifiers and qualifiers may be an attribute specifier list with no
7046 other specifiers or qualifiers.
7048 At present, the first parameter in a function prototype must have some
7049 type specifier that is not an attribute specifier; this resolves an
7050 ambiguity in the interpretation of @code{void f(int
7051 (__attribute__((foo)) x))}, but is subject to change. At present, if
7052 the parentheses of a function declarator contain only attributes then
7053 those attributes are ignored, rather than yielding an error or warning
7054 or implying a single parameter of type int, but this is subject to
7057 An attribute specifier list may appear immediately before a declarator
7058 (other than the first) in a comma-separated list of declarators in a
7059 declaration of more than one identifier using a single list of
7060 specifiers and qualifiers. Such attribute specifiers apply
7061 only to the identifier before whose declarator they appear. For
7065 __attribute__((noreturn)) void d0 (void),
7066 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
7071 the @code{noreturn} attribute applies to all the functions
7072 declared; the @code{format} attribute only applies to @code{d1}.
7074 An attribute specifier list may appear immediately before the comma,
7075 @code{=} or semicolon terminating the declaration of an identifier other
7076 than a function definition. Such attribute specifiers apply
7077 to the declared object or function. Where an
7078 assembler name for an object or function is specified (@pxref{Asm
7079 Labels}), the attribute must follow the @code{asm}
7082 An attribute specifier list may, in future, be permitted to appear after
7083 the declarator in a function definition (before any old-style parameter
7084 declarations or the function body).
7086 Attribute specifiers may be mixed with type qualifiers appearing inside
7087 the @code{[]} of a parameter array declarator, in the C99 construct by
7088 which such qualifiers are applied to the pointer to which the array is
7089 implicitly converted. Such attribute specifiers apply to the pointer,
7090 not to the array, but at present this is not implemented and they are
7093 An attribute specifier list may appear at the start of a nested
7094 declarator. At present, there are some limitations in this usage: the
7095 attributes correctly apply to the declarator, but for most individual
7096 attributes the semantics this implies are not implemented.
7097 When attribute specifiers follow the @code{*} of a pointer
7098 declarator, they may be mixed with any type qualifiers present.
7099 The following describes the formal semantics of this syntax. It makes the
7100 most sense if you are familiar with the formal specification of
7101 declarators in the ISO C standard.
7103 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
7104 D1}, where @code{T} contains declaration specifiers that specify a type
7105 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
7106 contains an identifier @var{ident}. The type specified for @var{ident}
7107 for derived declarators whose type does not include an attribute
7108 specifier is as in the ISO C standard.
7110 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7111 and the declaration @code{T D} specifies the type
7112 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7113 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7114 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7116 If @code{D1} has the form @code{*
7117 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7118 declaration @code{T D} specifies the type
7119 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7120 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7121 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7127 void (__attribute__((noreturn)) ****f) (void);
7131 specifies the type ``pointer to pointer to pointer to pointer to
7132 non-returning function returning @code{void}''. As another example,
7135 char *__attribute__((aligned(8))) *f;
7139 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7140 Note again that this does not work with most attributes; for example,
7141 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7142 is not yet supported.
7144 For compatibility with existing code written for compiler versions that
7145 did not implement attributes on nested declarators, some laxity is
7146 allowed in the placing of attributes. If an attribute that only applies
7147 to types is applied to a declaration, it is treated as applying to
7148 the type of that declaration. If an attribute that only applies to
7149 declarations is applied to the type of a declaration, it is treated
7150 as applying to that declaration; and, for compatibility with code
7151 placing the attributes immediately before the identifier declared, such
7152 an attribute applied to a function return type is treated as
7153 applying to the function type, and such an attribute applied to an array
7154 element type is treated as applying to the array type. If an
7155 attribute that only applies to function types is applied to a
7156 pointer-to-function type, it is treated as applying to the pointer
7157 target type; if such an attribute is applied to a function return type
7158 that is not a pointer-to-function type, it is treated as applying
7159 to the function type.
7161 @node Function Prototypes
7162 @section Prototypes and Old-Style Function Definitions
7163 @cindex function prototype declarations
7164 @cindex old-style function definitions
7165 @cindex promotion of formal parameters
7167 GNU C extends ISO C to allow a function prototype to override a later
7168 old-style non-prototype definition. Consider the following example:
7171 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7178 /* @r{Prototype function declaration.} */
7179 int isroot P((uid_t));
7181 /* @r{Old-style function definition.} */
7183 isroot (x) /* @r{??? lossage here ???} */
7190 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7191 not allow this example, because subword arguments in old-style
7192 non-prototype definitions are promoted. Therefore in this example the
7193 function definition's argument is really an @code{int}, which does not
7194 match the prototype argument type of @code{short}.
7196 This restriction of ISO C makes it hard to write code that is portable
7197 to traditional C compilers, because the programmer does not know
7198 whether the @code{uid_t} type is @code{short}, @code{int}, or
7199 @code{long}. Therefore, in cases like these GNU C allows a prototype
7200 to override a later old-style definition. More precisely, in GNU C, a
7201 function prototype argument type overrides the argument type specified
7202 by a later old-style definition if the former type is the same as the
7203 latter type before promotion. Thus in GNU C the above example is
7204 equivalent to the following:
7217 GNU C++ does not support old-style function definitions, so this
7218 extension is irrelevant.
7221 @section C++ Style Comments
7223 @cindex C++ comments
7224 @cindex comments, C++ style
7226 In GNU C, you may use C++ style comments, which start with @samp{//} and
7227 continue until the end of the line. Many other C implementations allow
7228 such comments, and they are included in the 1999 C standard. However,
7229 C++ style comments are not recognized if you specify an @option{-std}
7230 option specifying a version of ISO C before C99, or @option{-ansi}
7231 (equivalent to @option{-std=c90}).
7234 @section Dollar Signs in Identifier Names
7236 @cindex dollar signs in identifier names
7237 @cindex identifier names, dollar signs in
7239 In GNU C, you may normally use dollar signs in identifier names.
7240 This is because many traditional C implementations allow such identifiers.
7241 However, dollar signs in identifiers are not supported on a few target
7242 machines, typically because the target assembler does not allow them.
7244 @node Character Escapes
7245 @section The Character @key{ESC} in Constants
7247 You can use the sequence @samp{\e} in a string or character constant to
7248 stand for the ASCII character @key{ESC}.
7251 @section Inquiring on Alignment of Types or Variables
7253 @cindex type alignment
7254 @cindex variable alignment
7256 The keyword @code{__alignof__} allows you to inquire about how an object
7257 is aligned, or the minimum alignment usually required by a type. Its
7258 syntax is just like @code{sizeof}.
7260 For example, if the target machine requires a @code{double} value to be
7261 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7262 This is true on many RISC machines. On more traditional machine
7263 designs, @code{__alignof__ (double)} is 4 or even 2.
7265 Some machines never actually require alignment; they allow reference to any
7266 data type even at an odd address. For these machines, @code{__alignof__}
7267 reports the smallest alignment that GCC gives the data type, usually as
7268 mandated by the target ABI.
7270 If the operand of @code{__alignof__} is an lvalue rather than a type,
7271 its value is the required alignment for its type, taking into account
7272 any minimum alignment specified with GCC's @code{__attribute__}
7273 extension (@pxref{Variable Attributes}). For example, after this
7277 struct foo @{ int x; char y; @} foo1;
7281 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7282 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7284 It is an error to ask for the alignment of an incomplete type.
7288 @section An Inline Function is As Fast As a Macro
7289 @cindex inline functions
7290 @cindex integrating function code
7292 @cindex macros, inline alternative
7294 By declaring a function inline, you can direct GCC to make
7295 calls to that function faster. One way GCC can achieve this is to
7296 integrate that function's code into the code for its callers. This
7297 makes execution faster by eliminating the function-call overhead; in
7298 addition, if any of the actual argument values are constant, their
7299 known values may permit simplifications at compile time so that not
7300 all of the inline function's code needs to be included. The effect on
7301 code size is less predictable; object code may be larger or smaller
7302 with function inlining, depending on the particular case. You can
7303 also direct GCC to try to integrate all ``simple enough'' functions
7304 into their callers with the option @option{-finline-functions}.
7306 GCC implements three different semantics of declaring a function
7307 inline. One is available with @option{-std=gnu89} or
7308 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7309 on all inline declarations, another when
7310 @option{-std=c99}, @option{-std=c11},
7311 @option{-std=gnu99} or @option{-std=gnu11}
7312 (without @option{-fgnu89-inline}), and the third
7313 is used when compiling C++.
7315 To declare a function inline, use the @code{inline} keyword in its
7316 declaration, like this:
7326 If you are writing a header file to be included in ISO C90 programs, write
7327 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7329 The three types of inlining behave similarly in two important cases:
7330 when the @code{inline} keyword is used on a @code{static} function,
7331 like the example above, and when a function is first declared without
7332 using the @code{inline} keyword and then is defined with
7333 @code{inline}, like this:
7336 extern int inc (int *a);
7344 In both of these common cases, the program behaves the same as if you
7345 had not used the @code{inline} keyword, except for its speed.
7347 @cindex inline functions, omission of
7348 @opindex fkeep-inline-functions
7349 When a function is both inline and @code{static}, if all calls to the
7350 function are integrated into the caller, and the function's address is
7351 never used, then the function's own assembler code is never referenced.
7352 In this case, GCC does not actually output assembler code for the
7353 function, unless you specify the option @option{-fkeep-inline-functions}.
7354 If there is a nonintegrated call, then the function is compiled to
7355 assembler code as usual. The function must also be compiled as usual if
7356 the program refers to its address, because that can't be inlined.
7359 Note that certain usages in a function definition can make it unsuitable
7360 for inline substitution. Among these usages are: variadic functions,
7361 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7362 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7363 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7364 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7365 function marked @code{inline} could not be substituted, and gives the
7366 reason for the failure.
7368 @cindex automatic @code{inline} for C++ member fns
7369 @cindex @code{inline} automatic for C++ member fns
7370 @cindex member fns, automatically @code{inline}
7371 @cindex C++ member fns, automatically @code{inline}
7372 @opindex fno-default-inline
7373 As required by ISO C++, GCC considers member functions defined within
7374 the body of a class to be marked inline even if they are
7375 not explicitly declared with the @code{inline} keyword. You can
7376 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7377 Options,,Options Controlling C++ Dialect}.
7379 GCC does not inline any functions when not optimizing unless you specify
7380 the @samp{always_inline} attribute for the function, like this:
7383 /* @r{Prototype.} */
7384 inline void foo (const char) __attribute__((always_inline));
7387 The remainder of this section is specific to GNU C90 inlining.
7389 @cindex non-static inline function
7390 When an inline function is not @code{static}, then the compiler must assume
7391 that there may be calls from other source files; since a global symbol can
7392 be defined only once in any program, the function must not be defined in
7393 the other source files, so the calls therein cannot be integrated.
7394 Therefore, a non-@code{static} inline function is always compiled on its
7395 own in the usual fashion.
7397 If you specify both @code{inline} and @code{extern} in the function
7398 definition, then the definition is used only for inlining. In no case
7399 is the function compiled on its own, not even if you refer to its
7400 address explicitly. Such an address becomes an external reference, as
7401 if you had only declared the function, and had not defined it.
7403 This combination of @code{inline} and @code{extern} has almost the
7404 effect of a macro. The way to use it is to put a function definition in
7405 a header file with these keywords, and put another copy of the
7406 definition (lacking @code{inline} and @code{extern}) in a library file.
7407 The definition in the header file causes most calls to the function
7408 to be inlined. If any uses of the function remain, they refer to
7409 the single copy in the library.
7412 @section When is a Volatile Object Accessed?
7413 @cindex accessing volatiles
7414 @cindex volatile read
7415 @cindex volatile write
7416 @cindex volatile access
7418 C has the concept of volatile objects. These are normally accessed by
7419 pointers and used for accessing hardware or inter-thread
7420 communication. The standard encourages compilers to refrain from
7421 optimizations concerning accesses to volatile objects, but leaves it
7422 implementation defined as to what constitutes a volatile access. The
7423 minimum requirement is that at a sequence point all previous accesses
7424 to volatile objects have stabilized and no subsequent accesses have
7425 occurred. Thus an implementation is free to reorder and combine
7426 volatile accesses that occur between sequence points, but cannot do
7427 so for accesses across a sequence point. The use of volatile does
7428 not allow you to violate the restriction on updating objects multiple
7429 times between two sequence points.
7431 Accesses to non-volatile objects are not ordered with respect to
7432 volatile accesses. You cannot use a volatile object as a memory
7433 barrier to order a sequence of writes to non-volatile memory. For
7437 int *ptr = @var{something};
7439 *ptr = @var{something};
7444 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
7445 that the write to @var{*ptr} occurs by the time the update
7446 of @var{vobj} happens. If you need this guarantee, you must use
7447 a stronger memory barrier such as:
7450 int *ptr = @var{something};
7452 *ptr = @var{something};
7453 asm volatile ("" : : : "memory");
7457 A scalar volatile object is read when it is accessed in a void context:
7460 volatile int *src = @var{somevalue};
7464 Such expressions are rvalues, and GCC implements this as a
7465 read of the volatile object being pointed to.
7467 Assignments are also expressions and have an rvalue. However when
7468 assigning to a scalar volatile, the volatile object is not reread,
7469 regardless of whether the assignment expression's rvalue is used or
7470 not. If the assignment's rvalue is used, the value is that assigned
7471 to the volatile object. For instance, there is no read of @var{vobj}
7472 in all the following cases:
7477 vobj = @var{something};
7478 obj = vobj = @var{something};
7479 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
7480 obj = (@var{something}, vobj = @var{anotherthing});
7483 If you need to read the volatile object after an assignment has
7484 occurred, you must use a separate expression with an intervening
7487 As bit-fields are not individually addressable, volatile bit-fields may
7488 be implicitly read when written to, or when adjacent bit-fields are
7489 accessed. Bit-field operations may be optimized such that adjacent
7490 bit-fields are only partially accessed, if they straddle a storage unit
7491 boundary. For these reasons it is unwise to use volatile bit-fields to
7494 @node Using Assembly Language with C
7495 @section How to Use Inline Assembly Language in C Code
7496 @cindex @code{asm} keyword
7497 @cindex assembly language in C
7498 @cindex inline assembly language
7499 @cindex mixing assembly language and C
7501 The @code{asm} keyword allows you to embed assembler instructions
7502 within C code. GCC provides two forms of inline @code{asm}
7503 statements. A @dfn{basic @code{asm}} statement is one with no
7504 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
7505 statement (@pxref{Extended Asm}) includes one or more operands.
7506 The extended form is preferred for mixing C and assembly language
7507 within a function, but to include assembly language at
7508 top level you must use basic @code{asm}.
7510 You can also use the @code{asm} keyword to override the assembler name
7511 for a C symbol, or to place a C variable in a specific register.
7514 * Basic Asm:: Inline assembler without operands.
7515 * Extended Asm:: Inline assembler with operands.
7516 * Constraints:: Constraints for @code{asm} operands
7517 * Asm Labels:: Specifying the assembler name to use for a C symbol.
7518 * Explicit Register Variables:: Defining variables residing in specified
7520 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
7524 @subsection Basic Asm --- Assembler Instructions Without Operands
7525 @cindex basic @code{asm}
7526 @cindex assembly language in C, basic
7528 A basic @code{asm} statement has the following syntax:
7531 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
7534 The @code{asm} keyword is a GNU extension.
7535 When writing code that can be compiled with @option{-ansi} and the
7536 various @option{-std} options, use @code{__asm__} instead of
7537 @code{asm} (@pxref{Alternate Keywords}).
7539 @subsubheading Qualifiers
7542 The optional @code{volatile} qualifier has no effect.
7543 All basic @code{asm} blocks are implicitly volatile.
7546 @subsubheading Parameters
7549 @item AssemblerInstructions
7550 This is a literal string that specifies the assembler code. The string can
7551 contain any instructions recognized by the assembler, including directives.
7552 GCC does not parse the assembler instructions themselves and
7553 does not know what they mean or even whether they are valid assembler input.
7555 You may place multiple assembler instructions together in a single @code{asm}
7556 string, separated by the characters normally used in assembly code for the
7557 system. A combination that works in most places is a newline to break the
7558 line, plus a tab character (written as @samp{\n\t}).
7559 Some assemblers allow semicolons as a line separator. However,
7560 note that some assembler dialects use semicolons to start a comment.
7563 @subsubheading Remarks
7564 Using extended @code{asm} (@pxref{Extended Asm}) typically produces
7565 smaller, safer, and more efficient code, and in most cases it is a
7566 better solution than basic @code{asm}. However, there are two
7567 situations where only basic @code{asm} can be used:
7571 Extended @code{asm} statements have to be inside a C
7572 function, so to write inline assembly language at file scope (``top-level''),
7573 outside of C functions, you must use basic @code{asm}.
7574 You can use this technique to emit assembler directives,
7575 define assembly language macros that can be invoked elsewhere in the file,
7576 or write entire functions in assembly language.
7580 with the @code{naked} attribute also require basic @code{asm}
7581 (@pxref{Function Attributes}).
7584 Safely accessing C data and calling functions from basic @code{asm} is more
7585 complex than it may appear. To access C data, it is better to use extended
7588 Do not expect a sequence of @code{asm} statements to remain perfectly
7589 consecutive after compilation. If certain instructions need to remain
7590 consecutive in the output, put them in a single multi-instruction @code{asm}
7591 statement. Note that GCC's optimizers can move @code{asm} statements
7592 relative to other code, including across jumps.
7594 @code{asm} statements may not perform jumps into other @code{asm} statements.
7595 GCC does not know about these jumps, and therefore cannot take
7596 account of them when deciding how to optimize. Jumps from @code{asm} to C
7597 labels are only supported in extended @code{asm}.
7599 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7600 assembly code when optimizing. This can lead to unexpected duplicate
7601 symbol errors during compilation if your assembly code defines symbols or
7604 @strong{Warning:} The C standards do not specify semantics for @code{asm},
7605 making it a potential source of incompatibilities between compilers. These
7606 incompatibilities may not produce compiler warnings/errors.
7608 GCC does not parse basic @code{asm}'s @var{AssemblerInstructions}, which
7609 means there is no way to communicate to the compiler what is happening
7610 inside them. GCC has no visibility of symbols in the @code{asm} and may
7611 discard them as unreferenced. It also does not know about side effects of
7612 the assembler code, such as modifications to memory or registers. Unlike
7613 some compilers, GCC assumes that no changes to general purpose registers
7614 occur. This assumption may change in a future release.
7616 To avoid complications from future changes to the semantics and the
7617 compatibility issues between compilers, consider replacing basic @code{asm}
7618 with extended @code{asm}. See
7619 @uref{https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended, How to convert
7620 from basic asm to extended asm} for information about how to perform this
7623 The compiler copies the assembler instructions in a basic @code{asm}
7624 verbatim to the assembly language output file, without
7625 processing dialects or any of the @samp{%} operators that are available with
7626 extended @code{asm}. This results in minor differences between basic
7627 @code{asm} strings and extended @code{asm} templates. For example, to refer to
7628 registers you might use @samp{%eax} in basic @code{asm} and
7629 @samp{%%eax} in extended @code{asm}.
7631 On targets such as x86 that support multiple assembler dialects,
7632 all basic @code{asm} blocks use the assembler dialect specified by the
7633 @option{-masm} command-line option (@pxref{x86 Options}).
7634 Basic @code{asm} provides no
7635 mechanism to provide different assembler strings for different dialects.
7637 For basic @code{asm} with non-empty assembler string GCC assumes
7638 the assembler block does not change any general purpose registers,
7639 but it may read or write any globally accessible variable.
7641 Here is an example of basic @code{asm} for i386:
7644 /* Note that this code will not compile with -masm=intel */
7645 #define DebugBreak() asm("int $3")
7649 @subsection Extended Asm - Assembler Instructions with C Expression Operands
7650 @cindex extended @code{asm}
7651 @cindex assembly language in C, extended
7653 With extended @code{asm} you can read and write C variables from
7654 assembler and perform jumps from assembler code to C labels.
7655 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
7656 the operand parameters after the assembler template:
7659 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
7660 : @var{OutputOperands}
7661 @r{[} : @var{InputOperands}
7662 @r{[} : @var{Clobbers} @r{]} @r{]})
7664 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
7666 : @var{InputOperands}
7671 The @code{asm} keyword is a GNU extension.
7672 When writing code that can be compiled with @option{-ansi} and the
7673 various @option{-std} options, use @code{__asm__} instead of
7674 @code{asm} (@pxref{Alternate Keywords}).
7676 @subsubheading Qualifiers
7680 The typical use of extended @code{asm} statements is to manipulate input
7681 values to produce output values. However, your @code{asm} statements may
7682 also produce side effects. If so, you may need to use the @code{volatile}
7683 qualifier to disable certain optimizations. @xref{Volatile}.
7686 This qualifier informs the compiler that the @code{asm} statement may
7687 perform a jump to one of the labels listed in the @var{GotoLabels}.
7691 @subsubheading Parameters
7693 @item AssemblerTemplate
7694 This is a literal string that is the template for the assembler code. It is a
7695 combination of fixed text and tokens that refer to the input, output,
7696 and goto parameters. @xref{AssemblerTemplate}.
7698 @item OutputOperands
7699 A comma-separated list of the C variables modified by the instructions in the
7700 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
7703 A comma-separated list of C expressions read by the instructions in the
7704 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
7707 A comma-separated list of registers or other values changed by the
7708 @var{AssemblerTemplate}, beyond those listed as outputs.
7709 An empty list is permitted. @xref{Clobbers}.
7712 When you are using the @code{goto} form of @code{asm}, this section contains
7713 the list of all C labels to which the code in the
7714 @var{AssemblerTemplate} may jump.
7717 @code{asm} statements may not perform jumps into other @code{asm} statements,
7718 only to the listed @var{GotoLabels}.
7719 GCC's optimizers do not know about other jumps; therefore they cannot take
7720 account of them when deciding how to optimize.
7723 The total number of input + output + goto operands is limited to 30.
7725 @subsubheading Remarks
7726 The @code{asm} statement allows you to include assembly instructions directly
7727 within C code. This may help you to maximize performance in time-sensitive
7728 code or to access assembly instructions that are not readily available to C
7731 Note that extended @code{asm} statements must be inside a function. Only
7732 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
7733 Functions declared with the @code{naked} attribute also require basic
7734 @code{asm} (@pxref{Function Attributes}).
7736 While the uses of @code{asm} are many and varied, it may help to think of an
7737 @code{asm} statement as a series of low-level instructions that convert input
7738 parameters to output parameters. So a simple (if not particularly useful)
7739 example for i386 using @code{asm} might look like this:
7745 asm ("mov %1, %0\n\t"
7750 printf("%d\n", dst);
7753 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
7756 @subsubsection Volatile
7757 @cindex volatile @code{asm}
7758 @cindex @code{asm} volatile
7760 GCC's optimizers sometimes discard @code{asm} statements if they determine
7761 there is no need for the output variables. Also, the optimizers may move
7762 code out of loops if they believe that the code will always return the same
7763 result (i.e. none of its input values change between calls). Using the
7764 @code{volatile} qualifier disables these optimizations. @code{asm} statements
7765 that have no output operands, including @code{asm goto} statements,
7766 are implicitly volatile.
7768 This i386 code demonstrates a case that does not use (or require) the
7769 @code{volatile} qualifier. If it is performing assertion checking, this code
7770 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
7771 unreferenced by any code. As a result, the optimizers can discard the
7772 @code{asm} statement, which in turn removes the need for the entire
7773 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
7774 isn't needed you allow the optimizers to produce the most efficient code
7778 void DoCheck(uint32_t dwSomeValue)
7782 // Assumes dwSomeValue is not zero.
7792 The next example shows a case where the optimizers can recognize that the input
7793 (@code{dwSomeValue}) never changes during the execution of the function and can
7794 therefore move the @code{asm} outside the loop to produce more efficient code.
7795 Again, using @code{volatile} disables this type of optimization.
7798 void do_print(uint32_t dwSomeValue)
7802 for (uint32_t x=0; x < 5; x++)
7804 // Assumes dwSomeValue is not zero.
7810 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
7815 The following example demonstrates a case where you need to use the
7816 @code{volatile} qualifier.
7817 It uses the x86 @code{rdtsc} instruction, which reads
7818 the computer's time-stamp counter. Without the @code{volatile} qualifier,
7819 the optimizers might assume that the @code{asm} block will always return the
7820 same value and therefore optimize away the second call.
7825 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
7826 "shl $32, %%rdx\n\t" // Shift the upper bits left.
7827 "or %%rdx, %0" // 'Or' in the lower bits.
7832 printf("msr: %llx\n", msr);
7836 // Reprint the timestamp
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);
7847 GCC's optimizers do not treat this code like the non-volatile code in the
7848 earlier examples. They do not move it out of loops or omit it on the
7849 assumption that the result from a previous call is still valid.
7851 Note that the compiler can move even volatile @code{asm} instructions relative
7852 to other code, including across jump instructions. For example, on many
7853 targets there is a system register that controls the rounding mode of
7854 floating-point operations. Setting it with a volatile @code{asm}, as in the
7855 following PowerPC example, does not work reliably.
7858 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
7862 The compiler may move the addition back before the volatile @code{asm}. To
7863 make it work as expected, add an artificial dependency to the @code{asm} by
7864 referencing a variable in the subsequent code, for example:
7867 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
7871 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
7872 assembly code when optimizing. This can lead to unexpected duplicate symbol
7873 errors during compilation if your asm code defines symbols or labels.
7875 (@pxref{AssemblerTemplate}) may help resolve this problem.
7877 @anchor{AssemblerTemplate}
7878 @subsubsection Assembler Template
7879 @cindex @code{asm} assembler template
7881 An assembler template is a literal string containing assembler instructions.
7882 The compiler replaces tokens in the template that refer
7883 to inputs, outputs, and goto labels,
7884 and then outputs the resulting string to the assembler. The
7885 string can contain any instructions recognized by the assembler, including
7886 directives. GCC does not parse the assembler instructions
7887 themselves and does not know what they mean or even whether they are valid
7888 assembler input. However, it does count the statements
7889 (@pxref{Size of an asm}).
7891 You may place multiple assembler instructions together in a single @code{asm}
7892 string, separated by the characters normally used in assembly code for the
7893 system. A combination that works in most places is a newline to break the
7894 line, plus a tab character to move to the instruction field (written as
7896 Some assemblers allow semicolons as a line separator. However, note
7897 that some assembler dialects use semicolons to start a comment.
7899 Do not expect a sequence of @code{asm} statements to remain perfectly
7900 consecutive after compilation, even when you are using the @code{volatile}
7901 qualifier. If certain instructions need to remain consecutive in the output,
7902 put them in a single multi-instruction asm statement.
7904 Accessing data from C programs without using input/output operands (such as
7905 by using global symbols directly from the assembler template) may not work as
7906 expected. Similarly, calling functions directly from an assembler template
7907 requires a detailed understanding of the target assembler and ABI.
7909 Since GCC does not parse the assembler template,
7910 it has no visibility of any
7911 symbols it references. This may result in GCC discarding those symbols as
7912 unreferenced unless they are also listed as input, output, or goto operands.
7914 @subsubheading Special format strings
7916 In addition to the tokens described by the input, output, and goto operands,
7917 these tokens have special meanings in the assembler template:
7921 Outputs a single @samp{%} into the assembler code.
7924 Outputs a number that is unique to each instance of the @code{asm}
7925 statement in the entire compilation. This option is useful when creating local
7926 labels and referring to them multiple times in a single template that
7927 generates multiple assembler instructions.
7932 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
7933 into the assembler code. When unescaped, these characters have special
7934 meaning to indicate multiple assembler dialects, as described below.
7937 @subsubheading Multiple assembler dialects in @code{asm} templates
7939 On targets such as x86, GCC supports multiple assembler dialects.
7940 The @option{-masm} option controls which dialect GCC uses as its
7941 default for inline assembler. The target-specific documentation for the
7942 @option{-masm} option contains the list of supported dialects, as well as the
7943 default dialect if the option is not specified. This information may be
7944 important to understand, since assembler code that works correctly when
7945 compiled using one dialect will likely fail if compiled using another.
7948 If your code needs to support multiple assembler dialects (for example, if
7949 you are writing public headers that need to support a variety of compilation
7950 options), use constructs of this form:
7953 @{ dialect0 | dialect1 | dialect2... @}
7956 This construct outputs @code{dialect0}
7957 when using dialect #0 to compile the code,
7958 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
7959 braces than the number of dialects the compiler supports, the construct
7962 For example, if an x86 compiler supports two dialects
7963 (@samp{att}, @samp{intel}), an
7964 assembler template such as this:
7967 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
7971 is equivalent to one of
7974 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
7975 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
7978 Using that same compiler, this code:
7981 "xchg@{l@}\t@{%%@}ebx, %1"
7985 corresponds to either
7988 "xchgl\t%%ebx, %1" @r{/* att dialect */}
7989 "xchg\tebx, %1" @r{/* intel dialect */}
7992 There is no support for nesting dialect alternatives.
7994 @anchor{OutputOperands}
7995 @subsubsection Output Operands
7996 @cindex @code{asm} output operands
7998 An @code{asm} statement has zero or more output operands indicating the names
7999 of C variables modified by the assembler code.
8001 In this i386 example, @code{old} (referred to in the template string as
8002 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
8003 (@code{%2}) is an input:
8008 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
8009 "sbb %0,%0" // Use the CF to calculate old.
8010 : "=r" (old), "+rm" (*Base)
8017 Operands are separated by commas. Each operand has this format:
8020 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
8024 @item asmSymbolicName
8025 Specifies a symbolic name for the operand.
8026 Reference the name in the assembler template
8027 by enclosing it in square brackets
8028 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8029 that contains the definition. Any valid C variable name is acceptable,
8030 including names already defined in the surrounding code. No two operands
8031 within the same @code{asm} statement can use the same symbolic name.
8033 When not using an @var{asmSymbolicName}, use the (zero-based) position
8035 in the list of operands in the assembler template. For example if there are
8036 three output operands, use @samp{%0} in the template to refer to the first,
8037 @samp{%1} for the second, and @samp{%2} for the third.
8040 A string constant specifying constraints on the placement of the operand;
8041 @xref{Constraints}, for details.
8043 Output constraints must begin with either @samp{=} (a variable overwriting an
8044 existing value) or @samp{+} (when reading and writing). When using
8045 @samp{=}, do not assume the location contains the existing value
8046 on entry to the @code{asm}, except
8047 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
8049 After the prefix, there must be one or more additional constraints
8050 (@pxref{Constraints}) that describe where the value resides. Common
8051 constraints include @samp{r} for register and @samp{m} for memory.
8052 When you list more than one possible location (for example, @code{"=rm"}),
8053 the compiler chooses the most efficient one based on the current context.
8054 If you list as many alternates as the @code{asm} statement allows, you permit
8055 the optimizers to produce the best possible code.
8056 If you must use a specific register, but your Machine Constraints do not
8057 provide sufficient control to select the specific register you want,
8058 local register variables may provide a solution (@pxref{Local Register
8062 Specifies a C lvalue expression to hold the output, typically a variable name.
8063 The enclosing parentheses are a required part of the syntax.
8067 When the compiler selects the registers to use to
8068 represent the output operands, it does not use any of the clobbered registers
8071 Output operand expressions must be lvalues. The compiler cannot check whether
8072 the operands have data types that are reasonable for the instruction being
8073 executed. For output expressions that are not directly addressable (for
8074 example a bit-field), the constraint must allow a register. In that case, GCC
8075 uses the register as the output of the @code{asm}, and then stores that
8076 register into the output.
8078 Operands using the @samp{+} constraint modifier count as two operands
8079 (that is, both as input and output) towards the total maximum of 30 operands
8080 per @code{asm} statement.
8082 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
8083 operands that must not overlap an input. Otherwise,
8084 GCC may allocate the output operand in the same register as an unrelated
8085 input operand, on the assumption that the assembler code consumes its
8086 inputs before producing outputs. This assumption may be false if the assembler
8087 code actually consists of more than one instruction.
8089 The same problem can occur if one output parameter (@var{a}) allows a register
8090 constraint and another output parameter (@var{b}) allows a memory constraint.
8091 The code generated by GCC to access the memory address in @var{b} can contain
8092 registers which @emph{might} be shared by @var{a}, and GCC considers those
8093 registers to be inputs to the asm. As above, GCC assumes that such input
8094 registers are consumed before any outputs are written. This assumption may
8095 result in incorrect behavior if the asm writes to @var{a} before using
8096 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
8097 ensures that modifying @var{a} does not affect the address referenced by
8098 @var{b}. Otherwise, the location of @var{b}
8099 is undefined if @var{a} is modified before using @var{b}.
8101 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8102 instead of simply @samp{%2}). Typically these qualifiers are hardware
8103 dependent. The list of supported modifiers for x86 is found at
8104 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8106 If the C code that follows the @code{asm} makes no use of any of the output
8107 operands, use @code{volatile} for the @code{asm} statement to prevent the
8108 optimizers from discarding the @code{asm} statement as unneeded
8109 (see @ref{Volatile}).
8111 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
8112 references the first output operand as @code{%0} (were there a second, it
8113 would be @code{%1}, etc). The number of the first input operand is one greater
8114 than that of the last output operand. In this i386 example, that makes
8115 @code{Mask} referenced as @code{%1}:
8118 uint32_t Mask = 1234;
8127 That code overwrites the variable @code{Index} (@samp{=}),
8128 placing the value in a register (@samp{r}).
8129 Using the generic @samp{r} constraint instead of a constraint for a specific
8130 register allows the compiler to pick the register to use, which can result
8131 in more efficient code. This may not be possible if an assembler instruction
8132 requires a specific register.
8134 The following i386 example uses the @var{asmSymbolicName} syntax.
8136 same result as the code above, but some may consider it more readable or more
8137 maintainable since reordering index numbers is not necessary when adding or
8138 removing operands. The names @code{aIndex} and @code{aMask}
8139 are only used in this example to emphasize which
8140 names get used where.
8141 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8144 uint32_t Mask = 1234;
8147 asm ("bsfl %[aMask], %[aIndex]"
8148 : [aIndex] "=r" (Index)
8149 : [aMask] "r" (Mask)
8153 Here are some more examples of output operands.
8160 asm ("mov %[e], %[d]"
8165 Here, @code{d} may either be in a register or in memory. Since the compiler
8166 might already have the current value of the @code{uint32_t} location
8167 pointed to by @code{e}
8168 in a register, you can enable it to choose the best location
8169 for @code{d} by specifying both constraints.
8171 @anchor{FlagOutputOperands}
8172 @subsubsection Flag Output Operands
8173 @cindex @code{asm} flag output operands
8175 Some targets have a special register that holds the ``flags'' for the
8176 result of an operation or comparison. Normally, the contents of that
8177 register are either unmodifed by the asm, or the asm is considered to
8178 clobber the contents.
8180 On some targets, a special form of output operand exists by which
8181 conditions in the flags register may be outputs of the asm. The set of
8182 conditions supported are target specific, but the general rule is that
8183 the output variable must be a scalar integer, and the value is boolean.
8184 When supported, the target defines the preprocessor symbol
8185 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8187 Because of the special nature of the flag output operands, the constraint
8188 may not include alternatives.
8190 Most often, the target has only one flags register, and thus is an implied
8191 operand of many instructions. In this case, the operand should not be
8192 referenced within the assembler template via @code{%0} etc, as there's
8193 no corresponding text in the assembly language.
8197 The flag output constraints for the x86 family are of the form
8198 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8199 conditions defined in the ISA manual for @code{j@var{cc}} or
8204 ``above'' or unsigned greater than
8206 ``above or equal'' or unsigned greater than or equal
8208 ``below'' or unsigned less than
8210 ``below or equal'' or unsigned less than or equal
8215 ``equal'' or zero flag set
8219 signed greater than or equal
8223 signed less than or equal
8244 ``not'' @var{flag}, or inverted versions of those above
8249 @anchor{InputOperands}
8250 @subsubsection Input Operands
8251 @cindex @code{asm} input operands
8252 @cindex @code{asm} expressions
8254 Input operands make values from C variables and expressions available to the
8257 Operands are separated by commas. Each operand has this format:
8260 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8264 @item asmSymbolicName
8265 Specifies a symbolic name for the operand.
8266 Reference the name in the assembler template
8267 by enclosing it in square brackets
8268 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8269 that contains the definition. Any valid C variable name is acceptable,
8270 including names already defined in the surrounding code. No two operands
8271 within the same @code{asm} statement can use the same symbolic name.
8273 When not using an @var{asmSymbolicName}, use the (zero-based) position
8275 in the list of operands in the assembler template. For example if there are
8276 two output operands and three inputs,
8277 use @samp{%2} in the template to refer to the first input operand,
8278 @samp{%3} for the second, and @samp{%4} for the third.
8281 A string constant specifying constraints on the placement of the operand;
8282 @xref{Constraints}, for details.
8284 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8285 When you list more than one possible location (for example, @samp{"irm"}),
8286 the compiler chooses the most efficient one based on the current context.
8287 If you must use a specific register, but your Machine Constraints do not
8288 provide sufficient control to select the specific register you want,
8289 local register variables may provide a solution (@pxref{Local Register
8292 Input constraints can also be digits (for example, @code{"0"}). This indicates
8293 that the specified input must be in the same place as the output constraint
8294 at the (zero-based) index in the output constraint list.
8295 When using @var{asmSymbolicName} syntax for the output operands,
8296 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8299 This is the C variable or expression being passed to the @code{asm} statement
8300 as input. The enclosing parentheses are a required part of the syntax.
8304 When the compiler selects the registers to use to represent the input
8305 operands, it does not use any of the clobbered registers (@pxref{Clobbers}).
8307 If there are no output operands but there are input operands, place two
8308 consecutive colons where the output operands would go:
8311 __asm__ ("some instructions"
8313 : "r" (Offset / 8));
8316 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8317 (except for inputs tied to outputs). The compiler assumes that on exit from
8318 the @code{asm} statement these operands contain the same values as they
8319 had before executing the statement.
8320 It is @emph{not} possible to use clobbers
8321 to inform the compiler that the values in these inputs are changing. One
8322 common work-around is to tie the changing input variable to an output variable
8323 that never gets used. Note, however, that if the code that follows the
8324 @code{asm} statement makes no use of any of the output operands, the GCC
8325 optimizers may discard the @code{asm} statement as unneeded
8326 (see @ref{Volatile}).
8328 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8329 instead of simply @samp{%2}). Typically these qualifiers are hardware
8330 dependent. The list of supported modifiers for x86 is found at
8331 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8333 In this example using the fictitious @code{combine} instruction, the
8334 constraint @code{"0"} for input operand 1 says that it must occupy the same
8335 location as output operand 0. Only input operands may use numbers in
8336 constraints, and they must each refer to an output operand. Only a number (or
8337 the symbolic assembler name) in the constraint can guarantee that one operand
8338 is in the same place as another. The mere fact that @code{foo} is the value of
8339 both operands is not enough to guarantee that they are in the same place in
8340 the generated assembler code.
8343 asm ("combine %2, %0"
8345 : "0" (foo), "g" (bar));
8348 Here is an example using symbolic names.
8351 asm ("cmoveq %1, %2, %[result]"
8352 : [result] "=r"(result)
8353 : "r" (test), "r" (new), "[result]" (old));
8357 @subsubsection Clobbers
8358 @cindex @code{asm} clobbers
8360 While the compiler is aware of changes to entries listed in the output
8361 operands, the inline @code{asm} code may modify more than just the outputs. For
8362 example, calculations may require additional registers, or the processor may
8363 overwrite a register as a side effect of a particular assembler instruction.
8364 In order to inform the compiler of these changes, list them in the clobber
8365 list. Clobber list items are either register names or the special clobbers
8366 (listed below). Each clobber list item is a string constant
8367 enclosed in double quotes and separated by commas.
8369 Clobber descriptions may not in any way overlap with an input or output
8370 operand. For example, you may not have an operand describing a register class
8371 with one member when listing that register in the clobber list. Variables
8372 declared to live in specific registers (@pxref{Explicit Register
8373 Variables}) and used
8374 as @code{asm} input or output operands must have no part mentioned in the
8375 clobber description. In particular, there is no way to specify that input
8376 operands get modified without also specifying them as output operands.
8378 When the compiler selects which registers to use to represent input and output
8379 operands, it does not use any of the clobbered registers. As a result,
8380 clobbered registers are available for any use in the assembler code.
8382 Here is a realistic example for the VAX showing the use of clobbered
8386 asm volatile ("movc3 %0, %1, %2"
8388 : "g" (from), "g" (to), "g" (count)
8389 : "r0", "r1", "r2", "r3", "r4", "r5");
8392 Also, there are two special clobber arguments:
8396 The @code{"cc"} clobber indicates that the assembler code modifies the flags
8397 register. On some machines, GCC represents the condition codes as a specific
8398 hardware register; @code{"cc"} serves to name this register.
8399 On other machines, condition code handling is different,
8400 and specifying @code{"cc"} has no effect. But
8401 it is valid no matter what the target.
8404 The @code{"memory"} clobber tells the compiler that the assembly code
8406 reads or writes to items other than those listed in the input and output
8407 operands (for example, accessing the memory pointed to by one of the input
8408 parameters). To ensure memory contains correct values, GCC may need to flush
8409 specific register values to memory before executing the @code{asm}. Further,
8410 the compiler does not assume that any values read from memory before an
8411 @code{asm} remain unchanged after that @code{asm}; it reloads them as
8413 Using the @code{"memory"} clobber effectively forms a read/write
8414 memory barrier for the compiler.
8416 Note that this clobber does not prevent the @emph{processor} from doing
8417 speculative reads past the @code{asm} statement. To prevent that, you need
8418 processor-specific fence instructions.
8420 Flushing registers to memory has performance implications and may be an issue
8421 for time-sensitive code. You can use a trick to avoid this if the size of
8422 the memory being accessed is known at compile time. For example, if accessing
8423 ten bytes of a string, use a memory input like:
8425 @code{@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}}.
8430 @subsubsection Goto Labels
8431 @cindex @code{asm} goto labels
8433 @code{asm goto} allows assembly code to jump to one or more C labels. The
8434 @var{GotoLabels} section in an @code{asm goto} statement contains
8436 list of all C labels to which the assembler code may jump. GCC assumes that
8437 @code{asm} execution falls through to the next statement (if this is not the
8438 case, consider using the @code{__builtin_unreachable} intrinsic after the
8439 @code{asm} statement). Optimization of @code{asm goto} may be improved by
8440 using the @code{hot} and @code{cold} label attributes (@pxref{Label
8443 An @code{asm goto} statement cannot have outputs.
8444 This is due to an internal restriction of
8445 the compiler: control transfer instructions cannot have outputs.
8446 If the assembler code does modify anything, use the @code{"memory"} clobber
8448 optimizers to flush all register values to memory and reload them if
8449 necessary after the @code{asm} statement.
8451 Also note that an @code{asm goto} statement is always implicitly
8452 considered volatile.
8454 To reference a label in the assembler template,
8455 prefix it with @samp{%l} (lowercase @samp{L}) followed
8456 by its (zero-based) position in @var{GotoLabels} plus the number of input
8457 operands. For example, if the @code{asm} has three inputs and references two
8458 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
8460 Alternately, you can reference labels using the actual C label name enclosed
8461 in brackets. For example, to reference a label named @code{carry}, you can
8462 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
8463 section when using this approach.
8465 Here is an example of @code{asm goto} for i386:
8472 : "r" (p1), "r" (p2)
8482 The following example shows an @code{asm goto} that uses a memory clobber.
8488 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
8499 @anchor{x86Operandmodifiers}
8500 @subsubsection x86 Operand Modifiers
8502 References to input, output, and goto operands in the assembler template
8503 of extended @code{asm} statements can use
8504 modifiers to affect the way the operands are formatted in
8505 the code output to the assembler. For example, the
8506 following code uses the @samp{h} and @samp{b} modifiers for x86:
8510 asm volatile ("xchg %h0, %b0" : "+a" (num) );
8514 These modifiers generate this assembler code:
8520 The rest of this discussion uses the following code for illustrative purposes.
8529 asm volatile goto ("some assembler instructions here"
8531 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
8532 : /* No clobbers. */
8537 With no modifiers, this is what the output from the operands would be for the
8538 @samp{att} and @samp{intel} dialects of assembler:
8540 @multitable {Operand} {masm=att} {OFFSET FLAT:.L2}
8541 @headitem Operand @tab masm=att @tab masm=intel
8550 @tab @code{OFFSET FLAT:.L2}
8553 The table below shows the list of supported modifiers and their effects.
8555 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {masm=att} {masm=intel}
8556 @headitem Modifier @tab Description @tab Operand @tab @option{masm=att} @tab @option{masm=intel}
8558 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
8563 @tab Print the QImode name of the register.
8568 @tab Print the QImode name for a ``high'' register.
8573 @tab Print the HImode name of the register.
8578 @tab Print the SImode name of the register.
8583 @tab Print the DImode name of the register.
8588 @tab Print the label name with no punctuation.
8593 @tab Require a constant operand and print the constant expression with no punctuation.
8599 @anchor{x86floatingpointasmoperands}
8600 @subsubsection x86 Floating-Point @code{asm} Operands
8602 On x86 targets, there are several rules on the usage of stack-like registers
8603 in the operands of an @code{asm}. These rules apply only to the operands
8604 that are stack-like registers:
8608 Given a set of input registers that die in an @code{asm}, it is
8609 necessary to know which are implicitly popped by the @code{asm}, and
8610 which must be explicitly popped by GCC@.
8612 An input register that is implicitly popped by the @code{asm} must be
8613 explicitly clobbered, unless it is constrained to match an
8617 For any input register that is implicitly popped by an @code{asm}, it is
8618 necessary to know how to adjust the stack to compensate for the pop.
8619 If any non-popped input is closer to the top of the reg-stack than
8620 the implicitly popped register, it would not be possible to know what the
8621 stack looked like---it's not clear how the rest of the stack ``slides
8624 All implicitly popped input registers must be closer to the top of
8625 the reg-stack than any input that is not implicitly popped.
8627 It is possible that if an input dies in an @code{asm}, the compiler might
8628 use the input register for an output reload. Consider this example:
8631 asm ("foo" : "=t" (a) : "f" (b));
8635 This code says that input @code{b} is not popped by the @code{asm}, and that
8636 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
8637 deeper after the @code{asm} than it was before. But, it is possible that
8638 reload may think that it can use the same register for both the input and
8641 To prevent this from happening,
8642 if any input operand uses the @samp{f} constraint, all output register
8643 constraints must use the @samp{&} early-clobber modifier.
8645 The example above is correctly written as:
8648 asm ("foo" : "=&t" (a) : "f" (b));
8652 Some operands need to be in particular places on the stack. All
8653 output operands fall in this category---GCC has no other way to
8654 know which registers the outputs appear in unless you indicate
8655 this in the constraints.
8657 Output operands must specifically indicate which register an output
8658 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
8659 constraints must select a class with a single register.
8662 Output operands may not be ``inserted'' between existing stack registers.
8663 Since no 387 opcode uses a read/write operand, all output operands
8664 are dead before the @code{asm}, and are pushed by the @code{asm}.
8665 It makes no sense to push anywhere but the top of the reg-stack.
8667 Output operands must start at the top of the reg-stack: output
8668 operands may not ``skip'' a register.
8671 Some @code{asm} statements may need extra stack space for internal
8672 calculations. This can be guaranteed by clobbering stack registers
8673 unrelated to the inputs and outputs.
8678 takes one input, which is internally popped, and produces two outputs.
8681 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
8685 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
8686 and replaces them with one output. The @code{st(1)} clobber is necessary
8687 for the compiler to know that @code{fyl2xp1} pops both inputs.
8690 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
8698 @subsection Controlling Names Used in Assembler Code
8699 @cindex assembler names for identifiers
8700 @cindex names used in assembler code
8701 @cindex identifiers, names in assembler code
8703 You can specify the name to be used in the assembler code for a C
8704 function or variable by writing the @code{asm} (or @code{__asm__})
8705 keyword after the declarator.
8706 It is up to you to make sure that the assembler names you choose do not
8707 conflict with any other assembler symbols, or reference registers.
8709 @subsubheading Assembler names for data:
8711 This sample shows how to specify the assembler name for data:
8714 int foo asm ("myfoo") = 2;
8718 This specifies that the name to be used for the variable @code{foo} in
8719 the assembler code should be @samp{myfoo} rather than the usual
8722 On systems where an underscore is normally prepended to the name of a C
8723 variable, this feature allows you to define names for the
8724 linker that do not start with an underscore.
8726 GCC does not support using this feature with a non-static local variable
8727 since such variables do not have assembler names. If you are
8728 trying to put the variable in a particular register, see
8729 @ref{Explicit Register Variables}.
8731 @subsubheading Assembler names for functions:
8733 To specify the assembler name for functions, write a declaration for the
8734 function before its definition and put @code{asm} there, like this:
8737 int func (int x, int y) asm ("MYFUNC");
8739 int func (int x, int y)
8745 This specifies that the name to be used for the function @code{func} in
8746 the assembler code should be @code{MYFUNC}.
8748 @node Explicit Register Variables
8749 @subsection Variables in Specified Registers
8750 @anchor{Explicit Reg Vars}
8751 @cindex explicit register variables
8752 @cindex variables in specified registers
8753 @cindex specified registers
8755 GNU C allows you to associate specific hardware registers with C
8756 variables. In almost all cases, allowing the compiler to assign
8757 registers produces the best code. However under certain unusual
8758 circumstances, more precise control over the variable storage is
8761 Both global and local variables can be associated with a register. The
8762 consequences of performing this association are very different between
8763 the two, as explained in the sections below.
8766 * Global Register Variables:: Variables declared at global scope.
8767 * Local Register Variables:: Variables declared within a function.
8770 @node Global Register Variables
8771 @subsubsection Defining Global Register Variables
8772 @anchor{Global Reg Vars}
8773 @cindex global register variables
8774 @cindex registers, global variables in
8775 @cindex registers, global allocation
8777 You can define a global register variable and associate it with a specified
8781 register int *foo asm ("r12");
8785 Here @code{r12} is the name of the register that should be used. Note that
8786 this is the same syntax used for defining local register variables, but for
8787 a global variable the declaration appears outside a function. The
8788 @code{register} keyword is required, and cannot be combined with
8789 @code{static}. The register name must be a valid register name for the
8792 Registers are a scarce resource on most systems and allowing the
8793 compiler to manage their usage usually results in the best code. However,
8794 under special circumstances it can make sense to reserve some globally.
8795 For example this may be useful in programs such as programming language
8796 interpreters that have a couple of global variables that are accessed
8799 After defining a global register variable, for the current compilation
8803 @item The register is reserved entirely for this use, and will not be
8804 allocated for any other purpose.
8805 @item The register is not saved and restored by any functions.
8806 @item Stores into this register are never deleted even if they appear to be
8807 dead, but references may be deleted, moved or simplified.
8810 Note that these points @emph{only} apply to code that is compiled with the
8811 definition. The behavior of code that is merely linked in (for example
8812 code from libraries) is not affected.
8814 If you want to recompile source files that do not actually use your global
8815 register variable so they do not use the specified register for any other
8816 purpose, you need not actually add the global register declaration to
8817 their source code. It suffices to specify the compiler option
8818 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
8821 @subsubheading Declaring the variable
8823 Global register variables can not have initial values, because an
8824 executable file has no means to supply initial contents for a register.
8826 When selecting a register, choose one that is normally saved and
8827 restored by function calls on your machine. This ensures that code
8828 which is unaware of this reservation (such as library routines) will
8829 restore it before returning.
8831 On machines with register windows, be sure to choose a global
8832 register that is not affected magically by the function call mechanism.
8834 @subsubheading Using the variable
8836 @cindex @code{qsort}, and global register variables
8837 When calling routines that are not aware of the reservation, be
8838 cautious if those routines call back into code which uses them. As an
8839 example, if you call the system library version of @code{qsort}, it may
8840 clobber your registers during execution, but (if you have selected
8841 appropriate registers) it will restore them before returning. However
8842 it will @emph{not} restore them before calling @code{qsort}'s comparison
8843 function. As a result, global values will not reliably be available to
8844 the comparison function unless the @code{qsort} function itself is rebuilt.
8846 Similarly, it is not safe to access the global register variables from signal
8847 handlers or from more than one thread of control. Unless you recompile
8848 them specially for the task at hand, the system library routines may
8849 temporarily use the register for other things.
8851 @cindex register variable after @code{longjmp}
8852 @cindex global register after @code{longjmp}
8853 @cindex value after @code{longjmp}
8856 On most machines, @code{longjmp} restores to each global register
8857 variable the value it had at the time of the @code{setjmp}. On some
8858 machines, however, @code{longjmp} does not change the value of global
8859 register variables. To be portable, the function that called @code{setjmp}
8860 should make other arrangements to save the values of the global register
8861 variables, and to restore them in a @code{longjmp}. This way, the same
8862 thing happens regardless of what @code{longjmp} does.
8864 Eventually there may be a way of asking the compiler to choose a register
8865 automatically, but first we need to figure out how it should choose and
8866 how to enable you to guide the choice. No solution is evident.
8868 @node Local Register Variables
8869 @subsubsection Specifying Registers for Local Variables
8870 @anchor{Local Reg Vars}
8871 @cindex local variables, specifying registers
8872 @cindex specifying registers for local variables
8873 @cindex registers for local variables
8875 You can define a local register variable and associate it with a specified
8879 register int *foo asm ("r12");
8883 Here @code{r12} is the name of the register that should be used. Note
8884 that this is the same syntax used for defining global register variables,
8885 but for a local variable the declaration appears within a function. The
8886 @code{register} keyword is required, and cannot be combined with
8887 @code{static}. The register name must be a valid register name for the
8890 As with global register variables, it is recommended that you choose
8891 a register that is normally saved and restored by function calls on your
8892 machine, so that calls to library routines will not clobber it.
8894 The only supported use for this feature is to specify registers
8895 for input and output operands when calling Extended @code{asm}
8896 (@pxref{Extended Asm}). This may be necessary if the constraints for a
8897 particular machine don't provide sufficient control to select the desired
8898 register. To force an operand into a register, create a local variable
8899 and specify the register name after the variable's declaration. Then use
8900 the local variable for the @code{asm} operand and specify any constraint
8901 letter that matches the register:
8904 register int *p1 asm ("r0") = @dots{};
8905 register int *p2 asm ("r1") = @dots{};
8906 register int *result asm ("r0");
8907 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8910 @emph{Warning:} In the above example, be aware that a register (for example
8911 @code{r0}) can be call-clobbered by subsequent code, including function
8912 calls and library calls for arithmetic operators on other variables (for
8913 example the initialization of @code{p2}). In this case, use temporary
8914 variables for expressions between the register assignments:
8918 register int *p1 asm ("r0") = @dots{};
8919 register int *p2 asm ("r1") = t1;
8920 register int *result asm ("r0");
8921 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
8924 Defining a register variable does not reserve the register. Other than
8925 when invoking the Extended @code{asm}, the contents of the specified
8926 register are not guaranteed. For this reason, the following uses
8927 are explicitly @emph{not} supported. If they appear to work, it is only
8928 happenstance, and may stop working as intended due to (seemingly)
8929 unrelated changes in surrounding code, or even minor changes in the
8930 optimization of a future version of gcc:
8933 @item Passing parameters to or from Basic @code{asm}
8934 @item Passing parameters to or from Extended @code{asm} without using input
8936 @item Passing parameters to or from routines written in assembler (or
8937 other languages) using non-standard calling conventions.
8940 Some developers use Local Register Variables in an attempt to improve
8941 gcc's allocation of registers, especially in large functions. In this
8942 case the register name is essentially a hint to the register allocator.
8943 While in some instances this can generate better code, improvements are
8944 subject to the whims of the allocator/optimizers. Since there are no
8945 guarantees that your improvements won't be lost, this usage of Local
8946 Register Variables is discouraged.
8948 On the MIPS platform, there is related use for local register variables
8949 with slightly different characteristics (@pxref{MIPS Coprocessors,,
8950 Defining coprocessor specifics for MIPS targets, gccint,
8951 GNU Compiler Collection (GCC) Internals}).
8953 @node Size of an asm
8954 @subsection Size of an @code{asm}
8956 Some targets require that GCC track the size of each instruction used
8957 in order to generate correct code. Because the final length of the
8958 code produced by an @code{asm} statement is only known by the
8959 assembler, GCC must make an estimate as to how big it will be. It
8960 does this by counting the number of instructions in the pattern of the
8961 @code{asm} and multiplying that by the length of the longest
8962 instruction supported by that processor. (When working out the number
8963 of instructions, it assumes that any occurrence of a newline or of
8964 whatever statement separator character is supported by the assembler --
8965 typically @samp{;} --- indicates the end of an instruction.)
8967 Normally, GCC's estimate is adequate to ensure that correct
8968 code is generated, but it is possible to confuse the compiler if you use
8969 pseudo instructions or assembler macros that expand into multiple real
8970 instructions, or if you use assembler directives that expand to more
8971 space in the object file than is needed for a single instruction.
8972 If this happens then the assembler may produce a diagnostic saying that
8973 a label is unreachable.
8975 @node Alternate Keywords
8976 @section Alternate Keywords
8977 @cindex alternate keywords
8978 @cindex keywords, alternate
8980 @option{-ansi} and the various @option{-std} options disable certain
8981 keywords. This causes trouble when you want to use GNU C extensions, or
8982 a general-purpose header file that should be usable by all programs,
8983 including ISO C programs. The keywords @code{asm}, @code{typeof} and
8984 @code{inline} are not available in programs compiled with
8985 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
8986 program compiled with @option{-std=c99} or @option{-std=c11}). The
8988 @code{restrict} is only available when @option{-std=gnu99} (which will
8989 eventually be the default) or @option{-std=c99} (or the equivalent
8990 @option{-std=iso9899:1999}), or an option for a later standard
8993 The way to solve these problems is to put @samp{__} at the beginning and
8994 end of each problematical keyword. For example, use @code{__asm__}
8995 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
8997 Other C compilers won't accept these alternative keywords; if you want to
8998 compile with another compiler, you can define the alternate keywords as
8999 macros to replace them with the customary keywords. It looks like this:
9007 @findex __extension__
9009 @option{-pedantic} and other options cause warnings for many GNU C extensions.
9011 prevent such warnings within one expression by writing
9012 @code{__extension__} before the expression. @code{__extension__} has no
9013 effect aside from this.
9015 @node Incomplete Enums
9016 @section Incomplete @code{enum} Types
9018 You can define an @code{enum} tag without specifying its possible values.
9019 This results in an incomplete type, much like what you get if you write
9020 @code{struct foo} without describing the elements. A later declaration
9021 that does specify the possible values completes the type.
9023 You can't allocate variables or storage using the type while it is
9024 incomplete. However, you can work with pointers to that type.
9026 This extension may not be very useful, but it makes the handling of
9027 @code{enum} more consistent with the way @code{struct} and @code{union}
9030 This extension is not supported by GNU C++.
9032 @node Function Names
9033 @section Function Names as Strings
9034 @cindex @code{__func__} identifier
9035 @cindex @code{__FUNCTION__} identifier
9036 @cindex @code{__PRETTY_FUNCTION__} identifier
9038 GCC provides three magic constants that hold the name of the current
9039 function as a string. In C++11 and later modes, all three are treated
9040 as constant expressions and can be used in @code{constexpr} constexts.
9041 The first of these constants is @code{__func__}, which is part of
9044 The identifier @code{__func__} is implicitly declared by the translator
9045 as if, immediately following the opening brace of each function
9046 definition, the declaration
9049 static const char __func__[] = "function-name";
9053 appeared, where function-name is the name of the lexically-enclosing
9054 function. This name is the unadorned name of the function. As an
9055 extension, at file (or, in C++, namespace scope), @code{__func__}
9056 evaluates to the empty string.
9058 @code{__FUNCTION__} is another name for @code{__func__}, provided for
9059 backward compatibility with old versions of GCC.
9061 In C, @code{__PRETTY_FUNCTION__} is yet another name for
9062 @code{__func__}, except that at file (or, in C++, namespace scope),
9063 it evaluates to the string @code{"top level"}. In addition, in C++,
9064 @code{__PRETTY_FUNCTION__} contains the signature of the function as
9065 well as its bare name. For example, this program:
9068 extern "C" int printf (const char *, ...);
9074 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
9075 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
9093 __PRETTY_FUNCTION__ = void a::sub(int)
9096 These identifiers are variables, not preprocessor macros, and may not
9097 be used to initialize @code{char} arrays or be concatenated with string
9100 @node Return Address
9101 @section Getting the Return or Frame Address of a Function
9103 These functions may be used to get information about the callers of a
9106 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
9107 This function returns the return address of the current function, or of
9108 one of its callers. The @var{level} argument is number of frames to
9109 scan up the call stack. A value of @code{0} yields the return address
9110 of the current function, a value of @code{1} yields the return address
9111 of the caller of the current function, and so forth. When inlining
9112 the expected behavior is that the function returns the address of
9113 the function that is returned to. To work around this behavior use
9114 the @code{noinline} function attribute.
9116 The @var{level} argument must be a constant integer.
9118 On some machines it may be impossible to determine the return address of
9119 any function other than the current one; in such cases, or when the top
9120 of the stack has been reached, this function returns @code{0} or a
9121 random value. In addition, @code{__builtin_frame_address} may be used
9122 to determine if the top of the stack has been reached.
9124 Additional post-processing of the returned value may be needed, see
9125 @code{__builtin_extract_return_addr}.
9127 Calling this function with a nonzero argument can have unpredictable
9128 effects, including crashing the calling program. As a result, calls
9129 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9130 option is in effect. Such calls should only be made in debugging
9134 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9135 The address as returned by @code{__builtin_return_address} may have to be fed
9136 through this function to get the actual encoded address. For example, on the
9137 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9138 platforms an offset has to be added for the true next instruction to be
9141 If no fixup is needed, this function simply passes through @var{addr}.
9144 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9145 This function does the reverse of @code{__builtin_extract_return_addr}.
9148 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9149 This function is similar to @code{__builtin_return_address}, but it
9150 returns the address of the function frame rather than the return address
9151 of the function. Calling @code{__builtin_frame_address} with a value of
9152 @code{0} yields the frame address of the current function, a value of
9153 @code{1} yields the frame address of the caller of the current function,
9156 The frame is the area on the stack that holds local variables and saved
9157 registers. The frame address is normally the address of the first word
9158 pushed on to the stack by the function. However, the exact definition
9159 depends upon the processor and the calling convention. If the processor
9160 has a dedicated frame pointer register, and the function has a frame,
9161 then @code{__builtin_frame_address} returns the value of the frame
9164 On some machines it may be impossible to determine the frame address of
9165 any function other than the current one; in such cases, or when the top
9166 of the stack has been reached, this function returns @code{0} if
9167 the first frame pointer is properly initialized by the startup code.
9169 Calling this function with a nonzero argument can have unpredictable
9170 effects, including crashing the calling program. As a result, calls
9171 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9172 option is in effect. Such calls should only be made in debugging
9176 @node Vector Extensions
9177 @section Using Vector Instructions through Built-in Functions
9179 On some targets, the instruction set contains SIMD vector instructions which
9180 operate on multiple values contained in one large register at the same time.
9181 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9184 The first step in using these extensions is to provide the necessary data
9185 types. This should be done using an appropriate @code{typedef}:
9188 typedef int v4si __attribute__ ((vector_size (16)));
9192 The @code{int} type specifies the base type, while the attribute specifies
9193 the vector size for the variable, measured in bytes. For example, the
9194 declaration above causes the compiler to set the mode for the @code{v4si}
9195 type to be 16 bytes wide and divided into @code{int} sized units. For
9196 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9197 corresponding mode of @code{foo} is @acronym{V4SI}.
9199 The @code{vector_size} attribute is only applicable to integral and
9200 float scalars, although arrays, pointers, and function return values
9201 are allowed in conjunction with this construct. Only sizes that are
9202 a power of two are currently allowed.
9204 All the basic integer types can be used as base types, both as signed
9205 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9206 @code{long long}. In addition, @code{float} and @code{double} can be
9207 used to build floating-point vector types.
9209 Specifying a combination that is not valid for the current architecture
9210 causes GCC to synthesize the instructions using a narrower mode.
9211 For example, if you specify a variable of type @code{V4SI} and your
9212 architecture does not allow for this specific SIMD type, GCC
9213 produces code that uses 4 @code{SIs}.
9215 The types defined in this manner can be used with a subset of normal C
9216 operations. Currently, GCC allows using the following operators
9217 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9219 The operations behave like C++ @code{valarrays}. Addition is defined as
9220 the addition of the corresponding elements of the operands. For
9221 example, in the code below, each of the 4 elements in @var{a} is
9222 added to the corresponding 4 elements in @var{b} and the resulting
9223 vector is stored in @var{c}.
9226 typedef int v4si __attribute__ ((vector_size (16)));
9233 Subtraction, multiplication, division, and the logical operations
9234 operate in a similar manner. Likewise, the result of using the unary
9235 minus or complement operators on a vector type is a vector whose
9236 elements are the negative or complemented values of the corresponding
9237 elements in the operand.
9239 It is possible to use shifting operators @code{<<}, @code{>>} on
9240 integer-type vectors. The operation is defined as following: @code{@{a0,
9241 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9242 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9245 For convenience, it is allowed to use a binary vector operation
9246 where one operand is a scalar. In that case the compiler transforms
9247 the scalar operand into a vector where each element is the scalar from
9248 the operation. The transformation happens only if the scalar could be
9249 safely converted to the vector-element type.
9250 Consider the following code.
9253 typedef int v4si __attribute__ ((vector_size (16)));
9258 a = b + 1; /* a = b + @{1,1,1,1@}; */
9259 a = 2 * b; /* a = @{2,2,2,2@} * b; */
9261 a = l + a; /* Error, cannot convert long to int. */
9264 Vectors can be subscripted as if the vector were an array with
9265 the same number of elements and base type. Out of bound accesses
9266 invoke undefined behavior at run time. Warnings for out of bound
9267 accesses for vector subscription can be enabled with
9268 @option{-Warray-bounds}.
9270 Vector comparison is supported with standard comparison
9271 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
9272 vector expressions of integer-type or real-type. Comparison between
9273 integer-type vectors and real-type vectors are not supported. The
9274 result of the comparison is a vector of the same width and number of
9275 elements as the comparison operands with a signed integral element
9278 Vectors are compared element-wise producing 0 when comparison is false
9279 and -1 (constant of the appropriate type where all bits are set)
9280 otherwise. Consider the following example.
9283 typedef int v4si __attribute__ ((vector_size (16)));
9285 v4si a = @{1,2,3,4@};
9286 v4si b = @{3,2,1,4@};
9289 c = a > b; /* The result would be @{0, 0,-1, 0@} */
9290 c = a == b; /* The result would be @{0,-1, 0,-1@} */
9293 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
9294 @code{b} and @code{c} are vectors of the same type and @code{a} is an
9295 integer vector with the same number of elements of the same size as @code{b}
9296 and @code{c}, computes all three arguments and creates a vector
9297 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
9298 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
9299 As in the case of binary operations, this syntax is also accepted when
9300 one of @code{b} or @code{c} is a scalar that is then transformed into a
9301 vector. If both @code{b} and @code{c} are scalars and the type of
9302 @code{true?b:c} has the same size as the element type of @code{a}, then
9303 @code{b} and @code{c} are converted to a vector type whose elements have
9304 this type and with the same number of elements as @code{a}.
9306 In C++, the logic operators @code{!, &&, ||} are available for vectors.
9307 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
9308 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
9309 For mixed operations between a scalar @code{s} and a vector @code{v},
9310 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
9311 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
9313 Vector shuffling is available using functions
9314 @code{__builtin_shuffle (vec, mask)} and
9315 @code{__builtin_shuffle (vec0, vec1, mask)}.
9316 Both functions construct a permutation of elements from one or two
9317 vectors and return a vector of the same type as the input vector(s).
9318 The @var{mask} is an integral vector with the same width (@var{W})
9319 and element count (@var{N}) as the output vector.
9321 The elements of the input vectors are numbered in memory ordering of
9322 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
9323 elements of @var{mask} are considered modulo @var{N} in the single-operand
9324 case and modulo @math{2*@var{N}} in the two-operand case.
9326 Consider the following example,
9329 typedef int v4si __attribute__ ((vector_size (16)));
9331 v4si a = @{1,2,3,4@};
9332 v4si b = @{5,6,7,8@};
9333 v4si mask1 = @{0,1,1,3@};
9334 v4si mask2 = @{0,4,2,5@};
9337 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
9338 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
9341 Note that @code{__builtin_shuffle} is intentionally semantically
9342 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
9344 You can declare variables and use them in function calls and returns, as
9345 well as in assignments and some casts. You can specify a vector type as
9346 a return type for a function. Vector types can also be used as function
9347 arguments. It is possible to cast from one vector type to another,
9348 provided they are of the same size (in fact, you can also cast vectors
9349 to and from other datatypes of the same size).
9351 You cannot operate between vectors of different lengths or different
9352 signedness without a cast.
9355 @section Support for @code{offsetof}
9356 @findex __builtin_offsetof
9358 GCC implements for both C and C++ a syntactic extension to implement
9359 the @code{offsetof} macro.
9363 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
9365 offsetof_member_designator:
9367 | offsetof_member_designator "." @code{identifier}
9368 | offsetof_member_designator "[" @code{expr} "]"
9371 This extension is sufficient such that
9374 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
9378 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
9379 may be dependent. In either case, @var{member} may consist of a single
9380 identifier, or a sequence of member accesses and array references.
9382 @node __sync Builtins
9383 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
9385 The following built-in functions
9386 are intended to be compatible with those described
9387 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
9388 section 7.4. As such, they depart from normal GCC practice by not using
9389 the @samp{__builtin_} prefix and also by being overloaded so that they
9390 work on multiple types.
9392 The definition given in the Intel documentation allows only for the use of
9393 the types @code{int}, @code{long}, @code{long long} or their unsigned
9394 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
9395 size other than the C type @code{_Bool} or the C++ type @code{bool}.
9396 Operations on pointer arguments are performed as if the operands were
9397 of the @code{uintptr_t} type. That is, they are not scaled by the size
9398 of the type to which the pointer points.
9400 These functions are implemented in terms of the @samp{__atomic}
9401 builtins (@pxref{__atomic Builtins}). They should not be used for new
9402 code which should use the @samp{__atomic} builtins instead.
9404 Not all operations are supported by all target processors. If a particular
9405 operation cannot be implemented on the target processor, a warning is
9406 generated and a call to an external function is generated. The external
9407 function carries the same name as the built-in version,
9408 with an additional suffix
9409 @samp{_@var{n}} where @var{n} is the size of the data type.
9411 @c ??? Should we have a mechanism to suppress this warning? This is almost
9412 @c useful for implementing the operation under the control of an external
9415 In most cases, these built-in functions are considered a @dfn{full barrier}.
9417 no memory operand is moved across the operation, either forward or
9418 backward. Further, instructions are issued as necessary to prevent the
9419 processor from speculating loads across the operation and from queuing stores
9420 after the operation.
9422 All of the routines are described in the Intel documentation to take
9423 ``an optional list of variables protected by the memory barrier''. It's
9424 not clear what is meant by that; it could mean that @emph{only} the
9425 listed variables are protected, or it could mean a list of additional
9426 variables to be protected. The list is ignored by GCC which treats it as
9427 empty. GCC interprets an empty list as meaning that all globally
9428 accessible variables should be protected.
9431 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
9432 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
9433 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
9434 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
9435 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
9436 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
9437 @findex __sync_fetch_and_add
9438 @findex __sync_fetch_and_sub
9439 @findex __sync_fetch_and_or
9440 @findex __sync_fetch_and_and
9441 @findex __sync_fetch_and_xor
9442 @findex __sync_fetch_and_nand
9443 These built-in functions perform the operation suggested by the name, and
9444 returns the value that had previously been in memory. That is, operations
9445 on integer operands have the following semantics. Operations on pointer
9446 arguments are performed as if the operands were of the @code{uintptr_t}
9447 type. That is, they are not scaled by the size of the type to which
9451 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
9452 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
9455 The object pointed to by the first argument must be of integer or pointer
9456 type. It must not be a Boolean type.
9458 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
9459 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
9461 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
9462 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
9463 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
9464 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
9465 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
9466 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
9467 @findex __sync_add_and_fetch
9468 @findex __sync_sub_and_fetch
9469 @findex __sync_or_and_fetch
9470 @findex __sync_and_and_fetch
9471 @findex __sync_xor_and_fetch
9472 @findex __sync_nand_and_fetch
9473 These built-in functions perform the operation suggested by the name, and
9474 return the new value. That is, operations on integer operands have
9475 the following semantics. Operations on pointer operands are performed as
9476 if the operand's type were @code{uintptr_t}.
9479 @{ *ptr @var{op}= value; return *ptr; @}
9480 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
9483 The same constraints on arguments apply as for the corresponding
9484 @code{__sync_op_and_fetch} built-in functions.
9486 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
9487 as @code{*ptr = ~(*ptr & value)} instead of
9488 @code{*ptr = ~*ptr & value}.
9490 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9491 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
9492 @findex __sync_bool_compare_and_swap
9493 @findex __sync_val_compare_and_swap
9494 These built-in functions perform an atomic compare and swap.
9495 That is, if the current
9496 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
9499 The ``bool'' version returns true if the comparison is successful and
9500 @var{newval} is written. The ``val'' version returns the contents
9501 of @code{*@var{ptr}} before the operation.
9503 @item __sync_synchronize (...)
9504 @findex __sync_synchronize
9505 This built-in function issues a full memory barrier.
9507 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
9508 @findex __sync_lock_test_and_set
9509 This built-in function, as described by Intel, is not a traditional test-and-set
9510 operation, but rather an atomic exchange operation. It writes @var{value}
9511 into @code{*@var{ptr}}, and returns the previous contents of
9514 Many targets have only minimal support for such locks, and do not support
9515 a full exchange operation. In this case, a target may support reduced
9516 functionality here by which the @emph{only} valid value to store is the
9517 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
9518 is implementation defined.
9520 This built-in function is not a full barrier,
9521 but rather an @dfn{acquire barrier}.
9522 This means that references after the operation cannot move to (or be
9523 speculated to) before the operation, but previous memory stores may not
9524 be globally visible yet, and previous memory loads may not yet be
9527 @item void __sync_lock_release (@var{type} *ptr, ...)
9528 @findex __sync_lock_release
9529 This built-in function releases the lock acquired by
9530 @code{__sync_lock_test_and_set}.
9531 Normally this means writing the constant 0 to @code{*@var{ptr}}.
9533 This built-in function is not a full barrier,
9534 but rather a @dfn{release barrier}.
9535 This means that all previous memory stores are globally visible, and all
9536 previous memory loads have been satisfied, but following memory reads
9537 are not prevented from being speculated to before the barrier.
9540 @node __atomic Builtins
9541 @section Built-in Functions for Memory Model Aware Atomic Operations
9543 The following built-in functions approximately match the requirements
9544 for the C++11 memory model. They are all
9545 identified by being prefixed with @samp{__atomic} and most are
9546 overloaded so that they work with multiple types.
9548 These functions are intended to replace the legacy @samp{__sync}
9549 builtins. The main difference is that the memory order that is requested
9550 is a parameter to the functions. New code should always use the
9551 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
9553 Note that the @samp{__atomic} builtins assume that programs will
9554 conform to the C++11 memory model. In particular, they assume
9555 that programs are free of data races. See the C++11 standard for
9556 detailed requirements.
9558 The @samp{__atomic} builtins can be used with any integral scalar or
9559 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
9560 types are also allowed if @samp{__int128} (@pxref{__int128}) is
9561 supported by the architecture.
9563 The four non-arithmetic functions (load, store, exchange, and
9564 compare_exchange) all have a generic version as well. This generic
9565 version works on any data type. It uses the lock-free built-in function
9566 if the specific data type size makes that possible; otherwise, an
9567 external call is left to be resolved at run time. This external call is
9568 the same format with the addition of a @samp{size_t} parameter inserted
9569 as the first parameter indicating the size of the object being pointed to.
9570 All objects must be the same size.
9572 There are 6 different memory orders that can be specified. These map
9573 to the C++11 memory orders with the same names, see the C++11 standard
9574 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
9575 on atomic synchronization} for detailed definitions. Individual
9576 targets may also support additional memory orders for use on specific
9577 architectures. Refer to the target documentation for details of
9580 An atomic operation can both constrain code motion and
9581 be mapped to hardware instructions for synchronization between threads
9582 (e.g., a fence). To which extent this happens is controlled by the
9583 memory orders, which are listed here in approximately ascending order of
9584 strength. The description of each memory order is only meant to roughly
9585 illustrate the effects and is not a specification; see the C++11
9586 memory model for precise semantics.
9589 @item __ATOMIC_RELAXED
9590 Implies no inter-thread ordering constraints.
9591 @item __ATOMIC_CONSUME
9592 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
9593 memory order because of a deficiency in C++11's semantics for
9594 @code{memory_order_consume}.
9595 @item __ATOMIC_ACQUIRE
9596 Creates an inter-thread happens-before constraint from the release (or
9597 stronger) semantic store to this acquire load. Can prevent hoisting
9598 of code to before the operation.
9599 @item __ATOMIC_RELEASE
9600 Creates an inter-thread happens-before constraint to acquire (or stronger)
9601 semantic loads that read from this release store. Can prevent sinking
9602 of code to after the operation.
9603 @item __ATOMIC_ACQ_REL
9604 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
9605 @code{__ATOMIC_RELEASE}.
9606 @item __ATOMIC_SEQ_CST
9607 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
9610 Note that in the C++11 memory model, @emph{fences} (e.g.,
9611 @samp{__atomic_thread_fence}) take effect in combination with other
9612 atomic operations on specific memory locations (e.g., atomic loads);
9613 operations on specific memory locations do not necessarily affect other
9614 operations in the same way.
9616 Target architectures are encouraged to provide their own patterns for
9617 each of the atomic built-in functions. If no target is provided, the original
9618 non-memory model set of @samp{__sync} atomic built-in functions are
9619 used, along with any required synchronization fences surrounding it in
9620 order to achieve the proper behavior. Execution in this case is subject
9621 to the same restrictions as those built-in functions.
9623 If there is no pattern or mechanism to provide a lock-free instruction
9624 sequence, a call is made to an external routine with the same parameters
9625 to be resolved at run time.
9627 When implementing patterns for these built-in functions, the memory order
9628 parameter can be ignored as long as the pattern implements the most
9629 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
9630 orders execute correctly with this memory order but they may not execute as
9631 efficiently as they could with a more appropriate implementation of the
9632 relaxed requirements.
9634 Note that the C++11 standard allows for the memory order parameter to be
9635 determined at run time rather than at compile time. These built-in
9636 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
9637 than invoke a runtime library call or inline a switch statement. This is
9638 standard compliant, safe, and the simplest approach for now.
9640 The memory order parameter is a signed int, but only the lower 16 bits are
9641 reserved for the memory order. The remainder of the signed int is reserved
9642 for target use and should be 0. Use of the predefined atomic values
9643 ensures proper usage.
9645 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
9646 This built-in function implements an atomic load operation. It returns the
9647 contents of @code{*@var{ptr}}.
9649 The valid memory order variants are
9650 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9651 and @code{__ATOMIC_CONSUME}.
9655 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
9656 This is the generic version of an atomic load. It returns the
9657 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
9661 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
9662 This built-in function implements an atomic store operation. It writes
9663 @code{@var{val}} into @code{*@var{ptr}}.
9665 The valid memory order variants are
9666 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
9670 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
9671 This is the generic version of an atomic store. It stores the value
9672 of @code{*@var{val}} into @code{*@var{ptr}}.
9676 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
9677 This built-in function implements an atomic exchange operation. It writes
9678 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
9681 The valid memory order variants are
9682 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
9683 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
9687 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
9688 This is the generic version of an atomic exchange. It stores the
9689 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
9690 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
9694 @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)
9695 This built-in function implements an atomic compare and exchange operation.
9696 This compares the contents of @code{*@var{ptr}} with the contents of
9697 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
9698 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
9699 equal, the operation is a @emph{read} and the current contents of
9700 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
9701 for weak compare_exchange, which may fail spuriously, and false for
9702 the strong variation, which never fails spuriously. Many targets
9703 only offer the strong variation and ignore the parameter. When in doubt, use
9704 the strong variation.
9706 If @var{desired} is written into @code{*@var{ptr}} then true is returned
9707 and memory is affected according to the
9708 memory order specified by @var{success_memorder}. There are no
9709 restrictions on what memory order can be used here.
9711 Otherwise, false is returned and memory is affected according
9712 to @var{failure_memorder}. This memory order cannot be
9713 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
9714 stronger order than that specified by @var{success_memorder}.
9718 @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)
9719 This built-in function implements the generic version of
9720 @code{__atomic_compare_exchange}. The function is virtually identical to
9721 @code{__atomic_compare_exchange_n}, except the desired value is also a
9726 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
9727 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
9728 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
9729 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
9730 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
9731 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
9732 These built-in functions perform the operation suggested by the name, and
9733 return the result of the operation. Operations on pointer arguments are
9734 performed as if the operands were of the @code{uintptr_t} type. That is,
9735 they are not scaled by the size of the type to which the pointer points.
9738 @{ *ptr @var{op}= val; return *ptr; @}
9741 The object pointed to by the first argument must be of integer or pointer
9742 type. It must not be a Boolean type. All memory orders are valid.
9746 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
9747 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
9748 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
9749 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
9750 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
9751 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
9752 These built-in functions perform the operation suggested by the name, and
9753 return the value that had previously been in @code{*@var{ptr}}. Operations
9754 on pointer arguments are performed as if the operands were of
9755 the @code{uintptr_t} type. That is, they are not scaled by the size of
9756 the type to which the pointer points.
9759 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
9762 The same constraints on arguments apply as for the corresponding
9763 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
9767 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
9769 This built-in function performs an atomic test-and-set operation on
9770 the byte at @code{*@var{ptr}}. The byte is set to some implementation
9771 defined nonzero ``set'' value and the return value is @code{true} if and only
9772 if the previous contents were ``set''.
9773 It should be only used for operands of type @code{bool} or @code{char}. For
9774 other types only part of the value may be set.
9776 All memory orders are valid.
9780 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
9782 This built-in function performs an atomic clear operation on
9783 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
9784 It should be only used for operands of type @code{bool} or @code{char} and
9785 in conjunction with @code{__atomic_test_and_set}.
9786 For other types it may only clear partially. If the type is not @code{bool}
9787 prefer using @code{__atomic_store}.
9789 The valid memory order variants are
9790 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
9791 @code{__ATOMIC_RELEASE}.
9795 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
9797 This built-in function acts as a synchronization fence between threads
9798 based on the specified memory order.
9800 All memory orders are valid.
9804 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
9806 This built-in function acts as a synchronization fence between a thread
9807 and signal handlers based in the same thread.
9809 All memory orders are valid.
9813 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
9815 This built-in function returns true if objects of @var{size} bytes always
9816 generate lock-free atomic instructions for the target architecture.
9817 @var{size} must resolve to a compile-time constant and the result also
9818 resolves to a compile-time constant.
9820 @var{ptr} is an optional pointer to the object that may be used to determine
9821 alignment. A value of 0 indicates typical alignment should be used. The
9822 compiler may also ignore this parameter.
9825 if (__atomic_always_lock_free (sizeof (long long), 0))
9830 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
9832 This built-in function returns true if objects of @var{size} bytes always
9833 generate lock-free atomic instructions for the target architecture. If
9834 the built-in function is not known to be lock-free, a call is made to a
9835 runtime routine named @code{__atomic_is_lock_free}.
9837 @var{ptr} is an optional pointer to the object that may be used to determine
9838 alignment. A value of 0 indicates typical alignment should be used. The
9839 compiler may also ignore this parameter.
9842 @node Integer Overflow Builtins
9843 @section Built-in Functions to Perform Arithmetic with Overflow Checking
9845 The following built-in functions allow performing simple arithmetic operations
9846 together with checking whether the operations overflowed.
9848 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9849 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
9850 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
9851 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long int *res)
9852 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
9853 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9854 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9856 These built-in functions promote the first two operands into infinite precision signed
9857 type and perform addition on those promoted operands. The result is then
9858 cast to the type the third pointer argument points to and stored there.
9859 If the stored result is equal to the infinite precision result, the built-in
9860 functions return false, otherwise they return true. As the addition is
9861 performed in infinite signed precision, these built-in functions have fully defined
9862 behavior for all argument values.
9864 The first built-in function allows arbitrary integral types for operands and
9865 the result type must be pointer to some integral type other than enumerated or
9866 Boolean type, the rest of the built-in functions have explicit integer types.
9868 The compiler will attempt to use hardware instructions to implement
9869 these built-in functions where possible, like conditional jump on overflow
9870 after addition, conditional jump on carry etc.
9874 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9875 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
9876 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
9877 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long int *res)
9878 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
9879 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9880 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9882 These built-in functions are similar to the add overflow checking built-in
9883 functions above, except they perform subtraction, subtract the second argument
9884 from the first one, instead of addition.
9888 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
9889 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
9890 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
9891 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long int *res)
9892 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
9893 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
9894 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long int *res)
9896 These built-in functions are similar to the add overflow checking built-in
9897 functions above, except they perform multiplication, instead of addition.
9901 The following built-in functions allow checking if simple arithmetic operation
9904 @deftypefn {Built-in Function} bool __builtin_add_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9905 @deftypefnx {Built-in Function} bool __builtin_sub_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9906 @deftypefnx {Built-in Function} bool __builtin_mul_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
9908 These built-in functions are similar to @code{__builtin_add_overflow},
9909 @code{__builtin_sub_overflow}, or @code{__builtin_mul_overflow}, except that
9910 they don't store the result of the arithmetic operation anywhere and the
9911 last argument is not a pointer, but some expression with integral type other
9912 than enumerated or Boolean type.
9914 The built-in functions promote the first two operands into infinite precision signed type
9915 and perform addition on those promoted operands. The result is then
9916 cast to the type of the third argument. If the cast result is equal to the infinite
9917 precision result, the built-in functions return false, otherwise they return true.
9918 The value of the third argument is ignored, just the side-effects in the third argument
9919 are evaluated, and no integral argument promotions are performed on the last argument.
9920 If the third argument is a bit-field, the type used for the result cast has the
9921 precision and signedness of the given bit-field, rather than precision and signedness
9922 of the underlying type.
9924 For example, the following macro can be used to portably check, at
9925 compile-time, whether or not adding two constant integers will overflow,
9926 and perform the addition only when it is known to be safe and not to trigger
9927 a @option{-Woverflow} warning.
9930 #define INT_ADD_OVERFLOW_P(a, b) \
9931 __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0)
9935 C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B,
9936 D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0)
9940 The compiler will attempt to use hardware instructions to implement
9941 these built-in functions where possible, like conditional jump on overflow
9942 after addition, conditional jump on carry etc.
9946 @node x86 specific memory model extensions for transactional memory
9947 @section x86-Specific Memory Model Extensions for Transactional Memory
9949 The x86 architecture supports additional memory ordering flags
9950 to mark lock critical sections for hardware lock elision.
9951 These must be specified in addition to an existing memory order to
9955 @item __ATOMIC_HLE_ACQUIRE
9956 Start lock elision on a lock variable.
9957 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
9958 @item __ATOMIC_HLE_RELEASE
9959 End lock elision on a lock variable.
9960 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
9963 When a lock acquire fails, it is required for good performance to abort
9964 the transaction quickly. This can be done with a @code{_mm_pause}.
9967 #include <immintrin.h> // For _mm_pause
9971 /* Acquire lock with lock elision */
9972 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
9973 _mm_pause(); /* Abort failed transaction */
9975 /* Free lock with lock elision */
9976 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
9979 @node Object Size Checking
9980 @section Object Size Checking Built-in Functions
9981 @findex __builtin_object_size
9982 @findex __builtin___memcpy_chk
9983 @findex __builtin___mempcpy_chk
9984 @findex __builtin___memmove_chk
9985 @findex __builtin___memset_chk
9986 @findex __builtin___strcpy_chk
9987 @findex __builtin___stpcpy_chk
9988 @findex __builtin___strncpy_chk
9989 @findex __builtin___strcat_chk
9990 @findex __builtin___strncat_chk
9991 @findex __builtin___sprintf_chk
9992 @findex __builtin___snprintf_chk
9993 @findex __builtin___vsprintf_chk
9994 @findex __builtin___vsnprintf_chk
9995 @findex __builtin___printf_chk
9996 @findex __builtin___vprintf_chk
9997 @findex __builtin___fprintf_chk
9998 @findex __builtin___vfprintf_chk
10000 GCC implements a limited buffer overflow protection mechanism
10001 that can prevent some buffer overflow attacks.
10003 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
10004 is a built-in construct that returns a constant number of bytes from
10005 @var{ptr} to the end of the object @var{ptr} pointer points to
10006 (if known at compile time). @code{__builtin_object_size} never evaluates
10007 its arguments for side-effects. If there are any side-effects in them, it
10008 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10009 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
10010 point to and all of them are known at compile time, the returned number
10011 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
10012 0 and minimum if nonzero. If it is not possible to determine which objects
10013 @var{ptr} points to at compile time, @code{__builtin_object_size} should
10014 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10015 for @var{type} 2 or 3.
10017 @var{type} is an integer constant from 0 to 3. If the least significant
10018 bit is clear, objects are whole variables, if it is set, a closest
10019 surrounding subobject is considered the object a pointer points to.
10020 The second bit determines if maximum or minimum of remaining bytes
10024 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
10025 char *p = &var.buf1[1], *q = &var.b;
10027 /* Here the object p points to is var. */
10028 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
10029 /* The subobject p points to is var.buf1. */
10030 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
10031 /* The object q points to is var. */
10032 assert (__builtin_object_size (q, 0)
10033 == (char *) (&var + 1) - (char *) &var.b);
10034 /* The subobject q points to is var.b. */
10035 assert (__builtin_object_size (q, 1) == sizeof (var.b));
10039 There are built-in functions added for many common string operation
10040 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
10041 built-in is provided. This built-in has an additional last argument,
10042 which is the number of bytes remaining in object the @var{dest}
10043 argument points to or @code{(size_t) -1} if the size is not known.
10045 The built-in functions are optimized into the normal string functions
10046 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
10047 it is known at compile time that the destination object will not
10048 be overflown. If the compiler can determine at compile time the
10049 object will be always overflown, it issues a warning.
10051 The intended use can be e.g.@:
10055 #define bos0(dest) __builtin_object_size (dest, 0)
10056 #define memcpy(dest, src, n) \
10057 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
10061 /* It is unknown what object p points to, so this is optimized
10062 into plain memcpy - no checking is possible. */
10063 memcpy (p, "abcde", n);
10064 /* Destination is known and length too. It is known at compile
10065 time there will be no overflow. */
10066 memcpy (&buf[5], "abcde", 5);
10067 /* Destination is known, but the length is not known at compile time.
10068 This will result in __memcpy_chk call that can check for overflow
10070 memcpy (&buf[5], "abcde", n);
10071 /* Destination is known and it is known at compile time there will
10072 be overflow. There will be a warning and __memcpy_chk call that
10073 will abort the program at run time. */
10074 memcpy (&buf[6], "abcde", 5);
10077 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
10078 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
10079 @code{strcat} and @code{strncat}.
10081 There are also checking built-in functions for formatted output functions.
10083 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
10084 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10085 const char *fmt, ...);
10086 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
10088 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10089 const char *fmt, va_list ap);
10092 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
10093 etc.@: functions and can contain implementation specific flags on what
10094 additional security measures the checking function might take, such as
10095 handling @code{%n} differently.
10097 The @var{os} argument is the object size @var{s} points to, like in the
10098 other built-in functions. There is a small difference in the behavior
10099 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
10100 optimized into the non-checking functions only if @var{flag} is 0, otherwise
10101 the checking function is called with @var{os} argument set to
10102 @code{(size_t) -1}.
10104 In addition to this, there are checking built-in functions
10105 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
10106 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
10107 These have just one additional argument, @var{flag}, right before
10108 format string @var{fmt}. If the compiler is able to optimize them to
10109 @code{fputc} etc.@: functions, it does, otherwise the checking function
10110 is called and the @var{flag} argument passed to it.
10112 @node Pointer Bounds Checker builtins
10113 @section Pointer Bounds Checker Built-in Functions
10114 @cindex Pointer Bounds Checker builtins
10115 @findex __builtin___bnd_set_ptr_bounds
10116 @findex __builtin___bnd_narrow_ptr_bounds
10117 @findex __builtin___bnd_copy_ptr_bounds
10118 @findex __builtin___bnd_init_ptr_bounds
10119 @findex __builtin___bnd_null_ptr_bounds
10120 @findex __builtin___bnd_store_ptr_bounds
10121 @findex __builtin___bnd_chk_ptr_lbounds
10122 @findex __builtin___bnd_chk_ptr_ubounds
10123 @findex __builtin___bnd_chk_ptr_bounds
10124 @findex __builtin___bnd_get_ptr_lbound
10125 @findex __builtin___bnd_get_ptr_ubound
10127 GCC provides a set of built-in functions to control Pointer Bounds Checker
10128 instrumentation. Note that all Pointer Bounds Checker builtins can be used
10129 even if you compile with Pointer Bounds Checker off
10130 (@option{-fno-check-pointer-bounds}).
10131 The behavior may differ in such case as documented below.
10133 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
10135 This built-in function returns a new pointer with the value of @var{q}, and
10136 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
10137 Bounds Checker off, the built-in function just returns the first argument.
10140 extern void *__wrap_malloc (size_t n)
10142 void *p = (void *)__real_malloc (n);
10143 if (!p) return __builtin___bnd_null_ptr_bounds (p);
10144 return __builtin___bnd_set_ptr_bounds (p, n);
10150 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
10152 This built-in function returns a new pointer with the value of @var{p}
10153 and associates it with the narrowed bounds formed by the intersection
10154 of bounds associated with @var{q} and the bounds
10155 [@var{p}, @var{p} + @var{size} - 1].
10156 With Pointer Bounds Checker off, the built-in function just returns the first
10160 void init_objects (object *objs, size_t size)
10163 /* Initialize objects one-by-one passing pointers with bounds of
10164 an object, not the full array of objects. */
10165 for (i = 0; i < size; i++)
10166 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
10173 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
10175 This built-in function returns a new pointer with the value of @var{q},
10176 and associates it with the bounds already associated with pointer @var{r}.
10177 With Pointer Bounds Checker off, the built-in function just returns the first
10181 /* Here is a way to get pointer to object's field but
10182 still with the full object's bounds. */
10183 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10189 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10191 This built-in function returns a new pointer with the value of @var{q}, and
10192 associates it with INIT (allowing full memory access) bounds. With Pointer
10193 Bounds Checker off, the built-in function just returns the first argument.
10197 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10199 This built-in function returns a new pointer with the value of @var{q}, and
10200 associates it with NULL (allowing no memory access) bounds. With Pointer
10201 Bounds Checker off, the built-in function just returns the first argument.
10205 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10207 This built-in function stores the bounds associated with pointer @var{ptr_val}
10208 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10209 bounds from legacy code without touching the associated pointer's memory when
10210 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10211 function call is ignored.
10215 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10217 This built-in function checks if the pointer @var{q} is within the lower
10218 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10219 function call is ignored.
10222 extern void *__wrap_memset (void *dst, int c, size_t len)
10226 __builtin___bnd_chk_ptr_lbounds (dst);
10227 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10228 __real_memset (dst, c, len);
10236 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10238 This built-in function checks if the pointer @var{q} is within the upper
10239 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10240 function call is ignored.
10244 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10246 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10247 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10248 off, the built-in function call is ignored.
10251 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
10255 __bnd_chk_ptr_bounds (dst, n);
10256 __bnd_chk_ptr_bounds (src, n);
10257 __real_memcpy (dst, src, n);
10265 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
10267 This built-in function returns the lower bound associated
10268 with the pointer @var{q}, as a pointer value.
10269 This is useful for debugging using @code{printf}.
10270 With Pointer Bounds Checker off, the built-in function returns 0.
10273 void *lb = __builtin___bnd_get_ptr_lbound (q);
10274 void *ub = __builtin___bnd_get_ptr_ubound (q);
10275 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
10280 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
10282 This built-in function returns the upper bound (which is a pointer) associated
10283 with the pointer @var{q}. With Pointer Bounds Checker off,
10284 the built-in function returns -1.
10288 @node Cilk Plus Builtins
10289 @section Cilk Plus C/C++ Language Extension Built-in Functions
10291 GCC provides support for the following built-in reduction functions if Cilk Plus
10292 is enabled. Cilk Plus can be enabled using the @option{-fcilkplus} flag.
10295 @item @code{__sec_implicit_index}
10296 @item @code{__sec_reduce}
10297 @item @code{__sec_reduce_add}
10298 @item @code{__sec_reduce_all_nonzero}
10299 @item @code{__sec_reduce_all_zero}
10300 @item @code{__sec_reduce_any_nonzero}
10301 @item @code{__sec_reduce_any_zero}
10302 @item @code{__sec_reduce_max}
10303 @item @code{__sec_reduce_min}
10304 @item @code{__sec_reduce_max_ind}
10305 @item @code{__sec_reduce_min_ind}
10306 @item @code{__sec_reduce_mul}
10307 @item @code{__sec_reduce_mutating}
10310 Further details and examples about these built-in functions are described
10311 in the Cilk Plus language manual which can be found at
10312 @uref{http://www.cilkplus.org}.
10314 @node Other Builtins
10315 @section Other Built-in Functions Provided by GCC
10316 @cindex built-in functions
10317 @findex __builtin_alloca
10318 @findex __builtin_alloca_with_align
10319 @findex __builtin_call_with_static_chain
10320 @findex __builtin_fpclassify
10321 @findex __builtin_isfinite
10322 @findex __builtin_isnormal
10323 @findex __builtin_isgreater
10324 @findex __builtin_isgreaterequal
10325 @findex __builtin_isinf_sign
10326 @findex __builtin_isless
10327 @findex __builtin_islessequal
10328 @findex __builtin_islessgreater
10329 @findex __builtin_isunordered
10330 @findex __builtin_powi
10331 @findex __builtin_powif
10332 @findex __builtin_powil
10493 @findex fprintf_unlocked
10495 @findex fputs_unlocked
10603 @findex nexttowardf
10604 @findex nexttowardl
10612 @findex printf_unlocked
10642 @findex signbitd128
10643 @findex significand
10644 @findex significandf
10645 @findex significandl
10673 @findex strncasecmp
10716 GCC provides a large number of built-in functions other than the ones
10717 mentioned above. Some of these are for internal use in the processing
10718 of exceptions or variable-length argument lists and are not
10719 documented here because they may change from time to time; we do not
10720 recommend general use of these functions.
10722 The remaining functions are provided for optimization purposes.
10724 With the exception of built-ins that have library equivalents such as
10725 the standard C library functions discussed below, or that expand to
10726 library calls, GCC built-in functions are always expanded inline and
10727 thus do not have corresponding entry points and their address cannot
10728 be obtained. Attempting to use them in an expression other than
10729 a function call results in a compile-time error.
10731 @opindex fno-builtin
10732 GCC includes built-in versions of many of the functions in the standard
10733 C library. These functions come in two forms: one whose names start with
10734 the @code{__builtin_} prefix, and the other without. Both forms have the
10735 same type (including prototype), the same address (when their address is
10736 taken), and the same meaning as the C library functions even if you specify
10737 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
10738 functions are only optimized in certain cases; if they are not optimized in
10739 a particular case, a call to the library function is emitted.
10743 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
10744 @option{-std=c99} or @option{-std=c11}), the functions
10745 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
10746 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
10747 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
10748 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
10749 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
10750 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
10751 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
10752 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
10753 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
10754 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
10755 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
10756 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
10757 @code{signbitd64}, @code{signbitd128}, @code{significandf},
10758 @code{significandl}, @code{significand}, @code{sincosf},
10759 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
10760 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
10761 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
10762 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
10764 may be handled as built-in functions.
10765 All these functions have corresponding versions
10766 prefixed with @code{__builtin_}, which may be used even in strict C90
10769 The ISO C99 functions
10770 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
10771 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
10772 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
10773 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
10774 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
10775 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
10776 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
10777 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
10778 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
10779 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
10780 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
10781 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
10782 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
10783 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
10784 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
10785 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
10786 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
10787 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
10788 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
10789 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
10790 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
10791 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
10792 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
10793 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
10794 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
10795 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
10796 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
10797 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
10798 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
10799 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
10800 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
10801 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
10802 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
10803 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
10804 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
10805 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
10806 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
10807 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
10808 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
10809 are handled as built-in functions
10810 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10812 There are also built-in versions of the ISO C99 functions
10813 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
10814 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
10815 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
10816 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
10817 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
10818 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
10819 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
10820 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
10821 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
10822 that are recognized in any mode since ISO C90 reserves these names for
10823 the purpose to which ISO C99 puts them. All these functions have
10824 corresponding versions prefixed with @code{__builtin_}.
10826 There are also GNU extension functions @code{clog10}, @code{clog10f} and
10827 @code{clog10l} which names are reserved by ISO C99 for future use.
10828 All these functions have versions prefixed with @code{__builtin_}.
10830 The ISO C94 functions
10831 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
10832 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
10833 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
10835 are handled as built-in functions
10836 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
10838 The ISO C90 functions
10839 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
10840 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
10841 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
10842 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
10843 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
10844 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
10845 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
10846 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
10847 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
10848 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
10849 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
10850 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
10851 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
10852 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
10853 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
10854 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
10855 are all recognized as built-in functions unless
10856 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
10857 is specified for an individual function). All of these functions have
10858 corresponding versions prefixed with @code{__builtin_}.
10860 GCC provides built-in versions of the ISO C99 floating-point comparison
10861 macros that avoid raising exceptions for unordered operands. They have
10862 the same names as the standard macros ( @code{isgreater},
10863 @code{isgreaterequal}, @code{isless}, @code{islessequal},
10864 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
10865 prefixed. We intend for a library implementor to be able to simply
10866 @code{#define} each standard macro to its built-in equivalent.
10867 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
10868 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
10869 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
10870 built-in functions appear both with and without the @code{__builtin_} prefix.
10872 @deftypefn {Built-in Function} void *__builtin_alloca (size_t size)
10873 The @code{__builtin_alloca} function must be called at block scope.
10874 The function allocates an object @var{size} bytes large on the stack
10875 of the calling function. The object is aligned on the default stack
10876 alignment boundary for the target determined by the
10877 @code{__BIGGEST_ALIGNMENT__} macro. The @code{__builtin_alloca}
10878 function returns a pointer to the first byte of the allocated object.
10879 The lifetime of the allocated object ends just before the calling
10880 function returns to its caller. This is so even when
10881 @code{__builtin_alloca} is called within a nested block.
10883 For example, the following function allocates eight objects of @code{n}
10884 bytes each on the stack, storing a pointer to each in consecutive elements
10885 of the array @code{a}. It then passes the array to function @code{g}
10886 which can safely use the storage pointed to by each of the array elements.
10889 void f (unsigned n)
10892 for (int i = 0; i != 8; ++i)
10893 a [i] = __builtin_alloca (n);
10895 g (a, n); // @r{safe}
10899 Since the @code{__builtin_alloca} function doesn't validate its argument
10900 it is the responsibility of its caller to make sure the argument doesn't
10901 cause it to exceed the stack size limit.
10902 The @code{__builtin_alloca} function is provided to make it possible to
10903 allocate on the stack arrays of bytes with an upper bound that may be
10904 computed at run time. Since C99 Variable Length Arrays offer
10905 similar functionality under a portable, more convenient, and safer
10906 interface they are recommended instead, in both C99 and C++ programs
10907 where GCC provides them as an extension.
10908 @xref{Variable Length}, for details.
10912 @deftypefn {Built-in Function} void *__builtin_alloca_with_align (size_t size, size_t alignment)
10913 The @code{__builtin_alloca_with_align} function must be called at block
10914 scope. The function allocates an object @var{size} bytes large on
10915 the stack of the calling function. The allocated object is aligned on
10916 the boundary specified by the argument @var{alignment} whose unit is given
10917 in bits (not bytes). The @var{size} argument must be positive and not
10918 exceed the stack size limit. The @var{alignment} argument must be a constant
10919 integer expression that evaluates to a power of 2 greater than or equal to
10920 @code{CHAR_BIT} and less than some unspecified maximum. Invocations
10921 with other values are rejected with an error indicating the valid bounds.
10922 The function returns a pointer to the first byte of the allocated object.
10923 The lifetime of the allocated object ends at the end of the block in which
10924 the function was called. The allocated storage is released no later than
10925 just before the calling function returns to its caller, but may be released
10926 at the end of the block in which the function was called.
10928 For example, in the following function the call to @code{g} is unsafe
10929 because when @code{overalign} is non-zero, the space allocated by
10930 @code{__builtin_alloca_with_align} may have been released at the end
10931 of the @code{if} statement in which it was called.
10934 void f (unsigned n, bool overalign)
10938 p = __builtin_alloca_with_align (n, 64 /* bits */);
10940 p = __builtin_alloc (n);
10942 g (p, n); // @r{unsafe}
10946 Since the @code{__builtin_alloca_with_align} function doesn't validate its
10947 @var{size} argument it is the responsibility of its caller to make sure
10948 the argument doesn't cause it to exceed the stack size limit.
10949 The @code{__builtin_alloca_with_align} function is provided to make
10950 it possible to allocate on the stack overaligned arrays of bytes with
10951 an upper bound that may be computed at run time. Since C99
10952 Variable Length Arrays offer the same functionality under
10953 a portable, more convenient, and safer interface they are recommended
10954 instead, in both C99 and C++ programs where GCC provides them as
10955 an extension. @xref{Variable Length}, for details.
10959 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
10961 You can use the built-in function @code{__builtin_types_compatible_p} to
10962 determine whether two types are the same.
10964 This built-in function returns 1 if the unqualified versions of the
10965 types @var{type1} and @var{type2} (which are types, not expressions) are
10966 compatible, 0 otherwise. The result of this built-in function can be
10967 used in integer constant expressions.
10969 This built-in function ignores top level qualifiers (e.g., @code{const},
10970 @code{volatile}). For example, @code{int} is equivalent to @code{const
10973 The type @code{int[]} and @code{int[5]} are compatible. On the other
10974 hand, @code{int} and @code{char *} are not compatible, even if the size
10975 of their types, on the particular architecture are the same. Also, the
10976 amount of pointer indirection is taken into account when determining
10977 similarity. Consequently, @code{short *} is not similar to
10978 @code{short **}. Furthermore, two types that are typedefed are
10979 considered compatible if their underlying types are compatible.
10981 An @code{enum} type is not considered to be compatible with another
10982 @code{enum} type even if both are compatible with the same integer
10983 type; this is what the C standard specifies.
10984 For example, @code{enum @{foo, bar@}} is not similar to
10985 @code{enum @{hot, dog@}}.
10987 You typically use this function in code whose execution varies
10988 depending on the arguments' types. For example:
10993 typeof (x) tmp = (x); \
10994 if (__builtin_types_compatible_p (typeof (x), long double)) \
10995 tmp = foo_long_double (tmp); \
10996 else if (__builtin_types_compatible_p (typeof (x), double)) \
10997 tmp = foo_double (tmp); \
10998 else if (__builtin_types_compatible_p (typeof (x), float)) \
10999 tmp = foo_float (tmp); \
11006 @emph{Note:} This construct is only available for C@.
11010 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
11012 The @var{call_exp} expression must be a function call, and the
11013 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
11014 is passed to the function call in the target's static chain location.
11015 The result of builtin is the result of the function call.
11017 @emph{Note:} This builtin is only available for C@.
11018 This builtin can be used to call Go closures from C.
11022 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
11024 You can use the built-in function @code{__builtin_choose_expr} to
11025 evaluate code depending on the value of a constant expression. This
11026 built-in function returns @var{exp1} if @var{const_exp}, which is an
11027 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
11029 This built-in function is analogous to the @samp{? :} operator in C,
11030 except that the expression returned has its type unaltered by promotion
11031 rules. Also, the built-in function does not evaluate the expression
11032 that is not chosen. For example, if @var{const_exp} evaluates to true,
11033 @var{exp2} is not evaluated even if it has side-effects.
11035 This built-in function can return an lvalue if the chosen argument is an
11038 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
11039 type. Similarly, if @var{exp2} is returned, its return type is the same
11046 __builtin_choose_expr ( \
11047 __builtin_types_compatible_p (typeof (x), double), \
11049 __builtin_choose_expr ( \
11050 __builtin_types_compatible_p (typeof (x), float), \
11052 /* @r{The void expression results in a compile-time error} \
11053 @r{when assigning the result to something.} */ \
11057 @emph{Note:} This construct is only available for C@. Furthermore, the
11058 unused expression (@var{exp1} or @var{exp2} depending on the value of
11059 @var{const_exp}) may still generate syntax errors. This may change in
11064 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
11066 The built-in function @code{__builtin_complex} is provided for use in
11067 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
11068 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
11069 real binary floating-point type, and the result has the corresponding
11070 complex type with real and imaginary parts @var{real} and @var{imag}.
11071 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
11072 infinities, NaNs and negative zeros are involved.
11076 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
11077 You can use the built-in function @code{__builtin_constant_p} to
11078 determine if a value is known to be constant at compile time and hence
11079 that GCC can perform constant-folding on expressions involving that
11080 value. The argument of the function is the value to test. The function
11081 returns the integer 1 if the argument is known to be a compile-time
11082 constant and 0 if it is not known to be a compile-time constant. A
11083 return of 0 does not indicate that the value is @emph{not} a constant,
11084 but merely that GCC cannot prove it is a constant with the specified
11085 value of the @option{-O} option.
11087 You typically use this function in an embedded application where
11088 memory is a critical resource. If you have some complex calculation,
11089 you may want it to be folded if it involves constants, but need to call
11090 a function if it does not. For example:
11093 #define Scale_Value(X) \
11094 (__builtin_constant_p (X) \
11095 ? ((X) * SCALE + OFFSET) : Scale (X))
11098 You may use this built-in function in either a macro or an inline
11099 function. However, if you use it in an inlined function and pass an
11100 argument of the function as the argument to the built-in, GCC
11101 never returns 1 when you call the inline function with a string constant
11102 or compound literal (@pxref{Compound Literals}) and does not return 1
11103 when you pass a constant numeric value to the inline function unless you
11104 specify the @option{-O} option.
11106 You may also use @code{__builtin_constant_p} in initializers for static
11107 data. For instance, you can write
11110 static const int table[] = @{
11111 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
11117 This is an acceptable initializer even if @var{EXPRESSION} is not a
11118 constant expression, including the case where
11119 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
11120 folded to a constant but @var{EXPRESSION} contains operands that are
11121 not otherwise permitted in a static initializer (for example,
11122 @code{0 && foo ()}). GCC must be more conservative about evaluating the
11123 built-in in this case, because it has no opportunity to perform
11127 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
11128 @opindex fprofile-arcs
11129 You may use @code{__builtin_expect} to provide the compiler with
11130 branch prediction information. In general, you should prefer to
11131 use actual profile feedback for this (@option{-fprofile-arcs}), as
11132 programmers are notoriously bad at predicting how their programs
11133 actually perform. However, there are applications in which this
11134 data is hard to collect.
11136 The return value is the value of @var{exp}, which should be an integral
11137 expression. The semantics of the built-in are that it is expected that
11138 @var{exp} == @var{c}. For example:
11141 if (__builtin_expect (x, 0))
11146 indicates that we do not expect to call @code{foo}, since
11147 we expect @code{x} to be zero. Since you are limited to integral
11148 expressions for @var{exp}, you should use constructions such as
11151 if (__builtin_expect (ptr != NULL, 1))
11156 when testing pointer or floating-point values.
11159 @deftypefn {Built-in Function} void __builtin_trap (void)
11160 This function causes the program to exit abnormally. GCC implements
11161 this function by using a target-dependent mechanism (such as
11162 intentionally executing an illegal instruction) or by calling
11163 @code{abort}. The mechanism used may vary from release to release so
11164 you should not rely on any particular implementation.
11167 @deftypefn {Built-in Function} void __builtin_unreachable (void)
11168 If control flow reaches the point of the @code{__builtin_unreachable},
11169 the program is undefined. It is useful in situations where the
11170 compiler cannot deduce the unreachability of the code.
11172 One such case is immediately following an @code{asm} statement that
11173 either never terminates, or one that transfers control elsewhere
11174 and never returns. In this example, without the
11175 @code{__builtin_unreachable}, GCC issues a warning that control
11176 reaches the end of a non-void function. It also generates code
11177 to return after the @code{asm}.
11180 int f (int c, int v)
11188 asm("jmp error_handler");
11189 __builtin_unreachable ();
11195 Because the @code{asm} statement unconditionally transfers control out
11196 of the function, control never reaches the end of the function
11197 body. The @code{__builtin_unreachable} is in fact unreachable and
11198 communicates this fact to the compiler.
11200 Another use for @code{__builtin_unreachable} is following a call a
11201 function that never returns but that is not declared
11202 @code{__attribute__((noreturn))}, as in this example:
11205 void function_that_never_returns (void);
11215 function_that_never_returns ();
11216 __builtin_unreachable ();
11223 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
11224 This function returns its first argument, and allows the compiler
11225 to assume that the returned pointer is at least @var{align} bytes
11226 aligned. This built-in can have either two or three arguments,
11227 if it has three, the third argument should have integer type, and
11228 if it is nonzero means misalignment offset. For example:
11231 void *x = __builtin_assume_aligned (arg, 16);
11235 means that the compiler can assume @code{x}, set to @code{arg}, is at least
11236 16-byte aligned, while:
11239 void *x = __builtin_assume_aligned (arg, 32, 8);
11243 means that the compiler can assume for @code{x}, set to @code{arg}, that
11244 @code{(char *) x - 8} is 32-byte aligned.
11247 @deftypefn {Built-in Function} int __builtin_LINE ()
11248 This function is the equivalent of the preprocessor @code{__LINE__}
11249 macro and returns a constant integer expression that evaluates to
11250 the line number of the invocation of the built-in. When used as a C++
11251 default argument for a function @var{F}, it returns the line number
11252 of the call to @var{F}.
11255 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
11256 This function is the equivalent of the @code{__FUNCTION__} symbol
11257 and returns an address constant pointing to the name of the function
11258 from which the built-in was invoked, or the empty string if
11259 the invocation is not at function scope. When used as a C++ default
11260 argument for a function @var{F}, it returns the name of @var{F}'s
11261 caller or the empty string if the call was not made at function
11265 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
11266 This function is the equivalent of the preprocessor @code{__FILE__}
11267 macro and returns an address constant pointing to the file name
11268 containing the invocation of the built-in, or the empty string if
11269 the invocation is not at function scope. When used as a C++ default
11270 argument for a function @var{F}, it returns the file name of the call
11271 to @var{F} or the empty string if the call was not made at function
11274 For example, in the following, each call to function @code{foo} will
11275 print a line similar to @code{"file.c:123: foo: message"} with the name
11276 of the file and the line number of the @code{printf} call, the name of
11277 the function @code{foo}, followed by the word @code{message}.
11281 function (const char *func = __builtin_FUNCTION ())
11288 printf ("%s:%i: %s: message\n", file (), line (), function ());
11294 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
11295 This function is used to flush the processor's instruction cache for
11296 the region of memory between @var{begin} inclusive and @var{end}
11297 exclusive. Some targets require that the instruction cache be
11298 flushed, after modifying memory containing code, in order to obtain
11299 deterministic behavior.
11301 If the target does not require instruction cache flushes,
11302 @code{__builtin___clear_cache} has no effect. Otherwise either
11303 instructions are emitted in-line to clear the instruction cache or a
11304 call to the @code{__clear_cache} function in libgcc is made.
11307 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
11308 This function is used to minimize cache-miss latency by moving data into
11309 a cache before it is accessed.
11310 You can insert calls to @code{__builtin_prefetch} into code for which
11311 you know addresses of data in memory that is likely to be accessed soon.
11312 If the target supports them, data prefetch instructions are generated.
11313 If the prefetch is done early enough before the access then the data will
11314 be in the cache by the time it is accessed.
11316 The value of @var{addr} is the address of the memory to prefetch.
11317 There are two optional arguments, @var{rw} and @var{locality}.
11318 The value of @var{rw} is a compile-time constant one or zero; one
11319 means that the prefetch is preparing for a write to the memory address
11320 and zero, the default, means that the prefetch is preparing for a read.
11321 The value @var{locality} must be a compile-time constant integer between
11322 zero and three. A value of zero means that the data has no temporal
11323 locality, so it need not be left in the cache after the access. A value
11324 of three means that the data has a high degree of temporal locality and
11325 should be left in all levels of cache possible. Values of one and two
11326 mean, respectively, a low or moderate degree of temporal locality. The
11330 for (i = 0; i < n; i++)
11332 a[i] = a[i] + b[i];
11333 __builtin_prefetch (&a[i+j], 1, 1);
11334 __builtin_prefetch (&b[i+j], 0, 1);
11339 Data prefetch does not generate faults if @var{addr} is invalid, but
11340 the address expression itself must be valid. For example, a prefetch
11341 of @code{p->next} does not fault if @code{p->next} is not a valid
11342 address, but evaluation faults if @code{p} is not a valid address.
11344 If the target does not support data prefetch, the address expression
11345 is evaluated if it includes side effects but no other code is generated
11346 and GCC does not issue a warning.
11349 @deftypefn {Built-in Function} double __builtin_huge_val (void)
11350 Returns a positive infinity, if supported by the floating-point format,
11351 else @code{DBL_MAX}. This function is suitable for implementing the
11352 ISO C macro @code{HUGE_VAL}.
11355 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
11356 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
11359 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
11360 Similar to @code{__builtin_huge_val}, except the return
11361 type is @code{long double}.
11364 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
11365 This built-in implements the C99 fpclassify functionality. The first
11366 five int arguments should be the target library's notion of the
11367 possible FP classes and are used for return values. They must be
11368 constant values and they must appear in this order: @code{FP_NAN},
11369 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
11370 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
11371 to classify. GCC treats the last argument as type-generic, which
11372 means it does not do default promotion from float to double.
11375 @deftypefn {Built-in Function} double __builtin_inf (void)
11376 Similar to @code{__builtin_huge_val}, except a warning is generated
11377 if the target floating-point format does not support infinities.
11380 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
11381 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
11384 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
11385 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
11388 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
11389 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
11392 @deftypefn {Built-in Function} float __builtin_inff (void)
11393 Similar to @code{__builtin_inf}, except the return type is @code{float}.
11394 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
11397 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
11398 Similar to @code{__builtin_inf}, except the return
11399 type is @code{long double}.
11402 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
11403 Similar to @code{isinf}, except the return value is -1 for
11404 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
11405 Note while the parameter list is an
11406 ellipsis, this function only accepts exactly one floating-point
11407 argument. GCC treats this parameter as type-generic, which means it
11408 does not do default promotion from float to double.
11411 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
11412 This is an implementation of the ISO C99 function @code{nan}.
11414 Since ISO C99 defines this function in terms of @code{strtod}, which we
11415 do not implement, a description of the parsing is in order. The string
11416 is parsed as by @code{strtol}; that is, the base is recognized by
11417 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
11418 in the significand such that the least significant bit of the number
11419 is at the least significant bit of the significand. The number is
11420 truncated to fit the significand field provided. The significand is
11421 forced to be a quiet NaN@.
11423 This function, if given a string literal all of which would have been
11424 consumed by @code{strtol}, is evaluated early enough that it is considered a
11425 compile-time constant.
11428 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
11429 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
11432 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
11433 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
11436 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
11437 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
11440 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
11441 Similar to @code{__builtin_nan}, except the return type is @code{float}.
11444 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
11445 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
11448 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
11449 Similar to @code{__builtin_nan}, except the significand is forced
11450 to be a signaling NaN@. The @code{nans} function is proposed by
11451 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
11454 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
11455 Similar to @code{__builtin_nans}, except the return type is @code{float}.
11458 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
11459 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
11462 @deftypefn {Built-in Function} int __builtin_ffs (int x)
11463 Returns one plus the index of the least significant 1-bit of @var{x}, or
11464 if @var{x} is zero, returns zero.
11467 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
11468 Returns the number of leading 0-bits in @var{x}, starting at the most
11469 significant bit position. If @var{x} is 0, the result is undefined.
11472 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
11473 Returns the number of trailing 0-bits in @var{x}, starting at the least
11474 significant bit position. If @var{x} is 0, the result is undefined.
11477 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
11478 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
11479 number of bits following the most significant bit that are identical
11480 to it. There are no special cases for 0 or other values.
11483 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
11484 Returns the number of 1-bits in @var{x}.
11487 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
11488 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
11492 @deftypefn {Built-in Function} int __builtin_ffsl (long)
11493 Similar to @code{__builtin_ffs}, except the argument type is
11497 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
11498 Similar to @code{__builtin_clz}, except the argument type is
11499 @code{unsigned long}.
11502 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
11503 Similar to @code{__builtin_ctz}, except the argument type is
11504 @code{unsigned long}.
11507 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
11508 Similar to @code{__builtin_clrsb}, except the argument type is
11512 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
11513 Similar to @code{__builtin_popcount}, except the argument type is
11514 @code{unsigned long}.
11517 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
11518 Similar to @code{__builtin_parity}, except the argument type is
11519 @code{unsigned long}.
11522 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
11523 Similar to @code{__builtin_ffs}, except the argument type is
11527 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
11528 Similar to @code{__builtin_clz}, except the argument type is
11529 @code{unsigned long long}.
11532 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
11533 Similar to @code{__builtin_ctz}, except the argument type is
11534 @code{unsigned long long}.
11537 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
11538 Similar to @code{__builtin_clrsb}, except the argument type is
11542 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
11543 Similar to @code{__builtin_popcount}, except the argument type is
11544 @code{unsigned long long}.
11547 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
11548 Similar to @code{__builtin_parity}, except the argument type is
11549 @code{unsigned long long}.
11552 @deftypefn {Built-in Function} double __builtin_powi (double, int)
11553 Returns the first argument raised to the power of the second. Unlike the
11554 @code{pow} function no guarantees about precision and rounding are made.
11557 @deftypefn {Built-in Function} float __builtin_powif (float, int)
11558 Similar to @code{__builtin_powi}, except the argument and return types
11562 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
11563 Similar to @code{__builtin_powi}, except the argument and return types
11564 are @code{long double}.
11567 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
11568 Returns @var{x} with the order of the bytes reversed; for example,
11569 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
11573 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
11574 Similar to @code{__builtin_bswap16}, except the argument and return types
11578 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
11579 Similar to @code{__builtin_bswap32}, except the argument and return types
11583 @node Target Builtins
11584 @section Built-in Functions Specific to Particular Target Machines
11586 On some target machines, GCC supports many built-in functions specific
11587 to those machines. Generally these generate calls to specific machine
11588 instructions, but allow the compiler to schedule those calls.
11591 * AArch64 Built-in Functions::
11592 * Alpha Built-in Functions::
11593 * Altera Nios II Built-in Functions::
11594 * ARC Built-in Functions::
11595 * ARC SIMD Built-in Functions::
11596 * ARM iWMMXt Built-in Functions::
11597 * ARM C Language Extensions (ACLE)::
11598 * ARM Floating Point Status and Control Intrinsics::
11599 * AVR Built-in Functions::
11600 * Blackfin Built-in Functions::
11601 * FR-V Built-in Functions::
11602 * MIPS DSP Built-in Functions::
11603 * MIPS Paired-Single Support::
11604 * MIPS Loongson Built-in Functions::
11605 * MIPS SIMD Architecture (MSA) Support::
11606 * Other MIPS Built-in Functions::
11607 * MSP430 Built-in Functions::
11608 * NDS32 Built-in Functions::
11609 * picoChip Built-in Functions::
11610 * PowerPC Built-in Functions::
11611 * PowerPC AltiVec/VSX Built-in Functions::
11612 * PowerPC Hardware Transactional Memory Built-in Functions::
11613 * RX Built-in Functions::
11614 * S/390 System z Built-in Functions::
11615 * SH Built-in Functions::
11616 * SPARC VIS Built-in Functions::
11617 * SPU Built-in Functions::
11618 * TI C6X Built-in Functions::
11619 * TILE-Gx Built-in Functions::
11620 * TILEPro Built-in Functions::
11621 * x86 Built-in Functions::
11622 * x86 transactional memory intrinsics::
11625 @node AArch64 Built-in Functions
11626 @subsection AArch64 Built-in Functions
11628 These built-in functions are available for the AArch64 family of
11631 unsigned int __builtin_aarch64_get_fpcr ()
11632 void __builtin_aarch64_set_fpcr (unsigned int)
11633 unsigned int __builtin_aarch64_get_fpsr ()
11634 void __builtin_aarch64_set_fpsr (unsigned int)
11637 @node Alpha Built-in Functions
11638 @subsection Alpha Built-in Functions
11640 These built-in functions are available for the Alpha family of
11641 processors, depending on the command-line switches used.
11643 The following built-in functions are always available. They
11644 all generate the machine instruction that is part of the name.
11647 long __builtin_alpha_implver (void)
11648 long __builtin_alpha_rpcc (void)
11649 long __builtin_alpha_amask (long)
11650 long __builtin_alpha_cmpbge (long, long)
11651 long __builtin_alpha_extbl (long, long)
11652 long __builtin_alpha_extwl (long, long)
11653 long __builtin_alpha_extll (long, long)
11654 long __builtin_alpha_extql (long, long)
11655 long __builtin_alpha_extwh (long, long)
11656 long __builtin_alpha_extlh (long, long)
11657 long __builtin_alpha_extqh (long, long)
11658 long __builtin_alpha_insbl (long, long)
11659 long __builtin_alpha_inswl (long, long)
11660 long __builtin_alpha_insll (long, long)
11661 long __builtin_alpha_insql (long, long)
11662 long __builtin_alpha_inswh (long, long)
11663 long __builtin_alpha_inslh (long, long)
11664 long __builtin_alpha_insqh (long, long)
11665 long __builtin_alpha_mskbl (long, long)
11666 long __builtin_alpha_mskwl (long, long)
11667 long __builtin_alpha_mskll (long, long)
11668 long __builtin_alpha_mskql (long, long)
11669 long __builtin_alpha_mskwh (long, long)
11670 long __builtin_alpha_msklh (long, long)
11671 long __builtin_alpha_mskqh (long, long)
11672 long __builtin_alpha_umulh (long, long)
11673 long __builtin_alpha_zap (long, long)
11674 long __builtin_alpha_zapnot (long, long)
11677 The following built-in functions are always with @option{-mmax}
11678 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
11679 later. They all generate the machine instruction that is part
11683 long __builtin_alpha_pklb (long)
11684 long __builtin_alpha_pkwb (long)
11685 long __builtin_alpha_unpkbl (long)
11686 long __builtin_alpha_unpkbw (long)
11687 long __builtin_alpha_minub8 (long, long)
11688 long __builtin_alpha_minsb8 (long, long)
11689 long __builtin_alpha_minuw4 (long, long)
11690 long __builtin_alpha_minsw4 (long, long)
11691 long __builtin_alpha_maxub8 (long, long)
11692 long __builtin_alpha_maxsb8 (long, long)
11693 long __builtin_alpha_maxuw4 (long, long)
11694 long __builtin_alpha_maxsw4 (long, long)
11695 long __builtin_alpha_perr (long, long)
11698 The following built-in functions are always with @option{-mcix}
11699 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
11700 later. They all generate the machine instruction that is part
11704 long __builtin_alpha_cttz (long)
11705 long __builtin_alpha_ctlz (long)
11706 long __builtin_alpha_ctpop (long)
11709 The following built-in functions are available on systems that use the OSF/1
11710 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
11711 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
11712 @code{rdval} and @code{wrval}.
11715 void *__builtin_thread_pointer (void)
11716 void __builtin_set_thread_pointer (void *)
11719 @node Altera Nios II Built-in Functions
11720 @subsection Altera Nios II Built-in Functions
11722 These built-in functions are available for the Altera Nios II
11723 family of processors.
11725 The following built-in functions are always available. They
11726 all generate the machine instruction that is part of the name.
11729 int __builtin_ldbio (volatile const void *)
11730 int __builtin_ldbuio (volatile const void *)
11731 int __builtin_ldhio (volatile const void *)
11732 int __builtin_ldhuio (volatile const void *)
11733 int __builtin_ldwio (volatile const void *)
11734 void __builtin_stbio (volatile void *, int)
11735 void __builtin_sthio (volatile void *, int)
11736 void __builtin_stwio (volatile void *, int)
11737 void __builtin_sync (void)
11738 int __builtin_rdctl (int)
11739 int __builtin_rdprs (int, int)
11740 void __builtin_wrctl (int, int)
11741 void __builtin_flushd (volatile void *)
11742 void __builtin_flushda (volatile void *)
11743 int __builtin_wrpie (int);
11744 void __builtin_eni (int);
11745 int __builtin_ldex (volatile const void *)
11746 int __builtin_stex (volatile void *, int)
11747 int __builtin_ldsex (volatile const void *)
11748 int __builtin_stsex (volatile void *, int)
11751 The following built-in functions are always available. They
11752 all generate a Nios II Custom Instruction. The name of the
11753 function represents the types that the function takes and
11754 returns. The letter before the @code{n} is the return type
11755 or void if absent. The @code{n} represents the first parameter
11756 to all the custom instructions, the custom instruction number.
11757 The two letters after the @code{n} represent the up to two
11758 parameters to the function.
11760 The letters represent the following data types:
11763 @code{void} for return type and no parameter for parameter types.
11766 @code{int} for return type and parameter type
11769 @code{float} for return type and parameter type
11772 @code{void *} for return type and parameter type
11776 And the function names are:
11778 void __builtin_custom_n (void)
11779 void __builtin_custom_ni (int)
11780 void __builtin_custom_nf (float)
11781 void __builtin_custom_np (void *)
11782 void __builtin_custom_nii (int, int)
11783 void __builtin_custom_nif (int, float)
11784 void __builtin_custom_nip (int, void *)
11785 void __builtin_custom_nfi (float, int)
11786 void __builtin_custom_nff (float, float)
11787 void __builtin_custom_nfp (float, void *)
11788 void __builtin_custom_npi (void *, int)
11789 void __builtin_custom_npf (void *, float)
11790 void __builtin_custom_npp (void *, void *)
11791 int __builtin_custom_in (void)
11792 int __builtin_custom_ini (int)
11793 int __builtin_custom_inf (float)
11794 int __builtin_custom_inp (void *)
11795 int __builtin_custom_inii (int, int)
11796 int __builtin_custom_inif (int, float)
11797 int __builtin_custom_inip (int, void *)
11798 int __builtin_custom_infi (float, int)
11799 int __builtin_custom_inff (float, float)
11800 int __builtin_custom_infp (float, void *)
11801 int __builtin_custom_inpi (void *, int)
11802 int __builtin_custom_inpf (void *, float)
11803 int __builtin_custom_inpp (void *, void *)
11804 float __builtin_custom_fn (void)
11805 float __builtin_custom_fni (int)
11806 float __builtin_custom_fnf (float)
11807 float __builtin_custom_fnp (void *)
11808 float __builtin_custom_fnii (int, int)
11809 float __builtin_custom_fnif (int, float)
11810 float __builtin_custom_fnip (int, void *)
11811 float __builtin_custom_fnfi (float, int)
11812 float __builtin_custom_fnff (float, float)
11813 float __builtin_custom_fnfp (float, void *)
11814 float __builtin_custom_fnpi (void *, int)
11815 float __builtin_custom_fnpf (void *, float)
11816 float __builtin_custom_fnpp (void *, void *)
11817 void * __builtin_custom_pn (void)
11818 void * __builtin_custom_pni (int)
11819 void * __builtin_custom_pnf (float)
11820 void * __builtin_custom_pnp (void *)
11821 void * __builtin_custom_pnii (int, int)
11822 void * __builtin_custom_pnif (int, float)
11823 void * __builtin_custom_pnip (int, void *)
11824 void * __builtin_custom_pnfi (float, int)
11825 void * __builtin_custom_pnff (float, float)
11826 void * __builtin_custom_pnfp (float, void *)
11827 void * __builtin_custom_pnpi (void *, int)
11828 void * __builtin_custom_pnpf (void *, float)
11829 void * __builtin_custom_pnpp (void *, void *)
11832 @node ARC Built-in Functions
11833 @subsection ARC Built-in Functions
11835 The following built-in functions are provided for ARC targets. The
11836 built-ins generate the corresponding assembly instructions. In the
11837 examples given below, the generated code often requires an operand or
11838 result to be in a register. Where necessary further code will be
11839 generated to ensure this is true, but for brevity this is not
11840 described in each case.
11842 @emph{Note:} Using a built-in to generate an instruction not supported
11843 by a target may cause problems. At present the compiler is not
11844 guaranteed to detect such misuse, and as a result an internal compiler
11845 error may be generated.
11847 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
11848 Return 1 if @var{val} is known to have the byte alignment given
11849 by @var{alignval}, otherwise return 0.
11850 Note that this is different from
11852 __alignof__(*(char *)@var{val}) >= alignval
11854 because __alignof__ sees only the type of the dereference, whereas
11855 __builtin_arc_align uses alignment information from the pointer
11856 as well as from the pointed-to type.
11857 The information available will depend on optimization level.
11860 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
11867 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
11868 The operand is the number of a register to be read. Generates:
11870 mov @var{dest}, r@var{regno}
11872 where the value in @var{dest} will be the result returned from the
11876 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
11877 The first operand is the number of a register to be written, the
11878 second operand is a compile time constant to write into that
11879 register. Generates:
11881 mov r@var{regno}, @var{val}
11885 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
11886 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
11889 divaw @var{dest}, @var{a}, @var{b}
11891 where the value in @var{dest} will be the result returned from the
11895 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
11902 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
11903 The operand, @var{auxv}, is the address of an auxiliary register and
11904 must be a compile time constant. Generates:
11906 lr @var{dest}, [@var{auxr}]
11908 Where the value in @var{dest} will be the result returned from the
11912 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
11913 Only available with @option{-mmul64}. Generates:
11915 mul64 @var{a}, @var{b}
11919 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
11920 Only available with @option{-mmul64}. Generates:
11922 mulu64 @var{a}, @var{b}
11926 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
11933 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
11934 Only valid if the @samp{norm} instruction is available through the
11935 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11938 norm @var{dest}, @var{src}
11940 Where the value in @var{dest} will be the result returned from the
11944 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
11945 Only valid if the @samp{normw} instruction is available through the
11946 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
11949 normw @var{dest}, @var{src}
11951 Where the value in @var{dest} will be the result returned from the
11955 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
11962 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
11969 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
11970 The first argument, @var{auxv}, is the address of an auxiliary
11971 register, the second argument, @var{val}, is a compile time constant
11972 to be written to the register. Generates:
11974 sr @var{auxr}, [@var{val}]
11978 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
11979 Only valid with @option{-mswap}. Generates:
11981 swap @var{dest}, @var{src}
11983 Where the value in @var{dest} will be the result returned from the
11987 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
11994 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
11995 Only available with @option{-mcpu=ARC700}. Generates:
12001 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
12002 Only available with @option{-mcpu=ARC700}. Generates:
12008 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
12009 Only available with @option{-mcpu=ARC700}. Generates:
12015 The instructions generated by the following builtins are not
12016 considered as candidates for scheduling. They are not moved around by
12017 the compiler during scheduling, and thus can be expected to appear
12018 where they are put in the C code:
12020 __builtin_arc_brk()
12021 __builtin_arc_core_read()
12022 __builtin_arc_core_write()
12023 __builtin_arc_flag()
12025 __builtin_arc_sleep()
12027 __builtin_arc_swi()
12030 @node ARC SIMD Built-in Functions
12031 @subsection ARC SIMD Built-in Functions
12033 SIMD builtins provided by the compiler can be used to generate the
12034 vector instructions. This section describes the available builtins
12035 and their usage in programs. With the @option{-msimd} option, the
12036 compiler provides 128-bit vector types, which can be specified using
12037 the @code{vector_size} attribute. The header file @file{arc-simd.h}
12038 can be included to use the following predefined types:
12040 typedef int __v4si __attribute__((vector_size(16)));
12041 typedef short __v8hi __attribute__((vector_size(16)));
12044 These types can be used to define 128-bit variables. The built-in
12045 functions listed in the following section can be used on these
12046 variables to generate the vector operations.
12048 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
12049 @file{arc-simd.h} also provides equivalent macros called
12050 @code{_@var{someinsn}} that can be used for programming ease and
12051 improved readability. The following macros for DMA control are also
12054 #define _setup_dma_in_channel_reg _vdiwr
12055 #define _setup_dma_out_channel_reg _vdowr
12058 The following is a complete list of all the SIMD built-ins provided
12059 for ARC, grouped by calling signature.
12061 The following take two @code{__v8hi} arguments and return a
12062 @code{__v8hi} result:
12064 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
12065 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
12066 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
12067 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
12068 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
12069 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
12070 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
12071 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
12072 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
12073 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
12074 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
12075 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
12076 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
12077 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
12078 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
12079 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
12080 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
12081 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
12082 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
12083 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
12084 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
12085 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
12086 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
12087 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
12088 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
12089 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
12090 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
12091 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
12092 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
12093 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
12094 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
12095 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
12096 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
12097 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
12098 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
12099 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
12100 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
12101 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
12102 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
12103 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
12104 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
12105 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
12106 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
12107 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
12108 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
12109 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
12110 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
12111 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
12114 The following take one @code{__v8hi} and one @code{int} argument and return a
12115 @code{__v8hi} result:
12118 __v8hi __builtin_arc_vbaddw (__v8hi, int)
12119 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
12120 __v8hi __builtin_arc_vbminw (__v8hi, int)
12121 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
12122 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
12123 __v8hi __builtin_arc_vbmulw (__v8hi, int)
12124 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
12125 __v8hi __builtin_arc_vbsubw (__v8hi, int)
12128 The following take one @code{__v8hi} argument and one @code{int} argument which
12129 must be a 3-bit compile time constant indicating a register number
12130 I0-I7. They return a @code{__v8hi} result.
12132 __v8hi __builtin_arc_vasrw (__v8hi, const int)
12133 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
12134 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
12137 The following take one @code{__v8hi} argument and one @code{int}
12138 argument which must be a 6-bit compile time constant. They return a
12139 @code{__v8hi} result.
12141 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
12142 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
12143 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
12144 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
12145 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
12146 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
12147 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
12150 The following take one @code{__v8hi} argument and one @code{int} argument which
12151 must be a 8-bit compile time constant. They return a @code{__v8hi}
12154 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
12155 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
12156 __v8hi __builtin_arc_vmvw (__v8hi, const int)
12157 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
12160 The following take two @code{int} arguments, the second of which which
12161 must be a 8-bit compile time constant. They return a @code{__v8hi}
12164 __v8hi __builtin_arc_vmovaw (int, const int)
12165 __v8hi __builtin_arc_vmovw (int, const int)
12166 __v8hi __builtin_arc_vmovzw (int, const int)
12169 The following take a single @code{__v8hi} argument and return a
12170 @code{__v8hi} result:
12172 __v8hi __builtin_arc_vabsaw (__v8hi)
12173 __v8hi __builtin_arc_vabsw (__v8hi)
12174 __v8hi __builtin_arc_vaddsuw (__v8hi)
12175 __v8hi __builtin_arc_vexch1 (__v8hi)
12176 __v8hi __builtin_arc_vexch2 (__v8hi)
12177 __v8hi __builtin_arc_vexch4 (__v8hi)
12178 __v8hi __builtin_arc_vsignw (__v8hi)
12179 __v8hi __builtin_arc_vupbaw (__v8hi)
12180 __v8hi __builtin_arc_vupbw (__v8hi)
12181 __v8hi __builtin_arc_vupsbaw (__v8hi)
12182 __v8hi __builtin_arc_vupsbw (__v8hi)
12185 The following take two @code{int} arguments and return no result:
12187 void __builtin_arc_vdirun (int, int)
12188 void __builtin_arc_vdorun (int, int)
12191 The following take two @code{int} arguments and return no result. The
12192 first argument must a 3-bit compile time constant indicating one of
12193 the DR0-DR7 DMA setup channels:
12195 void __builtin_arc_vdiwr (const int, int)
12196 void __builtin_arc_vdowr (const int, int)
12199 The following take an @code{int} argument and return no result:
12201 void __builtin_arc_vendrec (int)
12202 void __builtin_arc_vrec (int)
12203 void __builtin_arc_vrecrun (int)
12204 void __builtin_arc_vrun (int)
12207 The following take a @code{__v8hi} argument and two @code{int}
12208 arguments and return a @code{__v8hi} result. The second argument must
12209 be a 3-bit compile time constants, indicating one the registers I0-I7,
12210 and the third argument must be an 8-bit compile time constant.
12212 @emph{Note:} Although the equivalent hardware instructions do not take
12213 an SIMD register as an operand, these builtins overwrite the relevant
12214 bits of the @code{__v8hi} register provided as the first argument with
12215 the value loaded from the @code{[Ib, u8]} location in the SDM.
12218 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
12219 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
12220 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
12221 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
12224 The following take two @code{int} arguments and return a @code{__v8hi}
12225 result. The first argument must be a 3-bit compile time constants,
12226 indicating one the registers I0-I7, and the second argument must be an
12227 8-bit compile time constant.
12230 __v8hi __builtin_arc_vld128 (const int, const int)
12231 __v8hi __builtin_arc_vld64w (const int, const int)
12234 The following take a @code{__v8hi} argument and two @code{int}
12235 arguments and return no result. The second argument must be a 3-bit
12236 compile time constants, indicating one the registers I0-I7, and the
12237 third argument must be an 8-bit compile time constant.
12240 void __builtin_arc_vst128 (__v8hi, const int, const int)
12241 void __builtin_arc_vst64 (__v8hi, const int, const int)
12244 The following take a @code{__v8hi} argument and three @code{int}
12245 arguments and return no result. The second argument must be a 3-bit
12246 compile-time constant, identifying the 16-bit sub-register to be
12247 stored, the third argument must be a 3-bit compile time constants,
12248 indicating one the registers I0-I7, and the fourth argument must be an
12249 8-bit compile time constant.
12252 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
12253 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
12256 @node ARM iWMMXt Built-in Functions
12257 @subsection ARM iWMMXt Built-in Functions
12259 These built-in functions are available for the ARM family of
12260 processors when the @option{-mcpu=iwmmxt} switch is used:
12263 typedef int v2si __attribute__ ((vector_size (8)));
12264 typedef short v4hi __attribute__ ((vector_size (8)));
12265 typedef char v8qi __attribute__ ((vector_size (8)));
12267 int __builtin_arm_getwcgr0 (void)
12268 void __builtin_arm_setwcgr0 (int)
12269 int __builtin_arm_getwcgr1 (void)
12270 void __builtin_arm_setwcgr1 (int)
12271 int __builtin_arm_getwcgr2 (void)
12272 void __builtin_arm_setwcgr2 (int)
12273 int __builtin_arm_getwcgr3 (void)
12274 void __builtin_arm_setwcgr3 (int)
12275 int __builtin_arm_textrmsb (v8qi, int)
12276 int __builtin_arm_textrmsh (v4hi, int)
12277 int __builtin_arm_textrmsw (v2si, int)
12278 int __builtin_arm_textrmub (v8qi, int)
12279 int __builtin_arm_textrmuh (v4hi, int)
12280 int __builtin_arm_textrmuw (v2si, int)
12281 v8qi __builtin_arm_tinsrb (v8qi, int, int)
12282 v4hi __builtin_arm_tinsrh (v4hi, int, int)
12283 v2si __builtin_arm_tinsrw (v2si, int, int)
12284 long long __builtin_arm_tmia (long long, int, int)
12285 long long __builtin_arm_tmiabb (long long, int, int)
12286 long long __builtin_arm_tmiabt (long long, int, int)
12287 long long __builtin_arm_tmiaph (long long, int, int)
12288 long long __builtin_arm_tmiatb (long long, int, int)
12289 long long __builtin_arm_tmiatt (long long, int, int)
12290 int __builtin_arm_tmovmskb (v8qi)
12291 int __builtin_arm_tmovmskh (v4hi)
12292 int __builtin_arm_tmovmskw (v2si)
12293 long long __builtin_arm_waccb (v8qi)
12294 long long __builtin_arm_wacch (v4hi)
12295 long long __builtin_arm_waccw (v2si)
12296 v8qi __builtin_arm_waddb (v8qi, v8qi)
12297 v8qi __builtin_arm_waddbss (v8qi, v8qi)
12298 v8qi __builtin_arm_waddbus (v8qi, v8qi)
12299 v4hi __builtin_arm_waddh (v4hi, v4hi)
12300 v4hi __builtin_arm_waddhss (v4hi, v4hi)
12301 v4hi __builtin_arm_waddhus (v4hi, v4hi)
12302 v2si __builtin_arm_waddw (v2si, v2si)
12303 v2si __builtin_arm_waddwss (v2si, v2si)
12304 v2si __builtin_arm_waddwus (v2si, v2si)
12305 v8qi __builtin_arm_walign (v8qi, v8qi, int)
12306 long long __builtin_arm_wand(long long, long long)
12307 long long __builtin_arm_wandn (long long, long long)
12308 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
12309 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
12310 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
12311 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
12312 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
12313 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
12314 v2si __builtin_arm_wcmpeqw (v2si, v2si)
12315 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
12316 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
12317 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
12318 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
12319 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
12320 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
12321 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
12322 long long __builtin_arm_wmacsz (v4hi, v4hi)
12323 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
12324 long long __builtin_arm_wmacuz (v4hi, v4hi)
12325 v4hi __builtin_arm_wmadds (v4hi, v4hi)
12326 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
12327 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
12328 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
12329 v2si __builtin_arm_wmaxsw (v2si, v2si)
12330 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
12331 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
12332 v2si __builtin_arm_wmaxuw (v2si, v2si)
12333 v8qi __builtin_arm_wminsb (v8qi, v8qi)
12334 v4hi __builtin_arm_wminsh (v4hi, v4hi)
12335 v2si __builtin_arm_wminsw (v2si, v2si)
12336 v8qi __builtin_arm_wminub (v8qi, v8qi)
12337 v4hi __builtin_arm_wminuh (v4hi, v4hi)
12338 v2si __builtin_arm_wminuw (v2si, v2si)
12339 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
12340 v4hi __builtin_arm_wmulul (v4hi, v4hi)
12341 v4hi __builtin_arm_wmulum (v4hi, v4hi)
12342 long long __builtin_arm_wor (long long, long long)
12343 v2si __builtin_arm_wpackdss (long long, long long)
12344 v2si __builtin_arm_wpackdus (long long, long long)
12345 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
12346 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
12347 v4hi __builtin_arm_wpackwss (v2si, v2si)
12348 v4hi __builtin_arm_wpackwus (v2si, v2si)
12349 long long __builtin_arm_wrord (long long, long long)
12350 long long __builtin_arm_wrordi (long long, int)
12351 v4hi __builtin_arm_wrorh (v4hi, long long)
12352 v4hi __builtin_arm_wrorhi (v4hi, int)
12353 v2si __builtin_arm_wrorw (v2si, long long)
12354 v2si __builtin_arm_wrorwi (v2si, int)
12355 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
12356 v2si __builtin_arm_wsadbz (v8qi, v8qi)
12357 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
12358 v2si __builtin_arm_wsadhz (v4hi, v4hi)
12359 v4hi __builtin_arm_wshufh (v4hi, int)
12360 long long __builtin_arm_wslld (long long, long long)
12361 long long __builtin_arm_wslldi (long long, int)
12362 v4hi __builtin_arm_wsllh (v4hi, long long)
12363 v4hi __builtin_arm_wsllhi (v4hi, int)
12364 v2si __builtin_arm_wsllw (v2si, long long)
12365 v2si __builtin_arm_wsllwi (v2si, int)
12366 long long __builtin_arm_wsrad (long long, long long)
12367 long long __builtin_arm_wsradi (long long, int)
12368 v4hi __builtin_arm_wsrah (v4hi, long long)
12369 v4hi __builtin_arm_wsrahi (v4hi, int)
12370 v2si __builtin_arm_wsraw (v2si, long long)
12371 v2si __builtin_arm_wsrawi (v2si, int)
12372 long long __builtin_arm_wsrld (long long, long long)
12373 long long __builtin_arm_wsrldi (long long, int)
12374 v4hi __builtin_arm_wsrlh (v4hi, long long)
12375 v4hi __builtin_arm_wsrlhi (v4hi, int)
12376 v2si __builtin_arm_wsrlw (v2si, long long)
12377 v2si __builtin_arm_wsrlwi (v2si, int)
12378 v8qi __builtin_arm_wsubb (v8qi, v8qi)
12379 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
12380 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
12381 v4hi __builtin_arm_wsubh (v4hi, v4hi)
12382 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
12383 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
12384 v2si __builtin_arm_wsubw (v2si, v2si)
12385 v2si __builtin_arm_wsubwss (v2si, v2si)
12386 v2si __builtin_arm_wsubwus (v2si, v2si)
12387 v4hi __builtin_arm_wunpckehsb (v8qi)
12388 v2si __builtin_arm_wunpckehsh (v4hi)
12389 long long __builtin_arm_wunpckehsw (v2si)
12390 v4hi __builtin_arm_wunpckehub (v8qi)
12391 v2si __builtin_arm_wunpckehuh (v4hi)
12392 long long __builtin_arm_wunpckehuw (v2si)
12393 v4hi __builtin_arm_wunpckelsb (v8qi)
12394 v2si __builtin_arm_wunpckelsh (v4hi)
12395 long long __builtin_arm_wunpckelsw (v2si)
12396 v4hi __builtin_arm_wunpckelub (v8qi)
12397 v2si __builtin_arm_wunpckeluh (v4hi)
12398 long long __builtin_arm_wunpckeluw (v2si)
12399 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
12400 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
12401 v2si __builtin_arm_wunpckihw (v2si, v2si)
12402 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
12403 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
12404 v2si __builtin_arm_wunpckilw (v2si, v2si)
12405 long long __builtin_arm_wxor (long long, long long)
12406 long long __builtin_arm_wzero ()
12410 @node ARM C Language Extensions (ACLE)
12411 @subsection ARM C Language Extensions (ACLE)
12413 GCC implements extensions for C as described in the ARM C Language
12414 Extensions (ACLE) specification, which can be found at
12415 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
12417 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
12418 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
12419 intrinsics can be found at
12420 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
12421 The built-in intrinsics for the Advanced SIMD extension are available when
12424 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
12425 back ends support CRC32 intrinsics from @file{arm_acle.h}. The ARM back end's
12426 16-bit floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
12427 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
12430 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
12431 availability of extensions.
12433 @node ARM Floating Point Status and Control Intrinsics
12434 @subsection ARM Floating Point Status and Control Intrinsics
12436 These built-in functions are available for the ARM family of
12437 processors with floating-point unit.
12440 unsigned int __builtin_arm_get_fpscr ()
12441 void __builtin_arm_set_fpscr (unsigned int)
12444 @node AVR Built-in Functions
12445 @subsection AVR Built-in Functions
12447 For each built-in function for AVR, there is an equally named,
12448 uppercase built-in macro defined. That way users can easily query if
12449 or if not a specific built-in is implemented or not. For example, if
12450 @code{__builtin_avr_nop} is available the macro
12451 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
12453 The following built-in functions map to the respective machine
12454 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
12455 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
12456 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
12457 as library call if no hardware multiplier is available.
12460 void __builtin_avr_nop (void)
12461 void __builtin_avr_sei (void)
12462 void __builtin_avr_cli (void)
12463 void __builtin_avr_sleep (void)
12464 void __builtin_avr_wdr (void)
12465 unsigned char __builtin_avr_swap (unsigned char)
12466 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
12467 int __builtin_avr_fmuls (char, char)
12468 int __builtin_avr_fmulsu (char, unsigned char)
12471 In order to delay execution for a specific number of cycles, GCC
12474 void __builtin_avr_delay_cycles (unsigned long ticks)
12478 @code{ticks} is the number of ticks to delay execution. Note that this
12479 built-in does not take into account the effect of interrupts that
12480 might increase delay time. @code{ticks} must be a compile-time
12481 integer constant; delays with a variable number of cycles are not supported.
12484 char __builtin_avr_flash_segment (const __memx void*)
12488 This built-in takes a byte address to the 24-bit
12489 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
12490 the number of the flash segment (the 64 KiB chunk) where the address
12491 points to. Counting starts at @code{0}.
12492 If the address does not point to flash memory, return @code{-1}.
12495 unsigned char __builtin_avr_insert_bits (unsigned long map, unsigned char bits, unsigned char val)
12499 Insert bits from @var{bits} into @var{val} and return the resulting
12500 value. The nibbles of @var{map} determine how the insertion is
12501 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
12503 @item If @var{X} is @code{0xf},
12504 then the @var{n}-th bit of @var{val} is returned unaltered.
12506 @item If X is in the range 0@dots{}7,
12507 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
12509 @item If X is in the range 8@dots{}@code{0xe},
12510 then the @var{n}-th result bit is undefined.
12514 One typical use case for this built-in is adjusting input and
12515 output values to non-contiguous port layouts. Some examples:
12518 // same as val, bits is unused
12519 __builtin_avr_insert_bits (0xffffffff, bits, val)
12523 // same as bits, val is unused
12524 __builtin_avr_insert_bits (0x76543210, bits, val)
12528 // same as rotating bits by 4
12529 __builtin_avr_insert_bits (0x32107654, bits, 0)
12533 // high nibble of result is the high nibble of val
12534 // low nibble of result is the low nibble of bits
12535 __builtin_avr_insert_bits (0xffff3210, bits, val)
12539 // reverse the bit order of bits
12540 __builtin_avr_insert_bits (0x01234567, bits, 0)
12543 @node Blackfin Built-in Functions
12544 @subsection Blackfin Built-in Functions
12546 Currently, there are two Blackfin-specific built-in functions. These are
12547 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
12548 using inline assembly; by using these built-in functions the compiler can
12549 automatically add workarounds for hardware errata involving these
12550 instructions. These functions are named as follows:
12553 void __builtin_bfin_csync (void)
12554 void __builtin_bfin_ssync (void)
12557 @node FR-V Built-in Functions
12558 @subsection FR-V Built-in Functions
12560 GCC provides many FR-V-specific built-in functions. In general,
12561 these functions are intended to be compatible with those described
12562 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
12563 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
12564 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
12565 pointer rather than by value.
12567 Most of the functions are named after specific FR-V instructions.
12568 Such functions are said to be ``directly mapped'' and are summarized
12569 here in tabular form.
12573 * Directly-mapped Integer Functions::
12574 * Directly-mapped Media Functions::
12575 * Raw read/write Functions::
12576 * Other Built-in Functions::
12579 @node Argument Types
12580 @subsubsection Argument Types
12582 The arguments to the built-in functions can be divided into three groups:
12583 register numbers, compile-time constants and run-time values. In order
12584 to make this classification clear at a glance, the arguments and return
12585 values are given the following pseudo types:
12587 @multitable @columnfractions .20 .30 .15 .35
12588 @item Pseudo type @tab Real C type @tab Constant? @tab Description
12589 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
12590 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
12591 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
12592 @item @code{uw2} @tab @code{unsigned long long} @tab No
12593 @tab an unsigned doubleword
12594 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
12595 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
12596 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
12597 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
12600 These pseudo types are not defined by GCC, they are simply a notational
12601 convenience used in this manual.
12603 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
12604 and @code{sw2} are evaluated at run time. They correspond to
12605 register operands in the underlying FR-V instructions.
12607 @code{const} arguments represent immediate operands in the underlying
12608 FR-V instructions. They must be compile-time constants.
12610 @code{acc} arguments are evaluated at compile time and specify the number
12611 of an accumulator register. For example, an @code{acc} argument of 2
12612 selects the ACC2 register.
12614 @code{iacc} arguments are similar to @code{acc} arguments but specify the
12615 number of an IACC register. See @pxref{Other Built-in Functions}
12618 @node Directly-mapped Integer Functions
12619 @subsubsection Directly-Mapped Integer Functions
12621 The functions listed below map directly to FR-V I-type instructions.
12623 @multitable @columnfractions .45 .32 .23
12624 @item Function prototype @tab Example usage @tab Assembly output
12625 @item @code{sw1 __ADDSS (sw1, sw1)}
12626 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
12627 @tab @code{ADDSS @var{a},@var{b},@var{c}}
12628 @item @code{sw1 __SCAN (sw1, sw1)}
12629 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
12630 @tab @code{SCAN @var{a},@var{b},@var{c}}
12631 @item @code{sw1 __SCUTSS (sw1)}
12632 @tab @code{@var{b} = __SCUTSS (@var{a})}
12633 @tab @code{SCUTSS @var{a},@var{b}}
12634 @item @code{sw1 __SLASS (sw1, sw1)}
12635 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
12636 @tab @code{SLASS @var{a},@var{b},@var{c}}
12637 @item @code{void __SMASS (sw1, sw1)}
12638 @tab @code{__SMASS (@var{a}, @var{b})}
12639 @tab @code{SMASS @var{a},@var{b}}
12640 @item @code{void __SMSSS (sw1, sw1)}
12641 @tab @code{__SMSSS (@var{a}, @var{b})}
12642 @tab @code{SMSSS @var{a},@var{b}}
12643 @item @code{void __SMU (sw1, sw1)}
12644 @tab @code{__SMU (@var{a}, @var{b})}
12645 @tab @code{SMU @var{a},@var{b}}
12646 @item @code{sw2 __SMUL (sw1, sw1)}
12647 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
12648 @tab @code{SMUL @var{a},@var{b},@var{c}}
12649 @item @code{sw1 __SUBSS (sw1, sw1)}
12650 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
12651 @tab @code{SUBSS @var{a},@var{b},@var{c}}
12652 @item @code{uw2 __UMUL (uw1, uw1)}
12653 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
12654 @tab @code{UMUL @var{a},@var{b},@var{c}}
12657 @node Directly-mapped Media Functions
12658 @subsubsection Directly-Mapped Media Functions
12660 The functions listed below map directly to FR-V M-type instructions.
12662 @multitable @columnfractions .45 .32 .23
12663 @item Function prototype @tab Example usage @tab Assembly output
12664 @item @code{uw1 __MABSHS (sw1)}
12665 @tab @code{@var{b} = __MABSHS (@var{a})}
12666 @tab @code{MABSHS @var{a},@var{b}}
12667 @item @code{void __MADDACCS (acc, acc)}
12668 @tab @code{__MADDACCS (@var{b}, @var{a})}
12669 @tab @code{MADDACCS @var{a},@var{b}}
12670 @item @code{sw1 __MADDHSS (sw1, sw1)}
12671 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
12672 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
12673 @item @code{uw1 __MADDHUS (uw1, uw1)}
12674 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
12675 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
12676 @item @code{uw1 __MAND (uw1, uw1)}
12677 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
12678 @tab @code{MAND @var{a},@var{b},@var{c}}
12679 @item @code{void __MASACCS (acc, acc)}
12680 @tab @code{__MASACCS (@var{b}, @var{a})}
12681 @tab @code{MASACCS @var{a},@var{b}}
12682 @item @code{uw1 __MAVEH (uw1, uw1)}
12683 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
12684 @tab @code{MAVEH @var{a},@var{b},@var{c}}
12685 @item @code{uw2 __MBTOH (uw1)}
12686 @tab @code{@var{b} = __MBTOH (@var{a})}
12687 @tab @code{MBTOH @var{a},@var{b}}
12688 @item @code{void __MBTOHE (uw1 *, uw1)}
12689 @tab @code{__MBTOHE (&@var{b}, @var{a})}
12690 @tab @code{MBTOHE @var{a},@var{b}}
12691 @item @code{void __MCLRACC (acc)}
12692 @tab @code{__MCLRACC (@var{a})}
12693 @tab @code{MCLRACC @var{a}}
12694 @item @code{void __MCLRACCA (void)}
12695 @tab @code{__MCLRACCA ()}
12696 @tab @code{MCLRACCA}
12697 @item @code{uw1 __Mcop1 (uw1, uw1)}
12698 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
12699 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
12700 @item @code{uw1 __Mcop2 (uw1, uw1)}
12701 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
12702 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
12703 @item @code{uw1 __MCPLHI (uw2, const)}
12704 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
12705 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
12706 @item @code{uw1 __MCPLI (uw2, const)}
12707 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
12708 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
12709 @item @code{void __MCPXIS (acc, sw1, sw1)}
12710 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
12711 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
12712 @item @code{void __MCPXIU (acc, uw1, uw1)}
12713 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
12714 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
12715 @item @code{void __MCPXRS (acc, sw1, sw1)}
12716 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
12717 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
12718 @item @code{void __MCPXRU (acc, uw1, uw1)}
12719 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
12720 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
12721 @item @code{uw1 __MCUT (acc, uw1)}
12722 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
12723 @tab @code{MCUT @var{a},@var{b},@var{c}}
12724 @item @code{uw1 __MCUTSS (acc, sw1)}
12725 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
12726 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
12727 @item @code{void __MDADDACCS (acc, acc)}
12728 @tab @code{__MDADDACCS (@var{b}, @var{a})}
12729 @tab @code{MDADDACCS @var{a},@var{b}}
12730 @item @code{void __MDASACCS (acc, acc)}
12731 @tab @code{__MDASACCS (@var{b}, @var{a})}
12732 @tab @code{MDASACCS @var{a},@var{b}}
12733 @item @code{uw2 __MDCUTSSI (acc, const)}
12734 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
12735 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
12736 @item @code{uw2 __MDPACKH (uw2, uw2)}
12737 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
12738 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
12739 @item @code{uw2 __MDROTLI (uw2, const)}
12740 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
12741 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
12742 @item @code{void __MDSUBACCS (acc, acc)}
12743 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
12744 @tab @code{MDSUBACCS @var{a},@var{b}}
12745 @item @code{void __MDUNPACKH (uw1 *, uw2)}
12746 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
12747 @tab @code{MDUNPACKH @var{a},@var{b}}
12748 @item @code{uw2 __MEXPDHD (uw1, const)}
12749 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
12750 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
12751 @item @code{uw1 __MEXPDHW (uw1, const)}
12752 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
12753 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
12754 @item @code{uw1 __MHDSETH (uw1, const)}
12755 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
12756 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
12757 @item @code{sw1 __MHDSETS (const)}
12758 @tab @code{@var{b} = __MHDSETS (@var{a})}
12759 @tab @code{MHDSETS #@var{a},@var{b}}
12760 @item @code{uw1 __MHSETHIH (uw1, const)}
12761 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
12762 @tab @code{MHSETHIH #@var{a},@var{b}}
12763 @item @code{sw1 __MHSETHIS (sw1, const)}
12764 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
12765 @tab @code{MHSETHIS #@var{a},@var{b}}
12766 @item @code{uw1 __MHSETLOH (uw1, const)}
12767 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
12768 @tab @code{MHSETLOH #@var{a},@var{b}}
12769 @item @code{sw1 __MHSETLOS (sw1, const)}
12770 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
12771 @tab @code{MHSETLOS #@var{a},@var{b}}
12772 @item @code{uw1 __MHTOB (uw2)}
12773 @tab @code{@var{b} = __MHTOB (@var{a})}
12774 @tab @code{MHTOB @var{a},@var{b}}
12775 @item @code{void __MMACHS (acc, sw1, sw1)}
12776 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
12777 @tab @code{MMACHS @var{a},@var{b},@var{c}}
12778 @item @code{void __MMACHU (acc, uw1, uw1)}
12779 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
12780 @tab @code{MMACHU @var{a},@var{b},@var{c}}
12781 @item @code{void __MMRDHS (acc, sw1, sw1)}
12782 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
12783 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
12784 @item @code{void __MMRDHU (acc, uw1, uw1)}
12785 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
12786 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
12787 @item @code{void __MMULHS (acc, sw1, sw1)}
12788 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
12789 @tab @code{MMULHS @var{a},@var{b},@var{c}}
12790 @item @code{void __MMULHU (acc, uw1, uw1)}
12791 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
12792 @tab @code{MMULHU @var{a},@var{b},@var{c}}
12793 @item @code{void __MMULXHS (acc, sw1, sw1)}
12794 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
12795 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
12796 @item @code{void __MMULXHU (acc, uw1, uw1)}
12797 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
12798 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
12799 @item @code{uw1 __MNOT (uw1)}
12800 @tab @code{@var{b} = __MNOT (@var{a})}
12801 @tab @code{MNOT @var{a},@var{b}}
12802 @item @code{uw1 __MOR (uw1, uw1)}
12803 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
12804 @tab @code{MOR @var{a},@var{b},@var{c}}
12805 @item @code{uw1 __MPACKH (uh, uh)}
12806 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
12807 @tab @code{MPACKH @var{a},@var{b},@var{c}}
12808 @item @code{sw2 __MQADDHSS (sw2, sw2)}
12809 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
12810 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
12811 @item @code{uw2 __MQADDHUS (uw2, uw2)}
12812 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
12813 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
12814 @item @code{void __MQCPXIS (acc, sw2, sw2)}
12815 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
12816 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
12817 @item @code{void __MQCPXIU (acc, uw2, uw2)}
12818 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
12819 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
12820 @item @code{void __MQCPXRS (acc, sw2, sw2)}
12821 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
12822 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
12823 @item @code{void __MQCPXRU (acc, uw2, uw2)}
12824 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
12825 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
12826 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
12827 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
12828 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
12829 @item @code{sw2 __MQLMTHS (sw2, sw2)}
12830 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
12831 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
12832 @item @code{void __MQMACHS (acc, sw2, sw2)}
12833 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
12834 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
12835 @item @code{void __MQMACHU (acc, uw2, uw2)}
12836 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
12837 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
12838 @item @code{void __MQMACXHS (acc, sw2, sw2)}
12839 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
12840 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
12841 @item @code{void __MQMULHS (acc, sw2, sw2)}
12842 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
12843 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
12844 @item @code{void __MQMULHU (acc, uw2, uw2)}
12845 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
12846 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
12847 @item @code{void __MQMULXHS (acc, sw2, sw2)}
12848 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
12849 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
12850 @item @code{void __MQMULXHU (acc, uw2, uw2)}
12851 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
12852 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
12853 @item @code{sw2 __MQSATHS (sw2, sw2)}
12854 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
12855 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
12856 @item @code{uw2 __MQSLLHI (uw2, int)}
12857 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
12858 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
12859 @item @code{sw2 __MQSRAHI (sw2, int)}
12860 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
12861 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
12862 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
12863 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
12864 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
12865 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
12866 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
12867 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
12868 @item @code{void __MQXMACHS (acc, sw2, sw2)}
12869 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
12870 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
12871 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
12872 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
12873 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
12874 @item @code{uw1 __MRDACC (acc)}
12875 @tab @code{@var{b} = __MRDACC (@var{a})}
12876 @tab @code{MRDACC @var{a},@var{b}}
12877 @item @code{uw1 __MRDACCG (acc)}
12878 @tab @code{@var{b} = __MRDACCG (@var{a})}
12879 @tab @code{MRDACCG @var{a},@var{b}}
12880 @item @code{uw1 __MROTLI (uw1, const)}
12881 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
12882 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
12883 @item @code{uw1 __MROTRI (uw1, const)}
12884 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
12885 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
12886 @item @code{sw1 __MSATHS (sw1, sw1)}
12887 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
12888 @tab @code{MSATHS @var{a},@var{b},@var{c}}
12889 @item @code{uw1 __MSATHU (uw1, uw1)}
12890 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
12891 @tab @code{MSATHU @var{a},@var{b},@var{c}}
12892 @item @code{uw1 __MSLLHI (uw1, const)}
12893 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
12894 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
12895 @item @code{sw1 __MSRAHI (sw1, const)}
12896 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
12897 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
12898 @item @code{uw1 __MSRLHI (uw1, const)}
12899 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
12900 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
12901 @item @code{void __MSUBACCS (acc, acc)}
12902 @tab @code{__MSUBACCS (@var{b}, @var{a})}
12903 @tab @code{MSUBACCS @var{a},@var{b}}
12904 @item @code{sw1 __MSUBHSS (sw1, sw1)}
12905 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
12906 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
12907 @item @code{uw1 __MSUBHUS (uw1, uw1)}
12908 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
12909 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
12910 @item @code{void __MTRAP (void)}
12911 @tab @code{__MTRAP ()}
12913 @item @code{uw2 __MUNPACKH (uw1)}
12914 @tab @code{@var{b} = __MUNPACKH (@var{a})}
12915 @tab @code{MUNPACKH @var{a},@var{b}}
12916 @item @code{uw1 __MWCUT (uw2, uw1)}
12917 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
12918 @tab @code{MWCUT @var{a},@var{b},@var{c}}
12919 @item @code{void __MWTACC (acc, uw1)}
12920 @tab @code{__MWTACC (@var{b}, @var{a})}
12921 @tab @code{MWTACC @var{a},@var{b}}
12922 @item @code{void __MWTACCG (acc, uw1)}
12923 @tab @code{__MWTACCG (@var{b}, @var{a})}
12924 @tab @code{MWTACCG @var{a},@var{b}}
12925 @item @code{uw1 __MXOR (uw1, uw1)}
12926 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
12927 @tab @code{MXOR @var{a},@var{b},@var{c}}
12930 @node Raw read/write Functions
12931 @subsubsection Raw Read/Write Functions
12933 This sections describes built-in functions related to read and write
12934 instructions to access memory. These functions generate
12935 @code{membar} instructions to flush the I/O load and stores where
12936 appropriate, as described in Fujitsu's manual described above.
12940 @item unsigned char __builtin_read8 (void *@var{data})
12941 @item unsigned short __builtin_read16 (void *@var{data})
12942 @item unsigned long __builtin_read32 (void *@var{data})
12943 @item unsigned long long __builtin_read64 (void *@var{data})
12945 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
12946 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
12947 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
12948 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
12951 @node Other Built-in Functions
12952 @subsubsection Other Built-in Functions
12954 This section describes built-in functions that are not named after
12955 a specific FR-V instruction.
12958 @item sw2 __IACCreadll (iacc @var{reg})
12959 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
12960 for future expansion and must be 0.
12962 @item sw1 __IACCreadl (iacc @var{reg})
12963 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
12964 Other values of @var{reg} are rejected as invalid.
12966 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
12967 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
12968 is reserved for future expansion and must be 0.
12970 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
12971 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
12972 is 1. Other values of @var{reg} are rejected as invalid.
12974 @item void __data_prefetch0 (const void *@var{x})
12975 Use the @code{dcpl} instruction to load the contents of address @var{x}
12976 into the data cache.
12978 @item void __data_prefetch (const void *@var{x})
12979 Use the @code{nldub} instruction to load the contents of address @var{x}
12980 into the data cache. The instruction is issued in slot I1@.
12983 @node MIPS DSP Built-in Functions
12984 @subsection MIPS DSP Built-in Functions
12986 The MIPS DSP Application-Specific Extension (ASE) includes new
12987 instructions that are designed to improve the performance of DSP and
12988 media applications. It provides instructions that operate on packed
12989 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
12991 GCC supports MIPS DSP operations using both the generic
12992 vector extensions (@pxref{Vector Extensions}) and a collection of
12993 MIPS-specific built-in functions. Both kinds of support are
12994 enabled by the @option{-mdsp} command-line option.
12996 Revision 2 of the ASE was introduced in the second half of 2006.
12997 This revision adds extra instructions to the original ASE, but is
12998 otherwise backwards-compatible with it. You can select revision 2
12999 using the command-line option @option{-mdspr2}; this option implies
13002 The SCOUNT and POS bits of the DSP control register are global. The
13003 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
13004 POS bits. During optimization, the compiler does not delete these
13005 instructions and it does not delete calls to functions containing
13006 these instructions.
13008 At present, GCC only provides support for operations on 32-bit
13009 vectors. The vector type associated with 8-bit integer data is
13010 usually called @code{v4i8}, the vector type associated with Q7
13011 is usually called @code{v4q7}, the vector type associated with 16-bit
13012 integer data is usually called @code{v2i16}, and the vector type
13013 associated with Q15 is usually called @code{v2q15}. They can be
13014 defined in C as follows:
13017 typedef signed char v4i8 __attribute__ ((vector_size(4)));
13018 typedef signed char v4q7 __attribute__ ((vector_size(4)));
13019 typedef short v2i16 __attribute__ ((vector_size(4)));
13020 typedef short v2q15 __attribute__ ((vector_size(4)));
13023 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
13024 initialized in the same way as aggregates. For example:
13027 v4i8 a = @{1, 2, 3, 4@};
13029 b = (v4i8) @{5, 6, 7, 8@};
13031 v2q15 c = @{0x0fcb, 0x3a75@};
13033 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
13036 @emph{Note:} The CPU's endianness determines the order in which values
13037 are packed. On little-endian targets, the first value is the least
13038 significant and the last value is the most significant. The opposite
13039 order applies to big-endian targets. For example, the code above
13040 sets the lowest byte of @code{a} to @code{1} on little-endian targets
13041 and @code{4} on big-endian targets.
13043 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
13044 representation. As shown in this example, the integer representation
13045 of a Q7 value can be obtained by multiplying the fractional value by
13046 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
13047 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
13050 The table below lists the @code{v4i8} and @code{v2q15} operations for which
13051 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
13052 and @code{c} and @code{d} are @code{v2q15} values.
13054 @multitable @columnfractions .50 .50
13055 @item C code @tab MIPS instruction
13056 @item @code{a + b} @tab @code{addu.qb}
13057 @item @code{c + d} @tab @code{addq.ph}
13058 @item @code{a - b} @tab @code{subu.qb}
13059 @item @code{c - d} @tab @code{subq.ph}
13062 The table below lists the @code{v2i16} operation for which
13063 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
13064 @code{v2i16} values.
13066 @multitable @columnfractions .50 .50
13067 @item C code @tab MIPS instruction
13068 @item @code{e * f} @tab @code{mul.ph}
13071 It is easier to describe the DSP built-in functions if we first define
13072 the following types:
13077 typedef unsigned int ui32;
13078 typedef long long a64;
13081 @code{q31} and @code{i32} are actually the same as @code{int}, but we
13082 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
13083 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
13084 @code{long long}, but we use @code{a64} to indicate values that are
13085 placed in one of the four DSP accumulators (@code{$ac0},
13086 @code{$ac1}, @code{$ac2} or @code{$ac3}).
13088 Also, some built-in functions prefer or require immediate numbers as
13089 parameters, because the corresponding DSP instructions accept both immediate
13090 numbers and register operands, or accept immediate numbers only. The
13091 immediate parameters are listed as follows.
13099 imm0_255: 0 to 255.
13100 imm_n32_31: -32 to 31.
13101 imm_n512_511: -512 to 511.
13104 The following built-in functions map directly to a particular MIPS DSP
13105 instruction. Please refer to the architecture specification
13106 for details on what each instruction does.
13109 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
13110 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
13111 q31 __builtin_mips_addq_s_w (q31, q31)
13112 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
13113 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
13114 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
13115 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
13116 q31 __builtin_mips_subq_s_w (q31, q31)
13117 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
13118 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
13119 i32 __builtin_mips_addsc (i32, i32)
13120 i32 __builtin_mips_addwc (i32, i32)
13121 i32 __builtin_mips_modsub (i32, i32)
13122 i32 __builtin_mips_raddu_w_qb (v4i8)
13123 v2q15 __builtin_mips_absq_s_ph (v2q15)
13124 q31 __builtin_mips_absq_s_w (q31)
13125 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
13126 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
13127 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
13128 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
13129 q31 __builtin_mips_preceq_w_phl (v2q15)
13130 q31 __builtin_mips_preceq_w_phr (v2q15)
13131 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
13132 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
13133 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
13134 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
13135 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
13136 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
13137 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
13138 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
13139 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
13140 v4i8 __builtin_mips_shll_qb (v4i8, i32)
13141 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
13142 v2q15 __builtin_mips_shll_ph (v2q15, i32)
13143 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
13144 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
13145 q31 __builtin_mips_shll_s_w (q31, imm0_31)
13146 q31 __builtin_mips_shll_s_w (q31, i32)
13147 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
13148 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
13149 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
13150 v2q15 __builtin_mips_shra_ph (v2q15, i32)
13151 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
13152 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
13153 q31 __builtin_mips_shra_r_w (q31, imm0_31)
13154 q31 __builtin_mips_shra_r_w (q31, i32)
13155 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
13156 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
13157 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
13158 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
13159 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
13160 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
13161 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
13162 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
13163 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
13164 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
13165 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
13166 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
13167 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
13168 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
13169 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
13170 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
13171 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
13172 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
13173 i32 __builtin_mips_bitrev (i32)
13174 i32 __builtin_mips_insv (i32, i32)
13175 v4i8 __builtin_mips_repl_qb (imm0_255)
13176 v4i8 __builtin_mips_repl_qb (i32)
13177 v2q15 __builtin_mips_repl_ph (imm_n512_511)
13178 v2q15 __builtin_mips_repl_ph (i32)
13179 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
13180 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
13181 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
13182 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
13183 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
13184 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
13185 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
13186 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
13187 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
13188 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
13189 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
13190 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
13191 i32 __builtin_mips_extr_w (a64, imm0_31)
13192 i32 __builtin_mips_extr_w (a64, i32)
13193 i32 __builtin_mips_extr_r_w (a64, imm0_31)
13194 i32 __builtin_mips_extr_s_h (a64, i32)
13195 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
13196 i32 __builtin_mips_extr_rs_w (a64, i32)
13197 i32 __builtin_mips_extr_s_h (a64, imm0_31)
13198 i32 __builtin_mips_extr_r_w (a64, i32)
13199 i32 __builtin_mips_extp (a64, imm0_31)
13200 i32 __builtin_mips_extp (a64, i32)
13201 i32 __builtin_mips_extpdp (a64, imm0_31)
13202 i32 __builtin_mips_extpdp (a64, i32)
13203 a64 __builtin_mips_shilo (a64, imm_n32_31)
13204 a64 __builtin_mips_shilo (a64, i32)
13205 a64 __builtin_mips_mthlip (a64, i32)
13206 void __builtin_mips_wrdsp (i32, imm0_63)
13207 i32 __builtin_mips_rddsp (imm0_63)
13208 i32 __builtin_mips_lbux (void *, i32)
13209 i32 __builtin_mips_lhx (void *, i32)
13210 i32 __builtin_mips_lwx (void *, i32)
13211 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
13212 i32 __builtin_mips_bposge32 (void)
13213 a64 __builtin_mips_madd (a64, i32, i32);
13214 a64 __builtin_mips_maddu (a64, ui32, ui32);
13215 a64 __builtin_mips_msub (a64, i32, i32);
13216 a64 __builtin_mips_msubu (a64, ui32, ui32);
13217 a64 __builtin_mips_mult (i32, i32);
13218 a64 __builtin_mips_multu (ui32, ui32);
13221 The following built-in functions map directly to a particular MIPS DSP REV 2
13222 instruction. Please refer to the architecture specification
13223 for details on what each instruction does.
13226 v4q7 __builtin_mips_absq_s_qb (v4q7);
13227 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
13228 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
13229 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
13230 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
13231 i32 __builtin_mips_append (i32, i32, imm0_31);
13232 i32 __builtin_mips_balign (i32, i32, imm0_3);
13233 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
13234 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
13235 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
13236 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
13237 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
13238 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
13239 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
13240 q31 __builtin_mips_mulq_rs_w (q31, q31);
13241 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
13242 q31 __builtin_mips_mulq_s_w (q31, q31);
13243 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
13244 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
13245 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
13246 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
13247 i32 __builtin_mips_prepend (i32, i32, imm0_31);
13248 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
13249 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
13250 v4i8 __builtin_mips_shra_qb (v4i8, i32);
13251 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
13252 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
13253 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
13254 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
13255 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
13256 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
13257 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
13258 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
13259 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
13260 q31 __builtin_mips_addqh_w (q31, q31);
13261 q31 __builtin_mips_addqh_r_w (q31, q31);
13262 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
13263 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
13264 q31 __builtin_mips_subqh_w (q31, q31);
13265 q31 __builtin_mips_subqh_r_w (q31, q31);
13266 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
13267 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
13268 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
13269 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
13270 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
13271 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
13275 @node MIPS Paired-Single Support
13276 @subsection MIPS Paired-Single Support
13278 The MIPS64 architecture includes a number of instructions that
13279 operate on pairs of single-precision floating-point values.
13280 Each pair is packed into a 64-bit floating-point register,
13281 with one element being designated the ``upper half'' and
13282 the other being designated the ``lower half''.
13284 GCC supports paired-single operations using both the generic
13285 vector extensions (@pxref{Vector Extensions}) and a collection of
13286 MIPS-specific built-in functions. Both kinds of support are
13287 enabled by the @option{-mpaired-single} command-line option.
13289 The vector type associated with paired-single values is usually
13290 called @code{v2sf}. It can be defined in C as follows:
13293 typedef float v2sf __attribute__ ((vector_size (8)));
13296 @code{v2sf} values are initialized in the same way as aggregates.
13300 v2sf a = @{1.5, 9.1@};
13303 b = (v2sf) @{e, f@};
13306 @emph{Note:} The CPU's endianness determines which value is stored in
13307 the upper half of a register and which value is stored in the lower half.
13308 On little-endian targets, the first value is the lower one and the second
13309 value is the upper one. The opposite order applies to big-endian targets.
13310 For example, the code above sets the lower half of @code{a} to
13311 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
13313 @node MIPS Loongson Built-in Functions
13314 @subsection MIPS Loongson Built-in Functions
13316 GCC provides intrinsics to access the SIMD instructions provided by the
13317 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
13318 available after inclusion of the @code{loongson.h} header file,
13319 operate on the following 64-bit vector types:
13322 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
13323 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
13324 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
13325 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
13326 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
13327 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
13330 The intrinsics provided are listed below; each is named after the
13331 machine instruction to which it corresponds, with suffixes added as
13332 appropriate to distinguish intrinsics that expand to the same machine
13333 instruction yet have different argument types. Refer to the architecture
13334 documentation for a description of the functionality of each
13338 int16x4_t packsswh (int32x2_t s, int32x2_t t);
13339 int8x8_t packsshb (int16x4_t s, int16x4_t t);
13340 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
13341 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
13342 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
13343 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
13344 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
13345 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
13346 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
13347 uint64_t paddd_u (uint64_t s, uint64_t t);
13348 int64_t paddd_s (int64_t s, int64_t t);
13349 int16x4_t paddsh (int16x4_t s, int16x4_t t);
13350 int8x8_t paddsb (int8x8_t s, int8x8_t t);
13351 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
13352 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
13353 uint64_t pandn_ud (uint64_t s, uint64_t t);
13354 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
13355 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
13356 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
13357 int64_t pandn_sd (int64_t s, int64_t t);
13358 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
13359 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
13360 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
13361 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
13362 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
13363 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
13364 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
13365 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
13366 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
13367 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
13368 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
13369 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
13370 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
13371 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
13372 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
13373 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
13374 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
13375 uint16x4_t pextrh_u (uint16x4_t s, int field);
13376 int16x4_t pextrh_s (int16x4_t s, int field);
13377 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
13378 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
13379 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
13380 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
13381 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
13382 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
13383 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
13384 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
13385 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
13386 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
13387 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
13388 int16x4_t pminsh (int16x4_t s, int16x4_t t);
13389 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
13390 uint8x8_t pmovmskb_u (uint8x8_t s);
13391 int8x8_t pmovmskb_s (int8x8_t s);
13392 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
13393 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
13394 int16x4_t pmullh (int16x4_t s, int16x4_t t);
13395 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
13396 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
13397 uint16x4_t biadd (uint8x8_t s);
13398 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
13399 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
13400 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
13401 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
13402 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
13403 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
13404 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
13405 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
13406 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
13407 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
13408 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
13409 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
13410 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
13411 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
13412 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
13413 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
13414 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
13415 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
13416 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
13417 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
13418 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
13419 uint64_t psubd_u (uint64_t s, uint64_t t);
13420 int64_t psubd_s (int64_t s, int64_t t);
13421 int16x4_t psubsh (int16x4_t s, int16x4_t t);
13422 int8x8_t psubsb (int8x8_t s, int8x8_t t);
13423 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
13424 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
13425 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
13426 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
13427 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
13428 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
13429 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
13430 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
13431 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
13432 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
13433 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
13434 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
13435 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
13436 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
13440 * Paired-Single Arithmetic::
13441 * Paired-Single Built-in Functions::
13442 * MIPS-3D Built-in Functions::
13445 @node Paired-Single Arithmetic
13446 @subsubsection Paired-Single Arithmetic
13448 The table below lists the @code{v2sf} operations for which hardware
13449 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
13450 values and @code{x} is an integral value.
13452 @multitable @columnfractions .50 .50
13453 @item C code @tab MIPS instruction
13454 @item @code{a + b} @tab @code{add.ps}
13455 @item @code{a - b} @tab @code{sub.ps}
13456 @item @code{-a} @tab @code{neg.ps}
13457 @item @code{a * b} @tab @code{mul.ps}
13458 @item @code{a * b + c} @tab @code{madd.ps}
13459 @item @code{a * b - c} @tab @code{msub.ps}
13460 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
13461 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
13462 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
13465 Note that the multiply-accumulate instructions can be disabled
13466 using the command-line option @code{-mno-fused-madd}.
13468 @node Paired-Single Built-in Functions
13469 @subsubsection Paired-Single Built-in Functions
13471 The following paired-single functions map directly to a particular
13472 MIPS instruction. Please refer to the architecture specification
13473 for details on what each instruction does.
13476 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
13477 Pair lower lower (@code{pll.ps}).
13479 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
13480 Pair upper lower (@code{pul.ps}).
13482 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
13483 Pair lower upper (@code{plu.ps}).
13485 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
13486 Pair upper upper (@code{puu.ps}).
13488 @item v2sf __builtin_mips_cvt_ps_s (float, float)
13489 Convert pair to paired single (@code{cvt.ps.s}).
13491 @item float __builtin_mips_cvt_s_pl (v2sf)
13492 Convert pair lower to single (@code{cvt.s.pl}).
13494 @item float __builtin_mips_cvt_s_pu (v2sf)
13495 Convert pair upper to single (@code{cvt.s.pu}).
13497 @item v2sf __builtin_mips_abs_ps (v2sf)
13498 Absolute value (@code{abs.ps}).
13500 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
13501 Align variable (@code{alnv.ps}).
13503 @emph{Note:} The value of the third parameter must be 0 or 4
13504 modulo 8, otherwise the result is unpredictable. Please read the
13505 instruction description for details.
13508 The following multi-instruction functions are also available.
13509 In each case, @var{cond} can be any of the 16 floating-point conditions:
13510 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13511 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
13512 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13515 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13516 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13517 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
13518 @code{movt.ps}/@code{movf.ps}).
13520 The @code{movt} functions return the value @var{x} computed by:
13523 c.@var{cond}.ps @var{cc},@var{a},@var{b}
13524 mov.ps @var{x},@var{c}
13525 movt.ps @var{x},@var{d},@var{cc}
13528 The @code{movf} functions are similar but use @code{movf.ps} instead
13531 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13532 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13533 Comparison of two paired-single values (@code{c.@var{cond}.ps},
13534 @code{bc1t}/@code{bc1f}).
13536 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13537 and return either the upper or lower half of the result. For example:
13541 if (__builtin_mips_upper_c_eq_ps (a, b))
13542 upper_halves_are_equal ();
13544 upper_halves_are_unequal ();
13546 if (__builtin_mips_lower_c_eq_ps (a, b))
13547 lower_halves_are_equal ();
13549 lower_halves_are_unequal ();
13553 @node MIPS-3D Built-in Functions
13554 @subsubsection MIPS-3D Built-in Functions
13556 The MIPS-3D Application-Specific Extension (ASE) includes additional
13557 paired-single instructions that are designed to improve the performance
13558 of 3D graphics operations. Support for these instructions is controlled
13559 by the @option{-mips3d} command-line option.
13561 The functions listed below map directly to a particular MIPS-3D
13562 instruction. Please refer to the architecture specification for
13563 more details on what each instruction does.
13566 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
13567 Reduction add (@code{addr.ps}).
13569 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
13570 Reduction multiply (@code{mulr.ps}).
13572 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
13573 Convert paired single to paired word (@code{cvt.pw.ps}).
13575 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
13576 Convert paired word to paired single (@code{cvt.ps.pw}).
13578 @item float __builtin_mips_recip1_s (float)
13579 @itemx double __builtin_mips_recip1_d (double)
13580 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
13581 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
13583 @item float __builtin_mips_recip2_s (float, float)
13584 @itemx double __builtin_mips_recip2_d (double, double)
13585 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
13586 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
13588 @item float __builtin_mips_rsqrt1_s (float)
13589 @itemx double __builtin_mips_rsqrt1_d (double)
13590 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
13591 Reduced-precision reciprocal square root (sequence step 1)
13592 (@code{rsqrt1.@var{fmt}}).
13594 @item float __builtin_mips_rsqrt2_s (float, float)
13595 @itemx double __builtin_mips_rsqrt2_d (double, double)
13596 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
13597 Reduced-precision reciprocal square root (sequence step 2)
13598 (@code{rsqrt2.@var{fmt}}).
13601 The following multi-instruction functions are also available.
13602 In each case, @var{cond} can be any of the 16 floating-point conditions:
13603 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
13604 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
13605 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
13608 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
13609 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
13610 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
13611 @code{bc1t}/@code{bc1f}).
13613 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
13614 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
13619 if (__builtin_mips_cabs_eq_s (a, b))
13625 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13626 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13627 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
13628 @code{bc1t}/@code{bc1f}).
13630 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
13631 and return either the upper or lower half of the result. For example:
13635 if (__builtin_mips_upper_cabs_eq_ps (a, b))
13636 upper_halves_are_equal ();
13638 upper_halves_are_unequal ();
13640 if (__builtin_mips_lower_cabs_eq_ps (a, b))
13641 lower_halves_are_equal ();
13643 lower_halves_are_unequal ();
13646 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13647 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13648 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
13649 @code{movt.ps}/@code{movf.ps}).
13651 The @code{movt} functions return the value @var{x} computed by:
13654 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
13655 mov.ps @var{x},@var{c}
13656 movt.ps @var{x},@var{d},@var{cc}
13659 The @code{movf} functions are similar but use @code{movf.ps} instead
13662 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13663 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13664 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13665 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
13666 Comparison of two paired-single values
13667 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13668 @code{bc1any2t}/@code{bc1any2f}).
13670 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
13671 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
13672 result is true and the @code{all} forms return true if both results are true.
13677 if (__builtin_mips_any_c_eq_ps (a, b))
13682 if (__builtin_mips_all_c_eq_ps (a, b))
13688 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13689 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13690 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13691 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
13692 Comparison of four paired-single values
13693 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
13694 @code{bc1any4t}/@code{bc1any4f}).
13696 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
13697 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
13698 The @code{any} forms return true if any of the four results are true
13699 and the @code{all} forms return true if all four results are true.
13704 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
13709 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
13716 @node MIPS SIMD Architecture (MSA) Support
13717 @subsection MIPS SIMD Architecture (MSA) Support
13720 * MIPS SIMD Architecture Built-in Functions::
13723 GCC provides intrinsics to access the SIMD instructions provided by the
13724 MSA MIPS SIMD Architecture. The interface is made available by including
13725 @code{<msa.h>} and using @option{-mmsa -mhard-float -mfp64 -mnan=2008}.
13726 For each @code{__builtin_msa_*}, there is a shortened name of the intrinsic,
13729 MSA implements 128-bit wide vector registers, operating on 8-, 16-, 32- and
13730 64-bit integer, 16- and 32-bit fixed-point, or 32- and 64-bit floating point
13731 data elements. The following vectors typedefs are included in @code{msa.h}:
13733 @item @code{v16i8}, a vector of sixteen signed 8-bit integers;
13734 @item @code{v16u8}, a vector of sixteen unsigned 8-bit integers;
13735 @item @code{v8i16}, a vector of eight signed 16-bit integers;
13736 @item @code{v8u16}, a vector of eight unsigned 16-bit integers;
13737 @item @code{v4i32}, a vector of four signed 32-bit integers;
13738 @item @code{v4u32}, a vector of four unsigned 32-bit integers;
13739 @item @code{v2i64}, a vector of two signed 64-bit integers;
13740 @item @code{v2u64}, a vector of two unsigned 64-bit integers;
13741 @item @code{v4f32}, a vector of four 32-bit floats;
13742 @item @code{v2f64}, a vector of two 64-bit doubles.
13745 Intructions and corresponding built-ins may have additional restrictions and/or
13746 input/output values manipulated:
13748 @item @code{imm0_1}, an integer literal in range 0 to 1;
13749 @item @code{imm0_3}, an integer literal in range 0 to 3;
13750 @item @code{imm0_7}, an integer literal in range 0 to 7;
13751 @item @code{imm0_15}, an integer literal in range 0 to 15;
13752 @item @code{imm0_31}, an integer literal in range 0 to 31;
13753 @item @code{imm0_63}, an integer literal in range 0 to 63;
13754 @item @code{imm0_255}, an integer literal in range 0 to 255;
13755 @item @code{imm_n16_15}, an integer literal in range -16 to 15;
13756 @item @code{imm_n512_511}, an integer literal in range -512 to 511;
13757 @item @code{imm_n1024_1022}, an integer literal in range -512 to 511 left
13758 shifted by 1 bit, i.e., -1024, -1022, @dots{}, 1020, 1022;
13759 @item @code{imm_n2048_2044}, an integer literal in range -512 to 511 left
13760 shifted by 2 bits, i.e., -2048, -2044, @dots{}, 2040, 2044;
13761 @item @code{imm_n4096_4088}, an integer literal in range -512 to 511 left
13762 shifted by 3 bits, i.e., -4096, -4088, @dots{}, 4080, 4088;
13763 @item @code{imm1_4}, an integer literal in range 1 to 4;
13764 @item @code{i32, i64, u32, u64, f32, f64}, defined as follows:
13770 #if __LONG_MAX__ == __LONG_LONG_MAX__
13773 typedef long long i64;
13776 typedef unsigned int u32;
13777 #if __LONG_MAX__ == __LONG_LONG_MAX__
13778 typedef unsigned long u64;
13780 typedef unsigned long long u64;
13783 typedef double f64;
13788 @node MIPS SIMD Architecture Built-in Functions
13789 @subsubsection MIPS SIMD Architecture Built-in Functions
13791 The intrinsics provided are listed below; each is named after the
13792 machine instruction.
13795 v16i8 __builtin_msa_add_a_b (v16i8, v16i8);
13796 v8i16 __builtin_msa_add_a_h (v8i16, v8i16);
13797 v4i32 __builtin_msa_add_a_w (v4i32, v4i32);
13798 v2i64 __builtin_msa_add_a_d (v2i64, v2i64);
13800 v16i8 __builtin_msa_adds_a_b (v16i8, v16i8);
13801 v8i16 __builtin_msa_adds_a_h (v8i16, v8i16);
13802 v4i32 __builtin_msa_adds_a_w (v4i32, v4i32);
13803 v2i64 __builtin_msa_adds_a_d (v2i64, v2i64);
13805 v16i8 __builtin_msa_adds_s_b (v16i8, v16i8);
13806 v8i16 __builtin_msa_adds_s_h (v8i16, v8i16);
13807 v4i32 __builtin_msa_adds_s_w (v4i32, v4i32);
13808 v2i64 __builtin_msa_adds_s_d (v2i64, v2i64);
13810 v16u8 __builtin_msa_adds_u_b (v16u8, v16u8);
13811 v8u16 __builtin_msa_adds_u_h (v8u16, v8u16);
13812 v4u32 __builtin_msa_adds_u_w (v4u32, v4u32);
13813 v2u64 __builtin_msa_adds_u_d (v2u64, v2u64);
13815 v16i8 __builtin_msa_addv_b (v16i8, v16i8);
13816 v8i16 __builtin_msa_addv_h (v8i16, v8i16);
13817 v4i32 __builtin_msa_addv_w (v4i32, v4i32);
13818 v2i64 __builtin_msa_addv_d (v2i64, v2i64);
13820 v16i8 __builtin_msa_addvi_b (v16i8, imm0_31);
13821 v8i16 __builtin_msa_addvi_h (v8i16, imm0_31);
13822 v4i32 __builtin_msa_addvi_w (v4i32, imm0_31);
13823 v2i64 __builtin_msa_addvi_d (v2i64, imm0_31);
13825 v16u8 __builtin_msa_and_v (v16u8, v16u8);
13827 v16u8 __builtin_msa_andi_b (v16u8, imm0_255);
13829 v16i8 __builtin_msa_asub_s_b (v16i8, v16i8);
13830 v8i16 __builtin_msa_asub_s_h (v8i16, v8i16);
13831 v4i32 __builtin_msa_asub_s_w (v4i32, v4i32);
13832 v2i64 __builtin_msa_asub_s_d (v2i64, v2i64);
13834 v16u8 __builtin_msa_asub_u_b (v16u8, v16u8);
13835 v8u16 __builtin_msa_asub_u_h (v8u16, v8u16);
13836 v4u32 __builtin_msa_asub_u_w (v4u32, v4u32);
13837 v2u64 __builtin_msa_asub_u_d (v2u64, v2u64);
13839 v16i8 __builtin_msa_ave_s_b (v16i8, v16i8);
13840 v8i16 __builtin_msa_ave_s_h (v8i16, v8i16);
13841 v4i32 __builtin_msa_ave_s_w (v4i32, v4i32);
13842 v2i64 __builtin_msa_ave_s_d (v2i64, v2i64);
13844 v16u8 __builtin_msa_ave_u_b (v16u8, v16u8);
13845 v8u16 __builtin_msa_ave_u_h (v8u16, v8u16);
13846 v4u32 __builtin_msa_ave_u_w (v4u32, v4u32);
13847 v2u64 __builtin_msa_ave_u_d (v2u64, v2u64);
13849 v16i8 __builtin_msa_aver_s_b (v16i8, v16i8);
13850 v8i16 __builtin_msa_aver_s_h (v8i16, v8i16);
13851 v4i32 __builtin_msa_aver_s_w (v4i32, v4i32);
13852 v2i64 __builtin_msa_aver_s_d (v2i64, v2i64);
13854 v16u8 __builtin_msa_aver_u_b (v16u8, v16u8);
13855 v8u16 __builtin_msa_aver_u_h (v8u16, v8u16);
13856 v4u32 __builtin_msa_aver_u_w (v4u32, v4u32);
13857 v2u64 __builtin_msa_aver_u_d (v2u64, v2u64);
13859 v16u8 __builtin_msa_bclr_b (v16u8, v16u8);
13860 v8u16 __builtin_msa_bclr_h (v8u16, v8u16);
13861 v4u32 __builtin_msa_bclr_w (v4u32, v4u32);
13862 v2u64 __builtin_msa_bclr_d (v2u64, v2u64);
13864 v16u8 __builtin_msa_bclri_b (v16u8, imm0_7);
13865 v8u16 __builtin_msa_bclri_h (v8u16, imm0_15);
13866 v4u32 __builtin_msa_bclri_w (v4u32, imm0_31);
13867 v2u64 __builtin_msa_bclri_d (v2u64, imm0_63);
13869 v16u8 __builtin_msa_binsl_b (v16u8, v16u8, v16u8);
13870 v8u16 __builtin_msa_binsl_h (v8u16, v8u16, v8u16);
13871 v4u32 __builtin_msa_binsl_w (v4u32, v4u32, v4u32);
13872 v2u64 __builtin_msa_binsl_d (v2u64, v2u64, v2u64);
13874 v16u8 __builtin_msa_binsli_b (v16u8, v16u8, imm0_7);
13875 v8u16 __builtin_msa_binsli_h (v8u16, v8u16, imm0_15);
13876 v4u32 __builtin_msa_binsli_w (v4u32, v4u32, imm0_31);
13877 v2u64 __builtin_msa_binsli_d (v2u64, v2u64, imm0_63);
13879 v16u8 __builtin_msa_binsr_b (v16u8, v16u8, v16u8);
13880 v8u16 __builtin_msa_binsr_h (v8u16, v8u16, v8u16);
13881 v4u32 __builtin_msa_binsr_w (v4u32, v4u32, v4u32);
13882 v2u64 __builtin_msa_binsr_d (v2u64, v2u64, v2u64);
13884 v16u8 __builtin_msa_binsri_b (v16u8, v16u8, imm0_7);
13885 v8u16 __builtin_msa_binsri_h (v8u16, v8u16, imm0_15);
13886 v4u32 __builtin_msa_binsri_w (v4u32, v4u32, imm0_31);
13887 v2u64 __builtin_msa_binsri_d (v2u64, v2u64, imm0_63);
13889 v16u8 __builtin_msa_bmnz_v (v16u8, v16u8, v16u8);
13891 v16u8 __builtin_msa_bmnzi_b (v16u8, v16u8, imm0_255);
13893 v16u8 __builtin_msa_bmz_v (v16u8, v16u8, v16u8);
13895 v16u8 __builtin_msa_bmzi_b (v16u8, v16u8, imm0_255);
13897 v16u8 __builtin_msa_bneg_b (v16u8, v16u8);
13898 v8u16 __builtin_msa_bneg_h (v8u16, v8u16);
13899 v4u32 __builtin_msa_bneg_w (v4u32, v4u32);
13900 v2u64 __builtin_msa_bneg_d (v2u64, v2u64);
13902 v16u8 __builtin_msa_bnegi_b (v16u8, imm0_7);
13903 v8u16 __builtin_msa_bnegi_h (v8u16, imm0_15);
13904 v4u32 __builtin_msa_bnegi_w (v4u32, imm0_31);
13905 v2u64 __builtin_msa_bnegi_d (v2u64, imm0_63);
13907 i32 __builtin_msa_bnz_b (v16u8);
13908 i32 __builtin_msa_bnz_h (v8u16);
13909 i32 __builtin_msa_bnz_w (v4u32);
13910 i32 __builtin_msa_bnz_d (v2u64);
13912 i32 __builtin_msa_bnz_v (v16u8);
13914 v16u8 __builtin_msa_bsel_v (v16u8, v16u8, v16u8);
13916 v16u8 __builtin_msa_bseli_b (v16u8, v16u8, imm0_255);
13918 v16u8 __builtin_msa_bset_b (v16u8, v16u8);
13919 v8u16 __builtin_msa_bset_h (v8u16, v8u16);
13920 v4u32 __builtin_msa_bset_w (v4u32, v4u32);
13921 v2u64 __builtin_msa_bset_d (v2u64, v2u64);
13923 v16u8 __builtin_msa_bseti_b (v16u8, imm0_7);
13924 v8u16 __builtin_msa_bseti_h (v8u16, imm0_15);
13925 v4u32 __builtin_msa_bseti_w (v4u32, imm0_31);
13926 v2u64 __builtin_msa_bseti_d (v2u64, imm0_63);
13928 i32 __builtin_msa_bz_b (v16u8);
13929 i32 __builtin_msa_bz_h (v8u16);
13930 i32 __builtin_msa_bz_w (v4u32);
13931 i32 __builtin_msa_bz_d (v2u64);
13933 i32 __builtin_msa_bz_v (v16u8);
13935 v16i8 __builtin_msa_ceq_b (v16i8, v16i8);
13936 v8i16 __builtin_msa_ceq_h (v8i16, v8i16);
13937 v4i32 __builtin_msa_ceq_w (v4i32, v4i32);
13938 v2i64 __builtin_msa_ceq_d (v2i64, v2i64);
13940 v16i8 __builtin_msa_ceqi_b (v16i8, imm_n16_15);
13941 v8i16 __builtin_msa_ceqi_h (v8i16, imm_n16_15);
13942 v4i32 __builtin_msa_ceqi_w (v4i32, imm_n16_15);
13943 v2i64 __builtin_msa_ceqi_d (v2i64, imm_n16_15);
13945 i32 __builtin_msa_cfcmsa (imm0_31);
13947 v16i8 __builtin_msa_cle_s_b (v16i8, v16i8);
13948 v8i16 __builtin_msa_cle_s_h (v8i16, v8i16);
13949 v4i32 __builtin_msa_cle_s_w (v4i32, v4i32);
13950 v2i64 __builtin_msa_cle_s_d (v2i64, v2i64);
13952 v16i8 __builtin_msa_cle_u_b (v16u8, v16u8);
13953 v8i16 __builtin_msa_cle_u_h (v8u16, v8u16);
13954 v4i32 __builtin_msa_cle_u_w (v4u32, v4u32);
13955 v2i64 __builtin_msa_cle_u_d (v2u64, v2u64);
13957 v16i8 __builtin_msa_clei_s_b (v16i8, imm_n16_15);
13958 v8i16 __builtin_msa_clei_s_h (v8i16, imm_n16_15);
13959 v4i32 __builtin_msa_clei_s_w (v4i32, imm_n16_15);
13960 v2i64 __builtin_msa_clei_s_d (v2i64, imm_n16_15);
13962 v16i8 __builtin_msa_clei_u_b (v16u8, imm0_31);
13963 v8i16 __builtin_msa_clei_u_h (v8u16, imm0_31);
13964 v4i32 __builtin_msa_clei_u_w (v4u32, imm0_31);
13965 v2i64 __builtin_msa_clei_u_d (v2u64, imm0_31);
13967 v16i8 __builtin_msa_clt_s_b (v16i8, v16i8);
13968 v8i16 __builtin_msa_clt_s_h (v8i16, v8i16);
13969 v4i32 __builtin_msa_clt_s_w (v4i32, v4i32);
13970 v2i64 __builtin_msa_clt_s_d (v2i64, v2i64);
13972 v16i8 __builtin_msa_clt_u_b (v16u8, v16u8);
13973 v8i16 __builtin_msa_clt_u_h (v8u16, v8u16);
13974 v4i32 __builtin_msa_clt_u_w (v4u32, v4u32);
13975 v2i64 __builtin_msa_clt_u_d (v2u64, v2u64);
13977 v16i8 __builtin_msa_clti_s_b (v16i8, imm_n16_15);
13978 v8i16 __builtin_msa_clti_s_h (v8i16, imm_n16_15);
13979 v4i32 __builtin_msa_clti_s_w (v4i32, imm_n16_15);
13980 v2i64 __builtin_msa_clti_s_d (v2i64, imm_n16_15);
13982 v16i8 __builtin_msa_clti_u_b (v16u8, imm0_31);
13983 v8i16 __builtin_msa_clti_u_h (v8u16, imm0_31);
13984 v4i32 __builtin_msa_clti_u_w (v4u32, imm0_31);
13985 v2i64 __builtin_msa_clti_u_d (v2u64, imm0_31);
13987 i32 __builtin_msa_copy_s_b (v16i8, imm0_15);
13988 i32 __builtin_msa_copy_s_h (v8i16, imm0_7);
13989 i32 __builtin_msa_copy_s_w (v4i32, imm0_3);
13990 i64 __builtin_msa_copy_s_d (v2i64, imm0_1);
13992 u32 __builtin_msa_copy_u_b (v16i8, imm0_15);
13993 u32 __builtin_msa_copy_u_h (v8i16, imm0_7);
13994 u32 __builtin_msa_copy_u_w (v4i32, imm0_3);
13995 u64 __builtin_msa_copy_u_d (v2i64, imm0_1);
13997 void __builtin_msa_ctcmsa (imm0_31, i32);
13999 v16i8 __builtin_msa_div_s_b (v16i8, v16i8);
14000 v8i16 __builtin_msa_div_s_h (v8i16, v8i16);
14001 v4i32 __builtin_msa_div_s_w (v4i32, v4i32);
14002 v2i64 __builtin_msa_div_s_d (v2i64, v2i64);
14004 v16u8 __builtin_msa_div_u_b (v16u8, v16u8);
14005 v8u16 __builtin_msa_div_u_h (v8u16, v8u16);
14006 v4u32 __builtin_msa_div_u_w (v4u32, v4u32);
14007 v2u64 __builtin_msa_div_u_d (v2u64, v2u64);
14009 v8i16 __builtin_msa_dotp_s_h (v16i8, v16i8);
14010 v4i32 __builtin_msa_dotp_s_w (v8i16, v8i16);
14011 v2i64 __builtin_msa_dotp_s_d (v4i32, v4i32);
14013 v8u16 __builtin_msa_dotp_u_h (v16u8, v16u8);
14014 v4u32 __builtin_msa_dotp_u_w (v8u16, v8u16);
14015 v2u64 __builtin_msa_dotp_u_d (v4u32, v4u32);
14017 v8i16 __builtin_msa_dpadd_s_h (v8i16, v16i8, v16i8);
14018 v4i32 __builtin_msa_dpadd_s_w (v4i32, v8i16, v8i16);
14019 v2i64 __builtin_msa_dpadd_s_d (v2i64, v4i32, v4i32);
14021 v8u16 __builtin_msa_dpadd_u_h (v8u16, v16u8, v16u8);
14022 v4u32 __builtin_msa_dpadd_u_w (v4u32, v8u16, v8u16);
14023 v2u64 __builtin_msa_dpadd_u_d (v2u64, v4u32, v4u32);
14025 v8i16 __builtin_msa_dpsub_s_h (v8i16, v16i8, v16i8);
14026 v4i32 __builtin_msa_dpsub_s_w (v4i32, v8i16, v8i16);
14027 v2i64 __builtin_msa_dpsub_s_d (v2i64, v4i32, v4i32);
14029 v8i16 __builtin_msa_dpsub_u_h (v8i16, v16u8, v16u8);
14030 v4i32 __builtin_msa_dpsub_u_w (v4i32, v8u16, v8u16);
14031 v2i64 __builtin_msa_dpsub_u_d (v2i64, v4u32, v4u32);
14033 v4f32 __builtin_msa_fadd_w (v4f32, v4f32);
14034 v2f64 __builtin_msa_fadd_d (v2f64, v2f64);
14036 v4i32 __builtin_msa_fcaf_w (v4f32, v4f32);
14037 v2i64 __builtin_msa_fcaf_d (v2f64, v2f64);
14039 v4i32 __builtin_msa_fceq_w (v4f32, v4f32);
14040 v2i64 __builtin_msa_fceq_d (v2f64, v2f64);
14042 v4i32 __builtin_msa_fclass_w (v4f32);
14043 v2i64 __builtin_msa_fclass_d (v2f64);
14045 v4i32 __builtin_msa_fcle_w (v4f32, v4f32);
14046 v2i64 __builtin_msa_fcle_d (v2f64, v2f64);
14048 v4i32 __builtin_msa_fclt_w (v4f32, v4f32);
14049 v2i64 __builtin_msa_fclt_d (v2f64, v2f64);
14051 v4i32 __builtin_msa_fcne_w (v4f32, v4f32);
14052 v2i64 __builtin_msa_fcne_d (v2f64, v2f64);
14054 v4i32 __builtin_msa_fcor_w (v4f32, v4f32);
14055 v2i64 __builtin_msa_fcor_d (v2f64, v2f64);
14057 v4i32 __builtin_msa_fcueq_w (v4f32, v4f32);
14058 v2i64 __builtin_msa_fcueq_d (v2f64, v2f64);
14060 v4i32 __builtin_msa_fcule_w (v4f32, v4f32);
14061 v2i64 __builtin_msa_fcule_d (v2f64, v2f64);
14063 v4i32 __builtin_msa_fcult_w (v4f32, v4f32);
14064 v2i64 __builtin_msa_fcult_d (v2f64, v2f64);
14066 v4i32 __builtin_msa_fcun_w (v4f32, v4f32);
14067 v2i64 __builtin_msa_fcun_d (v2f64, v2f64);
14069 v4i32 __builtin_msa_fcune_w (v4f32, v4f32);
14070 v2i64 __builtin_msa_fcune_d (v2f64, v2f64);
14072 v4f32 __builtin_msa_fdiv_w (v4f32, v4f32);
14073 v2f64 __builtin_msa_fdiv_d (v2f64, v2f64);
14075 v8i16 __builtin_msa_fexdo_h (v4f32, v4f32);
14076 v4f32 __builtin_msa_fexdo_w (v2f64, v2f64);
14078 v4f32 __builtin_msa_fexp2_w (v4f32, v4i32);
14079 v2f64 __builtin_msa_fexp2_d (v2f64, v2i64);
14081 v4f32 __builtin_msa_fexupl_w (v8i16);
14082 v2f64 __builtin_msa_fexupl_d (v4f32);
14084 v4f32 __builtin_msa_fexupr_w (v8i16);
14085 v2f64 __builtin_msa_fexupr_d (v4f32);
14087 v4f32 __builtin_msa_ffint_s_w (v4i32);
14088 v2f64 __builtin_msa_ffint_s_d (v2i64);
14090 v4f32 __builtin_msa_ffint_u_w (v4u32);
14091 v2f64 __builtin_msa_ffint_u_d (v2u64);
14093 v4f32 __builtin_msa_ffql_w (v8i16);
14094 v2f64 __builtin_msa_ffql_d (v4i32);
14096 v4f32 __builtin_msa_ffqr_w (v8i16);
14097 v2f64 __builtin_msa_ffqr_d (v4i32);
14099 v16i8 __builtin_msa_fill_b (i32);
14100 v8i16 __builtin_msa_fill_h (i32);
14101 v4i32 __builtin_msa_fill_w (i32);
14102 v2i64 __builtin_msa_fill_d (i64);
14104 v4f32 __builtin_msa_flog2_w (v4f32);
14105 v2f64 __builtin_msa_flog2_d (v2f64);
14107 v4f32 __builtin_msa_fmadd_w (v4f32, v4f32, v4f32);
14108 v2f64 __builtin_msa_fmadd_d (v2f64, v2f64, v2f64);
14110 v4f32 __builtin_msa_fmax_w (v4f32, v4f32);
14111 v2f64 __builtin_msa_fmax_d (v2f64, v2f64);
14113 v4f32 __builtin_msa_fmax_a_w (v4f32, v4f32);
14114 v2f64 __builtin_msa_fmax_a_d (v2f64, v2f64);
14116 v4f32 __builtin_msa_fmin_w (v4f32, v4f32);
14117 v2f64 __builtin_msa_fmin_d (v2f64, v2f64);
14119 v4f32 __builtin_msa_fmin_a_w (v4f32, v4f32);
14120 v2f64 __builtin_msa_fmin_a_d (v2f64, v2f64);
14122 v4f32 __builtin_msa_fmsub_w (v4f32, v4f32, v4f32);
14123 v2f64 __builtin_msa_fmsub_d (v2f64, v2f64, v2f64);
14125 v4f32 __builtin_msa_fmul_w (v4f32, v4f32);
14126 v2f64 __builtin_msa_fmul_d (v2f64, v2f64);
14128 v4f32 __builtin_msa_frint_w (v4f32);
14129 v2f64 __builtin_msa_frint_d (v2f64);
14131 v4f32 __builtin_msa_frcp_w (v4f32);
14132 v2f64 __builtin_msa_frcp_d (v2f64);
14134 v4f32 __builtin_msa_frsqrt_w (v4f32);
14135 v2f64 __builtin_msa_frsqrt_d (v2f64);
14137 v4i32 __builtin_msa_fsaf_w (v4f32, v4f32);
14138 v2i64 __builtin_msa_fsaf_d (v2f64, v2f64);
14140 v4i32 __builtin_msa_fseq_w (v4f32, v4f32);
14141 v2i64 __builtin_msa_fseq_d (v2f64, v2f64);
14143 v4i32 __builtin_msa_fsle_w (v4f32, v4f32);
14144 v2i64 __builtin_msa_fsle_d (v2f64, v2f64);
14146 v4i32 __builtin_msa_fslt_w (v4f32, v4f32);
14147 v2i64 __builtin_msa_fslt_d (v2f64, v2f64);
14149 v4i32 __builtin_msa_fsne_w (v4f32, v4f32);
14150 v2i64 __builtin_msa_fsne_d (v2f64, v2f64);
14152 v4i32 __builtin_msa_fsor_w (v4f32, v4f32);
14153 v2i64 __builtin_msa_fsor_d (v2f64, v2f64);
14155 v4f32 __builtin_msa_fsqrt_w (v4f32);
14156 v2f64 __builtin_msa_fsqrt_d (v2f64);
14158 v4f32 __builtin_msa_fsub_w (v4f32, v4f32);
14159 v2f64 __builtin_msa_fsub_d (v2f64, v2f64);
14161 v4i32 __builtin_msa_fsueq_w (v4f32, v4f32);
14162 v2i64 __builtin_msa_fsueq_d (v2f64, v2f64);
14164 v4i32 __builtin_msa_fsule_w (v4f32, v4f32);
14165 v2i64 __builtin_msa_fsule_d (v2f64, v2f64);
14167 v4i32 __builtin_msa_fsult_w (v4f32, v4f32);
14168 v2i64 __builtin_msa_fsult_d (v2f64, v2f64);
14170 v4i32 __builtin_msa_fsun_w (v4f32, v4f32);
14171 v2i64 __builtin_msa_fsun_d (v2f64, v2f64);
14173 v4i32 __builtin_msa_fsune_w (v4f32, v4f32);
14174 v2i64 __builtin_msa_fsune_d (v2f64, v2f64);
14176 v4i32 __builtin_msa_ftint_s_w (v4f32);
14177 v2i64 __builtin_msa_ftint_s_d (v2f64);
14179 v4u32 __builtin_msa_ftint_u_w (v4f32);
14180 v2u64 __builtin_msa_ftint_u_d (v2f64);
14182 v8i16 __builtin_msa_ftq_h (v4f32, v4f32);
14183 v4i32 __builtin_msa_ftq_w (v2f64, v2f64);
14185 v4i32 __builtin_msa_ftrunc_s_w (v4f32);
14186 v2i64 __builtin_msa_ftrunc_s_d (v2f64);
14188 v4u32 __builtin_msa_ftrunc_u_w (v4f32);
14189 v2u64 __builtin_msa_ftrunc_u_d (v2f64);
14191 v8i16 __builtin_msa_hadd_s_h (v16i8, v16i8);
14192 v4i32 __builtin_msa_hadd_s_w (v8i16, v8i16);
14193 v2i64 __builtin_msa_hadd_s_d (v4i32, v4i32);
14195 v8u16 __builtin_msa_hadd_u_h (v16u8, v16u8);
14196 v4u32 __builtin_msa_hadd_u_w (v8u16, v8u16);
14197 v2u64 __builtin_msa_hadd_u_d (v4u32, v4u32);
14199 v8i16 __builtin_msa_hsub_s_h (v16i8, v16i8);
14200 v4i32 __builtin_msa_hsub_s_w (v8i16, v8i16);
14201 v2i64 __builtin_msa_hsub_s_d (v4i32, v4i32);
14203 v8i16 __builtin_msa_hsub_u_h (v16u8, v16u8);
14204 v4i32 __builtin_msa_hsub_u_w (v8u16, v8u16);
14205 v2i64 __builtin_msa_hsub_u_d (v4u32, v4u32);
14207 v16i8 __builtin_msa_ilvev_b (v16i8, v16i8);
14208 v8i16 __builtin_msa_ilvev_h (v8i16, v8i16);
14209 v4i32 __builtin_msa_ilvev_w (v4i32, v4i32);
14210 v2i64 __builtin_msa_ilvev_d (v2i64, v2i64);
14212 v16i8 __builtin_msa_ilvl_b (v16i8, v16i8);
14213 v8i16 __builtin_msa_ilvl_h (v8i16, v8i16);
14214 v4i32 __builtin_msa_ilvl_w (v4i32, v4i32);
14215 v2i64 __builtin_msa_ilvl_d (v2i64, v2i64);
14217 v16i8 __builtin_msa_ilvod_b (v16i8, v16i8);
14218 v8i16 __builtin_msa_ilvod_h (v8i16, v8i16);
14219 v4i32 __builtin_msa_ilvod_w (v4i32, v4i32);
14220 v2i64 __builtin_msa_ilvod_d (v2i64, v2i64);
14222 v16i8 __builtin_msa_ilvr_b (v16i8, v16i8);
14223 v8i16 __builtin_msa_ilvr_h (v8i16, v8i16);
14224 v4i32 __builtin_msa_ilvr_w (v4i32, v4i32);
14225 v2i64 __builtin_msa_ilvr_d (v2i64, v2i64);
14227 v16i8 __builtin_msa_insert_b (v16i8, imm0_15, i32);
14228 v8i16 __builtin_msa_insert_h (v8i16, imm0_7, i32);
14229 v4i32 __builtin_msa_insert_w (v4i32, imm0_3, i32);
14230 v2i64 __builtin_msa_insert_d (v2i64, imm0_1, i64);
14232 v16i8 __builtin_msa_insve_b (v16i8, imm0_15, v16i8);
14233 v8i16 __builtin_msa_insve_h (v8i16, imm0_7, v8i16);
14234 v4i32 __builtin_msa_insve_w (v4i32, imm0_3, v4i32);
14235 v2i64 __builtin_msa_insve_d (v2i64, imm0_1, v2i64);
14237 v16i8 __builtin_msa_ld_b (void *, imm_n512_511);
14238 v8i16 __builtin_msa_ld_h (void *, imm_n1024_1022);
14239 v4i32 __builtin_msa_ld_w (void *, imm_n2048_2044);
14240 v2i64 __builtin_msa_ld_d (void *, imm_n4096_4088);
14242 v16i8 __builtin_msa_ldi_b (imm_n512_511);
14243 v8i16 __builtin_msa_ldi_h (imm_n512_511);
14244 v4i32 __builtin_msa_ldi_w (imm_n512_511);
14245 v2i64 __builtin_msa_ldi_d (imm_n512_511);
14247 v8i16 __builtin_msa_madd_q_h (v8i16, v8i16, v8i16);
14248 v4i32 __builtin_msa_madd_q_w (v4i32, v4i32, v4i32);
14250 v8i16 __builtin_msa_maddr_q_h (v8i16, v8i16, v8i16);
14251 v4i32 __builtin_msa_maddr_q_w (v4i32, v4i32, v4i32);
14253 v16i8 __builtin_msa_maddv_b (v16i8, v16i8, v16i8);
14254 v8i16 __builtin_msa_maddv_h (v8i16, v8i16, v8i16);
14255 v4i32 __builtin_msa_maddv_w (v4i32, v4i32, v4i32);
14256 v2i64 __builtin_msa_maddv_d (v2i64, v2i64, v2i64);
14258 v16i8 __builtin_msa_max_a_b (v16i8, v16i8);
14259 v8i16 __builtin_msa_max_a_h (v8i16, v8i16);
14260 v4i32 __builtin_msa_max_a_w (v4i32, v4i32);
14261 v2i64 __builtin_msa_max_a_d (v2i64, v2i64);
14263 v16i8 __builtin_msa_max_s_b (v16i8, v16i8);
14264 v8i16 __builtin_msa_max_s_h (v8i16, v8i16);
14265 v4i32 __builtin_msa_max_s_w (v4i32, v4i32);
14266 v2i64 __builtin_msa_max_s_d (v2i64, v2i64);
14268 v16u8 __builtin_msa_max_u_b (v16u8, v16u8);
14269 v8u16 __builtin_msa_max_u_h (v8u16, v8u16);
14270 v4u32 __builtin_msa_max_u_w (v4u32, v4u32);
14271 v2u64 __builtin_msa_max_u_d (v2u64, v2u64);
14273 v16i8 __builtin_msa_maxi_s_b (v16i8, imm_n16_15);
14274 v8i16 __builtin_msa_maxi_s_h (v8i16, imm_n16_15);
14275 v4i32 __builtin_msa_maxi_s_w (v4i32, imm_n16_15);
14276 v2i64 __builtin_msa_maxi_s_d (v2i64, imm_n16_15);
14278 v16u8 __builtin_msa_maxi_u_b (v16u8, imm0_31);
14279 v8u16 __builtin_msa_maxi_u_h (v8u16, imm0_31);
14280 v4u32 __builtin_msa_maxi_u_w (v4u32, imm0_31);
14281 v2u64 __builtin_msa_maxi_u_d (v2u64, imm0_31);
14283 v16i8 __builtin_msa_min_a_b (v16i8, v16i8);
14284 v8i16 __builtin_msa_min_a_h (v8i16, v8i16);
14285 v4i32 __builtin_msa_min_a_w (v4i32, v4i32);
14286 v2i64 __builtin_msa_min_a_d (v2i64, v2i64);
14288 v16i8 __builtin_msa_min_s_b (v16i8, v16i8);
14289 v8i16 __builtin_msa_min_s_h (v8i16, v8i16);
14290 v4i32 __builtin_msa_min_s_w (v4i32, v4i32);
14291 v2i64 __builtin_msa_min_s_d (v2i64, v2i64);
14293 v16u8 __builtin_msa_min_u_b (v16u8, v16u8);
14294 v8u16 __builtin_msa_min_u_h (v8u16, v8u16);
14295 v4u32 __builtin_msa_min_u_w (v4u32, v4u32);
14296 v2u64 __builtin_msa_min_u_d (v2u64, v2u64);
14298 v16i8 __builtin_msa_mini_s_b (v16i8, imm_n16_15);
14299 v8i16 __builtin_msa_mini_s_h (v8i16, imm_n16_15);
14300 v4i32 __builtin_msa_mini_s_w (v4i32, imm_n16_15);
14301 v2i64 __builtin_msa_mini_s_d (v2i64, imm_n16_15);
14303 v16u8 __builtin_msa_mini_u_b (v16u8, imm0_31);
14304 v8u16 __builtin_msa_mini_u_h (v8u16, imm0_31);
14305 v4u32 __builtin_msa_mini_u_w (v4u32, imm0_31);
14306 v2u64 __builtin_msa_mini_u_d (v2u64, imm0_31);
14308 v16i8 __builtin_msa_mod_s_b (v16i8, v16i8);
14309 v8i16 __builtin_msa_mod_s_h (v8i16, v8i16);
14310 v4i32 __builtin_msa_mod_s_w (v4i32, v4i32);
14311 v2i64 __builtin_msa_mod_s_d (v2i64, v2i64);
14313 v16u8 __builtin_msa_mod_u_b (v16u8, v16u8);
14314 v8u16 __builtin_msa_mod_u_h (v8u16, v8u16);
14315 v4u32 __builtin_msa_mod_u_w (v4u32, v4u32);
14316 v2u64 __builtin_msa_mod_u_d (v2u64, v2u64);
14318 v16i8 __builtin_msa_move_v (v16i8);
14320 v8i16 __builtin_msa_msub_q_h (v8i16, v8i16, v8i16);
14321 v4i32 __builtin_msa_msub_q_w (v4i32, v4i32, v4i32);
14323 v8i16 __builtin_msa_msubr_q_h (v8i16, v8i16, v8i16);
14324 v4i32 __builtin_msa_msubr_q_w (v4i32, v4i32, v4i32);
14326 v16i8 __builtin_msa_msubv_b (v16i8, v16i8, v16i8);
14327 v8i16 __builtin_msa_msubv_h (v8i16, v8i16, v8i16);
14328 v4i32 __builtin_msa_msubv_w (v4i32, v4i32, v4i32);
14329 v2i64 __builtin_msa_msubv_d (v2i64, v2i64, v2i64);
14331 v8i16 __builtin_msa_mul_q_h (v8i16, v8i16);
14332 v4i32 __builtin_msa_mul_q_w (v4i32, v4i32);
14334 v8i16 __builtin_msa_mulr_q_h (v8i16, v8i16);
14335 v4i32 __builtin_msa_mulr_q_w (v4i32, v4i32);
14337 v16i8 __builtin_msa_mulv_b (v16i8, v16i8);
14338 v8i16 __builtin_msa_mulv_h (v8i16, v8i16);
14339 v4i32 __builtin_msa_mulv_w (v4i32, v4i32);
14340 v2i64 __builtin_msa_mulv_d (v2i64, v2i64);
14342 v16i8 __builtin_msa_nloc_b (v16i8);
14343 v8i16 __builtin_msa_nloc_h (v8i16);
14344 v4i32 __builtin_msa_nloc_w (v4i32);
14345 v2i64 __builtin_msa_nloc_d (v2i64);
14347 v16i8 __builtin_msa_nlzc_b (v16i8);
14348 v8i16 __builtin_msa_nlzc_h (v8i16);
14349 v4i32 __builtin_msa_nlzc_w (v4i32);
14350 v2i64 __builtin_msa_nlzc_d (v2i64);
14352 v16u8 __builtin_msa_nor_v (v16u8, v16u8);
14354 v16u8 __builtin_msa_nori_b (v16u8, imm0_255);
14356 v16u8 __builtin_msa_or_v (v16u8, v16u8);
14358 v16u8 __builtin_msa_ori_b (v16u8, imm0_255);
14360 v16i8 __builtin_msa_pckev_b (v16i8, v16i8);
14361 v8i16 __builtin_msa_pckev_h (v8i16, v8i16);
14362 v4i32 __builtin_msa_pckev_w (v4i32, v4i32);
14363 v2i64 __builtin_msa_pckev_d (v2i64, v2i64);
14365 v16i8 __builtin_msa_pckod_b (v16i8, v16i8);
14366 v8i16 __builtin_msa_pckod_h (v8i16, v8i16);
14367 v4i32 __builtin_msa_pckod_w (v4i32, v4i32);
14368 v2i64 __builtin_msa_pckod_d (v2i64, v2i64);
14370 v16i8 __builtin_msa_pcnt_b (v16i8);
14371 v8i16 __builtin_msa_pcnt_h (v8i16);
14372 v4i32 __builtin_msa_pcnt_w (v4i32);
14373 v2i64 __builtin_msa_pcnt_d (v2i64);
14375 v16i8 __builtin_msa_sat_s_b (v16i8, imm0_7);
14376 v8i16 __builtin_msa_sat_s_h (v8i16, imm0_15);
14377 v4i32 __builtin_msa_sat_s_w (v4i32, imm0_31);
14378 v2i64 __builtin_msa_sat_s_d (v2i64, imm0_63);
14380 v16u8 __builtin_msa_sat_u_b (v16u8, imm0_7);
14381 v8u16 __builtin_msa_sat_u_h (v8u16, imm0_15);
14382 v4u32 __builtin_msa_sat_u_w (v4u32, imm0_31);
14383 v2u64 __builtin_msa_sat_u_d (v2u64, imm0_63);
14385 v16i8 __builtin_msa_shf_b (v16i8, imm0_255);
14386 v8i16 __builtin_msa_shf_h (v8i16, imm0_255);
14387 v4i32 __builtin_msa_shf_w (v4i32, imm0_255);
14389 v16i8 __builtin_msa_sld_b (v16i8, v16i8, i32);
14390 v8i16 __builtin_msa_sld_h (v8i16, v8i16, i32);
14391 v4i32 __builtin_msa_sld_w (v4i32, v4i32, i32);
14392 v2i64 __builtin_msa_sld_d (v2i64, v2i64, i32);
14394 v16i8 __builtin_msa_sldi_b (v16i8, v16i8, imm0_15);
14395 v8i16 __builtin_msa_sldi_h (v8i16, v8i16, imm0_7);
14396 v4i32 __builtin_msa_sldi_w (v4i32, v4i32, imm0_3);
14397 v2i64 __builtin_msa_sldi_d (v2i64, v2i64, imm0_1);
14399 v16i8 __builtin_msa_sll_b (v16i8, v16i8);
14400 v8i16 __builtin_msa_sll_h (v8i16, v8i16);
14401 v4i32 __builtin_msa_sll_w (v4i32, v4i32);
14402 v2i64 __builtin_msa_sll_d (v2i64, v2i64);
14404 v16i8 __builtin_msa_slli_b (v16i8, imm0_7);
14405 v8i16 __builtin_msa_slli_h (v8i16, imm0_15);
14406 v4i32 __builtin_msa_slli_w (v4i32, imm0_31);
14407 v2i64 __builtin_msa_slli_d (v2i64, imm0_63);
14409 v16i8 __builtin_msa_splat_b (v16i8, i32);
14410 v8i16 __builtin_msa_splat_h (v8i16, i32);
14411 v4i32 __builtin_msa_splat_w (v4i32, i32);
14412 v2i64 __builtin_msa_splat_d (v2i64, i32);
14414 v16i8 __builtin_msa_splati_b (v16i8, imm0_15);
14415 v8i16 __builtin_msa_splati_h (v8i16, imm0_7);
14416 v4i32 __builtin_msa_splati_w (v4i32, imm0_3);
14417 v2i64 __builtin_msa_splati_d (v2i64, imm0_1);
14419 v16i8 __builtin_msa_sra_b (v16i8, v16i8);
14420 v8i16 __builtin_msa_sra_h (v8i16, v8i16);
14421 v4i32 __builtin_msa_sra_w (v4i32, v4i32);
14422 v2i64 __builtin_msa_sra_d (v2i64, v2i64);
14424 v16i8 __builtin_msa_srai_b (v16i8, imm0_7);
14425 v8i16 __builtin_msa_srai_h (v8i16, imm0_15);
14426 v4i32 __builtin_msa_srai_w (v4i32, imm0_31);
14427 v2i64 __builtin_msa_srai_d (v2i64, imm0_63);
14429 v16i8 __builtin_msa_srar_b (v16i8, v16i8);
14430 v8i16 __builtin_msa_srar_h (v8i16, v8i16);
14431 v4i32 __builtin_msa_srar_w (v4i32, v4i32);
14432 v2i64 __builtin_msa_srar_d (v2i64, v2i64);
14434 v16i8 __builtin_msa_srari_b (v16i8, imm0_7);
14435 v8i16 __builtin_msa_srari_h (v8i16, imm0_15);
14436 v4i32 __builtin_msa_srari_w (v4i32, imm0_31);
14437 v2i64 __builtin_msa_srari_d (v2i64, imm0_63);
14439 v16i8 __builtin_msa_srl_b (v16i8, v16i8);
14440 v8i16 __builtin_msa_srl_h (v8i16, v8i16);
14441 v4i32 __builtin_msa_srl_w (v4i32, v4i32);
14442 v2i64 __builtin_msa_srl_d (v2i64, v2i64);
14444 v16i8 __builtin_msa_srli_b (v16i8, imm0_7);
14445 v8i16 __builtin_msa_srli_h (v8i16, imm0_15);
14446 v4i32 __builtin_msa_srli_w (v4i32, imm0_31);
14447 v2i64 __builtin_msa_srli_d (v2i64, imm0_63);
14449 v16i8 __builtin_msa_srlr_b (v16i8, v16i8);
14450 v8i16 __builtin_msa_srlr_h (v8i16, v8i16);
14451 v4i32 __builtin_msa_srlr_w (v4i32, v4i32);
14452 v2i64 __builtin_msa_srlr_d (v2i64, v2i64);
14454 v16i8 __builtin_msa_srlri_b (v16i8, imm0_7);
14455 v8i16 __builtin_msa_srlri_h (v8i16, imm0_15);
14456 v4i32 __builtin_msa_srlri_w (v4i32, imm0_31);
14457 v2i64 __builtin_msa_srlri_d (v2i64, imm0_63);
14459 void __builtin_msa_st_b (v16i8, void *, imm_n512_511);
14460 void __builtin_msa_st_h (v8i16, void *, imm_n1024_1022);
14461 void __builtin_msa_st_w (v4i32, void *, imm_n2048_2044);
14462 void __builtin_msa_st_d (v2i64, void *, imm_n4096_4088);
14464 v16i8 __builtin_msa_subs_s_b (v16i8, v16i8);
14465 v8i16 __builtin_msa_subs_s_h (v8i16, v8i16);
14466 v4i32 __builtin_msa_subs_s_w (v4i32, v4i32);
14467 v2i64 __builtin_msa_subs_s_d (v2i64, v2i64);
14469 v16u8 __builtin_msa_subs_u_b (v16u8, v16u8);
14470 v8u16 __builtin_msa_subs_u_h (v8u16, v8u16);
14471 v4u32 __builtin_msa_subs_u_w (v4u32, v4u32);
14472 v2u64 __builtin_msa_subs_u_d (v2u64, v2u64);
14474 v16u8 __builtin_msa_subsus_u_b (v16u8, v16i8);
14475 v8u16 __builtin_msa_subsus_u_h (v8u16, v8i16);
14476 v4u32 __builtin_msa_subsus_u_w (v4u32, v4i32);
14477 v2u64 __builtin_msa_subsus_u_d (v2u64, v2i64);
14479 v16i8 __builtin_msa_subsuu_s_b (v16u8, v16u8);
14480 v8i16 __builtin_msa_subsuu_s_h (v8u16, v8u16);
14481 v4i32 __builtin_msa_subsuu_s_w (v4u32, v4u32);
14482 v2i64 __builtin_msa_subsuu_s_d (v2u64, v2u64);
14484 v16i8 __builtin_msa_subv_b (v16i8, v16i8);
14485 v8i16 __builtin_msa_subv_h (v8i16, v8i16);
14486 v4i32 __builtin_msa_subv_w (v4i32, v4i32);
14487 v2i64 __builtin_msa_subv_d (v2i64, v2i64);
14489 v16i8 __builtin_msa_subvi_b (v16i8, imm0_31);
14490 v8i16 __builtin_msa_subvi_h (v8i16, imm0_31);
14491 v4i32 __builtin_msa_subvi_w (v4i32, imm0_31);
14492 v2i64 __builtin_msa_subvi_d (v2i64, imm0_31);
14494 v16i8 __builtin_msa_vshf_b (v16i8, v16i8, v16i8);
14495 v8i16 __builtin_msa_vshf_h (v8i16, v8i16, v8i16);
14496 v4i32 __builtin_msa_vshf_w (v4i32, v4i32, v4i32);
14497 v2i64 __builtin_msa_vshf_d (v2i64, v2i64, v2i64);
14499 v16u8 __builtin_msa_xor_v (v16u8, v16u8);
14501 v16u8 __builtin_msa_xori_b (v16u8, imm0_255);
14504 @node Other MIPS Built-in Functions
14505 @subsection Other MIPS Built-in Functions
14507 GCC provides other MIPS-specific built-in functions:
14510 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
14511 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
14512 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
14513 when this function is available.
14515 @item unsigned int __builtin_mips_get_fcsr (void)
14516 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
14517 Get and set the contents of the floating-point control and status register
14518 (FPU control register 31). These functions are only available in hard-float
14519 code but can be called in both MIPS16 and non-MIPS16 contexts.
14521 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
14522 register except the condition codes, which GCC assumes are preserved.
14525 @node MSP430 Built-in Functions
14526 @subsection MSP430 Built-in Functions
14528 GCC provides a couple of special builtin functions to aid in the
14529 writing of interrupt handlers in C.
14532 @item __bic_SR_register_on_exit (int @var{mask})
14533 This clears the indicated bits in the saved copy of the status register
14534 currently residing on the stack. This only works inside interrupt
14535 handlers and the changes to the status register will only take affect
14536 once the handler returns.
14538 @item __bis_SR_register_on_exit (int @var{mask})
14539 This sets the indicated bits in the saved copy of the status register
14540 currently residing on the stack. This only works inside interrupt
14541 handlers and the changes to the status register will only take affect
14542 once the handler returns.
14544 @item __delay_cycles (long long @var{cycles})
14545 This inserts an instruction sequence that takes exactly @var{cycles}
14546 cycles (between 0 and about 17E9) to complete. The inserted sequence
14547 may use jumps, loops, or no-ops, and does not interfere with any other
14548 instructions. Note that @var{cycles} must be a compile-time constant
14549 integer - that is, you must pass a number, not a variable that may be
14550 optimized to a constant later. The number of cycles delayed by this
14554 @node NDS32 Built-in Functions
14555 @subsection NDS32 Built-in Functions
14557 These built-in functions are available for the NDS32 target:
14559 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
14560 Insert an ISYNC instruction into the instruction stream where
14561 @var{addr} is an instruction address for serialization.
14564 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
14565 Insert an ISB instruction into the instruction stream.
14568 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
14569 Return the content of a system register which is mapped by @var{sr}.
14572 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
14573 Return the content of a user space register which is mapped by @var{usr}.
14576 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
14577 Move the @var{value} to a system register which is mapped by @var{sr}.
14580 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
14581 Move the @var{value} to a user space register which is mapped by @var{usr}.
14584 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
14585 Enable global interrupt.
14588 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
14589 Disable global interrupt.
14592 @node picoChip Built-in Functions
14593 @subsection picoChip Built-in Functions
14595 GCC provides an interface to selected machine instructions from the
14596 picoChip instruction set.
14599 @item int __builtin_sbc (int @var{value})
14600 Sign bit count. Return the number of consecutive bits in @var{value}
14601 that have the same value as the sign bit. The result is the number of
14602 leading sign bits minus one, giving the number of redundant sign bits in
14605 @item int __builtin_byteswap (int @var{value})
14606 Byte swap. Return the result of swapping the upper and lower bytes of
14609 @item int __builtin_brev (int @var{value})
14610 Bit reversal. Return the result of reversing the bits in
14611 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
14614 @item int __builtin_adds (int @var{x}, int @var{y})
14615 Saturating addition. Return the result of adding @var{x} and @var{y},
14616 storing the value 32767 if the result overflows.
14618 @item int __builtin_subs (int @var{x}, int @var{y})
14619 Saturating subtraction. Return the result of subtracting @var{y} from
14620 @var{x}, storing the value @minus{}32768 if the result overflows.
14622 @item void __builtin_halt (void)
14623 Halt. The processor stops execution. This built-in is useful for
14624 implementing assertions.
14628 @node PowerPC Built-in Functions
14629 @subsection PowerPC Built-in Functions
14631 The following built-in functions are always available and can be used to
14632 check the PowerPC target platform type:
14634 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
14635 This function is a @code{nop} on the PowerPC platform and is included solely
14636 to maintain API compatibility with the x86 builtins.
14639 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
14640 This function returns a value of @code{1} if the run-time CPU is of type
14641 @var{cpuname} and returns @code{0} otherwise. The following CPU names can be
14646 IBM POWER9 Server CPU.
14648 IBM POWER8 Server CPU.
14650 IBM POWER7 Server CPU.
14652 IBM POWER6 Server CPU (RAW mode).
14654 IBM POWER6 Server CPU (Architected mode).
14656 IBM POWER5+ Server CPU.
14658 IBM POWER5 Server CPU.
14660 IBM 970 Server CPU (ie, Apple G5).
14662 IBM POWER4 Server CPU.
14664 IBM A2 64-bit Embedded CPU
14666 IBM PowerPC 476FP 32-bit Embedded CPU.
14668 IBM PowerPC 464 32-bit Embedded CPU.
14670 PowerPC 440 32-bit Embedded CPU.
14672 PowerPC 405 32-bit Embedded CPU.
14674 IBM PowerPC Cell Broadband Engine Architecture CPU.
14677 Here is an example:
14679 if (__builtin_cpu_is ("power8"))
14681 do_power8 (); // POWER8 specific implementation.
14685 do_generic (); // Generic implementation.
14690 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
14691 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
14692 feature @var{feature} and returns @code{0} otherwise. The following features can be
14697 4xx CPU has a Multiply Accumulator.
14699 CPU has a SIMD/Vector Unit.
14701 CPU supports ISA 2.05 (eg, POWER6)
14703 CPU supports ISA 2.06 (eg, POWER7)
14705 CPU supports ISA 2.07 (eg, POWER8)
14707 CPU supports ISA 3.0 (eg, POWER9)
14709 CPU supports the set of compatible performance monitoring events.
14711 CPU supports the Embedded ISA category.
14713 CPU has a CELL broadband engine.
14715 CPU has a decimal floating point unit.
14717 CPU supports the data stream control register.
14719 CPU supports event base branching.
14721 CPU has a SPE double precision floating point unit.
14723 CPU has a SPE single precision floating point unit.
14725 CPU has a floating point unit.
14727 CPU has hardware transaction memory instructions.
14729 Kernel aborts hardware transactions when a syscall is made.
14731 CPU supports icache snooping capabilities.
14733 CPU supports 128-bit IEEE binary floating point instructions.
14735 CPU supports the integer select instruction.
14737 CPU has a memory management unit.
14739 CPU does not have a timebase (eg, 601 and 403gx).
14741 CPU supports the PA Semi 6T CORE ISA.
14743 CPU supports ISA 2.00 (eg, POWER4)
14745 CPU supports ISA 2.02 (eg, POWER5)
14747 CPU supports ISA 2.03 (eg, POWER5+)
14749 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
14751 CPU supports 32-bit mode execution.
14753 CPU supports the old POWER ISA (eg, 601)
14755 CPU supports 64-bit mode execution.
14757 CPU supports a little-endian mode that uses address swizzling.
14759 CPU support simultaneous multi-threading.
14761 CPU has a signal processing extension unit.
14763 CPU supports the target address register.
14765 CPU supports true little-endian mode.
14767 CPU has unified I/D cache.
14769 CPU supports the vector cryptography instructions.
14771 CPU supports the vector-scalar extension.
14774 Here is an example:
14776 if (__builtin_cpu_supports ("fpu"))
14778 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
14782 dst = __fadd (src1, src2); // Software FP addition function.
14787 These built-in functions are available for the PowerPC family of
14790 float __builtin_recipdivf (float, float);
14791 float __builtin_rsqrtf (float);
14792 double __builtin_recipdiv (double, double);
14793 double __builtin_rsqrt (double);
14794 uint64_t __builtin_ppc_get_timebase ();
14795 unsigned long __builtin_ppc_mftb ();
14796 double __builtin_unpack_longdouble (long double, int);
14797 long double __builtin_pack_longdouble (double, double);
14800 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
14801 @code{__builtin_rsqrtf} functions generate multiple instructions to
14802 implement the reciprocal sqrt functionality using reciprocal sqrt
14803 estimate instructions.
14805 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
14806 functions generate multiple instructions to implement division using
14807 the reciprocal estimate instructions.
14809 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
14810 functions generate instructions to read the Time Base Register. The
14811 @code{__builtin_ppc_get_timebase} function may generate multiple
14812 instructions and always returns the 64 bits of the Time Base Register.
14813 The @code{__builtin_ppc_mftb} function always generates one instruction and
14814 returns the Time Base Register value as an unsigned long, throwing away
14815 the most significant word on 32-bit environments.
14817 Additional built-in functions are available for the 64-bit PowerPC
14818 family of processors, for efficient use of 128-bit floating point
14819 (@code{__float128}) values.
14821 The following floating-point built-in functions are available with
14822 @code{-mfloat128} and Altivec support. All of them implement the
14823 function that is part of the name.
14826 __float128 __builtin_fabsq (__float128)
14827 __float128 __builtin_copysignq (__float128, __float128)
14830 The following built-in functions are available with @code{-mfloat128}
14831 and Altivec support.
14834 @item __float128 __builtin_infq (void)
14835 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
14836 @findex __builtin_infq
14838 @item __float128 __builtin_huge_valq (void)
14839 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
14840 @findex __builtin_huge_valq
14842 @item __float128 __builtin_nanq (void)
14843 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
14844 @findex __builtin_nanq
14846 @item __float128 __builtin_nansq (void)
14847 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
14848 @findex __builtin_nansq
14851 The following built-in functions are available for the PowerPC family
14852 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
14853 or @option{-mpopcntd}):
14855 long __builtin_bpermd (long, long);
14856 int __builtin_divwe (int, int);
14857 int __builtin_divweo (int, int);
14858 unsigned int __builtin_divweu (unsigned int, unsigned int);
14859 unsigned int __builtin_divweuo (unsigned int, unsigned int);
14860 long __builtin_divde (long, long);
14861 long __builtin_divdeo (long, long);
14862 unsigned long __builtin_divdeu (unsigned long, unsigned long);
14863 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
14864 unsigned int cdtbcd (unsigned int);
14865 unsigned int cbcdtd (unsigned int);
14866 unsigned int addg6s (unsigned int, unsigned int);
14869 The @code{__builtin_divde}, @code{__builtin_divdeo},
14870 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
14871 64-bit environment support ISA 2.06 or later.
14873 The following built-in functions are available for the PowerPC family
14874 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
14876 long long __builtin_darn (void);
14877 long long __builtin_darn_raw (void);
14878 int __builtin_darn_32 (void);
14880 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal64 value);
14881 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal128 value);
14882 int __builtin_dfp_dtstsfi_lt_dd (unsigned int comparison, _Decimal64 value);
14883 int __builtin_dfp_dtstsfi_lt_td (unsigned int comparison, _Decimal128 value);
14885 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal64 value);
14886 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal128 value);
14887 int __builtin_dfp_dtstsfi_gt_dd (unsigned int comparison, _Decimal64 value);
14888 int __builtin_dfp_dtstsfi_gt_td (unsigned int comparison, _Decimal128 value);
14890 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal64 value);
14891 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal128 value);
14892 int __builtin_dfp_dtstsfi_eq_dd (unsigned int comparison, _Decimal64 value);
14893 int __builtin_dfp_dtstsfi_eq_td (unsigned int comparison, _Decimal128 value);
14895 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal64 value);
14896 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal128 value);
14897 int __builtin_dfp_dtstsfi_ov_dd (unsigned int comparison, _Decimal64 value);
14898 int __builtin_dfp_dtstsfi_ov_td (unsigned int comparison, _Decimal128 value);
14901 The @code{__builtin_darn} and @code{__builtin_darn_raw}
14902 functions require a
14903 64-bit environment supporting ISA 3.0 or later.
14904 The @code{__builtin_darn} function provides a 64-bit conditioned
14905 random number. The @code{__builtin_darn_raw} function provides a
14906 64-bit raw random number. The @code{__builtin_darn_32} function
14907 provides a 32-bit random number.
14909 The @code{__builtin_dfp_dtstsfi_lt} function returns a non-zero value
14910 if and only if the number of signficant digits of its @code{value} argument
14911 is less than its @code{comparison} argument. The
14912 @code{__builtin_dfp_dtstsfi_lt_dd} and
14913 @code{__builtin_dfp_dtstsfi_lt_td} functions behave similarly, but
14914 require that the type of the @code{value} argument be
14915 @code{__Decimal64} and @code{__Decimal128} respectively.
14917 The @code{__builtin_dfp_dtstsfi_gt} function returns a non-zero value
14918 if and only if the number of signficant digits of its @code{value} argument
14919 is greater than its @code{comparison} argument. The
14920 @code{__builtin_dfp_dtstsfi_gt_dd} and
14921 @code{__builtin_dfp_dtstsfi_gt_td} functions behave similarly, but
14922 require that the type of the @code{value} argument be
14923 @code{__Decimal64} and @code{__Decimal128} respectively.
14925 The @code{__builtin_dfp_dtstsfi_eq} function returns a non-zero value
14926 if and only if the number of signficant digits of its @code{value} argument
14927 equals its @code{comparison} argument. The
14928 @code{__builtin_dfp_dtstsfi_eq_dd} and
14929 @code{__builtin_dfp_dtstsfi_eq_td} functions behave similarly, but
14930 require that the type of the @code{value} argument be
14931 @code{__Decimal64} and @code{__Decimal128} respectively.
14933 The @code{__builtin_dfp_dtstsfi_ov} function returns a non-zero value
14934 if and only if its @code{value} argument has an undefined number of
14935 significant digits, such as when @code{value} is an encoding of @code{NaN}.
14936 The @code{__builtin_dfp_dtstsfi_ov_dd} and
14937 @code{__builtin_dfp_dtstsfi_ov_td} functions behave similarly, but
14938 require that the type of the @code{value} argument be
14939 @code{__Decimal64} and @code{__Decimal128} respectively.
14941 The following built-in functions are available for the PowerPC family
14942 of processors when hardware decimal floating point
14943 (@option{-mhard-dfp}) is available:
14945 _Decimal64 __builtin_dxex (_Decimal64);
14946 _Decimal128 __builtin_dxexq (_Decimal128);
14947 _Decimal64 __builtin_ddedpd (int, _Decimal64);
14948 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
14949 _Decimal64 __builtin_denbcd (int, _Decimal64);
14950 _Decimal128 __builtin_denbcdq (int, _Decimal128);
14951 _Decimal64 __builtin_diex (_Decimal64, _Decimal64);
14952 _Decimal128 _builtin_diexq (_Decimal128, _Decimal128);
14953 _Decimal64 __builtin_dscli (_Decimal64, int);
14954 _Decimal128 __builtin_dscliq (_Decimal128, int);
14955 _Decimal64 __builtin_dscri (_Decimal64, int);
14956 _Decimal128 __builtin_dscriq (_Decimal128, int);
14957 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
14958 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
14961 The following built-in functions are available for the PowerPC family
14962 of processors when the Vector Scalar (vsx) instruction set is
14965 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
14966 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
14967 unsigned long long);
14970 @node PowerPC AltiVec/VSX Built-in Functions
14971 @subsection PowerPC AltiVec Built-in Functions
14973 GCC provides an interface for the PowerPC family of processors to access
14974 the AltiVec operations described in Motorola's AltiVec Programming
14975 Interface Manual. The interface is made available by including
14976 @code{<altivec.h>} and using @option{-maltivec} and
14977 @option{-mabi=altivec}. The interface supports the following vector
14981 vector unsigned char
14985 vector unsigned short
14986 vector signed short
14990 vector unsigned int
14996 If @option{-mvsx} is used the following additional vector types are
15000 vector unsigned long
15005 The long types are only implemented for 64-bit code generation, and
15006 the long type is only used in the floating point/integer conversion
15009 GCC's implementation of the high-level language interface available from
15010 C and C++ code differs from Motorola's documentation in several ways.
15015 A vector constant is a list of constant expressions within curly braces.
15018 A vector initializer requires no cast if the vector constant is of the
15019 same type as the variable it is initializing.
15022 If @code{signed} or @code{unsigned} is omitted, the signedness of the
15023 vector type is the default signedness of the base type. The default
15024 varies depending on the operating system, so a portable program should
15025 always specify the signedness.
15028 Compiling with @option{-maltivec} adds keywords @code{__vector},
15029 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
15030 @code{bool}. When compiling ISO C, the context-sensitive substitution
15031 of the keywords @code{vector}, @code{pixel} and @code{bool} is
15032 disabled. To use them, you must include @code{<altivec.h>} instead.
15035 GCC allows using a @code{typedef} name as the type specifier for a
15039 For C, overloaded functions are implemented with macros so the following
15043 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
15047 Since @code{vec_add} is a macro, the vector constant in the example
15048 is treated as four separate arguments. Wrap the entire argument in
15049 parentheses for this to work.
15052 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
15053 Internally, GCC uses built-in functions to achieve the functionality in
15054 the aforementioned header file, but they are not supported and are
15055 subject to change without notice.
15057 The following interfaces are supported for the generic and specific
15058 AltiVec operations and the AltiVec predicates. In cases where there
15059 is a direct mapping between generic and specific operations, only the
15060 generic names are shown here, although the specific operations can also
15063 Arguments that are documented as @code{const int} require literal
15064 integral values within the range required for that operation.
15067 vector signed char vec_abs (vector signed char);
15068 vector signed short vec_abs (vector signed short);
15069 vector signed int vec_abs (vector signed int);
15070 vector float vec_abs (vector float);
15072 vector signed char vec_abss (vector signed char);
15073 vector signed short vec_abss (vector signed short);
15074 vector signed int vec_abss (vector signed int);
15076 vector signed char vec_add (vector bool char, vector signed char);
15077 vector signed char vec_add (vector signed char, vector bool char);
15078 vector signed char vec_add (vector signed char, vector signed char);
15079 vector unsigned char vec_add (vector bool char, vector unsigned char);
15080 vector unsigned char vec_add (vector unsigned char, vector bool char);
15081 vector unsigned char vec_add (vector unsigned char,
15082 vector unsigned char);
15083 vector signed short vec_add (vector bool short, vector signed short);
15084 vector signed short vec_add (vector signed short, vector bool short);
15085 vector signed short vec_add (vector signed short, vector signed short);
15086 vector unsigned short vec_add (vector bool short,
15087 vector unsigned short);
15088 vector unsigned short vec_add (vector unsigned short,
15089 vector bool short);
15090 vector unsigned short vec_add (vector unsigned short,
15091 vector unsigned short);
15092 vector signed int vec_add (vector bool int, vector signed int);
15093 vector signed int vec_add (vector signed int, vector bool int);
15094 vector signed int vec_add (vector signed int, vector signed int);
15095 vector unsigned int vec_add (vector bool int, vector unsigned int);
15096 vector unsigned int vec_add (vector unsigned int, vector bool int);
15097 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
15098 vector float vec_add (vector float, vector float);
15100 vector float vec_vaddfp (vector float, vector float);
15102 vector signed int vec_vadduwm (vector bool int, vector signed int);
15103 vector signed int vec_vadduwm (vector signed int, vector bool int);
15104 vector signed int vec_vadduwm (vector signed int, vector signed int);
15105 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
15106 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
15107 vector unsigned int vec_vadduwm (vector unsigned int,
15108 vector unsigned int);
15110 vector signed short vec_vadduhm (vector bool short,
15111 vector signed short);
15112 vector signed short vec_vadduhm (vector signed short,
15113 vector bool short);
15114 vector signed short vec_vadduhm (vector signed short,
15115 vector signed short);
15116 vector unsigned short vec_vadduhm (vector bool short,
15117 vector unsigned short);
15118 vector unsigned short vec_vadduhm (vector unsigned short,
15119 vector bool short);
15120 vector unsigned short vec_vadduhm (vector unsigned short,
15121 vector unsigned short);
15123 vector signed char vec_vaddubm (vector bool char, vector signed char);
15124 vector signed char vec_vaddubm (vector signed char, vector bool char);
15125 vector signed char vec_vaddubm (vector signed char, vector signed char);
15126 vector unsigned char vec_vaddubm (vector bool char,
15127 vector unsigned char);
15128 vector unsigned char vec_vaddubm (vector unsigned char,
15130 vector unsigned char vec_vaddubm (vector unsigned char,
15131 vector unsigned char);
15133 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
15135 vector unsigned char vec_adds (vector bool char, vector unsigned char);
15136 vector unsigned char vec_adds (vector unsigned char, vector bool char);
15137 vector unsigned char vec_adds (vector unsigned char,
15138 vector unsigned char);
15139 vector signed char vec_adds (vector bool char, vector signed char);
15140 vector signed char vec_adds (vector signed char, vector bool char);
15141 vector signed char vec_adds (vector signed char, vector signed char);
15142 vector unsigned short vec_adds (vector bool short,
15143 vector unsigned short);
15144 vector unsigned short vec_adds (vector unsigned short,
15145 vector bool short);
15146 vector unsigned short vec_adds (vector unsigned short,
15147 vector unsigned short);
15148 vector signed short vec_adds (vector bool short, vector signed short);
15149 vector signed short vec_adds (vector signed short, vector bool short);
15150 vector signed short vec_adds (vector signed short, vector signed short);
15151 vector unsigned int vec_adds (vector bool int, vector unsigned int);
15152 vector unsigned int vec_adds (vector unsigned int, vector bool int);
15153 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
15154 vector signed int vec_adds (vector bool int, vector signed int);
15155 vector signed int vec_adds (vector signed int, vector bool int);
15156 vector signed int vec_adds (vector signed int, vector signed int);
15158 vector signed int vec_vaddsws (vector bool int, vector signed int);
15159 vector signed int vec_vaddsws (vector signed int, vector bool int);
15160 vector signed int vec_vaddsws (vector signed int, vector signed int);
15162 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
15163 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
15164 vector unsigned int vec_vadduws (vector unsigned int,
15165 vector unsigned int);
15167 vector signed short vec_vaddshs (vector bool short,
15168 vector signed short);
15169 vector signed short vec_vaddshs (vector signed short,
15170 vector bool short);
15171 vector signed short vec_vaddshs (vector signed short,
15172 vector signed short);
15174 vector unsigned short vec_vadduhs (vector bool short,
15175 vector unsigned short);
15176 vector unsigned short vec_vadduhs (vector unsigned short,
15177 vector bool short);
15178 vector unsigned short vec_vadduhs (vector unsigned short,
15179 vector unsigned short);
15181 vector signed char vec_vaddsbs (vector bool char, vector signed char);
15182 vector signed char vec_vaddsbs (vector signed char, vector bool char);
15183 vector signed char vec_vaddsbs (vector signed char, vector signed char);
15185 vector unsigned char vec_vaddubs (vector bool char,
15186 vector unsigned char);
15187 vector unsigned char vec_vaddubs (vector unsigned char,
15189 vector unsigned char vec_vaddubs (vector unsigned char,
15190 vector unsigned char);
15192 vector float vec_and (vector float, vector float);
15193 vector float vec_and (vector float, vector bool int);
15194 vector float vec_and (vector bool int, vector float);
15195 vector bool int vec_and (vector bool int, vector bool int);
15196 vector signed int vec_and (vector bool int, vector signed int);
15197 vector signed int vec_and (vector signed int, vector bool int);
15198 vector signed int vec_and (vector signed int, vector signed int);
15199 vector unsigned int vec_and (vector bool int, vector unsigned int);
15200 vector unsigned int vec_and (vector unsigned int, vector bool int);
15201 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
15202 vector bool short vec_and (vector bool short, vector bool short);
15203 vector signed short vec_and (vector bool short, vector signed short);
15204 vector signed short vec_and (vector signed short, vector bool short);
15205 vector signed short vec_and (vector signed short, vector signed short);
15206 vector unsigned short vec_and (vector bool short,
15207 vector unsigned short);
15208 vector unsigned short vec_and (vector unsigned short,
15209 vector bool short);
15210 vector unsigned short vec_and (vector unsigned short,
15211 vector unsigned short);
15212 vector signed char vec_and (vector bool char, vector signed char);
15213 vector bool char vec_and (vector bool char, vector bool char);
15214 vector signed char vec_and (vector signed char, vector bool char);
15215 vector signed char vec_and (vector signed char, vector signed char);
15216 vector unsigned char vec_and (vector bool char, vector unsigned char);
15217 vector unsigned char vec_and (vector unsigned char, vector bool char);
15218 vector unsigned char vec_and (vector unsigned char,
15219 vector unsigned char);
15221 vector float vec_andc (vector float, vector float);
15222 vector float vec_andc (vector float, vector bool int);
15223 vector float vec_andc (vector bool int, vector float);
15224 vector bool int vec_andc (vector bool int, vector bool int);
15225 vector signed int vec_andc (vector bool int, vector signed int);
15226 vector signed int vec_andc (vector signed int, vector bool int);
15227 vector signed int vec_andc (vector signed int, vector signed int);
15228 vector unsigned int vec_andc (vector bool int, vector unsigned int);
15229 vector unsigned int vec_andc (vector unsigned int, vector bool int);
15230 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
15231 vector bool short vec_andc (vector bool short, vector bool short);
15232 vector signed short vec_andc (vector bool short, vector signed short);
15233 vector signed short vec_andc (vector signed short, vector bool short);
15234 vector signed short vec_andc (vector signed short, vector signed short);
15235 vector unsigned short vec_andc (vector bool short,
15236 vector unsigned short);
15237 vector unsigned short vec_andc (vector unsigned short,
15238 vector bool short);
15239 vector unsigned short vec_andc (vector unsigned short,
15240 vector unsigned short);
15241 vector signed char vec_andc (vector bool char, vector signed char);
15242 vector bool char vec_andc (vector bool char, vector bool char);
15243 vector signed char vec_andc (vector signed char, vector bool char);
15244 vector signed char vec_andc (vector signed char, vector signed char);
15245 vector unsigned char vec_andc (vector bool char, vector unsigned char);
15246 vector unsigned char vec_andc (vector unsigned char, vector bool char);
15247 vector unsigned char vec_andc (vector unsigned char,
15248 vector unsigned char);
15250 vector unsigned char vec_avg (vector unsigned char,
15251 vector unsigned char);
15252 vector signed char vec_avg (vector signed char, vector signed char);
15253 vector unsigned short vec_avg (vector unsigned short,
15254 vector unsigned short);
15255 vector signed short vec_avg (vector signed short, vector signed short);
15256 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
15257 vector signed int vec_avg (vector signed int, vector signed int);
15259 vector signed int vec_vavgsw (vector signed int, vector signed int);
15261 vector unsigned int vec_vavguw (vector unsigned int,
15262 vector unsigned int);
15264 vector signed short vec_vavgsh (vector signed short,
15265 vector signed short);
15267 vector unsigned short vec_vavguh (vector unsigned short,
15268 vector unsigned short);
15270 vector signed char vec_vavgsb (vector signed char, vector signed char);
15272 vector unsigned char vec_vavgub (vector unsigned char,
15273 vector unsigned char);
15275 vector float vec_copysign (vector float);
15277 vector float vec_ceil (vector float);
15279 vector signed int vec_cmpb (vector float, vector float);
15281 vector bool char vec_cmpeq (vector signed char, vector signed char);
15282 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
15283 vector bool short vec_cmpeq (vector signed short, vector signed short);
15284 vector bool short vec_cmpeq (vector unsigned short,
15285 vector unsigned short);
15286 vector bool int vec_cmpeq (vector signed int, vector signed int);
15287 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
15288 vector bool int vec_cmpeq (vector float, vector float);
15290 vector bool int vec_vcmpeqfp (vector float, vector float);
15292 vector bool int vec_vcmpequw (vector signed int, vector signed int);
15293 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
15295 vector bool short vec_vcmpequh (vector signed short,
15296 vector signed short);
15297 vector bool short vec_vcmpequh (vector unsigned short,
15298 vector unsigned short);
15300 vector bool char vec_vcmpequb (vector signed char, vector signed char);
15301 vector bool char vec_vcmpequb (vector unsigned char,
15302 vector unsigned char);
15304 vector bool int vec_cmpge (vector float, vector float);
15306 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
15307 vector bool char vec_cmpgt (vector signed char, vector signed char);
15308 vector bool short vec_cmpgt (vector unsigned short,
15309 vector unsigned short);
15310 vector bool short vec_cmpgt (vector signed short, vector signed short);
15311 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
15312 vector bool int vec_cmpgt (vector signed int, vector signed int);
15313 vector bool int vec_cmpgt (vector float, vector float);
15315 vector bool int vec_vcmpgtfp (vector float, vector float);
15317 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
15319 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
15321 vector bool short vec_vcmpgtsh (vector signed short,
15322 vector signed short);
15324 vector bool short vec_vcmpgtuh (vector unsigned short,
15325 vector unsigned short);
15327 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
15329 vector bool char vec_vcmpgtub (vector unsigned char,
15330 vector unsigned char);
15332 vector bool int vec_cmple (vector float, vector float);
15334 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
15335 vector bool char vec_cmplt (vector signed char, vector signed char);
15336 vector bool short vec_cmplt (vector unsigned short,
15337 vector unsigned short);
15338 vector bool short vec_cmplt (vector signed short, vector signed short);
15339 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
15340 vector bool int vec_cmplt (vector signed int, vector signed int);
15341 vector bool int vec_cmplt (vector float, vector float);
15343 vector float vec_cpsgn (vector float, vector float);
15345 vector float vec_ctf (vector unsigned int, const int);
15346 vector float vec_ctf (vector signed int, const int);
15347 vector double vec_ctf (vector unsigned long, const int);
15348 vector double vec_ctf (vector signed long, const int);
15350 vector float vec_vcfsx (vector signed int, const int);
15352 vector float vec_vcfux (vector unsigned int, const int);
15354 vector signed int vec_cts (vector float, const int);
15355 vector signed long vec_cts (vector double, const int);
15357 vector unsigned int vec_ctu (vector float, const int);
15358 vector unsigned long vec_ctu (vector double, const int);
15360 void vec_dss (const int);
15362 void vec_dssall (void);
15364 void vec_dst (const vector unsigned char *, int, const int);
15365 void vec_dst (const vector signed char *, int, const int);
15366 void vec_dst (const vector bool char *, int, const int);
15367 void vec_dst (const vector unsigned short *, int, const int);
15368 void vec_dst (const vector signed short *, int, const int);
15369 void vec_dst (const vector bool short *, int, const int);
15370 void vec_dst (const vector pixel *, int, const int);
15371 void vec_dst (const vector unsigned int *, int, const int);
15372 void vec_dst (const vector signed int *, int, const int);
15373 void vec_dst (const vector bool int *, int, const int);
15374 void vec_dst (const vector float *, int, const int);
15375 void vec_dst (const unsigned char *, int, const int);
15376 void vec_dst (const signed char *, int, const int);
15377 void vec_dst (const unsigned short *, int, const int);
15378 void vec_dst (const short *, int, const int);
15379 void vec_dst (const unsigned int *, int, const int);
15380 void vec_dst (const int *, int, const int);
15381 void vec_dst (const unsigned long *, int, const int);
15382 void vec_dst (const long *, int, const int);
15383 void vec_dst (const float *, int, const int);
15385 void vec_dstst (const vector unsigned char *, int, const int);
15386 void vec_dstst (const vector signed char *, int, const int);
15387 void vec_dstst (const vector bool char *, int, const int);
15388 void vec_dstst (const vector unsigned short *, int, const int);
15389 void vec_dstst (const vector signed short *, int, const int);
15390 void vec_dstst (const vector bool short *, int, const int);
15391 void vec_dstst (const vector pixel *, int, const int);
15392 void vec_dstst (const vector unsigned int *, int, const int);
15393 void vec_dstst (const vector signed int *, int, const int);
15394 void vec_dstst (const vector bool int *, int, const int);
15395 void vec_dstst (const vector float *, int, const int);
15396 void vec_dstst (const unsigned char *, int, const int);
15397 void vec_dstst (const signed char *, int, const int);
15398 void vec_dstst (const unsigned short *, int, const int);
15399 void vec_dstst (const short *, int, const int);
15400 void vec_dstst (const unsigned int *, int, const int);
15401 void vec_dstst (const int *, int, const int);
15402 void vec_dstst (const unsigned long *, int, const int);
15403 void vec_dstst (const long *, int, const int);
15404 void vec_dstst (const float *, int, const int);
15406 void vec_dststt (const vector unsigned char *, int, const int);
15407 void vec_dststt (const vector signed char *, int, const int);
15408 void vec_dststt (const vector bool char *, int, const int);
15409 void vec_dststt (const vector unsigned short *, int, const int);
15410 void vec_dststt (const vector signed short *, int, const int);
15411 void vec_dststt (const vector bool short *, int, const int);
15412 void vec_dststt (const vector pixel *, int, const int);
15413 void vec_dststt (const vector unsigned int *, int, const int);
15414 void vec_dststt (const vector signed int *, int, const int);
15415 void vec_dststt (const vector bool int *, int, const int);
15416 void vec_dststt (const vector float *, int, const int);
15417 void vec_dststt (const unsigned char *, int, const int);
15418 void vec_dststt (const signed char *, int, const int);
15419 void vec_dststt (const unsigned short *, int, const int);
15420 void vec_dststt (const short *, int, const int);
15421 void vec_dststt (const unsigned int *, int, const int);
15422 void vec_dststt (const int *, int, const int);
15423 void vec_dststt (const unsigned long *, int, const int);
15424 void vec_dststt (const long *, int, const int);
15425 void vec_dststt (const float *, int, const int);
15427 void vec_dstt (const vector unsigned char *, int, const int);
15428 void vec_dstt (const vector signed char *, int, const int);
15429 void vec_dstt (const vector bool char *, int, const int);
15430 void vec_dstt (const vector unsigned short *, int, const int);
15431 void vec_dstt (const vector signed short *, int, const int);
15432 void vec_dstt (const vector bool short *, int, const int);
15433 void vec_dstt (const vector pixel *, int, const int);
15434 void vec_dstt (const vector unsigned int *, int, const int);
15435 void vec_dstt (const vector signed int *, int, const int);
15436 void vec_dstt (const vector bool int *, int, const int);
15437 void vec_dstt (const vector float *, int, const int);
15438 void vec_dstt (const unsigned char *, int, const int);
15439 void vec_dstt (const signed char *, int, const int);
15440 void vec_dstt (const unsigned short *, int, const int);
15441 void vec_dstt (const short *, int, const int);
15442 void vec_dstt (const unsigned int *, int, const int);
15443 void vec_dstt (const int *, int, const int);
15444 void vec_dstt (const unsigned long *, int, const int);
15445 void vec_dstt (const long *, int, const int);
15446 void vec_dstt (const float *, int, const int);
15448 vector float vec_expte (vector float);
15450 vector float vec_floor (vector float);
15452 vector float vec_ld (int, const vector float *);
15453 vector float vec_ld (int, const float *);
15454 vector bool int vec_ld (int, const vector bool int *);
15455 vector signed int vec_ld (int, const vector signed int *);
15456 vector signed int vec_ld (int, const int *);
15457 vector signed int vec_ld (int, const long *);
15458 vector unsigned int vec_ld (int, const vector unsigned int *);
15459 vector unsigned int vec_ld (int, const unsigned int *);
15460 vector unsigned int vec_ld (int, const unsigned long *);
15461 vector bool short vec_ld (int, const vector bool short *);
15462 vector pixel vec_ld (int, const vector pixel *);
15463 vector signed short vec_ld (int, const vector signed short *);
15464 vector signed short vec_ld (int, const short *);
15465 vector unsigned short vec_ld (int, const vector unsigned short *);
15466 vector unsigned short vec_ld (int, const unsigned short *);
15467 vector bool char vec_ld (int, const vector bool char *);
15468 vector signed char vec_ld (int, const vector signed char *);
15469 vector signed char vec_ld (int, const signed char *);
15470 vector unsigned char vec_ld (int, const vector unsigned char *);
15471 vector unsigned char vec_ld (int, const unsigned char *);
15473 vector signed char vec_lde (int, const signed char *);
15474 vector unsigned char vec_lde (int, const unsigned char *);
15475 vector signed short vec_lde (int, const short *);
15476 vector unsigned short vec_lde (int, const unsigned short *);
15477 vector float vec_lde (int, const float *);
15478 vector signed int vec_lde (int, const int *);
15479 vector unsigned int vec_lde (int, const unsigned int *);
15480 vector signed int vec_lde (int, const long *);
15481 vector unsigned int vec_lde (int, const unsigned long *);
15483 vector float vec_lvewx (int, float *);
15484 vector signed int vec_lvewx (int, int *);
15485 vector unsigned int vec_lvewx (int, unsigned int *);
15486 vector signed int vec_lvewx (int, long *);
15487 vector unsigned int vec_lvewx (int, unsigned long *);
15489 vector signed short vec_lvehx (int, short *);
15490 vector unsigned short vec_lvehx (int, unsigned short *);
15492 vector signed char vec_lvebx (int, char *);
15493 vector unsigned char vec_lvebx (int, unsigned char *);
15495 vector float vec_ldl (int, const vector float *);
15496 vector float vec_ldl (int, const float *);
15497 vector bool int vec_ldl (int, const vector bool int *);
15498 vector signed int vec_ldl (int, const vector signed int *);
15499 vector signed int vec_ldl (int, const int *);
15500 vector signed int vec_ldl (int, const long *);
15501 vector unsigned int vec_ldl (int, const vector unsigned int *);
15502 vector unsigned int vec_ldl (int, const unsigned int *);
15503 vector unsigned int vec_ldl (int, const unsigned long *);
15504 vector bool short vec_ldl (int, const vector bool short *);
15505 vector pixel vec_ldl (int, const vector pixel *);
15506 vector signed short vec_ldl (int, const vector signed short *);
15507 vector signed short vec_ldl (int, const short *);
15508 vector unsigned short vec_ldl (int, const vector unsigned short *);
15509 vector unsigned short vec_ldl (int, const unsigned short *);
15510 vector bool char vec_ldl (int, const vector bool char *);
15511 vector signed char vec_ldl (int, const vector signed char *);
15512 vector signed char vec_ldl (int, const signed char *);
15513 vector unsigned char vec_ldl (int, const vector unsigned char *);
15514 vector unsigned char vec_ldl (int, const unsigned char *);
15516 vector float vec_loge (vector float);
15518 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
15519 vector unsigned char vec_lvsl (int, const volatile signed char *);
15520 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
15521 vector unsigned char vec_lvsl (int, const volatile short *);
15522 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
15523 vector unsigned char vec_lvsl (int, const volatile int *);
15524 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
15525 vector unsigned char vec_lvsl (int, const volatile long *);
15526 vector unsigned char vec_lvsl (int, const volatile float *);
15528 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
15529 vector unsigned char vec_lvsr (int, const volatile signed char *);
15530 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
15531 vector unsigned char vec_lvsr (int, const volatile short *);
15532 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
15533 vector unsigned char vec_lvsr (int, const volatile int *);
15534 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
15535 vector unsigned char vec_lvsr (int, const volatile long *);
15536 vector unsigned char vec_lvsr (int, const volatile float *);
15538 vector float vec_madd (vector float, vector float, vector float);
15540 vector signed short vec_madds (vector signed short,
15541 vector signed short,
15542 vector signed short);
15544 vector unsigned char vec_max (vector bool char, vector unsigned char);
15545 vector unsigned char vec_max (vector unsigned char, vector bool char);
15546 vector unsigned char vec_max (vector unsigned char,
15547 vector unsigned char);
15548 vector signed char vec_max (vector bool char, vector signed char);
15549 vector signed char vec_max (vector signed char, vector bool char);
15550 vector signed char vec_max (vector signed char, vector signed char);
15551 vector unsigned short vec_max (vector bool short,
15552 vector unsigned short);
15553 vector unsigned short vec_max (vector unsigned short,
15554 vector bool short);
15555 vector unsigned short vec_max (vector unsigned short,
15556 vector unsigned short);
15557 vector signed short vec_max (vector bool short, vector signed short);
15558 vector signed short vec_max (vector signed short, vector bool short);
15559 vector signed short vec_max (vector signed short, vector signed short);
15560 vector unsigned int vec_max (vector bool int, vector unsigned int);
15561 vector unsigned int vec_max (vector unsigned int, vector bool int);
15562 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
15563 vector signed int vec_max (vector bool int, vector signed int);
15564 vector signed int vec_max (vector signed int, vector bool int);
15565 vector signed int vec_max (vector signed int, vector signed int);
15566 vector float vec_max (vector float, vector float);
15568 vector float vec_vmaxfp (vector float, vector float);
15570 vector signed int vec_vmaxsw (vector bool int, vector signed int);
15571 vector signed int vec_vmaxsw (vector signed int, vector bool int);
15572 vector signed int vec_vmaxsw (vector signed int, vector signed int);
15574 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
15575 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
15576 vector unsigned int vec_vmaxuw (vector unsigned int,
15577 vector unsigned int);
15579 vector signed short vec_vmaxsh (vector bool short, vector signed short);
15580 vector signed short vec_vmaxsh (vector signed short, vector bool short);
15581 vector signed short vec_vmaxsh (vector signed short,
15582 vector signed short);
15584 vector unsigned short vec_vmaxuh (vector bool short,
15585 vector unsigned short);
15586 vector unsigned short vec_vmaxuh (vector unsigned short,
15587 vector bool short);
15588 vector unsigned short vec_vmaxuh (vector unsigned short,
15589 vector unsigned short);
15591 vector signed char vec_vmaxsb (vector bool char, vector signed char);
15592 vector signed char vec_vmaxsb (vector signed char, vector bool char);
15593 vector signed char vec_vmaxsb (vector signed char, vector signed char);
15595 vector unsigned char vec_vmaxub (vector bool char,
15596 vector unsigned char);
15597 vector unsigned char vec_vmaxub (vector unsigned char,
15599 vector unsigned char vec_vmaxub (vector unsigned char,
15600 vector unsigned char);
15602 vector bool char vec_mergeh (vector bool char, vector bool char);
15603 vector signed char vec_mergeh (vector signed char, vector signed char);
15604 vector unsigned char vec_mergeh (vector unsigned char,
15605 vector unsigned char);
15606 vector bool short vec_mergeh (vector bool short, vector bool short);
15607 vector pixel vec_mergeh (vector pixel, vector pixel);
15608 vector signed short vec_mergeh (vector signed short,
15609 vector signed short);
15610 vector unsigned short vec_mergeh (vector unsigned short,
15611 vector unsigned short);
15612 vector float vec_mergeh (vector float, vector float);
15613 vector bool int vec_mergeh (vector bool int, vector bool int);
15614 vector signed int vec_mergeh (vector signed int, vector signed int);
15615 vector unsigned int vec_mergeh (vector unsigned int,
15616 vector unsigned int);
15618 vector float vec_vmrghw (vector float, vector float);
15619 vector bool int vec_vmrghw (vector bool int, vector bool int);
15620 vector signed int vec_vmrghw (vector signed int, vector signed int);
15621 vector unsigned int vec_vmrghw (vector unsigned int,
15622 vector unsigned int);
15624 vector bool short vec_vmrghh (vector bool short, vector bool short);
15625 vector signed short vec_vmrghh (vector signed short,
15626 vector signed short);
15627 vector unsigned short vec_vmrghh (vector unsigned short,
15628 vector unsigned short);
15629 vector pixel vec_vmrghh (vector pixel, vector pixel);
15631 vector bool char vec_vmrghb (vector bool char, vector bool char);
15632 vector signed char vec_vmrghb (vector signed char, vector signed char);
15633 vector unsigned char vec_vmrghb (vector unsigned char,
15634 vector unsigned char);
15636 vector bool char vec_mergel (vector bool char, vector bool char);
15637 vector signed char vec_mergel (vector signed char, vector signed char);
15638 vector unsigned char vec_mergel (vector unsigned char,
15639 vector unsigned char);
15640 vector bool short vec_mergel (vector bool short, vector bool short);
15641 vector pixel vec_mergel (vector pixel, vector pixel);
15642 vector signed short vec_mergel (vector signed short,
15643 vector signed short);
15644 vector unsigned short vec_mergel (vector unsigned short,
15645 vector unsigned short);
15646 vector float vec_mergel (vector float, vector float);
15647 vector bool int vec_mergel (vector bool int, vector bool int);
15648 vector signed int vec_mergel (vector signed int, vector signed int);
15649 vector unsigned int vec_mergel (vector unsigned int,
15650 vector unsigned int);
15652 vector float vec_vmrglw (vector float, vector float);
15653 vector signed int vec_vmrglw (vector signed int, vector signed int);
15654 vector unsigned int vec_vmrglw (vector unsigned int,
15655 vector unsigned int);
15656 vector bool int vec_vmrglw (vector bool int, vector bool int);
15658 vector bool short vec_vmrglh (vector bool short, vector bool short);
15659 vector signed short vec_vmrglh (vector signed short,
15660 vector signed short);
15661 vector unsigned short vec_vmrglh (vector unsigned short,
15662 vector unsigned short);
15663 vector pixel vec_vmrglh (vector pixel, vector pixel);
15665 vector bool char vec_vmrglb (vector bool char, vector bool char);
15666 vector signed char vec_vmrglb (vector signed char, vector signed char);
15667 vector unsigned char vec_vmrglb (vector unsigned char,
15668 vector unsigned char);
15670 vector unsigned short vec_mfvscr (void);
15672 vector unsigned char vec_min (vector bool char, vector unsigned char);
15673 vector unsigned char vec_min (vector unsigned char, vector bool char);
15674 vector unsigned char vec_min (vector unsigned char,
15675 vector unsigned char);
15676 vector signed char vec_min (vector bool char, vector signed char);
15677 vector signed char vec_min (vector signed char, vector bool char);
15678 vector signed char vec_min (vector signed char, vector signed char);
15679 vector unsigned short vec_min (vector bool short,
15680 vector unsigned short);
15681 vector unsigned short vec_min (vector unsigned short,
15682 vector bool short);
15683 vector unsigned short vec_min (vector unsigned short,
15684 vector unsigned short);
15685 vector signed short vec_min (vector bool short, vector signed short);
15686 vector signed short vec_min (vector signed short, vector bool short);
15687 vector signed short vec_min (vector signed short, vector signed short);
15688 vector unsigned int vec_min (vector bool int, vector unsigned int);
15689 vector unsigned int vec_min (vector unsigned int, vector bool int);
15690 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
15691 vector signed int vec_min (vector bool int, vector signed int);
15692 vector signed int vec_min (vector signed int, vector bool int);
15693 vector signed int vec_min (vector signed int, vector signed int);
15694 vector float vec_min (vector float, vector float);
15696 vector float vec_vminfp (vector float, vector float);
15698 vector signed int vec_vminsw (vector bool int, vector signed int);
15699 vector signed int vec_vminsw (vector signed int, vector bool int);
15700 vector signed int vec_vminsw (vector signed int, vector signed int);
15702 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
15703 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
15704 vector unsigned int vec_vminuw (vector unsigned int,
15705 vector unsigned int);
15707 vector signed short vec_vminsh (vector bool short, vector signed short);
15708 vector signed short vec_vminsh (vector signed short, vector bool short);
15709 vector signed short vec_vminsh (vector signed short,
15710 vector signed short);
15712 vector unsigned short vec_vminuh (vector bool short,
15713 vector unsigned short);
15714 vector unsigned short vec_vminuh (vector unsigned short,
15715 vector bool short);
15716 vector unsigned short vec_vminuh (vector unsigned short,
15717 vector unsigned short);
15719 vector signed char vec_vminsb (vector bool char, vector signed char);
15720 vector signed char vec_vminsb (vector signed char, vector bool char);
15721 vector signed char vec_vminsb (vector signed char, vector signed char);
15723 vector unsigned char vec_vminub (vector bool char,
15724 vector unsigned char);
15725 vector unsigned char vec_vminub (vector unsigned char,
15727 vector unsigned char vec_vminub (vector unsigned char,
15728 vector unsigned char);
15730 vector signed short vec_mladd (vector signed short,
15731 vector signed short,
15732 vector signed short);
15733 vector signed short vec_mladd (vector signed short,
15734 vector unsigned short,
15735 vector unsigned short);
15736 vector signed short vec_mladd (vector unsigned short,
15737 vector signed short,
15738 vector signed short);
15739 vector unsigned short vec_mladd (vector unsigned short,
15740 vector unsigned short,
15741 vector unsigned short);
15743 vector signed short vec_mradds (vector signed short,
15744 vector signed short,
15745 vector signed short);
15747 vector unsigned int vec_msum (vector unsigned char,
15748 vector unsigned char,
15749 vector unsigned int);
15750 vector signed int vec_msum (vector signed char,
15751 vector unsigned char,
15752 vector signed int);
15753 vector unsigned int vec_msum (vector unsigned short,
15754 vector unsigned short,
15755 vector unsigned int);
15756 vector signed int vec_msum (vector signed short,
15757 vector signed short,
15758 vector signed int);
15760 vector signed int vec_vmsumshm (vector signed short,
15761 vector signed short,
15762 vector signed int);
15764 vector unsigned int vec_vmsumuhm (vector unsigned short,
15765 vector unsigned short,
15766 vector unsigned int);
15768 vector signed int vec_vmsummbm (vector signed char,
15769 vector unsigned char,
15770 vector signed int);
15772 vector unsigned int vec_vmsumubm (vector unsigned char,
15773 vector unsigned char,
15774 vector unsigned int);
15776 vector unsigned int vec_msums (vector unsigned short,
15777 vector unsigned short,
15778 vector unsigned int);
15779 vector signed int vec_msums (vector signed short,
15780 vector signed short,
15781 vector signed int);
15783 vector signed int vec_vmsumshs (vector signed short,
15784 vector signed short,
15785 vector signed int);
15787 vector unsigned int vec_vmsumuhs (vector unsigned short,
15788 vector unsigned short,
15789 vector unsigned int);
15791 void vec_mtvscr (vector signed int);
15792 void vec_mtvscr (vector unsigned int);
15793 void vec_mtvscr (vector bool int);
15794 void vec_mtvscr (vector signed short);
15795 void vec_mtvscr (vector unsigned short);
15796 void vec_mtvscr (vector bool short);
15797 void vec_mtvscr (vector pixel);
15798 void vec_mtvscr (vector signed char);
15799 void vec_mtvscr (vector unsigned char);
15800 void vec_mtvscr (vector bool char);
15802 vector unsigned short vec_mule (vector unsigned char,
15803 vector unsigned char);
15804 vector signed short vec_mule (vector signed char,
15805 vector signed char);
15806 vector unsigned int vec_mule (vector unsigned short,
15807 vector unsigned short);
15808 vector signed int vec_mule (vector signed short, vector signed short);
15810 vector signed int vec_vmulesh (vector signed short,
15811 vector signed short);
15813 vector unsigned int vec_vmuleuh (vector unsigned short,
15814 vector unsigned short);
15816 vector signed short vec_vmulesb (vector signed char,
15817 vector signed char);
15819 vector unsigned short vec_vmuleub (vector unsigned char,
15820 vector unsigned char);
15822 vector unsigned short vec_mulo (vector unsigned char,
15823 vector unsigned char);
15824 vector signed short vec_mulo (vector signed char, vector signed char);
15825 vector unsigned int vec_mulo (vector unsigned short,
15826 vector unsigned short);
15827 vector signed int vec_mulo (vector signed short, vector signed short);
15829 vector signed int vec_vmulosh (vector signed short,
15830 vector signed short);
15832 vector unsigned int vec_vmulouh (vector unsigned short,
15833 vector unsigned short);
15835 vector signed short vec_vmulosb (vector signed char,
15836 vector signed char);
15838 vector unsigned short vec_vmuloub (vector unsigned char,
15839 vector unsigned char);
15841 vector float vec_nmsub (vector float, vector float, vector float);
15843 vector float vec_nor (vector float, vector float);
15844 vector signed int vec_nor (vector signed int, vector signed int);
15845 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
15846 vector bool int vec_nor (vector bool int, vector bool int);
15847 vector signed short vec_nor (vector signed short, vector signed short);
15848 vector unsigned short vec_nor (vector unsigned short,
15849 vector unsigned short);
15850 vector bool short vec_nor (vector bool short, vector bool short);
15851 vector signed char vec_nor (vector signed char, vector signed char);
15852 vector unsigned char vec_nor (vector unsigned char,
15853 vector unsigned char);
15854 vector bool char vec_nor (vector bool char, vector bool char);
15856 vector float vec_or (vector float, vector float);
15857 vector float vec_or (vector float, vector bool int);
15858 vector float vec_or (vector bool int, vector float);
15859 vector bool int vec_or (vector bool int, vector bool int);
15860 vector signed int vec_or (vector bool int, vector signed int);
15861 vector signed int vec_or (vector signed int, vector bool int);
15862 vector signed int vec_or (vector signed int, vector signed int);
15863 vector unsigned int vec_or (vector bool int, vector unsigned int);
15864 vector unsigned int vec_or (vector unsigned int, vector bool int);
15865 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
15866 vector bool short vec_or (vector bool short, vector bool short);
15867 vector signed short vec_or (vector bool short, vector signed short);
15868 vector signed short vec_or (vector signed short, vector bool short);
15869 vector signed short vec_or (vector signed short, vector signed short);
15870 vector unsigned short vec_or (vector bool short, vector unsigned short);
15871 vector unsigned short vec_or (vector unsigned short, vector bool short);
15872 vector unsigned short vec_or (vector unsigned short,
15873 vector unsigned short);
15874 vector signed char vec_or (vector bool char, vector signed char);
15875 vector bool char vec_or (vector bool char, vector bool char);
15876 vector signed char vec_or (vector signed char, vector bool char);
15877 vector signed char vec_or (vector signed char, vector signed char);
15878 vector unsigned char vec_or (vector bool char, vector unsigned char);
15879 vector unsigned char vec_or (vector unsigned char, vector bool char);
15880 vector unsigned char vec_or (vector unsigned char,
15881 vector unsigned char);
15883 vector signed char vec_pack (vector signed short, vector signed short);
15884 vector unsigned char vec_pack (vector unsigned short,
15885 vector unsigned short);
15886 vector bool char vec_pack (vector bool short, vector bool short);
15887 vector signed short vec_pack (vector signed int, vector signed int);
15888 vector unsigned short vec_pack (vector unsigned int,
15889 vector unsigned int);
15890 vector bool short vec_pack (vector bool int, vector bool int);
15892 vector bool short vec_vpkuwum (vector bool int, vector bool int);
15893 vector signed short vec_vpkuwum (vector signed int, vector signed int);
15894 vector unsigned short vec_vpkuwum (vector unsigned int,
15895 vector unsigned int);
15897 vector bool char vec_vpkuhum (vector bool short, vector bool short);
15898 vector signed char vec_vpkuhum (vector signed short,
15899 vector signed short);
15900 vector unsigned char vec_vpkuhum (vector unsigned short,
15901 vector unsigned short);
15903 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
15905 vector unsigned char vec_packs (vector unsigned short,
15906 vector unsigned short);
15907 vector signed char vec_packs (vector signed short, vector signed short);
15908 vector unsigned short vec_packs (vector unsigned int,
15909 vector unsigned int);
15910 vector signed short vec_packs (vector signed int, vector signed int);
15912 vector signed short vec_vpkswss (vector signed int, vector signed int);
15914 vector unsigned short vec_vpkuwus (vector unsigned int,
15915 vector unsigned int);
15917 vector signed char vec_vpkshss (vector signed short,
15918 vector signed short);
15920 vector unsigned char vec_vpkuhus (vector unsigned short,
15921 vector unsigned short);
15923 vector unsigned char vec_packsu (vector unsigned short,
15924 vector unsigned short);
15925 vector unsigned char vec_packsu (vector signed short,
15926 vector signed short);
15927 vector unsigned short vec_packsu (vector unsigned int,
15928 vector unsigned int);
15929 vector unsigned short vec_packsu (vector signed int, vector signed int);
15931 vector unsigned short vec_vpkswus (vector signed int,
15932 vector signed int);
15934 vector unsigned char vec_vpkshus (vector signed short,
15935 vector signed short);
15937 vector float vec_perm (vector float,
15939 vector unsigned char);
15940 vector signed int vec_perm (vector signed int,
15942 vector unsigned char);
15943 vector unsigned int vec_perm (vector unsigned int,
15944 vector unsigned int,
15945 vector unsigned char);
15946 vector bool int vec_perm (vector bool int,
15948 vector unsigned char);
15949 vector signed short vec_perm (vector signed short,
15950 vector signed short,
15951 vector unsigned char);
15952 vector unsigned short vec_perm (vector unsigned short,
15953 vector unsigned short,
15954 vector unsigned char);
15955 vector bool short vec_perm (vector bool short,
15957 vector unsigned char);
15958 vector pixel vec_perm (vector pixel,
15960 vector unsigned char);
15961 vector signed char vec_perm (vector signed char,
15962 vector signed char,
15963 vector unsigned char);
15964 vector unsigned char vec_perm (vector unsigned char,
15965 vector unsigned char,
15966 vector unsigned char);
15967 vector bool char vec_perm (vector bool char,
15969 vector unsigned char);
15971 vector float vec_re (vector float);
15973 vector signed char vec_rl (vector signed char,
15974 vector unsigned char);
15975 vector unsigned char vec_rl (vector unsigned char,
15976 vector unsigned char);
15977 vector signed short vec_rl (vector signed short, vector unsigned short);
15978 vector unsigned short vec_rl (vector unsigned short,
15979 vector unsigned short);
15980 vector signed int vec_rl (vector signed int, vector unsigned int);
15981 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
15983 vector signed int vec_vrlw (vector signed int, vector unsigned int);
15984 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
15986 vector signed short vec_vrlh (vector signed short,
15987 vector unsigned short);
15988 vector unsigned short vec_vrlh (vector unsigned short,
15989 vector unsigned short);
15991 vector signed char vec_vrlb (vector signed char, vector unsigned char);
15992 vector unsigned char vec_vrlb (vector unsigned char,
15993 vector unsigned char);
15995 vector float vec_round (vector float);
15997 vector float vec_recip (vector float, vector float);
15999 vector float vec_rsqrt (vector float);
16001 vector float vec_rsqrte (vector float);
16003 vector float vec_sel (vector float, vector float, vector bool int);
16004 vector float vec_sel (vector float, vector float, vector unsigned int);
16005 vector signed int vec_sel (vector signed int,
16008 vector signed int vec_sel (vector signed int,
16010 vector unsigned int);
16011 vector unsigned int vec_sel (vector unsigned int,
16012 vector unsigned int,
16014 vector unsigned int vec_sel (vector unsigned int,
16015 vector unsigned int,
16016 vector unsigned int);
16017 vector bool int vec_sel (vector bool int,
16020 vector bool int vec_sel (vector bool int,
16022 vector unsigned int);
16023 vector signed short vec_sel (vector signed short,
16024 vector signed short,
16025 vector bool short);
16026 vector signed short vec_sel (vector signed short,
16027 vector signed short,
16028 vector unsigned short);
16029 vector unsigned short vec_sel (vector unsigned short,
16030 vector unsigned short,
16031 vector bool short);
16032 vector unsigned short vec_sel (vector unsigned short,
16033 vector unsigned short,
16034 vector unsigned short);
16035 vector bool short vec_sel (vector bool short,
16037 vector bool short);
16038 vector bool short vec_sel (vector bool short,
16040 vector unsigned short);
16041 vector signed char vec_sel (vector signed char,
16042 vector signed char,
16044 vector signed char vec_sel (vector signed char,
16045 vector signed char,
16046 vector unsigned char);
16047 vector unsigned char vec_sel (vector unsigned char,
16048 vector unsigned char,
16050 vector unsigned char vec_sel (vector unsigned char,
16051 vector unsigned char,
16052 vector unsigned char);
16053 vector bool char vec_sel (vector bool char,
16056 vector bool char vec_sel (vector bool char,
16058 vector unsigned char);
16060 vector signed char vec_sl (vector signed char,
16061 vector unsigned char);
16062 vector unsigned char vec_sl (vector unsigned char,
16063 vector unsigned char);
16064 vector signed short vec_sl (vector signed short, vector unsigned short);
16065 vector unsigned short vec_sl (vector unsigned short,
16066 vector unsigned short);
16067 vector signed int vec_sl (vector signed int, vector unsigned int);
16068 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
16070 vector signed int vec_vslw (vector signed int, vector unsigned int);
16071 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
16073 vector signed short vec_vslh (vector signed short,
16074 vector unsigned short);
16075 vector unsigned short vec_vslh (vector unsigned short,
16076 vector unsigned short);
16078 vector signed char vec_vslb (vector signed char, vector unsigned char);
16079 vector unsigned char vec_vslb (vector unsigned char,
16080 vector unsigned char);
16082 vector float vec_sld (vector float, vector float, const int);
16083 vector signed int vec_sld (vector signed int,
16086 vector unsigned int vec_sld (vector unsigned int,
16087 vector unsigned int,
16089 vector bool int vec_sld (vector bool int,
16092 vector signed short vec_sld (vector signed short,
16093 vector signed short,
16095 vector unsigned short vec_sld (vector unsigned short,
16096 vector unsigned short,
16098 vector bool short vec_sld (vector bool short,
16101 vector pixel vec_sld (vector pixel,
16104 vector signed char vec_sld (vector signed char,
16105 vector signed char,
16107 vector unsigned char vec_sld (vector unsigned char,
16108 vector unsigned char,
16110 vector bool char vec_sld (vector bool char,
16114 vector signed int vec_sll (vector signed int,
16115 vector unsigned int);
16116 vector signed int vec_sll (vector signed int,
16117 vector unsigned short);
16118 vector signed int vec_sll (vector signed int,
16119 vector unsigned char);
16120 vector unsigned int vec_sll (vector unsigned int,
16121 vector unsigned int);
16122 vector unsigned int vec_sll (vector unsigned int,
16123 vector unsigned short);
16124 vector unsigned int vec_sll (vector unsigned int,
16125 vector unsigned char);
16126 vector bool int vec_sll (vector bool int,
16127 vector unsigned int);
16128 vector bool int vec_sll (vector bool int,
16129 vector unsigned short);
16130 vector bool int vec_sll (vector bool int,
16131 vector unsigned char);
16132 vector signed short vec_sll (vector signed short,
16133 vector unsigned int);
16134 vector signed short vec_sll (vector signed short,
16135 vector unsigned short);
16136 vector signed short vec_sll (vector signed short,
16137 vector unsigned char);
16138 vector unsigned short vec_sll (vector unsigned short,
16139 vector unsigned int);
16140 vector unsigned short vec_sll (vector unsigned short,
16141 vector unsigned short);
16142 vector unsigned short vec_sll (vector unsigned short,
16143 vector unsigned char);
16144 vector bool short vec_sll (vector bool short, vector unsigned int);
16145 vector bool short vec_sll (vector bool short, vector unsigned short);
16146 vector bool short vec_sll (vector bool short, vector unsigned char);
16147 vector pixel vec_sll (vector pixel, vector unsigned int);
16148 vector pixel vec_sll (vector pixel, vector unsigned short);
16149 vector pixel vec_sll (vector pixel, vector unsigned char);
16150 vector signed char vec_sll (vector signed char, vector unsigned int);
16151 vector signed char vec_sll (vector signed char, vector unsigned short);
16152 vector signed char vec_sll (vector signed char, vector unsigned char);
16153 vector unsigned char vec_sll (vector unsigned char,
16154 vector unsigned int);
16155 vector unsigned char vec_sll (vector unsigned char,
16156 vector unsigned short);
16157 vector unsigned char vec_sll (vector unsigned char,
16158 vector unsigned char);
16159 vector bool char vec_sll (vector bool char, vector unsigned int);
16160 vector bool char vec_sll (vector bool char, vector unsigned short);
16161 vector bool char vec_sll (vector bool char, vector unsigned char);
16163 vector float vec_slo (vector float, vector signed char);
16164 vector float vec_slo (vector float, vector unsigned char);
16165 vector signed int vec_slo (vector signed int, vector signed char);
16166 vector signed int vec_slo (vector signed int, vector unsigned char);
16167 vector unsigned int vec_slo (vector unsigned int, vector signed char);
16168 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
16169 vector signed short vec_slo (vector signed short, vector signed char);
16170 vector signed short vec_slo (vector signed short, vector unsigned char);
16171 vector unsigned short vec_slo (vector unsigned short,
16172 vector signed char);
16173 vector unsigned short vec_slo (vector unsigned short,
16174 vector unsigned char);
16175 vector pixel vec_slo (vector pixel, vector signed char);
16176 vector pixel vec_slo (vector pixel, vector unsigned char);
16177 vector signed char vec_slo (vector signed char, vector signed char);
16178 vector signed char vec_slo (vector signed char, vector unsigned char);
16179 vector unsigned char vec_slo (vector unsigned char, vector signed char);
16180 vector unsigned char vec_slo (vector unsigned char,
16181 vector unsigned char);
16183 vector signed char vec_splat (vector signed char, const int);
16184 vector unsigned char vec_splat (vector unsigned char, const int);
16185 vector bool char vec_splat (vector bool char, const int);
16186 vector signed short vec_splat (vector signed short, const int);
16187 vector unsigned short vec_splat (vector unsigned short, const int);
16188 vector bool short vec_splat (vector bool short, const int);
16189 vector pixel vec_splat (vector pixel, const int);
16190 vector float vec_splat (vector float, const int);
16191 vector signed int vec_splat (vector signed int, const int);
16192 vector unsigned int vec_splat (vector unsigned int, const int);
16193 vector bool int vec_splat (vector bool int, const int);
16194 vector signed long vec_splat (vector signed long, const int);
16195 vector unsigned long vec_splat (vector unsigned long, const int);
16197 vector signed char vec_splats (signed char);
16198 vector unsigned char vec_splats (unsigned char);
16199 vector signed short vec_splats (signed short);
16200 vector unsigned short vec_splats (unsigned short);
16201 vector signed int vec_splats (signed int);
16202 vector unsigned int vec_splats (unsigned int);
16203 vector float vec_splats (float);
16205 vector float vec_vspltw (vector float, const int);
16206 vector signed int vec_vspltw (vector signed int, const int);
16207 vector unsigned int vec_vspltw (vector unsigned int, const int);
16208 vector bool int vec_vspltw (vector bool int, const int);
16210 vector bool short vec_vsplth (vector bool short, const int);
16211 vector signed short vec_vsplth (vector signed short, const int);
16212 vector unsigned short vec_vsplth (vector unsigned short, const int);
16213 vector pixel vec_vsplth (vector pixel, const int);
16215 vector signed char vec_vspltb (vector signed char, const int);
16216 vector unsigned char vec_vspltb (vector unsigned char, const int);
16217 vector bool char vec_vspltb (vector bool char, const int);
16219 vector signed char vec_splat_s8 (const int);
16221 vector signed short vec_splat_s16 (const int);
16223 vector signed int vec_splat_s32 (const int);
16225 vector unsigned char vec_splat_u8 (const int);
16227 vector unsigned short vec_splat_u16 (const int);
16229 vector unsigned int vec_splat_u32 (const int);
16231 vector signed char vec_sr (vector signed char, vector unsigned char);
16232 vector unsigned char vec_sr (vector unsigned char,
16233 vector unsigned char);
16234 vector signed short vec_sr (vector signed short,
16235 vector unsigned short);
16236 vector unsigned short vec_sr (vector unsigned short,
16237 vector unsigned short);
16238 vector signed int vec_sr (vector signed int, vector unsigned int);
16239 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
16241 vector signed int vec_vsrw (vector signed int, vector unsigned int);
16242 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
16244 vector signed short vec_vsrh (vector signed short,
16245 vector unsigned short);
16246 vector unsigned short vec_vsrh (vector unsigned short,
16247 vector unsigned short);
16249 vector signed char vec_vsrb (vector signed char, vector unsigned char);
16250 vector unsigned char vec_vsrb (vector unsigned char,
16251 vector unsigned char);
16253 vector signed char vec_sra (vector signed char, vector unsigned char);
16254 vector unsigned char vec_sra (vector unsigned char,
16255 vector unsigned char);
16256 vector signed short vec_sra (vector signed short,
16257 vector unsigned short);
16258 vector unsigned short vec_sra (vector unsigned short,
16259 vector unsigned short);
16260 vector signed int vec_sra (vector signed int, vector unsigned int);
16261 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
16263 vector signed int vec_vsraw (vector signed int, vector unsigned int);
16264 vector unsigned int vec_vsraw (vector unsigned int,
16265 vector unsigned int);
16267 vector signed short vec_vsrah (vector signed short,
16268 vector unsigned short);
16269 vector unsigned short vec_vsrah (vector unsigned short,
16270 vector unsigned short);
16272 vector signed char vec_vsrab (vector signed char, vector unsigned char);
16273 vector unsigned char vec_vsrab (vector unsigned char,
16274 vector unsigned char);
16276 vector signed int vec_srl (vector signed int, vector unsigned int);
16277 vector signed int vec_srl (vector signed int, vector unsigned short);
16278 vector signed int vec_srl (vector signed int, vector unsigned char);
16279 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
16280 vector unsigned int vec_srl (vector unsigned int,
16281 vector unsigned short);
16282 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
16283 vector bool int vec_srl (vector bool int, vector unsigned int);
16284 vector bool int vec_srl (vector bool int, vector unsigned short);
16285 vector bool int vec_srl (vector bool int, vector unsigned char);
16286 vector signed short vec_srl (vector signed short, vector unsigned int);
16287 vector signed short vec_srl (vector signed short,
16288 vector unsigned short);
16289 vector signed short vec_srl (vector signed short, vector unsigned char);
16290 vector unsigned short vec_srl (vector unsigned short,
16291 vector unsigned int);
16292 vector unsigned short vec_srl (vector unsigned short,
16293 vector unsigned short);
16294 vector unsigned short vec_srl (vector unsigned short,
16295 vector unsigned char);
16296 vector bool short vec_srl (vector bool short, vector unsigned int);
16297 vector bool short vec_srl (vector bool short, vector unsigned short);
16298 vector bool short vec_srl (vector bool short, vector unsigned char);
16299 vector pixel vec_srl (vector pixel, vector unsigned int);
16300 vector pixel vec_srl (vector pixel, vector unsigned short);
16301 vector pixel vec_srl (vector pixel, vector unsigned char);
16302 vector signed char vec_srl (vector signed char, vector unsigned int);
16303 vector signed char vec_srl (vector signed char, vector unsigned short);
16304 vector signed char vec_srl (vector signed char, vector unsigned char);
16305 vector unsigned char vec_srl (vector unsigned char,
16306 vector unsigned int);
16307 vector unsigned char vec_srl (vector unsigned char,
16308 vector unsigned short);
16309 vector unsigned char vec_srl (vector unsigned char,
16310 vector unsigned char);
16311 vector bool char vec_srl (vector bool char, vector unsigned int);
16312 vector bool char vec_srl (vector bool char, vector unsigned short);
16313 vector bool char vec_srl (vector bool char, vector unsigned char);
16315 vector float vec_sro (vector float, vector signed char);
16316 vector float vec_sro (vector float, vector unsigned char);
16317 vector signed int vec_sro (vector signed int, vector signed char);
16318 vector signed int vec_sro (vector signed int, vector unsigned char);
16319 vector unsigned int vec_sro (vector unsigned int, vector signed char);
16320 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
16321 vector signed short vec_sro (vector signed short, vector signed char);
16322 vector signed short vec_sro (vector signed short, vector unsigned char);
16323 vector unsigned short vec_sro (vector unsigned short,
16324 vector signed char);
16325 vector unsigned short vec_sro (vector unsigned short,
16326 vector unsigned char);
16327 vector pixel vec_sro (vector pixel, vector signed char);
16328 vector pixel vec_sro (vector pixel, vector unsigned char);
16329 vector signed char vec_sro (vector signed char, vector signed char);
16330 vector signed char vec_sro (vector signed char, vector unsigned char);
16331 vector unsigned char vec_sro (vector unsigned char, vector signed char);
16332 vector unsigned char vec_sro (vector unsigned char,
16333 vector unsigned char);
16335 void vec_st (vector float, int, vector float *);
16336 void vec_st (vector float, int, float *);
16337 void vec_st (vector signed int, int, vector signed int *);
16338 void vec_st (vector signed int, int, int *);
16339 void vec_st (vector unsigned int, int, vector unsigned int *);
16340 void vec_st (vector unsigned int, int, unsigned int *);
16341 void vec_st (vector bool int, int, vector bool int *);
16342 void vec_st (vector bool int, int, unsigned int *);
16343 void vec_st (vector bool int, int, int *);
16344 void vec_st (vector signed short, int, vector signed short *);
16345 void vec_st (vector signed short, int, short *);
16346 void vec_st (vector unsigned short, int, vector unsigned short *);
16347 void vec_st (vector unsigned short, int, unsigned short *);
16348 void vec_st (vector bool short, int, vector bool short *);
16349 void vec_st (vector bool short, int, unsigned short *);
16350 void vec_st (vector pixel, int, vector pixel *);
16351 void vec_st (vector pixel, int, unsigned short *);
16352 void vec_st (vector pixel, int, short *);
16353 void vec_st (vector bool short, int, short *);
16354 void vec_st (vector signed char, int, vector signed char *);
16355 void vec_st (vector signed char, int, signed char *);
16356 void vec_st (vector unsigned char, int, vector unsigned char *);
16357 void vec_st (vector unsigned char, int, unsigned char *);
16358 void vec_st (vector bool char, int, vector bool char *);
16359 void vec_st (vector bool char, int, unsigned char *);
16360 void vec_st (vector bool char, int, signed char *);
16362 void vec_ste (vector signed char, int, signed char *);
16363 void vec_ste (vector unsigned char, int, unsigned char *);
16364 void vec_ste (vector bool char, int, signed char *);
16365 void vec_ste (vector bool char, int, unsigned char *);
16366 void vec_ste (vector signed short, int, short *);
16367 void vec_ste (vector unsigned short, int, unsigned short *);
16368 void vec_ste (vector bool short, int, short *);
16369 void vec_ste (vector bool short, int, unsigned short *);
16370 void vec_ste (vector pixel, int, short *);
16371 void vec_ste (vector pixel, int, unsigned short *);
16372 void vec_ste (vector float, int, float *);
16373 void vec_ste (vector signed int, int, int *);
16374 void vec_ste (vector unsigned int, int, unsigned int *);
16375 void vec_ste (vector bool int, int, int *);
16376 void vec_ste (vector bool int, int, unsigned int *);
16378 void vec_stvewx (vector float, int, float *);
16379 void vec_stvewx (vector signed int, int, int *);
16380 void vec_stvewx (vector unsigned int, int, unsigned int *);
16381 void vec_stvewx (vector bool int, int, int *);
16382 void vec_stvewx (vector bool int, int, unsigned int *);
16384 void vec_stvehx (vector signed short, int, short *);
16385 void vec_stvehx (vector unsigned short, int, unsigned short *);
16386 void vec_stvehx (vector bool short, int, short *);
16387 void vec_stvehx (vector bool short, int, unsigned short *);
16388 void vec_stvehx (vector pixel, int, short *);
16389 void vec_stvehx (vector pixel, int, unsigned short *);
16391 void vec_stvebx (vector signed char, int, signed char *);
16392 void vec_stvebx (vector unsigned char, int, unsigned char *);
16393 void vec_stvebx (vector bool char, int, signed char *);
16394 void vec_stvebx (vector bool char, int, unsigned char *);
16396 void vec_stl (vector float, int, vector float *);
16397 void vec_stl (vector float, int, float *);
16398 void vec_stl (vector signed int, int, vector signed int *);
16399 void vec_stl (vector signed int, int, int *);
16400 void vec_stl (vector unsigned int, int, vector unsigned int *);
16401 void vec_stl (vector unsigned int, int, unsigned int *);
16402 void vec_stl (vector bool int, int, vector bool int *);
16403 void vec_stl (vector bool int, int, unsigned int *);
16404 void vec_stl (vector bool int, int, int *);
16405 void vec_stl (vector signed short, int, vector signed short *);
16406 void vec_stl (vector signed short, int, short *);
16407 void vec_stl (vector unsigned short, int, vector unsigned short *);
16408 void vec_stl (vector unsigned short, int, unsigned short *);
16409 void vec_stl (vector bool short, int, vector bool short *);
16410 void vec_stl (vector bool short, int, unsigned short *);
16411 void vec_stl (vector bool short, int, short *);
16412 void vec_stl (vector pixel, int, vector pixel *);
16413 void vec_stl (vector pixel, int, unsigned short *);
16414 void vec_stl (vector pixel, int, short *);
16415 void vec_stl (vector signed char, int, vector signed char *);
16416 void vec_stl (vector signed char, int, signed char *);
16417 void vec_stl (vector unsigned char, int, vector unsigned char *);
16418 void vec_stl (vector unsigned char, int, unsigned char *);
16419 void vec_stl (vector bool char, int, vector bool char *);
16420 void vec_stl (vector bool char, int, unsigned char *);
16421 void vec_stl (vector bool char, int, signed char *);
16423 vector signed char vec_sub (vector bool char, vector signed char);
16424 vector signed char vec_sub (vector signed char, vector bool char);
16425 vector signed char vec_sub (vector signed char, vector signed char);
16426 vector unsigned char vec_sub (vector bool char, vector unsigned char);
16427 vector unsigned char vec_sub (vector unsigned char, vector bool char);
16428 vector unsigned char vec_sub (vector unsigned char,
16429 vector unsigned char);
16430 vector signed short vec_sub (vector bool short, vector signed short);
16431 vector signed short vec_sub (vector signed short, vector bool short);
16432 vector signed short vec_sub (vector signed short, vector signed short);
16433 vector unsigned short vec_sub (vector bool short,
16434 vector unsigned short);
16435 vector unsigned short vec_sub (vector unsigned short,
16436 vector bool short);
16437 vector unsigned short vec_sub (vector unsigned short,
16438 vector unsigned short);
16439 vector signed int vec_sub (vector bool int, vector signed int);
16440 vector signed int vec_sub (vector signed int, vector bool int);
16441 vector signed int vec_sub (vector signed int, vector signed int);
16442 vector unsigned int vec_sub (vector bool int, vector unsigned int);
16443 vector unsigned int vec_sub (vector unsigned int, vector bool int);
16444 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
16445 vector float vec_sub (vector float, vector float);
16447 vector float vec_vsubfp (vector float, vector float);
16449 vector signed int vec_vsubuwm (vector bool int, vector signed int);
16450 vector signed int vec_vsubuwm (vector signed int, vector bool int);
16451 vector signed int vec_vsubuwm (vector signed int, vector signed int);
16452 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
16453 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
16454 vector unsigned int vec_vsubuwm (vector unsigned int,
16455 vector unsigned int);
16457 vector signed short vec_vsubuhm (vector bool short,
16458 vector signed short);
16459 vector signed short vec_vsubuhm (vector signed short,
16460 vector bool short);
16461 vector signed short vec_vsubuhm (vector signed short,
16462 vector signed short);
16463 vector unsigned short vec_vsubuhm (vector bool short,
16464 vector unsigned short);
16465 vector unsigned short vec_vsubuhm (vector unsigned short,
16466 vector bool short);
16467 vector unsigned short vec_vsubuhm (vector unsigned short,
16468 vector unsigned short);
16470 vector signed char vec_vsububm (vector bool char, vector signed char);
16471 vector signed char vec_vsububm (vector signed char, vector bool char);
16472 vector signed char vec_vsububm (vector signed char, vector signed char);
16473 vector unsigned char vec_vsububm (vector bool char,
16474 vector unsigned char);
16475 vector unsigned char vec_vsububm (vector unsigned char,
16477 vector unsigned char vec_vsububm (vector unsigned char,
16478 vector unsigned char);
16480 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
16482 vector unsigned char vec_subs (vector bool char, vector unsigned char);
16483 vector unsigned char vec_subs (vector unsigned char, vector bool char);
16484 vector unsigned char vec_subs (vector unsigned char,
16485 vector unsigned char);
16486 vector signed char vec_subs (vector bool char, vector signed char);
16487 vector signed char vec_subs (vector signed char, vector bool char);
16488 vector signed char vec_subs (vector signed char, vector signed char);
16489 vector unsigned short vec_subs (vector bool short,
16490 vector unsigned short);
16491 vector unsigned short vec_subs (vector unsigned short,
16492 vector bool short);
16493 vector unsigned short vec_subs (vector unsigned short,
16494 vector unsigned short);
16495 vector signed short vec_subs (vector bool short, vector signed short);
16496 vector signed short vec_subs (vector signed short, vector bool short);
16497 vector signed short vec_subs (vector signed short, vector signed short);
16498 vector unsigned int vec_subs (vector bool int, vector unsigned int);
16499 vector unsigned int vec_subs (vector unsigned int, vector bool int);
16500 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
16501 vector signed int vec_subs (vector bool int, vector signed int);
16502 vector signed int vec_subs (vector signed int, vector bool int);
16503 vector signed int vec_subs (vector signed int, vector signed int);
16505 vector signed int vec_vsubsws (vector bool int, vector signed int);
16506 vector signed int vec_vsubsws (vector signed int, vector bool int);
16507 vector signed int vec_vsubsws (vector signed int, vector signed int);
16509 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
16510 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
16511 vector unsigned int vec_vsubuws (vector unsigned int,
16512 vector unsigned int);
16514 vector signed short vec_vsubshs (vector bool short,
16515 vector signed short);
16516 vector signed short vec_vsubshs (vector signed short,
16517 vector bool short);
16518 vector signed short vec_vsubshs (vector signed short,
16519 vector signed short);
16521 vector unsigned short vec_vsubuhs (vector bool short,
16522 vector unsigned short);
16523 vector unsigned short vec_vsubuhs (vector unsigned short,
16524 vector bool short);
16525 vector unsigned short vec_vsubuhs (vector unsigned short,
16526 vector unsigned short);
16528 vector signed char vec_vsubsbs (vector bool char, vector signed char);
16529 vector signed char vec_vsubsbs (vector signed char, vector bool char);
16530 vector signed char vec_vsubsbs (vector signed char, vector signed char);
16532 vector unsigned char vec_vsububs (vector bool char,
16533 vector unsigned char);
16534 vector unsigned char vec_vsububs (vector unsigned char,
16536 vector unsigned char vec_vsububs (vector unsigned char,
16537 vector unsigned char);
16539 vector unsigned int vec_sum4s (vector unsigned char,
16540 vector unsigned int);
16541 vector signed int vec_sum4s (vector signed char, vector signed int);
16542 vector signed int vec_sum4s (vector signed short, vector signed int);
16544 vector signed int vec_vsum4shs (vector signed short, vector signed int);
16546 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
16548 vector unsigned int vec_vsum4ubs (vector unsigned char,
16549 vector unsigned int);
16551 vector signed int vec_sum2s (vector signed int, vector signed int);
16553 vector signed int vec_sums (vector signed int, vector signed int);
16555 vector float vec_trunc (vector float);
16557 vector signed short vec_unpackh (vector signed char);
16558 vector bool short vec_unpackh (vector bool char);
16559 vector signed int vec_unpackh (vector signed short);
16560 vector bool int vec_unpackh (vector bool short);
16561 vector unsigned int vec_unpackh (vector pixel);
16563 vector bool int vec_vupkhsh (vector bool short);
16564 vector signed int vec_vupkhsh (vector signed short);
16566 vector unsigned int vec_vupkhpx (vector pixel);
16568 vector bool short vec_vupkhsb (vector bool char);
16569 vector signed short vec_vupkhsb (vector signed char);
16571 vector signed short vec_unpackl (vector signed char);
16572 vector bool short vec_unpackl (vector bool char);
16573 vector unsigned int vec_unpackl (vector pixel);
16574 vector signed int vec_unpackl (vector signed short);
16575 vector bool int vec_unpackl (vector bool short);
16577 vector unsigned int vec_vupklpx (vector pixel);
16579 vector bool int vec_vupklsh (vector bool short);
16580 vector signed int vec_vupklsh (vector signed short);
16582 vector bool short vec_vupklsb (vector bool char);
16583 vector signed short vec_vupklsb (vector signed char);
16585 vector float vec_xor (vector float, vector float);
16586 vector float vec_xor (vector float, vector bool int);
16587 vector float vec_xor (vector bool int, vector float);
16588 vector bool int vec_xor (vector bool int, vector bool int);
16589 vector signed int vec_xor (vector bool int, vector signed int);
16590 vector signed int vec_xor (vector signed int, vector bool int);
16591 vector signed int vec_xor (vector signed int, vector signed int);
16592 vector unsigned int vec_xor (vector bool int, vector unsigned int);
16593 vector unsigned int vec_xor (vector unsigned int, vector bool int);
16594 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
16595 vector bool short vec_xor (vector bool short, vector bool short);
16596 vector signed short vec_xor (vector bool short, vector signed short);
16597 vector signed short vec_xor (vector signed short, vector bool short);
16598 vector signed short vec_xor (vector signed short, vector signed short);
16599 vector unsigned short vec_xor (vector bool short,
16600 vector unsigned short);
16601 vector unsigned short vec_xor (vector unsigned short,
16602 vector bool short);
16603 vector unsigned short vec_xor (vector unsigned short,
16604 vector unsigned short);
16605 vector signed char vec_xor (vector bool char, vector signed char);
16606 vector bool char vec_xor (vector bool char, vector bool char);
16607 vector signed char vec_xor (vector signed char, vector bool char);
16608 vector signed char vec_xor (vector signed char, vector signed char);
16609 vector unsigned char vec_xor (vector bool char, vector unsigned char);
16610 vector unsigned char vec_xor (vector unsigned char, vector bool char);
16611 vector unsigned char vec_xor (vector unsigned char,
16612 vector unsigned char);
16614 int vec_all_eq (vector signed char, vector bool char);
16615 int vec_all_eq (vector signed char, vector signed char);
16616 int vec_all_eq (vector unsigned char, vector bool char);
16617 int vec_all_eq (vector unsigned char, vector unsigned char);
16618 int vec_all_eq (vector bool char, vector bool char);
16619 int vec_all_eq (vector bool char, vector unsigned char);
16620 int vec_all_eq (vector bool char, vector signed char);
16621 int vec_all_eq (vector signed short, vector bool short);
16622 int vec_all_eq (vector signed short, vector signed short);
16623 int vec_all_eq (vector unsigned short, vector bool short);
16624 int vec_all_eq (vector unsigned short, vector unsigned short);
16625 int vec_all_eq (vector bool short, vector bool short);
16626 int vec_all_eq (vector bool short, vector unsigned short);
16627 int vec_all_eq (vector bool short, vector signed short);
16628 int vec_all_eq (vector pixel, vector pixel);
16629 int vec_all_eq (vector signed int, vector bool int);
16630 int vec_all_eq (vector signed int, vector signed int);
16631 int vec_all_eq (vector unsigned int, vector bool int);
16632 int vec_all_eq (vector unsigned int, vector unsigned int);
16633 int vec_all_eq (vector bool int, vector bool int);
16634 int vec_all_eq (vector bool int, vector unsigned int);
16635 int vec_all_eq (vector bool int, vector signed int);
16636 int vec_all_eq (vector float, vector float);
16638 int vec_all_ge (vector bool char, vector unsigned char);
16639 int vec_all_ge (vector unsigned char, vector bool char);
16640 int vec_all_ge (vector unsigned char, vector unsigned char);
16641 int vec_all_ge (vector bool char, vector signed char);
16642 int vec_all_ge (vector signed char, vector bool char);
16643 int vec_all_ge (vector signed char, vector signed char);
16644 int vec_all_ge (vector bool short, vector unsigned short);
16645 int vec_all_ge (vector unsigned short, vector bool short);
16646 int vec_all_ge (vector unsigned short, vector unsigned short);
16647 int vec_all_ge (vector signed short, vector signed short);
16648 int vec_all_ge (vector bool short, vector signed short);
16649 int vec_all_ge (vector signed short, vector bool short);
16650 int vec_all_ge (vector bool int, vector unsigned int);
16651 int vec_all_ge (vector unsigned int, vector bool int);
16652 int vec_all_ge (vector unsigned int, vector unsigned int);
16653 int vec_all_ge (vector bool int, vector signed int);
16654 int vec_all_ge (vector signed int, vector bool int);
16655 int vec_all_ge (vector signed int, vector signed int);
16656 int vec_all_ge (vector float, vector float);
16658 int vec_all_gt (vector bool char, vector unsigned char);
16659 int vec_all_gt (vector unsigned char, vector bool char);
16660 int vec_all_gt (vector unsigned char, vector unsigned char);
16661 int vec_all_gt (vector bool char, vector signed char);
16662 int vec_all_gt (vector signed char, vector bool char);
16663 int vec_all_gt (vector signed char, vector signed char);
16664 int vec_all_gt (vector bool short, vector unsigned short);
16665 int vec_all_gt (vector unsigned short, vector bool short);
16666 int vec_all_gt (vector unsigned short, vector unsigned short);
16667 int vec_all_gt (vector bool short, vector signed short);
16668 int vec_all_gt (vector signed short, vector bool short);
16669 int vec_all_gt (vector signed short, vector signed short);
16670 int vec_all_gt (vector bool int, vector unsigned int);
16671 int vec_all_gt (vector unsigned int, vector bool int);
16672 int vec_all_gt (vector unsigned int, vector unsigned int);
16673 int vec_all_gt (vector bool int, vector signed int);
16674 int vec_all_gt (vector signed int, vector bool int);
16675 int vec_all_gt (vector signed int, vector signed int);
16676 int vec_all_gt (vector float, vector float);
16678 int vec_all_in (vector float, vector float);
16680 int vec_all_le (vector bool char, vector unsigned char);
16681 int vec_all_le (vector unsigned char, vector bool char);
16682 int vec_all_le (vector unsigned char, vector unsigned char);
16683 int vec_all_le (vector bool char, vector signed char);
16684 int vec_all_le (vector signed char, vector bool char);
16685 int vec_all_le (vector signed char, vector signed char);
16686 int vec_all_le (vector bool short, vector unsigned short);
16687 int vec_all_le (vector unsigned short, vector bool short);
16688 int vec_all_le (vector unsigned short, vector unsigned short);
16689 int vec_all_le (vector bool short, vector signed short);
16690 int vec_all_le (vector signed short, vector bool short);
16691 int vec_all_le (vector signed short, vector signed short);
16692 int vec_all_le (vector bool int, vector unsigned int);
16693 int vec_all_le (vector unsigned int, vector bool int);
16694 int vec_all_le (vector unsigned int, vector unsigned int);
16695 int vec_all_le (vector bool int, vector signed int);
16696 int vec_all_le (vector signed int, vector bool int);
16697 int vec_all_le (vector signed int, vector signed int);
16698 int vec_all_le (vector float, vector float);
16700 int vec_all_lt (vector bool char, vector unsigned char);
16701 int vec_all_lt (vector unsigned char, vector bool char);
16702 int vec_all_lt (vector unsigned char, vector unsigned char);
16703 int vec_all_lt (vector bool char, vector signed char);
16704 int vec_all_lt (vector signed char, vector bool char);
16705 int vec_all_lt (vector signed char, vector signed char);
16706 int vec_all_lt (vector bool short, vector unsigned short);
16707 int vec_all_lt (vector unsigned short, vector bool short);
16708 int vec_all_lt (vector unsigned short, vector unsigned short);
16709 int vec_all_lt (vector bool short, vector signed short);
16710 int vec_all_lt (vector signed short, vector bool short);
16711 int vec_all_lt (vector signed short, vector signed short);
16712 int vec_all_lt (vector bool int, vector unsigned int);
16713 int vec_all_lt (vector unsigned int, vector bool int);
16714 int vec_all_lt (vector unsigned int, vector unsigned int);
16715 int vec_all_lt (vector bool int, vector signed int);
16716 int vec_all_lt (vector signed int, vector bool int);
16717 int vec_all_lt (vector signed int, vector signed int);
16718 int vec_all_lt (vector float, vector float);
16720 int vec_all_nan (vector float);
16722 int vec_all_ne (vector signed char, vector bool char);
16723 int vec_all_ne (vector signed char, vector signed char);
16724 int vec_all_ne (vector unsigned char, vector bool char);
16725 int vec_all_ne (vector unsigned char, vector unsigned char);
16726 int vec_all_ne (vector bool char, vector bool char);
16727 int vec_all_ne (vector bool char, vector unsigned char);
16728 int vec_all_ne (vector bool char, vector signed char);
16729 int vec_all_ne (vector signed short, vector bool short);
16730 int vec_all_ne (vector signed short, vector signed short);
16731 int vec_all_ne (vector unsigned short, vector bool short);
16732 int vec_all_ne (vector unsigned short, vector unsigned short);
16733 int vec_all_ne (vector bool short, vector bool short);
16734 int vec_all_ne (vector bool short, vector unsigned short);
16735 int vec_all_ne (vector bool short, vector signed short);
16736 int vec_all_ne (vector pixel, vector pixel);
16737 int vec_all_ne (vector signed int, vector bool int);
16738 int vec_all_ne (vector signed int, vector signed int);
16739 int vec_all_ne (vector unsigned int, vector bool int);
16740 int vec_all_ne (vector unsigned int, vector unsigned int);
16741 int vec_all_ne (vector bool int, vector bool int);
16742 int vec_all_ne (vector bool int, vector unsigned int);
16743 int vec_all_ne (vector bool int, vector signed int);
16744 int vec_all_ne (vector float, vector float);
16746 int vec_all_nge (vector float, vector float);
16748 int vec_all_ngt (vector float, vector float);
16750 int vec_all_nle (vector float, vector float);
16752 int vec_all_nlt (vector float, vector float);
16754 int vec_all_numeric (vector float);
16756 int vec_any_eq (vector signed char, vector bool char);
16757 int vec_any_eq (vector signed char, vector signed char);
16758 int vec_any_eq (vector unsigned char, vector bool char);
16759 int vec_any_eq (vector unsigned char, vector unsigned char);
16760 int vec_any_eq (vector bool char, vector bool char);
16761 int vec_any_eq (vector bool char, vector unsigned char);
16762 int vec_any_eq (vector bool char, vector signed char);
16763 int vec_any_eq (vector signed short, vector bool short);
16764 int vec_any_eq (vector signed short, vector signed short);
16765 int vec_any_eq (vector unsigned short, vector bool short);
16766 int vec_any_eq (vector unsigned short, vector unsigned short);
16767 int vec_any_eq (vector bool short, vector bool short);
16768 int vec_any_eq (vector bool short, vector unsigned short);
16769 int vec_any_eq (vector bool short, vector signed short);
16770 int vec_any_eq (vector pixel, vector pixel);
16771 int vec_any_eq (vector signed int, vector bool int);
16772 int vec_any_eq (vector signed int, vector signed int);
16773 int vec_any_eq (vector unsigned int, vector bool int);
16774 int vec_any_eq (vector unsigned int, vector unsigned int);
16775 int vec_any_eq (vector bool int, vector bool int);
16776 int vec_any_eq (vector bool int, vector unsigned int);
16777 int vec_any_eq (vector bool int, vector signed int);
16778 int vec_any_eq (vector float, vector float);
16780 int vec_any_ge (vector signed char, vector bool char);
16781 int vec_any_ge (vector unsigned char, vector bool char);
16782 int vec_any_ge (vector unsigned char, vector unsigned char);
16783 int vec_any_ge (vector signed char, vector signed char);
16784 int vec_any_ge (vector bool char, vector unsigned char);
16785 int vec_any_ge (vector bool char, vector signed char);
16786 int vec_any_ge (vector unsigned short, vector bool short);
16787 int vec_any_ge (vector unsigned short, vector unsigned short);
16788 int vec_any_ge (vector signed short, vector signed short);
16789 int vec_any_ge (vector signed short, vector bool short);
16790 int vec_any_ge (vector bool short, vector unsigned short);
16791 int vec_any_ge (vector bool short, vector signed short);
16792 int vec_any_ge (vector signed int, vector bool int);
16793 int vec_any_ge (vector unsigned int, vector bool int);
16794 int vec_any_ge (vector unsigned int, vector unsigned int);
16795 int vec_any_ge (vector signed int, vector signed int);
16796 int vec_any_ge (vector bool int, vector unsigned int);
16797 int vec_any_ge (vector bool int, vector signed int);
16798 int vec_any_ge (vector float, vector float);
16800 int vec_any_gt (vector bool char, vector unsigned char);
16801 int vec_any_gt (vector unsigned char, vector bool char);
16802 int vec_any_gt (vector unsigned char, vector unsigned char);
16803 int vec_any_gt (vector bool char, vector signed char);
16804 int vec_any_gt (vector signed char, vector bool char);
16805 int vec_any_gt (vector signed char, vector signed char);
16806 int vec_any_gt (vector bool short, vector unsigned short);
16807 int vec_any_gt (vector unsigned short, vector bool short);
16808 int vec_any_gt (vector unsigned short, vector unsigned short);
16809 int vec_any_gt (vector bool short, vector signed short);
16810 int vec_any_gt (vector signed short, vector bool short);
16811 int vec_any_gt (vector signed short, vector signed short);
16812 int vec_any_gt (vector bool int, vector unsigned int);
16813 int vec_any_gt (vector unsigned int, vector bool int);
16814 int vec_any_gt (vector unsigned int, vector unsigned int);
16815 int vec_any_gt (vector bool int, vector signed int);
16816 int vec_any_gt (vector signed int, vector bool int);
16817 int vec_any_gt (vector signed int, vector signed int);
16818 int vec_any_gt (vector float, vector float);
16820 int vec_any_le (vector bool char, vector unsigned char);
16821 int vec_any_le (vector unsigned char, vector bool char);
16822 int vec_any_le (vector unsigned char, vector unsigned char);
16823 int vec_any_le (vector bool char, vector signed char);
16824 int vec_any_le (vector signed char, vector bool char);
16825 int vec_any_le (vector signed char, vector signed char);
16826 int vec_any_le (vector bool short, vector unsigned short);
16827 int vec_any_le (vector unsigned short, vector bool short);
16828 int vec_any_le (vector unsigned short, vector unsigned short);
16829 int vec_any_le (vector bool short, vector signed short);
16830 int vec_any_le (vector signed short, vector bool short);
16831 int vec_any_le (vector signed short, vector signed short);
16832 int vec_any_le (vector bool int, vector unsigned int);
16833 int vec_any_le (vector unsigned int, vector bool int);
16834 int vec_any_le (vector unsigned int, vector unsigned int);
16835 int vec_any_le (vector bool int, vector signed int);
16836 int vec_any_le (vector signed int, vector bool int);
16837 int vec_any_le (vector signed int, vector signed int);
16838 int vec_any_le (vector float, vector float);
16840 int vec_any_lt (vector bool char, vector unsigned char);
16841 int vec_any_lt (vector unsigned char, vector bool char);
16842 int vec_any_lt (vector unsigned char, vector unsigned char);
16843 int vec_any_lt (vector bool char, vector signed char);
16844 int vec_any_lt (vector signed char, vector bool char);
16845 int vec_any_lt (vector signed char, vector signed char);
16846 int vec_any_lt (vector bool short, vector unsigned short);
16847 int vec_any_lt (vector unsigned short, vector bool short);
16848 int vec_any_lt (vector unsigned short, vector unsigned short);
16849 int vec_any_lt (vector bool short, vector signed short);
16850 int vec_any_lt (vector signed short, vector bool short);
16851 int vec_any_lt (vector signed short, vector signed short);
16852 int vec_any_lt (vector bool int, vector unsigned int);
16853 int vec_any_lt (vector unsigned int, vector bool int);
16854 int vec_any_lt (vector unsigned int, vector unsigned int);
16855 int vec_any_lt (vector bool int, vector signed int);
16856 int vec_any_lt (vector signed int, vector bool int);
16857 int vec_any_lt (vector signed int, vector signed int);
16858 int vec_any_lt (vector float, vector float);
16860 int vec_any_nan (vector float);
16862 int vec_any_ne (vector signed char, vector bool char);
16863 int vec_any_ne (vector signed char, vector signed char);
16864 int vec_any_ne (vector unsigned char, vector bool char);
16865 int vec_any_ne (vector unsigned char, vector unsigned char);
16866 int vec_any_ne (vector bool char, vector bool char);
16867 int vec_any_ne (vector bool char, vector unsigned char);
16868 int vec_any_ne (vector bool char, vector signed char);
16869 int vec_any_ne (vector signed short, vector bool short);
16870 int vec_any_ne (vector signed short, vector signed short);
16871 int vec_any_ne (vector unsigned short, vector bool short);
16872 int vec_any_ne (vector unsigned short, vector unsigned short);
16873 int vec_any_ne (vector bool short, vector bool short);
16874 int vec_any_ne (vector bool short, vector unsigned short);
16875 int vec_any_ne (vector bool short, vector signed short);
16876 int vec_any_ne (vector pixel, vector pixel);
16877 int vec_any_ne (vector signed int, vector bool int);
16878 int vec_any_ne (vector signed int, vector signed int);
16879 int vec_any_ne (vector unsigned int, vector bool int);
16880 int vec_any_ne (vector unsigned int, vector unsigned int);
16881 int vec_any_ne (vector bool int, vector bool int);
16882 int vec_any_ne (vector bool int, vector unsigned int);
16883 int vec_any_ne (vector bool int, vector signed int);
16884 int vec_any_ne (vector float, vector float);
16886 int vec_any_nge (vector float, vector float);
16888 int vec_any_ngt (vector float, vector float);
16890 int vec_any_nle (vector float, vector float);
16892 int vec_any_nlt (vector float, vector float);
16894 int vec_any_numeric (vector float);
16896 int vec_any_out (vector float, vector float);
16899 If the vector/scalar (VSX) instruction set is available, the following
16900 additional functions are available:
16903 vector double vec_abs (vector double);
16904 vector double vec_add (vector double, vector double);
16905 vector double vec_and (vector double, vector double);
16906 vector double vec_and (vector double, vector bool long);
16907 vector double vec_and (vector bool long, vector double);
16908 vector long vec_and (vector long, vector long);
16909 vector long vec_and (vector long, vector bool long);
16910 vector long vec_and (vector bool long, vector long);
16911 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
16912 vector unsigned long vec_and (vector unsigned long, vector bool long);
16913 vector unsigned long vec_and (vector bool long, vector unsigned long);
16914 vector double vec_andc (vector double, vector double);
16915 vector double vec_andc (vector double, vector bool long);
16916 vector double vec_andc (vector bool long, vector double);
16917 vector long vec_andc (vector long, vector long);
16918 vector long vec_andc (vector long, vector bool long);
16919 vector long vec_andc (vector bool long, vector long);
16920 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
16921 vector unsigned long vec_andc (vector unsigned long, vector bool long);
16922 vector unsigned long vec_andc (vector bool long, vector unsigned long);
16923 vector double vec_ceil (vector double);
16924 vector bool long vec_cmpeq (vector double, vector double);
16925 vector bool long vec_cmpge (vector double, vector double);
16926 vector bool long vec_cmpgt (vector double, vector double);
16927 vector bool long vec_cmple (vector double, vector double);
16928 vector bool long vec_cmplt (vector double, vector double);
16929 vector double vec_cpsgn (vector double, vector double);
16930 vector float vec_div (vector float, vector float);
16931 vector double vec_div (vector double, vector double);
16932 vector long vec_div (vector long, vector long);
16933 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
16934 vector double vec_floor (vector double);
16935 vector double vec_ld (int, const vector double *);
16936 vector double vec_ld (int, const double *);
16937 vector double vec_ldl (int, const vector double *);
16938 vector double vec_ldl (int, const double *);
16939 vector unsigned char vec_lvsl (int, const volatile double *);
16940 vector unsigned char vec_lvsr (int, const volatile double *);
16941 vector double vec_madd (vector double, vector double, vector double);
16942 vector double vec_max (vector double, vector double);
16943 vector signed long vec_mergeh (vector signed long, vector signed long);
16944 vector signed long vec_mergeh (vector signed long, vector bool long);
16945 vector signed long vec_mergeh (vector bool long, vector signed long);
16946 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
16947 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
16948 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
16949 vector signed long vec_mergel (vector signed long, vector signed long);
16950 vector signed long vec_mergel (vector signed long, vector bool long);
16951 vector signed long vec_mergel (vector bool long, vector signed long);
16952 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
16953 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
16954 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
16955 vector double vec_min (vector double, vector double);
16956 vector float vec_msub (vector float, vector float, vector float);
16957 vector double vec_msub (vector double, vector double, vector double);
16958 vector float vec_mul (vector float, vector float);
16959 vector double vec_mul (vector double, vector double);
16960 vector long vec_mul (vector long, vector long);
16961 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
16962 vector float vec_nearbyint (vector float);
16963 vector double vec_nearbyint (vector double);
16964 vector float vec_nmadd (vector float, vector float, vector float);
16965 vector double vec_nmadd (vector double, vector double, vector double);
16966 vector double vec_nmsub (vector double, vector double, vector double);
16967 vector double vec_nor (vector double, vector double);
16968 vector long vec_nor (vector long, vector long);
16969 vector long vec_nor (vector long, vector bool long);
16970 vector long vec_nor (vector bool long, vector long);
16971 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
16972 vector unsigned long vec_nor (vector unsigned long, vector bool long);
16973 vector unsigned long vec_nor (vector bool long, vector unsigned long);
16974 vector double vec_or (vector double, vector double);
16975 vector double vec_or (vector double, vector bool long);
16976 vector double vec_or (vector bool long, vector double);
16977 vector long vec_or (vector long, vector long);
16978 vector long vec_or (vector long, vector bool long);
16979 vector long vec_or (vector bool long, vector long);
16980 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
16981 vector unsigned long vec_or (vector unsigned long, vector bool long);
16982 vector unsigned long vec_or (vector bool long, vector unsigned long);
16983 vector double vec_perm (vector double, vector double, vector unsigned char);
16984 vector long vec_perm (vector long, vector long, vector unsigned char);
16985 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
16986 vector unsigned char);
16987 vector double vec_rint (vector double);
16988 vector double vec_recip (vector double, vector double);
16989 vector double vec_rsqrt (vector double);
16990 vector double vec_rsqrte (vector double);
16991 vector double vec_sel (vector double, vector double, vector bool long);
16992 vector double vec_sel (vector double, vector double, vector unsigned long);
16993 vector long vec_sel (vector long, vector long, vector long);
16994 vector long vec_sel (vector long, vector long, vector unsigned long);
16995 vector long vec_sel (vector long, vector long, vector bool long);
16996 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
16998 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
16999 vector unsigned long);
17000 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
17002 vector double vec_splats (double);
17003 vector signed long vec_splats (signed long);
17004 vector unsigned long vec_splats (unsigned long);
17005 vector float vec_sqrt (vector float);
17006 vector double vec_sqrt (vector double);
17007 void vec_st (vector double, int, vector double *);
17008 void vec_st (vector double, int, double *);
17009 vector double vec_sub (vector double, vector double);
17010 vector double vec_trunc (vector double);
17011 vector double vec_xl (int, vector double *);
17012 vector double vec_xl (int, double *);
17013 vector long long vec_xl (int, vector long long *);
17014 vector long long vec_xl (int, long long *);
17015 vector unsigned long long vec_xl (int, vector unsigned long long *);
17016 vector unsigned long long vec_xl (int, unsigned long long *);
17017 vector float vec_xl (int, vector float *);
17018 vector float vec_xl (int, float *);
17019 vector int vec_xl (int, vector int *);
17020 vector int vec_xl (int, int *);
17021 vector unsigned int vec_xl (int, vector unsigned int *);
17022 vector unsigned int vec_xl (int, unsigned int *);
17023 vector double vec_xor (vector double, vector double);
17024 vector double vec_xor (vector double, vector bool long);
17025 vector double vec_xor (vector bool long, vector double);
17026 vector long vec_xor (vector long, vector long);
17027 vector long vec_xor (vector long, vector bool long);
17028 vector long vec_xor (vector bool long, vector long);
17029 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
17030 vector unsigned long vec_xor (vector unsigned long, vector bool long);
17031 vector unsigned long vec_xor (vector bool long, vector unsigned long);
17032 void vec_xst (vector double, int, vector double *);
17033 void vec_xst (vector double, int, double *);
17034 void vec_xst (vector long long, int, vector long long *);
17035 void vec_xst (vector long long, int, long long *);
17036 void vec_xst (vector unsigned long long, int, vector unsigned long long *);
17037 void vec_xst (vector unsigned long long, int, unsigned long long *);
17038 void vec_xst (vector float, int, vector float *);
17039 void vec_xst (vector float, int, float *);
17040 void vec_xst (vector int, int, vector int *);
17041 void vec_xst (vector int, int, int *);
17042 void vec_xst (vector unsigned int, int, vector unsigned int *);
17043 void vec_xst (vector unsigned int, int, unsigned int *);
17044 int vec_all_eq (vector double, vector double);
17045 int vec_all_ge (vector double, vector double);
17046 int vec_all_gt (vector double, vector double);
17047 int vec_all_le (vector double, vector double);
17048 int vec_all_lt (vector double, vector double);
17049 int vec_all_nan (vector double);
17050 int vec_all_ne (vector double, vector double);
17051 int vec_all_nge (vector double, vector double);
17052 int vec_all_ngt (vector double, vector double);
17053 int vec_all_nle (vector double, vector double);
17054 int vec_all_nlt (vector double, vector double);
17055 int vec_all_numeric (vector double);
17056 int vec_any_eq (vector double, vector double);
17057 int vec_any_ge (vector double, vector double);
17058 int vec_any_gt (vector double, vector double);
17059 int vec_any_le (vector double, vector double);
17060 int vec_any_lt (vector double, vector double);
17061 int vec_any_nan (vector double);
17062 int vec_any_ne (vector double, vector double);
17063 int vec_any_nge (vector double, vector double);
17064 int vec_any_ngt (vector double, vector double);
17065 int vec_any_nle (vector double, vector double);
17066 int vec_any_nlt (vector double, vector double);
17067 int vec_any_numeric (vector double);
17069 vector double vec_vsx_ld (int, const vector double *);
17070 vector double vec_vsx_ld (int, const double *);
17071 vector float vec_vsx_ld (int, const vector float *);
17072 vector float vec_vsx_ld (int, const float *);
17073 vector bool int vec_vsx_ld (int, const vector bool int *);
17074 vector signed int vec_vsx_ld (int, const vector signed int *);
17075 vector signed int vec_vsx_ld (int, const int *);
17076 vector signed int vec_vsx_ld (int, const long *);
17077 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
17078 vector unsigned int vec_vsx_ld (int, const unsigned int *);
17079 vector unsigned int vec_vsx_ld (int, const unsigned long *);
17080 vector bool short vec_vsx_ld (int, const vector bool short *);
17081 vector pixel vec_vsx_ld (int, const vector pixel *);
17082 vector signed short vec_vsx_ld (int, const vector signed short *);
17083 vector signed short vec_vsx_ld (int, const short *);
17084 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
17085 vector unsigned short vec_vsx_ld (int, const unsigned short *);
17086 vector bool char vec_vsx_ld (int, const vector bool char *);
17087 vector signed char vec_vsx_ld (int, const vector signed char *);
17088 vector signed char vec_vsx_ld (int, const signed char *);
17089 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
17090 vector unsigned char vec_vsx_ld (int, const unsigned char *);
17092 void vec_vsx_st (vector double, int, vector double *);
17093 void vec_vsx_st (vector double, int, double *);
17094 void vec_vsx_st (vector float, int, vector float *);
17095 void vec_vsx_st (vector float, int, float *);
17096 void vec_vsx_st (vector signed int, int, vector signed int *);
17097 void vec_vsx_st (vector signed int, int, int *);
17098 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
17099 void vec_vsx_st (vector unsigned int, int, unsigned int *);
17100 void vec_vsx_st (vector bool int, int, vector bool int *);
17101 void vec_vsx_st (vector bool int, int, unsigned int *);
17102 void vec_vsx_st (vector bool int, int, int *);
17103 void vec_vsx_st (vector signed short, int, vector signed short *);
17104 void vec_vsx_st (vector signed short, int, short *);
17105 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
17106 void vec_vsx_st (vector unsigned short, int, unsigned short *);
17107 void vec_vsx_st (vector bool short, int, vector bool short *);
17108 void vec_vsx_st (vector bool short, int, unsigned short *);
17109 void vec_vsx_st (vector pixel, int, vector pixel *);
17110 void vec_vsx_st (vector pixel, int, unsigned short *);
17111 void vec_vsx_st (vector pixel, int, short *);
17112 void vec_vsx_st (vector bool short, int, short *);
17113 void vec_vsx_st (vector signed char, int, vector signed char *);
17114 void vec_vsx_st (vector signed char, int, signed char *);
17115 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
17116 void vec_vsx_st (vector unsigned char, int, unsigned char *);
17117 void vec_vsx_st (vector bool char, int, vector bool char *);
17118 void vec_vsx_st (vector bool char, int, unsigned char *);
17119 void vec_vsx_st (vector bool char, int, signed char *);
17121 vector double vec_xxpermdi (vector double, vector double, int);
17122 vector float vec_xxpermdi (vector float, vector float, int);
17123 vector long long vec_xxpermdi (vector long long, vector long long, int);
17124 vector unsigned long long vec_xxpermdi (vector unsigned long long,
17125 vector unsigned long long, int);
17126 vector int vec_xxpermdi (vector int, vector int, int);
17127 vector unsigned int vec_xxpermdi (vector unsigned int,
17128 vector unsigned int, int);
17129 vector short vec_xxpermdi (vector short, vector short, int);
17130 vector unsigned short vec_xxpermdi (vector unsigned short,
17131 vector unsigned short, int);
17132 vector signed char vec_xxpermdi (vector signed char, vector signed char, int);
17133 vector unsigned char vec_xxpermdi (vector unsigned char,
17134 vector unsigned char, int);
17136 vector double vec_xxsldi (vector double, vector double, int);
17137 vector float vec_xxsldi (vector float, vector float, int);
17138 vector long long vec_xxsldi (vector long long, vector long long, int);
17139 vector unsigned long long vec_xxsldi (vector unsigned long long,
17140 vector unsigned long long, int);
17141 vector int vec_xxsldi (vector int, vector int, int);
17142 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
17143 vector short vec_xxsldi (vector short, vector short, int);
17144 vector unsigned short vec_xxsldi (vector unsigned short,
17145 vector unsigned short, int);
17146 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
17147 vector unsigned char vec_xxsldi (vector unsigned char,
17148 vector unsigned char, int);
17151 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
17152 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
17153 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
17154 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
17155 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
17157 If the ISA 2.07 additions to the vector/scalar (power8-vector)
17158 instruction set are available, the following additional functions are
17159 available for both 32-bit and 64-bit targets. For 64-bit targets, you
17160 can use @var{vector long} instead of @var{vector long long},
17161 @var{vector bool long} instead of @var{vector bool long long}, and
17162 @var{vector unsigned long} instead of @var{vector unsigned long long}.
17165 vector long long vec_abs (vector long long);
17167 vector long long vec_add (vector long long, vector long long);
17168 vector unsigned long long vec_add (vector unsigned long long,
17169 vector unsigned long long);
17171 int vec_all_eq (vector long long, vector long long);
17172 int vec_all_eq (vector unsigned long long, vector unsigned long long);
17173 int vec_all_ge (vector long long, vector long long);
17174 int vec_all_ge (vector unsigned long long, vector unsigned long long);
17175 int vec_all_gt (vector long long, vector long long);
17176 int vec_all_gt (vector unsigned long long, vector unsigned long long);
17177 int vec_all_le (vector long long, vector long long);
17178 int vec_all_le (vector unsigned long long, vector unsigned long long);
17179 int vec_all_lt (vector long long, vector long long);
17180 int vec_all_lt (vector unsigned long long, vector unsigned long long);
17181 int vec_all_ne (vector long long, vector long long);
17182 int vec_all_ne (vector unsigned long long, vector unsigned long long);
17184 int vec_any_eq (vector long long, vector long long);
17185 int vec_any_eq (vector unsigned long long, vector unsigned long long);
17186 int vec_any_ge (vector long long, vector long long);
17187 int vec_any_ge (vector unsigned long long, vector unsigned long long);
17188 int vec_any_gt (vector long long, vector long long);
17189 int vec_any_gt (vector unsigned long long, vector unsigned long long);
17190 int vec_any_le (vector long long, vector long long);
17191 int vec_any_le (vector unsigned long long, vector unsigned long long);
17192 int vec_any_lt (vector long long, vector long long);
17193 int vec_any_lt (vector unsigned long long, vector unsigned long long);
17194 int vec_any_ne (vector long long, vector long long);
17195 int vec_any_ne (vector unsigned long long, vector unsigned long long);
17197 vector long long vec_eqv (vector long long, vector long long);
17198 vector long long vec_eqv (vector bool long long, vector long long);
17199 vector long long vec_eqv (vector long long, vector bool long long);
17200 vector unsigned long long vec_eqv (vector unsigned long long,
17201 vector unsigned long long);
17202 vector unsigned long long vec_eqv (vector bool long long,
17203 vector unsigned long long);
17204 vector unsigned long long vec_eqv (vector unsigned long long,
17205 vector bool long long);
17206 vector int vec_eqv (vector int, vector int);
17207 vector int vec_eqv (vector bool int, vector int);
17208 vector int vec_eqv (vector int, vector bool int);
17209 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
17210 vector unsigned int vec_eqv (vector bool unsigned int,
17211 vector unsigned int);
17212 vector unsigned int vec_eqv (vector unsigned int,
17213 vector bool unsigned int);
17214 vector short vec_eqv (vector short, vector short);
17215 vector short vec_eqv (vector bool short, vector short);
17216 vector short vec_eqv (vector short, vector bool short);
17217 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
17218 vector unsigned short vec_eqv (vector bool unsigned short,
17219 vector unsigned short);
17220 vector unsigned short vec_eqv (vector unsigned short,
17221 vector bool unsigned short);
17222 vector signed char vec_eqv (vector signed char, vector signed char);
17223 vector signed char vec_eqv (vector bool signed char, vector signed char);
17224 vector signed char vec_eqv (vector signed char, vector bool signed char);
17225 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
17226 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
17227 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
17229 vector long long vec_max (vector long long, vector long long);
17230 vector unsigned long long vec_max (vector unsigned long long,
17231 vector unsigned long long);
17233 vector signed int vec_mergee (vector signed int, vector signed int);
17234 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
17235 vector bool int vec_mergee (vector bool int, vector bool int);
17237 vector signed int vec_mergeo (vector signed int, vector signed int);
17238 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
17239 vector bool int vec_mergeo (vector bool int, vector bool int);
17241 vector long long vec_min (vector long long, vector long long);
17242 vector unsigned long long vec_min (vector unsigned long long,
17243 vector unsigned long long);
17245 vector long long vec_nand (vector long long, vector long long);
17246 vector long long vec_nand (vector bool long long, vector long long);
17247 vector long long vec_nand (vector long long, vector bool long long);
17248 vector unsigned long long vec_nand (vector unsigned long long,
17249 vector unsigned long long);
17250 vector unsigned long long vec_nand (vector bool long long,
17251 vector unsigned long long);
17252 vector unsigned long long vec_nand (vector unsigned long long,
17253 vector bool long long);
17254 vector int vec_nand (vector int, vector int);
17255 vector int vec_nand (vector bool int, vector int);
17256 vector int vec_nand (vector int, vector bool int);
17257 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
17258 vector unsigned int vec_nand (vector bool unsigned int,
17259 vector unsigned int);
17260 vector unsigned int vec_nand (vector unsigned int,
17261 vector bool unsigned int);
17262 vector short vec_nand (vector short, vector short);
17263 vector short vec_nand (vector bool short, vector short);
17264 vector short vec_nand (vector short, vector bool short);
17265 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
17266 vector unsigned short vec_nand (vector bool unsigned short,
17267 vector unsigned short);
17268 vector unsigned short vec_nand (vector unsigned short,
17269 vector bool unsigned short);
17270 vector signed char vec_nand (vector signed char, vector signed char);
17271 vector signed char vec_nand (vector bool signed char, vector signed char);
17272 vector signed char vec_nand (vector signed char, vector bool signed char);
17273 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
17274 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
17275 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
17277 vector long long vec_orc (vector long long, vector long long);
17278 vector long long vec_orc (vector bool long long, vector long long);
17279 vector long long vec_orc (vector long long, vector bool long long);
17280 vector unsigned long long vec_orc (vector unsigned long long,
17281 vector unsigned long long);
17282 vector unsigned long long vec_orc (vector bool long long,
17283 vector unsigned long long);
17284 vector unsigned long long vec_orc (vector unsigned long long,
17285 vector bool long long);
17286 vector int vec_orc (vector int, vector int);
17287 vector int vec_orc (vector bool int, vector int);
17288 vector int vec_orc (vector int, vector bool int);
17289 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
17290 vector unsigned int vec_orc (vector bool unsigned int,
17291 vector unsigned int);
17292 vector unsigned int vec_orc (vector unsigned int,
17293 vector bool unsigned int);
17294 vector short vec_orc (vector short, vector short);
17295 vector short vec_orc (vector bool short, vector short);
17296 vector short vec_orc (vector short, vector bool short);
17297 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
17298 vector unsigned short vec_orc (vector bool unsigned short,
17299 vector unsigned short);
17300 vector unsigned short vec_orc (vector unsigned short,
17301 vector bool unsigned short);
17302 vector signed char vec_orc (vector signed char, vector signed char);
17303 vector signed char vec_orc (vector bool signed char, vector signed char);
17304 vector signed char vec_orc (vector signed char, vector bool signed char);
17305 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
17306 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
17307 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
17309 vector int vec_pack (vector long long, vector long long);
17310 vector unsigned int vec_pack (vector unsigned long long,
17311 vector unsigned long long);
17312 vector bool int vec_pack (vector bool long long, vector bool long long);
17314 vector int vec_packs (vector long long, vector long long);
17315 vector unsigned int vec_packs (vector unsigned long long,
17316 vector unsigned long long);
17318 vector unsigned int vec_packsu (vector long long, vector long long);
17319 vector unsigned int vec_packsu (vector unsigned long long,
17320 vector unsigned long long);
17322 vector long long vec_rl (vector long long,
17323 vector unsigned long long);
17324 vector long long vec_rl (vector unsigned long long,
17325 vector unsigned long long);
17327 vector long long vec_sl (vector long long, vector unsigned long long);
17328 vector long long vec_sl (vector unsigned long long,
17329 vector unsigned long long);
17331 vector long long vec_sr (vector long long, vector unsigned long long);
17332 vector unsigned long long char vec_sr (vector unsigned long long,
17333 vector unsigned long long);
17335 vector long long vec_sra (vector long long, vector unsigned long long);
17336 vector unsigned long long vec_sra (vector unsigned long long,
17337 vector unsigned long long);
17339 vector long long vec_sub (vector long long, vector long long);
17340 vector unsigned long long vec_sub (vector unsigned long long,
17341 vector unsigned long long);
17343 vector long long vec_unpackh (vector int);
17344 vector unsigned long long vec_unpackh (vector unsigned int);
17346 vector long long vec_unpackl (vector int);
17347 vector unsigned long long vec_unpackl (vector unsigned int);
17349 vector long long vec_vaddudm (vector long long, vector long long);
17350 vector long long vec_vaddudm (vector bool long long, vector long long);
17351 vector long long vec_vaddudm (vector long long, vector bool long long);
17352 vector unsigned long long vec_vaddudm (vector unsigned long long,
17353 vector unsigned long long);
17354 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
17355 vector unsigned long long);
17356 vector unsigned long long vec_vaddudm (vector unsigned long long,
17357 vector bool unsigned long long);
17359 vector long long vec_vbpermq (vector signed char, vector signed char);
17360 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
17362 vector long long vec_cntlz (vector long long);
17363 vector unsigned long long vec_cntlz (vector unsigned long long);
17364 vector int vec_cntlz (vector int);
17365 vector unsigned int vec_cntlz (vector int);
17366 vector short vec_cntlz (vector short);
17367 vector unsigned short vec_cntlz (vector unsigned short);
17368 vector signed char vec_cntlz (vector signed char);
17369 vector unsigned char vec_cntlz (vector unsigned char);
17371 vector long long vec_vclz (vector long long);
17372 vector unsigned long long vec_vclz (vector unsigned long long);
17373 vector int vec_vclz (vector int);
17374 vector unsigned int vec_vclz (vector int);
17375 vector short vec_vclz (vector short);
17376 vector unsigned short vec_vclz (vector unsigned short);
17377 vector signed char vec_vclz (vector signed char);
17378 vector unsigned char vec_vclz (vector unsigned char);
17380 vector signed char vec_vclzb (vector signed char);
17381 vector unsigned char vec_vclzb (vector unsigned char);
17383 vector long long vec_vclzd (vector long long);
17384 vector unsigned long long vec_vclzd (vector unsigned long long);
17386 vector short vec_vclzh (vector short);
17387 vector unsigned short vec_vclzh (vector unsigned short);
17389 vector int vec_vclzw (vector int);
17390 vector unsigned int vec_vclzw (vector int);
17392 vector signed char vec_vgbbd (vector signed char);
17393 vector unsigned char vec_vgbbd (vector unsigned char);
17395 vector long long vec_vmaxsd (vector long long, vector long long);
17397 vector unsigned long long vec_vmaxud (vector unsigned long long,
17398 unsigned vector long long);
17400 vector long long vec_vminsd (vector long long, vector long long);
17402 vector unsigned long long vec_vminud (vector long long,
17405 vector int vec_vpksdss (vector long long, vector long long);
17406 vector unsigned int vec_vpksdss (vector long long, vector long long);
17408 vector unsigned int vec_vpkudus (vector unsigned long long,
17409 vector unsigned long long);
17411 vector int vec_vpkudum (vector long long, vector long long);
17412 vector unsigned int vec_vpkudum (vector unsigned long long,
17413 vector unsigned long long);
17414 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
17416 vector long long vec_vpopcnt (vector long long);
17417 vector unsigned long long vec_vpopcnt (vector unsigned long long);
17418 vector int vec_vpopcnt (vector int);
17419 vector unsigned int vec_vpopcnt (vector int);
17420 vector short vec_vpopcnt (vector short);
17421 vector unsigned short vec_vpopcnt (vector unsigned short);
17422 vector signed char vec_vpopcnt (vector signed char);
17423 vector unsigned char vec_vpopcnt (vector unsigned char);
17425 vector signed char vec_vpopcntb (vector signed char);
17426 vector unsigned char vec_vpopcntb (vector unsigned char);
17428 vector long long vec_vpopcntd (vector long long);
17429 vector unsigned long long vec_vpopcntd (vector unsigned long long);
17431 vector short vec_vpopcnth (vector short);
17432 vector unsigned short vec_vpopcnth (vector unsigned short);
17434 vector int vec_vpopcntw (vector int);
17435 vector unsigned int vec_vpopcntw (vector int);
17437 vector long long vec_vrld (vector long long, vector unsigned long long);
17438 vector unsigned long long vec_vrld (vector unsigned long long,
17439 vector unsigned long long);
17441 vector long long vec_vsld (vector long long, vector unsigned long long);
17442 vector long long vec_vsld (vector unsigned long long,
17443 vector unsigned long long);
17445 vector long long vec_vsrad (vector long long, vector unsigned long long);
17446 vector unsigned long long vec_vsrad (vector unsigned long long,
17447 vector unsigned long long);
17449 vector long long vec_vsrd (vector long long, vector unsigned long long);
17450 vector unsigned long long char vec_vsrd (vector unsigned long long,
17451 vector unsigned long long);
17453 vector long long vec_vsubudm (vector long long, vector long long);
17454 vector long long vec_vsubudm (vector bool long long, vector long long);
17455 vector long long vec_vsubudm (vector long long, vector bool long long);
17456 vector unsigned long long vec_vsubudm (vector unsigned long long,
17457 vector unsigned long long);
17458 vector unsigned long long vec_vsubudm (vector bool long long,
17459 vector unsigned long long);
17460 vector unsigned long long vec_vsubudm (vector unsigned long long,
17461 vector bool long long);
17463 vector long long vec_vupkhsw (vector int);
17464 vector unsigned long long vec_vupkhsw (vector unsigned int);
17466 vector long long vec_vupklsw (vector int);
17467 vector unsigned long long vec_vupklsw (vector int);
17470 If the ISA 2.07 additions to the vector/scalar (power8-vector)
17471 instruction set are available, the following additional functions are
17472 available for 64-bit targets. New vector types
17473 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
17474 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
17477 The normal vector extract, and set operations work on
17478 @var{vector __int128_t} and @var{vector __uint128_t} types,
17479 but the index value must be 0.
17482 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
17483 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
17485 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
17486 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
17488 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
17489 vector __int128_t);
17490 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
17491 vector __uint128_t);
17493 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
17494 vector __int128_t);
17495 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
17496 vector __uint128_t);
17498 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
17499 vector __int128_t);
17500 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
17501 vector __uint128_t);
17503 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
17504 vector __int128_t);
17505 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
17506 vector __uint128_t);
17508 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
17509 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
17511 __int128_t vec_vsubuqm (__int128_t, __int128_t);
17512 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
17514 vector __int128_t __builtin_bcdadd (vector __int128_t, vector__int128_t);
17515 int __builtin_bcdadd_lt (vector __int128_t, vector__int128_t);
17516 int __builtin_bcdadd_eq (vector __int128_t, vector__int128_t);
17517 int __builtin_bcdadd_gt (vector __int128_t, vector__int128_t);
17518 int __builtin_bcdadd_ov (vector __int128_t, vector__int128_t);
17519 vector __int128_t bcdsub (vector __int128_t, vector__int128_t);
17520 int __builtin_bcdsub_lt (vector __int128_t, vector__int128_t);
17521 int __builtin_bcdsub_eq (vector __int128_t, vector__int128_t);
17522 int __builtin_bcdsub_gt (vector __int128_t, vector__int128_t);
17523 int __builtin_bcdsub_ov (vector __int128_t, vector__int128_t);
17526 If the ISA 3.0 additions to the vector/scalar (power9-vector)
17527 instruction set are available:
17530 vector long long vec_vctz (vector long long);
17531 vector unsigned long long vec_vctz (vector unsigned long long);
17532 vector int vec_vctz (vector int);
17533 vector unsigned int vec_vctz (vector int);
17534 vector short vec_vctz (vector short);
17535 vector unsigned short vec_vctz (vector unsigned short);
17536 vector signed char vec_vctz (vector signed char);
17537 vector unsigned char vec_vctz (vector unsigned char);
17539 vector signed char vec_vctzb (vector signed char);
17540 vector unsigned char vec_vctzb (vector unsigned char);
17542 vector long long vec_vctzd (vector long long);
17543 vector unsigned long long vec_vctzd (vector unsigned long long);
17545 vector short vec_vctzh (vector short);
17546 vector unsigned short vec_vctzh (vector unsigned short);
17548 vector int vec_vctzw (vector int);
17549 vector unsigned int vec_vctzw (vector int);
17551 vector int vec_vprtyb (vector int);
17552 vector unsigned int vec_vprtyb (vector unsigned int);
17553 vector long long vec_vprtyb (vector long long);
17554 vector unsigned long long vec_vprtyb (vector unsigned long long);
17556 vector int vec_vprtybw (vector int);
17557 vector unsigned int vec_vprtybw (vector unsigned int);
17559 vector long long vec_vprtybd (vector long long);
17560 vector unsigned long long vec_vprtybd (vector unsigned long long);
17564 If the ISA 3.0 additions to the vector/scalar (power9-vector)
17565 instruction set are available for 64-bit targets:
17568 vector long vec_vprtyb (vector long);
17569 vector unsigned long vec_vprtyb (vector unsigned long);
17570 vector __int128_t vec_vprtyb (vector __int128_t);
17571 vector __uint128_t vec_vprtyb (vector __uint128_t);
17573 vector long vec_vprtybd (vector long);
17574 vector unsigned long vec_vprtybd (vector unsigned long);
17576 vector __int128_t vec_vprtybq (vector __int128_t);
17577 vector __uint128_t vec_vprtybd (vector __uint128_t);
17580 The following built-in vector functions are available for the PowerPC family
17581 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9})
17582 or with @option{-mpower9-vector}:
17584 __vector unsigned char
17585 vec_slv (__vector unsigned char src, __vector unsigned char shift_distance);
17586 __vector unsigned char
17587 vec_srv (__vector unsigned char src, __vector unsigned char shift_distance);
17590 The @code{vec_slv} and @code{vec_srv} functions operate on
17591 all of the bytes of their @code{src} and @code{shift_distance}
17592 arguments in parallel. The behavior of the @code{vec_slv} is as if
17593 there existed a temporary array of 17 unsigned characters
17594 @code{slv_array} within which elements 0 through 15 are the same as
17595 the entries in the @code{src} array and element 16 equals 0. The
17596 result returned from the @code{vec_slv} function is a
17597 @code{__vector} of 16 unsigned characters within which element
17598 @code{i} is computed using the C expression
17599 @code{0xff & (*((unsigned short *)(slv_array + i)) << (0x07 &
17600 shift_distance[i]))},
17601 with this resulting value coerced to the @code{unsigned char} type.
17602 The behavior of the @code{vec_srv} is as if
17603 there existed a temporary array of 17 unsigned characters
17604 @code{srv_array} within which element 0 equals zero and
17605 elements 1 through 16 equal the elements 0 through 15 of
17606 the @code{src} array. The
17607 result returned from the @code{vec_srv} function is a
17608 @code{__vector} of 16 unsigned characters within which element
17609 @code{i} is computed using the C expression
17610 @code{0xff & (*((unsigned short *)(srv_array + i)) >>
17611 (0x07 & shift_distance[i]))},
17612 with this resulting value coerced to the @code{unsigned char} type.
17614 The following built-in functions are available for the PowerPC family
17615 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9})
17616 or with @option{-mpower9-vector}:
17618 __vector unsigned char
17619 vec_absd (__vector unsigned char arg1, __vector unsigned char arg2);
17620 __vector unsigned short
17621 vec_absd (__vector unsigned short arg1, __vector unsigned short arg2);
17622 __vector unsigned int
17623 vec_absd (__vector unsigned int arg1, __vector unsigned int arg2);
17625 __vector unsigned char
17626 vec_absdb (__vector unsigned char arg1, __vector unsigned char arg2);
17627 __vector unsigned short
17628 vec_absdh (__vector unsigned short arg1, __vector unsigned short arg2);
17629 __vector unsigned int
17630 vec_absdw (__vector unsigned int arg1, __vector unsigned int arg2);
17633 The @code{vec_absd}, @code{vec_absdb}, @code{vec_absdh}, and
17634 @code{vec_absdw} built-in functions each computes the absolute
17635 differences of the pairs of vector elements supplied in its two vector
17636 arguments, placing the absolute differences into the corresponding
17637 elements of the vector result.
17639 If the cryptographic instructions are enabled (@option{-mcrypto} or
17640 @option{-mcpu=power8}), the following builtins are enabled.
17643 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
17645 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
17646 vector unsigned long long);
17648 vector unsigned long long __builtin_crypto_vcipherlast
17649 (vector unsigned long long,
17650 vector unsigned long long);
17652 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
17653 vector unsigned long long);
17655 vector unsigned long long __builtin_crypto_vncipherlast
17656 (vector unsigned long long,
17657 vector unsigned long long);
17659 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
17660 vector unsigned char,
17661 vector unsigned char);
17663 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
17664 vector unsigned short,
17665 vector unsigned short);
17667 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
17668 vector unsigned int,
17669 vector unsigned int);
17671 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
17672 vector unsigned long long,
17673 vector unsigned long long);
17675 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
17676 vector unsigned char);
17678 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
17679 vector unsigned short);
17681 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
17682 vector unsigned int);
17684 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
17685 vector unsigned long long);
17687 vector unsigned long long __builtin_crypto_vshasigmad
17688 (vector unsigned long long, int, int);
17690 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
17694 The second argument to the @var{__builtin_crypto_vshasigmad} and
17695 @var{__builtin_crypto_vshasigmaw} builtin functions must be a constant
17696 integer that is 0 or 1. The third argument to these builtin functions
17697 must be a constant integer in the range of 0 to 15.
17699 If the ISA 3.0 additions to the vector/scalar (power9-vector)
17700 instruction set are available, the following additional functions are
17701 available for both 32-bit and 64-bit targets.
17703 vector short vec_xl (int, vector short *);
17704 vector short vec_xl (int, short *);
17705 vector unsigned short vec_xl (int, vector unsigned short *);
17706 vector unsigned short vec_xl (int, unsigned short *);
17707 vector char vec_xl (int, vector char *);
17708 vector char vec_xl (int, char *);
17709 vector unsigned char vec_xl (int, vector unsigned char *);
17710 vector unsigned char vec_xl (int, unsigned char *);
17712 void vec_xst (vector short, int, vector short *);
17713 void vec_xst (vector short, int, short *);
17714 void vec_xst (vector unsigned short, int, vector unsigned short *);
17715 void vec_xst (vector unsigned short, int, unsigned short *);
17716 void vec_xst (vector char, int, vector char *);
17717 void vec_xst (vector char, int, char *);
17718 void vec_xst (vector unsigned char, int, vector unsigned char *);
17719 void vec_xst (vector unsigned char, int, unsigned char *);
17721 @node PowerPC Hardware Transactional Memory Built-in Functions
17722 @subsection PowerPC Hardware Transactional Memory Built-in Functions
17723 GCC provides two interfaces for accessing the Hardware Transactional
17724 Memory (HTM) instructions available on some of the PowerPC family
17725 of processors (eg, POWER8). The two interfaces come in a low level
17726 interface, consisting of built-in functions specific to PowerPC and a
17727 higher level interface consisting of inline functions that are common
17728 between PowerPC and S/390.
17730 @subsubsection PowerPC HTM Low Level Built-in Functions
17732 The following low level built-in functions are available with
17733 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
17734 They all generate the machine instruction that is part of the name.
17736 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
17737 the full 4-bit condition register value set by their associated hardware
17738 instruction. The header file @code{htmintrin.h} defines some macros that can
17739 be used to decipher the return value. The @code{__builtin_tbegin} builtin
17740 returns a simple true or false value depending on whether a transaction was
17741 successfully started or not. The arguments of the builtins match exactly the
17742 type and order of the associated hardware instruction's operands, except for
17743 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
17744 Refer to the ISA manual for a description of each instruction's operands.
17747 unsigned int __builtin_tbegin (unsigned int)
17748 unsigned int __builtin_tend (unsigned int)
17750 unsigned int __builtin_tabort (unsigned int)
17751 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
17752 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
17753 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
17754 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
17756 unsigned int __builtin_tcheck (void)
17757 unsigned int __builtin_treclaim (unsigned int)
17758 unsigned int __builtin_trechkpt (void)
17759 unsigned int __builtin_tsr (unsigned int)
17762 In addition to the above HTM built-ins, we have added built-ins for
17763 some common extended mnemonics of the HTM instructions:
17766 unsigned int __builtin_tendall (void)
17767 unsigned int __builtin_tresume (void)
17768 unsigned int __builtin_tsuspend (void)
17771 Note that the semantics of the above HTM builtins are required to mimic
17772 the locking semantics used for critical sections. Builtins that are used
17773 to create a new transaction or restart a suspended transaction must have
17774 lock acquisition like semantics while those builtins that end or suspend a
17775 transaction must have lock release like semantics. Specifically, this must
17776 mimic lock semantics as specified by C++11, for example: Lock acquisition is
17777 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
17778 that returns 0, and lock release is as-if an execution of
17779 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
17780 implicit implementation-defined lock used for all transactions. The HTM
17781 instructions associated with with the builtins inherently provide the
17782 correct acquisition and release hardware barriers required. However,
17783 the compiler must also be prohibited from moving loads and stores across
17784 the builtins in a way that would violate their semantics. This has been
17785 accomplished by adding memory barriers to the associated HTM instructions
17786 (which is a conservative approach to provide acquire and release semantics).
17787 Earlier versions of the compiler did not treat the HTM instructions as
17788 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
17789 be used to determine whether the current compiler treats HTM instructions
17790 as memory barriers or not. This allows the user to explicitly add memory
17791 barriers to their code when using an older version of the compiler.
17793 The following set of built-in functions are available to gain access
17794 to the HTM specific special purpose registers.
17797 unsigned long __builtin_get_texasr (void)
17798 unsigned long __builtin_get_texasru (void)
17799 unsigned long __builtin_get_tfhar (void)
17800 unsigned long __builtin_get_tfiar (void)
17802 void __builtin_set_texasr (unsigned long);
17803 void __builtin_set_texasru (unsigned long);
17804 void __builtin_set_tfhar (unsigned long);
17805 void __builtin_set_tfiar (unsigned long);
17808 Example usage of these low level built-in functions may look like:
17811 #include <htmintrin.h>
17813 int num_retries = 10;
17817 if (__builtin_tbegin (0))
17819 /* Transaction State Initiated. */
17820 if (is_locked (lock))
17821 __builtin_tabort (0);
17822 ... transaction code...
17823 __builtin_tend (0);
17828 /* Transaction State Failed. Use locks if the transaction
17829 failure is "persistent" or we've tried too many times. */
17830 if (num_retries-- <= 0
17831 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
17833 acquire_lock (lock);
17834 ... non transactional fallback path...
17835 release_lock (lock);
17842 One final built-in function has been added that returns the value of
17843 the 2-bit Transaction State field of the Machine Status Register (MSR)
17844 as stored in @code{CR0}.
17847 unsigned long __builtin_ttest (void)
17850 This built-in can be used to determine the current transaction state
17851 using the following code example:
17854 #include <htmintrin.h>
17856 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
17858 if (tx_state == _HTM_TRANSACTIONAL)
17860 /* Code to use in transactional state. */
17862 else if (tx_state == _HTM_NONTRANSACTIONAL)
17864 /* Code to use in non-transactional state. */
17866 else if (tx_state == _HTM_SUSPENDED)
17868 /* Code to use in transaction suspended state. */
17872 @subsubsection PowerPC HTM High Level Inline Functions
17874 The following high level HTM interface is made available by including
17875 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
17876 where CPU is `power8' or later. This interface is common between PowerPC
17877 and S/390, allowing users to write one HTM source implementation that
17878 can be compiled and executed on either system.
17881 long __TM_simple_begin (void)
17882 long __TM_begin (void* const TM_buff)
17883 long __TM_end (void)
17884 void __TM_abort (void)
17885 void __TM_named_abort (unsigned char const code)
17886 void __TM_resume (void)
17887 void __TM_suspend (void)
17889 long __TM_is_user_abort (void* const TM_buff)
17890 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
17891 long __TM_is_illegal (void* const TM_buff)
17892 long __TM_is_footprint_exceeded (void* const TM_buff)
17893 long __TM_nesting_depth (void* const TM_buff)
17894 long __TM_is_nested_too_deep(void* const TM_buff)
17895 long __TM_is_conflict(void* const TM_buff)
17896 long __TM_is_failure_persistent(void* const TM_buff)
17897 long __TM_failure_address(void* const TM_buff)
17898 long long __TM_failure_code(void* const TM_buff)
17901 Using these common set of HTM inline functions, we can create
17902 a more portable version of the HTM example in the previous
17903 section that will work on either PowerPC or S/390:
17906 #include <htmxlintrin.h>
17908 int num_retries = 10;
17909 TM_buff_type TM_buff;
17913 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
17915 /* Transaction State Initiated. */
17916 if (is_locked (lock))
17918 ... transaction code...
17924 /* Transaction State Failed. Use locks if the transaction
17925 failure is "persistent" or we've tried too many times. */
17926 if (num_retries-- <= 0
17927 || __TM_is_failure_persistent (TM_buff))
17929 acquire_lock (lock);
17930 ... non transactional fallback path...
17931 release_lock (lock);
17938 @node RX Built-in Functions
17939 @subsection RX Built-in Functions
17940 GCC supports some of the RX instructions which cannot be expressed in
17941 the C programming language via the use of built-in functions. The
17942 following functions are supported:
17944 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
17945 Generates the @code{brk} machine instruction.
17948 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
17949 Generates the @code{clrpsw} machine instruction to clear the specified
17950 bit in the processor status word.
17953 @deftypefn {Built-in Function} void __builtin_rx_int (int)
17954 Generates the @code{int} machine instruction to generate an interrupt
17955 with the specified value.
17958 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
17959 Generates the @code{machi} machine instruction to add the result of
17960 multiplying the top 16 bits of the two arguments into the
17964 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
17965 Generates the @code{maclo} machine instruction to add the result of
17966 multiplying the bottom 16 bits of the two arguments into the
17970 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
17971 Generates the @code{mulhi} machine instruction to place the result of
17972 multiplying the top 16 bits of the two arguments into the
17976 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
17977 Generates the @code{mullo} machine instruction to place the result of
17978 multiplying the bottom 16 bits of the two arguments into the
17982 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
17983 Generates the @code{mvfachi} machine instruction to read the top
17984 32 bits of the accumulator.
17987 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
17988 Generates the @code{mvfacmi} machine instruction to read the middle
17989 32 bits of the accumulator.
17992 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
17993 Generates the @code{mvfc} machine instruction which reads the control
17994 register specified in its argument and returns its value.
17997 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
17998 Generates the @code{mvtachi} machine instruction to set the top
17999 32 bits of the accumulator.
18002 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
18003 Generates the @code{mvtaclo} machine instruction to set the bottom
18004 32 bits of the accumulator.
18007 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
18008 Generates the @code{mvtc} machine instruction which sets control
18009 register number @code{reg} to @code{val}.
18012 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
18013 Generates the @code{mvtipl} machine instruction set the interrupt
18017 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
18018 Generates the @code{racw} machine instruction to round the accumulator
18019 according to the specified mode.
18022 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
18023 Generates the @code{revw} machine instruction which swaps the bytes in
18024 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
18025 and also bits 16--23 occupy bits 24--31 and vice versa.
18028 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
18029 Generates the @code{rmpa} machine instruction which initiates a
18030 repeated multiply and accumulate sequence.
18033 @deftypefn {Built-in Function} void __builtin_rx_round (float)
18034 Generates the @code{round} machine instruction which returns the
18035 floating-point argument rounded according to the current rounding mode
18036 set in the floating-point status word register.
18039 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
18040 Generates the @code{sat} machine instruction which returns the
18041 saturated value of the argument.
18044 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
18045 Generates the @code{setpsw} machine instruction to set the specified
18046 bit in the processor status word.
18049 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
18050 Generates the @code{wait} machine instruction.
18053 @node S/390 System z Built-in Functions
18054 @subsection S/390 System z Built-in Functions
18055 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
18056 Generates the @code{tbegin} machine instruction starting a
18057 non-constrained hardware transaction. If the parameter is non-NULL the
18058 memory area is used to store the transaction diagnostic buffer and
18059 will be passed as first operand to @code{tbegin}. This buffer can be
18060 defined using the @code{struct __htm_tdb} C struct defined in
18061 @code{htmintrin.h} and must reside on a double-word boundary. The
18062 second tbegin operand is set to @code{0xff0c}. This enables
18063 save/restore of all GPRs and disables aborts for FPR and AR
18064 manipulations inside the transaction body. The condition code set by
18065 the tbegin instruction is returned as integer value. The tbegin
18066 instruction by definition overwrites the content of all FPRs. The
18067 compiler will generate code which saves and restores the FPRs. For
18068 soft-float code it is recommended to used the @code{*_nofloat}
18069 variant. In order to prevent a TDB from being written it is required
18070 to pass a constant zero value as parameter. Passing a zero value
18071 through a variable is not sufficient. Although modifications of
18072 access registers inside the transaction will not trigger an
18073 transaction abort it is not supported to actually modify them. Access
18074 registers do not get saved when entering a transaction. They will have
18075 undefined state when reaching the abort code.
18078 Macros for the possible return codes of tbegin are defined in the
18079 @code{htmintrin.h} header file:
18082 @item _HTM_TBEGIN_STARTED
18083 @code{tbegin} has been executed as part of normal processing. The
18084 transaction body is supposed to be executed.
18085 @item _HTM_TBEGIN_INDETERMINATE
18086 The transaction was aborted due to an indeterminate condition which
18087 might be persistent.
18088 @item _HTM_TBEGIN_TRANSIENT
18089 The transaction aborted due to a transient failure. The transaction
18090 should be re-executed in that case.
18091 @item _HTM_TBEGIN_PERSISTENT
18092 The transaction aborted due to a persistent failure. Re-execution
18093 under same circumstances will not be productive.
18096 @defmac _HTM_FIRST_USER_ABORT_CODE
18097 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
18098 specifies the first abort code which can be used for
18099 @code{__builtin_tabort}. Values below this threshold are reserved for
18103 @deftp {Data type} {struct __htm_tdb}
18104 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
18105 the structure of the transaction diagnostic block as specified in the
18106 Principles of Operation manual chapter 5-91.
18109 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
18110 Same as @code{__builtin_tbegin} but without FPR saves and restores.
18111 Using this variant in code making use of FPRs will leave the FPRs in
18112 undefined state when entering the transaction abort handler code.
18115 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
18116 In addition to @code{__builtin_tbegin} a loop for transient failures
18117 is generated. If tbegin returns a condition code of 2 the transaction
18118 will be retried as often as specified in the second argument. The
18119 perform processor assist instruction is used to tell the CPU about the
18120 number of fails so far.
18123 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
18124 Same as @code{__builtin_tbegin_retry} but without FPR saves and
18125 restores. Using this variant in code making use of FPRs will leave
18126 the FPRs in undefined state when entering the transaction abort
18130 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
18131 Generates the @code{tbeginc} machine instruction starting a constrained
18132 hardware transaction. The second operand is set to @code{0xff08}.
18135 @deftypefn {Built-in Function} int __builtin_tend (void)
18136 Generates the @code{tend} machine instruction finishing a transaction
18137 and making the changes visible to other threads. The condition code
18138 generated by tend is returned as integer value.
18141 @deftypefn {Built-in Function} void __builtin_tabort (int)
18142 Generates the @code{tabort} machine instruction with the specified
18143 abort code. Abort codes from 0 through 255 are reserved and will
18144 result in an error message.
18147 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
18148 Generates the @code{ppa rX,rY,1} machine instruction. Where the
18149 integer parameter is loaded into rX and a value of zero is loaded into
18150 rY. The integer parameter specifies the number of times the
18151 transaction repeatedly aborted.
18154 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
18155 Generates the @code{etnd} machine instruction. The current nesting
18156 depth is returned as integer value. For a nesting depth of 0 the code
18157 is not executed as part of an transaction.
18160 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
18162 Generates the @code{ntstg} machine instruction. The second argument
18163 is written to the first arguments location. The store operation will
18164 not be rolled-back in case of an transaction abort.
18167 @node SH Built-in Functions
18168 @subsection SH Built-in Functions
18169 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
18170 families of processors:
18172 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
18173 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
18174 used by system code that manages threads and execution contexts. The compiler
18175 normally does not generate code that modifies the contents of @samp{GBR} and
18176 thus the value is preserved across function calls. Changing the @samp{GBR}
18177 value in user code must be done with caution, since the compiler might use
18178 @samp{GBR} in order to access thread local variables.
18182 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
18183 Returns the value that is currently set in the @samp{GBR} register.
18184 Memory loads and stores that use the thread pointer as a base address are
18185 turned into @samp{GBR} based displacement loads and stores, if possible.
18193 int get_tcb_value (void)
18195 // Generate @samp{mov.l @@(8,gbr),r0} instruction
18196 return ((my_tcb*)__builtin_thread_pointer ())->c;
18202 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
18203 Returns the value that is currently set in the @samp{FPSCR} register.
18206 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
18207 Sets the @samp{FPSCR} register to the specified value @var{val}, while
18208 preserving the current values of the FR, SZ and PR bits.
18211 @node SPARC VIS Built-in Functions
18212 @subsection SPARC VIS Built-in Functions
18214 GCC supports SIMD operations on the SPARC using both the generic vector
18215 extensions (@pxref{Vector Extensions}) as well as built-in functions for
18216 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
18217 switch, the VIS extension is exposed as the following built-in functions:
18220 typedef int v1si __attribute__ ((vector_size (4)));
18221 typedef int v2si __attribute__ ((vector_size (8)));
18222 typedef short v4hi __attribute__ ((vector_size (8)));
18223 typedef short v2hi __attribute__ ((vector_size (4)));
18224 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
18225 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
18227 void __builtin_vis_write_gsr (int64_t);
18228 int64_t __builtin_vis_read_gsr (void);
18230 void * __builtin_vis_alignaddr (void *, long);
18231 void * __builtin_vis_alignaddrl (void *, long);
18232 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
18233 v2si __builtin_vis_faligndatav2si (v2si, v2si);
18234 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
18235 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
18237 v4hi __builtin_vis_fexpand (v4qi);
18239 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
18240 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
18241 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
18242 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
18243 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
18244 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
18245 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
18247 v4qi __builtin_vis_fpack16 (v4hi);
18248 v8qi __builtin_vis_fpack32 (v2si, v8qi);
18249 v2hi __builtin_vis_fpackfix (v2si);
18250 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
18252 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
18254 long __builtin_vis_edge8 (void *, void *);
18255 long __builtin_vis_edge8l (void *, void *);
18256 long __builtin_vis_edge16 (void *, void *);
18257 long __builtin_vis_edge16l (void *, void *);
18258 long __builtin_vis_edge32 (void *, void *);
18259 long __builtin_vis_edge32l (void *, void *);
18261 long __builtin_vis_fcmple16 (v4hi, v4hi);
18262 long __builtin_vis_fcmple32 (v2si, v2si);
18263 long __builtin_vis_fcmpne16 (v4hi, v4hi);
18264 long __builtin_vis_fcmpne32 (v2si, v2si);
18265 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
18266 long __builtin_vis_fcmpgt32 (v2si, v2si);
18267 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
18268 long __builtin_vis_fcmpeq32 (v2si, v2si);
18270 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
18271 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
18272 v2si __builtin_vis_fpadd32 (v2si, v2si);
18273 v1si __builtin_vis_fpadd32s (v1si, v1si);
18274 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
18275 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
18276 v2si __builtin_vis_fpsub32 (v2si, v2si);
18277 v1si __builtin_vis_fpsub32s (v1si, v1si);
18279 long __builtin_vis_array8 (long, long);
18280 long __builtin_vis_array16 (long, long);
18281 long __builtin_vis_array32 (long, long);
18284 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
18285 functions also become available:
18288 long __builtin_vis_bmask (long, long);
18289 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
18290 v2si __builtin_vis_bshufflev2si (v2si, v2si);
18291 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
18292 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
18294 long __builtin_vis_edge8n (void *, void *);
18295 long __builtin_vis_edge8ln (void *, void *);
18296 long __builtin_vis_edge16n (void *, void *);
18297 long __builtin_vis_edge16ln (void *, void *);
18298 long __builtin_vis_edge32n (void *, void *);
18299 long __builtin_vis_edge32ln (void *, void *);
18302 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
18303 functions also become available:
18306 void __builtin_vis_cmask8 (long);
18307 void __builtin_vis_cmask16 (long);
18308 void __builtin_vis_cmask32 (long);
18310 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
18312 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
18313 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
18314 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
18315 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
18316 v2si __builtin_vis_fsll16 (v2si, v2si);
18317 v2si __builtin_vis_fslas16 (v2si, v2si);
18318 v2si __builtin_vis_fsrl16 (v2si, v2si);
18319 v2si __builtin_vis_fsra16 (v2si, v2si);
18321 long __builtin_vis_pdistn (v8qi, v8qi);
18323 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
18325 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
18326 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
18328 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
18329 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
18330 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
18331 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
18332 v2si __builtin_vis_fpadds32 (v2si, v2si);
18333 v1si __builtin_vis_fpadds32s (v1si, v1si);
18334 v2si __builtin_vis_fpsubs32 (v2si, v2si);
18335 v1si __builtin_vis_fpsubs32s (v1si, v1si);
18337 long __builtin_vis_fucmple8 (v8qi, v8qi);
18338 long __builtin_vis_fucmpne8 (v8qi, v8qi);
18339 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
18340 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
18342 float __builtin_vis_fhadds (float, float);
18343 double __builtin_vis_fhaddd (double, double);
18344 float __builtin_vis_fhsubs (float, float);
18345 double __builtin_vis_fhsubd (double, double);
18346 float __builtin_vis_fnhadds (float, float);
18347 double __builtin_vis_fnhaddd (double, double);
18349 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
18350 int64_t __builtin_vis_xmulx (int64_t, int64_t);
18351 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
18354 When you use the @option{-mvis4} switch, the VIS version 4.0 built-in
18355 functions also become available:
18358 v8qi __builtin_vis_fpadd8 (v8qi, v8qi);
18359 v8qi __builtin_vis_fpadds8 (v8qi, v8qi);
18360 v8qi __builtin_vis_fpaddus8 (v8qi, v8qi);
18361 v4hi __builtin_vis_fpaddus16 (v4hi, v4hi);
18363 v8qi __builtin_vis_fpsub8 (v8qi, v8qi);
18364 v8qi __builtin_vis_fpsubs8 (v8qi, v8qi);
18365 v8qi __builtin_vis_fpsubus8 (v8qi, v8qi);
18366 v4hi __builtin_vis_fpsubus16 (v4hi, v4hi);
18368 long __builtin_vis_fpcmple8 (v8qi, v8qi);
18369 long __builtin_vis_fpcmpgt8 (v8qi, v8qi);
18370 long __builtin_vis_fpcmpule16 (v4hi, v4hi);
18371 long __builtin_vis_fpcmpugt16 (v4hi, v4hi);
18372 long __builtin_vis_fpcmpule32 (v2si, v2si);
18373 long __builtin_vis_fpcmpugt32 (v2si, v2si);
18375 v8qi __builtin_vis_fpmax8 (v8qi, v8qi);
18376 v4hi __builtin_vis_fpmax16 (v4hi, v4hi);
18377 v2si __builtin_vis_fpmax32 (v2si, v2si);
18379 v8qi __builtin_vis_fpmaxu8 (v8qi, v8qi);
18380 v4hi __builtin_vis_fpmaxu16 (v4hi, v4hi);
18381 v2si __builtin_vis_fpmaxu32 (v2si, v2si);
18384 v8qi __builtin_vis_fpmin8 (v8qi, v8qi);
18385 v4hi __builtin_vis_fpmin16 (v4hi, v4hi);
18386 v2si __builtin_vis_fpmin32 (v2si, v2si);
18388 v8qi __builtin_vis_fpminu8 (v8qi, v8qi);
18389 v4hi __builtin_vis_fpminu16 (v4hi, v4hi);
18390 v2si __builtin_vis_fpminu32 (v2si, v2si);
18393 @node SPU Built-in Functions
18394 @subsection SPU Built-in Functions
18396 GCC provides extensions for the SPU processor as described in the
18397 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
18398 found at @uref{http://cell.scei.co.jp/} or
18399 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
18400 implementation differs in several ways.
18405 The optional extension of specifying vector constants in parentheses is
18409 A vector initializer requires no cast if the vector constant is of the
18410 same type as the variable it is initializing.
18413 If @code{signed} or @code{unsigned} is omitted, the signedness of the
18414 vector type is the default signedness of the base type. The default
18415 varies depending on the operating system, so a portable program should
18416 always specify the signedness.
18419 By default, the keyword @code{__vector} is added. The macro
18420 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
18424 GCC allows using a @code{typedef} name as the type specifier for a
18428 For C, overloaded functions are implemented with macros so the following
18432 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
18436 Since @code{spu_add} is a macro, the vector constant in the example
18437 is treated as four separate arguments. Wrap the entire argument in
18438 parentheses for this to work.
18441 The extended version of @code{__builtin_expect} is not supported.
18445 @emph{Note:} Only the interface described in the aforementioned
18446 specification is supported. Internally, GCC uses built-in functions to
18447 implement the required functionality, but these are not supported and
18448 are subject to change without notice.
18450 @node TI C6X Built-in Functions
18451 @subsection TI C6X Built-in Functions
18453 GCC provides intrinsics to access certain instructions of the TI C6X
18454 processors. These intrinsics, listed below, are available after
18455 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
18456 to C6X instructions.
18460 int _sadd (int, int)
18461 int _ssub (int, int)
18462 int _sadd2 (int, int)
18463 int _ssub2 (int, int)
18464 long long _mpy2 (int, int)
18465 long long _smpy2 (int, int)
18466 int _add4 (int, int)
18467 int _sub4 (int, int)
18468 int _saddu4 (int, int)
18470 int _smpy (int, int)
18471 int _smpyh (int, int)
18472 int _smpyhl (int, int)
18473 int _smpylh (int, int)
18475 int _sshl (int, int)
18476 int _subc (int, int)
18478 int _avg2 (int, int)
18479 int _avgu4 (int, int)
18481 int _clrr (int, int)
18482 int _extr (int, int)
18483 int _extru (int, int)
18489 @node TILE-Gx Built-in Functions
18490 @subsection TILE-Gx Built-in Functions
18492 GCC provides intrinsics to access every instruction of the TILE-Gx
18493 processor. The intrinsics are of the form:
18497 unsigned long long __insn_@var{op} (...)
18501 Where @var{op} is the name of the instruction. Refer to the ISA manual
18502 for the complete list of instructions.
18504 GCC also provides intrinsics to directly access the network registers.
18505 The intrinsics are:
18509 unsigned long long __tile_idn0_receive (void)
18510 unsigned long long __tile_idn1_receive (void)
18511 unsigned long long __tile_udn0_receive (void)
18512 unsigned long long __tile_udn1_receive (void)
18513 unsigned long long __tile_udn2_receive (void)
18514 unsigned long long __tile_udn3_receive (void)
18515 void __tile_idn_send (unsigned long long)
18516 void __tile_udn_send (unsigned long long)
18520 The intrinsic @code{void __tile_network_barrier (void)} is used to
18521 guarantee that no network operations before it are reordered with
18524 @node TILEPro Built-in Functions
18525 @subsection TILEPro Built-in Functions
18527 GCC provides intrinsics to access every instruction of the TILEPro
18528 processor. The intrinsics are of the form:
18532 unsigned __insn_@var{op} (...)
18537 where @var{op} is the name of the instruction. Refer to the ISA manual
18538 for the complete list of instructions.
18540 GCC also provides intrinsics to directly access the network registers.
18541 The intrinsics are:
18545 unsigned __tile_idn0_receive (void)
18546 unsigned __tile_idn1_receive (void)
18547 unsigned __tile_sn_receive (void)
18548 unsigned __tile_udn0_receive (void)
18549 unsigned __tile_udn1_receive (void)
18550 unsigned __tile_udn2_receive (void)
18551 unsigned __tile_udn3_receive (void)
18552 void __tile_idn_send (unsigned)
18553 void __tile_sn_send (unsigned)
18554 void __tile_udn_send (unsigned)
18558 The intrinsic @code{void __tile_network_barrier (void)} is used to
18559 guarantee that no network operations before it are reordered with
18562 @node x86 Built-in Functions
18563 @subsection x86 Built-in Functions
18565 These built-in functions are available for the x86-32 and x86-64 family
18566 of computers, depending on the command-line switches used.
18568 If you specify command-line switches such as @option{-msse},
18569 the compiler could use the extended instruction sets even if the built-ins
18570 are not used explicitly in the program. For this reason, applications
18571 that perform run-time CPU detection must compile separate files for each
18572 supported architecture, using the appropriate flags. In particular,
18573 the file containing the CPU detection code should be compiled without
18576 The following machine modes are available for use with MMX built-in functions
18577 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
18578 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
18579 vector of eight 8-bit integers. Some of the built-in functions operate on
18580 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
18582 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
18583 of two 32-bit floating-point values.
18585 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
18586 floating-point values. Some instructions use a vector of four 32-bit
18587 integers, these use @code{V4SI}. Finally, some instructions operate on an
18588 entire vector register, interpreting it as a 128-bit integer, these use mode
18591 The x86-32 and x86-64 family of processors use additional built-in
18592 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
18593 floating point and @code{TC} 128-bit complex floating-point values.
18595 The following floating-point built-in functions are always available. All
18596 of them implement the function that is part of the name.
18599 __float128 __builtin_fabsq (__float128)
18600 __float128 __builtin_copysignq (__float128, __float128)
18603 The following built-in functions are always available.
18606 @item __float128 __builtin_infq (void)
18607 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
18608 @findex __builtin_infq
18610 @item __float128 __builtin_huge_valq (void)
18611 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
18612 @findex __builtin_huge_valq
18614 @item __float128 __builtin_nanq (void)
18615 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
18616 @findex __builtin_nanq
18618 @item __float128 __builtin_nansq (void)
18619 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
18620 @findex __builtin_nansq
18623 The following built-in function is always available.
18626 @item void __builtin_ia32_pause (void)
18627 Generates the @code{pause} machine instruction with a compiler memory
18631 The following built-in functions are always available and can be used to
18632 check the target platform type.
18634 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
18635 This function runs the CPU detection code to check the type of CPU and the
18636 features supported. This built-in function needs to be invoked along with the built-in functions
18637 to check CPU type and features, @code{__builtin_cpu_is} and
18638 @code{__builtin_cpu_supports}, only when used in a function that is
18639 executed before any constructors are called. The CPU detection code is
18640 automatically executed in a very high priority constructor.
18642 For example, this function has to be used in @code{ifunc} resolvers that
18643 check for CPU type using the built-in functions @code{__builtin_cpu_is}
18644 and @code{__builtin_cpu_supports}, or in constructors on targets that
18645 don't support constructor priority.
18648 static void (*resolve_memcpy (void)) (void)
18650 // ifunc resolvers fire before constructors, explicitly call the init
18652 __builtin_cpu_init ();
18653 if (__builtin_cpu_supports ("ssse3"))
18654 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
18656 return default_memcpy;
18659 void *memcpy (void *, const void *, size_t)
18660 __attribute__ ((ifunc ("resolve_memcpy")));
18665 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
18666 This function returns a positive integer if the run-time CPU
18667 is of type @var{cpuname}
18668 and returns @code{0} otherwise. The following CPU names can be detected:
18684 Intel Core i7 Nehalem CPU.
18687 Intel Core i7 Westmere CPU.
18690 Intel Core i7 Sandy Bridge CPU.
18696 AMD Family 10h CPU.
18699 AMD Family 10h Barcelona CPU.
18702 AMD Family 10h Shanghai CPU.
18705 AMD Family 10h Istanbul CPU.
18708 AMD Family 14h CPU.
18711 AMD Family 15h CPU.
18714 AMD Family 15h Bulldozer version 1.
18717 AMD Family 15h Bulldozer version 2.
18720 AMD Family 15h Bulldozer version 3.
18723 AMD Family 15h Bulldozer version 4.
18726 AMD Family 16h CPU.
18729 AMD Family 17h CPU.
18732 Here is an example:
18734 if (__builtin_cpu_is ("corei7"))
18736 do_corei7 (); // Core i7 specific implementation.
18740 do_generic (); // Generic implementation.
18745 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
18746 This function returns a positive integer if the run-time CPU
18747 supports @var{feature}
18748 and returns @code{0} otherwise. The following features can be detected:
18756 POPCNT instruction.
18764 SSSE3 instructions.
18766 SSE4.1 instructions.
18768 SSE4.2 instructions.
18774 AVX512F instructions.
18777 Here is an example:
18779 if (__builtin_cpu_supports ("popcnt"))
18781 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
18785 count = generic_countbits (n); //generic implementation.
18791 The following built-in functions are made available by @option{-mmmx}.
18792 All of them generate the machine instruction that is part of the name.
18795 v8qi __builtin_ia32_paddb (v8qi, v8qi)
18796 v4hi __builtin_ia32_paddw (v4hi, v4hi)
18797 v2si __builtin_ia32_paddd (v2si, v2si)
18798 v8qi __builtin_ia32_psubb (v8qi, v8qi)
18799 v4hi __builtin_ia32_psubw (v4hi, v4hi)
18800 v2si __builtin_ia32_psubd (v2si, v2si)
18801 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
18802 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
18803 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
18804 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
18805 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
18806 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
18807 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
18808 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
18809 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
18810 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
18811 di __builtin_ia32_pand (di, di)
18812 di __builtin_ia32_pandn (di,di)
18813 di __builtin_ia32_por (di, di)
18814 di __builtin_ia32_pxor (di, di)
18815 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
18816 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
18817 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
18818 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
18819 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
18820 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
18821 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
18822 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
18823 v2si __builtin_ia32_punpckhdq (v2si, v2si)
18824 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
18825 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
18826 v2si __builtin_ia32_punpckldq (v2si, v2si)
18827 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
18828 v4hi __builtin_ia32_packssdw (v2si, v2si)
18829 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
18831 v4hi __builtin_ia32_psllw (v4hi, v4hi)
18832 v2si __builtin_ia32_pslld (v2si, v2si)
18833 v1di __builtin_ia32_psllq (v1di, v1di)
18834 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
18835 v2si __builtin_ia32_psrld (v2si, v2si)
18836 v1di __builtin_ia32_psrlq (v1di, v1di)
18837 v4hi __builtin_ia32_psraw (v4hi, v4hi)
18838 v2si __builtin_ia32_psrad (v2si, v2si)
18839 v4hi __builtin_ia32_psllwi (v4hi, int)
18840 v2si __builtin_ia32_pslldi (v2si, int)
18841 v1di __builtin_ia32_psllqi (v1di, int)
18842 v4hi __builtin_ia32_psrlwi (v4hi, int)
18843 v2si __builtin_ia32_psrldi (v2si, int)
18844 v1di __builtin_ia32_psrlqi (v1di, int)
18845 v4hi __builtin_ia32_psrawi (v4hi, int)
18846 v2si __builtin_ia32_psradi (v2si, int)
18850 The following built-in functions are made available either with
18851 @option{-msse}, or with a combination of @option{-m3dnow} and
18852 @option{-march=athlon}. All of them generate the machine
18853 instruction that is part of the name.
18856 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
18857 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
18858 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
18859 v1di __builtin_ia32_psadbw (v8qi, v8qi)
18860 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
18861 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
18862 v8qi __builtin_ia32_pminub (v8qi, v8qi)
18863 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
18864 int __builtin_ia32_pmovmskb (v8qi)
18865 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
18866 void __builtin_ia32_movntq (di *, di)
18867 void __builtin_ia32_sfence (void)
18870 The following built-in functions are available when @option{-msse} is used.
18871 All of them generate the machine instruction that is part of the name.
18874 int __builtin_ia32_comieq (v4sf, v4sf)
18875 int __builtin_ia32_comineq (v4sf, v4sf)
18876 int __builtin_ia32_comilt (v4sf, v4sf)
18877 int __builtin_ia32_comile (v4sf, v4sf)
18878 int __builtin_ia32_comigt (v4sf, v4sf)
18879 int __builtin_ia32_comige (v4sf, v4sf)
18880 int __builtin_ia32_ucomieq (v4sf, v4sf)
18881 int __builtin_ia32_ucomineq (v4sf, v4sf)
18882 int __builtin_ia32_ucomilt (v4sf, v4sf)
18883 int __builtin_ia32_ucomile (v4sf, v4sf)
18884 int __builtin_ia32_ucomigt (v4sf, v4sf)
18885 int __builtin_ia32_ucomige (v4sf, v4sf)
18886 v4sf __builtin_ia32_addps (v4sf, v4sf)
18887 v4sf __builtin_ia32_subps (v4sf, v4sf)
18888 v4sf __builtin_ia32_mulps (v4sf, v4sf)
18889 v4sf __builtin_ia32_divps (v4sf, v4sf)
18890 v4sf __builtin_ia32_addss (v4sf, v4sf)
18891 v4sf __builtin_ia32_subss (v4sf, v4sf)
18892 v4sf __builtin_ia32_mulss (v4sf, v4sf)
18893 v4sf __builtin_ia32_divss (v4sf, v4sf)
18894 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
18895 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
18896 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
18897 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
18898 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
18899 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
18900 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
18901 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
18902 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
18903 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
18904 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
18905 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
18906 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
18907 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
18908 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
18909 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
18910 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
18911 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
18912 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
18913 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
18914 v4sf __builtin_ia32_maxps (v4sf, v4sf)
18915 v4sf __builtin_ia32_maxss (v4sf, v4sf)
18916 v4sf __builtin_ia32_minps (v4sf, v4sf)
18917 v4sf __builtin_ia32_minss (v4sf, v4sf)
18918 v4sf __builtin_ia32_andps (v4sf, v4sf)
18919 v4sf __builtin_ia32_andnps (v4sf, v4sf)
18920 v4sf __builtin_ia32_orps (v4sf, v4sf)
18921 v4sf __builtin_ia32_xorps (v4sf, v4sf)
18922 v4sf __builtin_ia32_movss (v4sf, v4sf)
18923 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
18924 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
18925 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
18926 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
18927 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
18928 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
18929 v2si __builtin_ia32_cvtps2pi (v4sf)
18930 int __builtin_ia32_cvtss2si (v4sf)
18931 v2si __builtin_ia32_cvttps2pi (v4sf)
18932 int __builtin_ia32_cvttss2si (v4sf)
18933 v4sf __builtin_ia32_rcpps (v4sf)
18934 v4sf __builtin_ia32_rsqrtps (v4sf)
18935 v4sf __builtin_ia32_sqrtps (v4sf)
18936 v4sf __builtin_ia32_rcpss (v4sf)
18937 v4sf __builtin_ia32_rsqrtss (v4sf)
18938 v4sf __builtin_ia32_sqrtss (v4sf)
18939 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
18940 void __builtin_ia32_movntps (float *, v4sf)
18941 int __builtin_ia32_movmskps (v4sf)
18944 The following built-in functions are available when @option{-msse} is used.
18947 @item v4sf __builtin_ia32_loadups (float *)
18948 Generates the @code{movups} machine instruction as a load from memory.
18949 @item void __builtin_ia32_storeups (float *, v4sf)
18950 Generates the @code{movups} machine instruction as a store to memory.
18951 @item v4sf __builtin_ia32_loadss (float *)
18952 Generates the @code{movss} machine instruction as a load from memory.
18953 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
18954 Generates the @code{movhps} machine instruction as a load from memory.
18955 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
18956 Generates the @code{movlps} machine instruction as a load from memory
18957 @item void __builtin_ia32_storehps (v2sf *, v4sf)
18958 Generates the @code{movhps} machine instruction as a store to memory.
18959 @item void __builtin_ia32_storelps (v2sf *, v4sf)
18960 Generates the @code{movlps} machine instruction as a store to memory.
18963 The following built-in functions are available when @option{-msse2} is used.
18964 All of them generate the machine instruction that is part of the name.
18967 int __builtin_ia32_comisdeq (v2df, v2df)
18968 int __builtin_ia32_comisdlt (v2df, v2df)
18969 int __builtin_ia32_comisdle (v2df, v2df)
18970 int __builtin_ia32_comisdgt (v2df, v2df)
18971 int __builtin_ia32_comisdge (v2df, v2df)
18972 int __builtin_ia32_comisdneq (v2df, v2df)
18973 int __builtin_ia32_ucomisdeq (v2df, v2df)
18974 int __builtin_ia32_ucomisdlt (v2df, v2df)
18975 int __builtin_ia32_ucomisdle (v2df, v2df)
18976 int __builtin_ia32_ucomisdgt (v2df, v2df)
18977 int __builtin_ia32_ucomisdge (v2df, v2df)
18978 int __builtin_ia32_ucomisdneq (v2df, v2df)
18979 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
18980 v2df __builtin_ia32_cmpltpd (v2df, v2df)
18981 v2df __builtin_ia32_cmplepd (v2df, v2df)
18982 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
18983 v2df __builtin_ia32_cmpgepd (v2df, v2df)
18984 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
18985 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
18986 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
18987 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
18988 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
18989 v2df __builtin_ia32_cmpngepd (v2df, v2df)
18990 v2df __builtin_ia32_cmpordpd (v2df, v2df)
18991 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
18992 v2df __builtin_ia32_cmpltsd (v2df, v2df)
18993 v2df __builtin_ia32_cmplesd (v2df, v2df)
18994 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
18995 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
18996 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
18997 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
18998 v2df __builtin_ia32_cmpordsd (v2df, v2df)
18999 v2di __builtin_ia32_paddq (v2di, v2di)
19000 v2di __builtin_ia32_psubq (v2di, v2di)
19001 v2df __builtin_ia32_addpd (v2df, v2df)
19002 v2df __builtin_ia32_subpd (v2df, v2df)
19003 v2df __builtin_ia32_mulpd (v2df, v2df)
19004 v2df __builtin_ia32_divpd (v2df, v2df)
19005 v2df __builtin_ia32_addsd (v2df, v2df)
19006 v2df __builtin_ia32_subsd (v2df, v2df)
19007 v2df __builtin_ia32_mulsd (v2df, v2df)
19008 v2df __builtin_ia32_divsd (v2df, v2df)
19009 v2df __builtin_ia32_minpd (v2df, v2df)
19010 v2df __builtin_ia32_maxpd (v2df, v2df)
19011 v2df __builtin_ia32_minsd (v2df, v2df)
19012 v2df __builtin_ia32_maxsd (v2df, v2df)
19013 v2df __builtin_ia32_andpd (v2df, v2df)
19014 v2df __builtin_ia32_andnpd (v2df, v2df)
19015 v2df __builtin_ia32_orpd (v2df, v2df)
19016 v2df __builtin_ia32_xorpd (v2df, v2df)
19017 v2df __builtin_ia32_movsd (v2df, v2df)
19018 v2df __builtin_ia32_unpckhpd (v2df, v2df)
19019 v2df __builtin_ia32_unpcklpd (v2df, v2df)
19020 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
19021 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
19022 v4si __builtin_ia32_paddd128 (v4si, v4si)
19023 v2di __builtin_ia32_paddq128 (v2di, v2di)
19024 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
19025 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
19026 v4si __builtin_ia32_psubd128 (v4si, v4si)
19027 v2di __builtin_ia32_psubq128 (v2di, v2di)
19028 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
19029 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
19030 v2di __builtin_ia32_pand128 (v2di, v2di)
19031 v2di __builtin_ia32_pandn128 (v2di, v2di)
19032 v2di __builtin_ia32_por128 (v2di, v2di)
19033 v2di __builtin_ia32_pxor128 (v2di, v2di)
19034 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
19035 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
19036 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
19037 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
19038 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
19039 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
19040 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
19041 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
19042 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
19043 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
19044 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
19045 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
19046 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
19047 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
19048 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
19049 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
19050 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
19051 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
19052 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
19053 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
19054 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
19055 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
19056 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
19057 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
19058 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
19059 v2df __builtin_ia32_loadupd (double *)
19060 void __builtin_ia32_storeupd (double *, v2df)
19061 v2df __builtin_ia32_loadhpd (v2df, double const *)
19062 v2df __builtin_ia32_loadlpd (v2df, double const *)
19063 int __builtin_ia32_movmskpd (v2df)
19064 int __builtin_ia32_pmovmskb128 (v16qi)
19065 void __builtin_ia32_movnti (int *, int)
19066 void __builtin_ia32_movnti64 (long long int *, long long int)
19067 void __builtin_ia32_movntpd (double *, v2df)
19068 void __builtin_ia32_movntdq (v2df *, v2df)
19069 v4si __builtin_ia32_pshufd (v4si, int)
19070 v8hi __builtin_ia32_pshuflw (v8hi, int)
19071 v8hi __builtin_ia32_pshufhw (v8hi, int)
19072 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
19073 v2df __builtin_ia32_sqrtpd (v2df)
19074 v2df __builtin_ia32_sqrtsd (v2df)
19075 v2df __builtin_ia32_shufpd (v2df, v2df, int)
19076 v2df __builtin_ia32_cvtdq2pd (v4si)
19077 v4sf __builtin_ia32_cvtdq2ps (v4si)
19078 v4si __builtin_ia32_cvtpd2dq (v2df)
19079 v2si __builtin_ia32_cvtpd2pi (v2df)
19080 v4sf __builtin_ia32_cvtpd2ps (v2df)
19081 v4si __builtin_ia32_cvttpd2dq (v2df)
19082 v2si __builtin_ia32_cvttpd2pi (v2df)
19083 v2df __builtin_ia32_cvtpi2pd (v2si)
19084 int __builtin_ia32_cvtsd2si (v2df)
19085 int __builtin_ia32_cvttsd2si (v2df)
19086 long long __builtin_ia32_cvtsd2si64 (v2df)
19087 long long __builtin_ia32_cvttsd2si64 (v2df)
19088 v4si __builtin_ia32_cvtps2dq (v4sf)
19089 v2df __builtin_ia32_cvtps2pd (v4sf)
19090 v4si __builtin_ia32_cvttps2dq (v4sf)
19091 v2df __builtin_ia32_cvtsi2sd (v2df, int)
19092 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
19093 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
19094 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
19095 void __builtin_ia32_clflush (const void *)
19096 void __builtin_ia32_lfence (void)
19097 void __builtin_ia32_mfence (void)
19098 v16qi __builtin_ia32_loaddqu (const char *)
19099 void __builtin_ia32_storedqu (char *, v16qi)
19100 v1di __builtin_ia32_pmuludq (v2si, v2si)
19101 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
19102 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
19103 v4si __builtin_ia32_pslld128 (v4si, v4si)
19104 v2di __builtin_ia32_psllq128 (v2di, v2di)
19105 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
19106 v4si __builtin_ia32_psrld128 (v4si, v4si)
19107 v2di __builtin_ia32_psrlq128 (v2di, v2di)
19108 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
19109 v4si __builtin_ia32_psrad128 (v4si, v4si)
19110 v2di __builtin_ia32_pslldqi128 (v2di, int)
19111 v8hi __builtin_ia32_psllwi128 (v8hi, int)
19112 v4si __builtin_ia32_pslldi128 (v4si, int)
19113 v2di __builtin_ia32_psllqi128 (v2di, int)
19114 v2di __builtin_ia32_psrldqi128 (v2di, int)
19115 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
19116 v4si __builtin_ia32_psrldi128 (v4si, int)
19117 v2di __builtin_ia32_psrlqi128 (v2di, int)
19118 v8hi __builtin_ia32_psrawi128 (v8hi, int)
19119 v4si __builtin_ia32_psradi128 (v4si, int)
19120 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
19121 v2di __builtin_ia32_movq128 (v2di)
19124 The following built-in functions are available when @option{-msse3} is used.
19125 All of them generate the machine instruction that is part of the name.
19128 v2df __builtin_ia32_addsubpd (v2df, v2df)
19129 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
19130 v2df __builtin_ia32_haddpd (v2df, v2df)
19131 v4sf __builtin_ia32_haddps (v4sf, v4sf)
19132 v2df __builtin_ia32_hsubpd (v2df, v2df)
19133 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
19134 v16qi __builtin_ia32_lddqu (char const *)
19135 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
19136 v4sf __builtin_ia32_movshdup (v4sf)
19137 v4sf __builtin_ia32_movsldup (v4sf)
19138 void __builtin_ia32_mwait (unsigned int, unsigned int)
19141 The following built-in functions are available when @option{-mssse3} is used.
19142 All of them generate the machine instruction that is part of the name.
19145 v2si __builtin_ia32_phaddd (v2si, v2si)
19146 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
19147 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
19148 v2si __builtin_ia32_phsubd (v2si, v2si)
19149 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
19150 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
19151 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
19152 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
19153 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
19154 v8qi __builtin_ia32_psignb (v8qi, v8qi)
19155 v2si __builtin_ia32_psignd (v2si, v2si)
19156 v4hi __builtin_ia32_psignw (v4hi, v4hi)
19157 v1di __builtin_ia32_palignr (v1di, v1di, int)
19158 v8qi __builtin_ia32_pabsb (v8qi)
19159 v2si __builtin_ia32_pabsd (v2si)
19160 v4hi __builtin_ia32_pabsw (v4hi)
19163 The following built-in functions are available when @option{-mssse3} is used.
19164 All of them generate the machine instruction that is part of the name.
19167 v4si __builtin_ia32_phaddd128 (v4si, v4si)
19168 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
19169 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
19170 v4si __builtin_ia32_phsubd128 (v4si, v4si)
19171 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
19172 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
19173 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
19174 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
19175 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
19176 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
19177 v4si __builtin_ia32_psignd128 (v4si, v4si)
19178 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
19179 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
19180 v16qi __builtin_ia32_pabsb128 (v16qi)
19181 v4si __builtin_ia32_pabsd128 (v4si)
19182 v8hi __builtin_ia32_pabsw128 (v8hi)
19185 The following built-in functions are available when @option{-msse4.1} is
19186 used. All of them generate the machine instruction that is part of the
19190 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
19191 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
19192 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
19193 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
19194 v2df __builtin_ia32_dppd (v2df, v2df, const int)
19195 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
19196 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
19197 v2di __builtin_ia32_movntdqa (v2di *);
19198 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
19199 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
19200 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
19201 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
19202 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
19203 v8hi __builtin_ia32_phminposuw128 (v8hi)
19204 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
19205 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
19206 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
19207 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
19208 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
19209 v4si __builtin_ia32_pminsd128 (v4si, v4si)
19210 v4si __builtin_ia32_pminud128 (v4si, v4si)
19211 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
19212 v4si __builtin_ia32_pmovsxbd128 (v16qi)
19213 v2di __builtin_ia32_pmovsxbq128 (v16qi)
19214 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
19215 v2di __builtin_ia32_pmovsxdq128 (v4si)
19216 v4si __builtin_ia32_pmovsxwd128 (v8hi)
19217 v2di __builtin_ia32_pmovsxwq128 (v8hi)
19218 v4si __builtin_ia32_pmovzxbd128 (v16qi)
19219 v2di __builtin_ia32_pmovzxbq128 (v16qi)
19220 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
19221 v2di __builtin_ia32_pmovzxdq128 (v4si)
19222 v4si __builtin_ia32_pmovzxwd128 (v8hi)
19223 v2di __builtin_ia32_pmovzxwq128 (v8hi)
19224 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
19225 v4si __builtin_ia32_pmulld128 (v4si, v4si)
19226 int __builtin_ia32_ptestc128 (v2di, v2di)
19227 int __builtin_ia32_ptestnzc128 (v2di, v2di)
19228 int __builtin_ia32_ptestz128 (v2di, v2di)
19229 v2df __builtin_ia32_roundpd (v2df, const int)
19230 v4sf __builtin_ia32_roundps (v4sf, const int)
19231 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
19232 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
19235 The following built-in functions are available when @option{-msse4.1} is
19239 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
19240 Generates the @code{insertps} machine instruction.
19241 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
19242 Generates the @code{pextrb} machine instruction.
19243 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
19244 Generates the @code{pinsrb} machine instruction.
19245 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
19246 Generates the @code{pinsrd} machine instruction.
19247 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
19248 Generates the @code{pinsrq} machine instruction in 64bit mode.
19251 The following built-in functions are changed to generate new SSE4.1
19252 instructions when @option{-msse4.1} is used.
19255 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
19256 Generates the @code{extractps} machine instruction.
19257 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
19258 Generates the @code{pextrd} machine instruction.
19259 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
19260 Generates the @code{pextrq} machine instruction in 64bit mode.
19263 The following built-in functions are available when @option{-msse4.2} is
19264 used. All of them generate the machine instruction that is part of the
19268 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
19269 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
19270 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
19271 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
19272 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
19273 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
19274 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
19275 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
19276 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
19277 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
19278 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
19279 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
19280 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
19281 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
19282 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
19285 The following built-in functions are available when @option{-msse4.2} is
19289 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
19290 Generates the @code{crc32b} machine instruction.
19291 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
19292 Generates the @code{crc32w} machine instruction.
19293 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
19294 Generates the @code{crc32l} machine instruction.
19295 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
19296 Generates the @code{crc32q} machine instruction.
19299 The following built-in functions are changed to generate new SSE4.2
19300 instructions when @option{-msse4.2} is used.
19303 @item int __builtin_popcount (unsigned int)
19304 Generates the @code{popcntl} machine instruction.
19305 @item int __builtin_popcountl (unsigned long)
19306 Generates the @code{popcntl} or @code{popcntq} machine instruction,
19307 depending on the size of @code{unsigned long}.
19308 @item int __builtin_popcountll (unsigned long long)
19309 Generates the @code{popcntq} machine instruction.
19312 The following built-in functions are available when @option{-mavx} is
19313 used. All of them generate the machine instruction that is part of the
19317 v4df __builtin_ia32_addpd256 (v4df,v4df)
19318 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
19319 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
19320 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
19321 v4df __builtin_ia32_andnpd256 (v4df,v4df)
19322 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
19323 v4df __builtin_ia32_andpd256 (v4df,v4df)
19324 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
19325 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
19326 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
19327 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
19328 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
19329 v2df __builtin_ia32_cmppd (v2df,v2df,int)
19330 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
19331 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
19332 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
19333 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
19334 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
19335 v4df __builtin_ia32_cvtdq2pd256 (v4si)
19336 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
19337 v4si __builtin_ia32_cvtpd2dq256 (v4df)
19338 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
19339 v8si __builtin_ia32_cvtps2dq256 (v8sf)
19340 v4df __builtin_ia32_cvtps2pd256 (v4sf)
19341 v4si __builtin_ia32_cvttpd2dq256 (v4df)
19342 v8si __builtin_ia32_cvttps2dq256 (v8sf)
19343 v4df __builtin_ia32_divpd256 (v4df,v4df)
19344 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
19345 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
19346 v4df __builtin_ia32_haddpd256 (v4df,v4df)
19347 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
19348 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
19349 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
19350 v32qi __builtin_ia32_lddqu256 (pcchar)
19351 v32qi __builtin_ia32_loaddqu256 (pcchar)
19352 v4df __builtin_ia32_loadupd256 (pcdouble)
19353 v8sf __builtin_ia32_loadups256 (pcfloat)
19354 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
19355 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
19356 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
19357 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
19358 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
19359 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
19360 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
19361 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
19362 v4df __builtin_ia32_maxpd256 (v4df,v4df)
19363 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
19364 v4df __builtin_ia32_minpd256 (v4df,v4df)
19365 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
19366 v4df __builtin_ia32_movddup256 (v4df)
19367 int __builtin_ia32_movmskpd256 (v4df)
19368 int __builtin_ia32_movmskps256 (v8sf)
19369 v8sf __builtin_ia32_movshdup256 (v8sf)
19370 v8sf __builtin_ia32_movsldup256 (v8sf)
19371 v4df __builtin_ia32_mulpd256 (v4df,v4df)
19372 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
19373 v4df __builtin_ia32_orpd256 (v4df,v4df)
19374 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
19375 v2df __builtin_ia32_pd_pd256 (v4df)
19376 v4df __builtin_ia32_pd256_pd (v2df)
19377 v4sf __builtin_ia32_ps_ps256 (v8sf)
19378 v8sf __builtin_ia32_ps256_ps (v4sf)
19379 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
19380 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
19381 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
19382 v8sf __builtin_ia32_rcpps256 (v8sf)
19383 v4df __builtin_ia32_roundpd256 (v4df,int)
19384 v8sf __builtin_ia32_roundps256 (v8sf,int)
19385 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
19386 v8sf __builtin_ia32_rsqrtps256 (v8sf)
19387 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
19388 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
19389 v4si __builtin_ia32_si_si256 (v8si)
19390 v8si __builtin_ia32_si256_si (v4si)
19391 v4df __builtin_ia32_sqrtpd256 (v4df)
19392 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
19393 v8sf __builtin_ia32_sqrtps256 (v8sf)
19394 void __builtin_ia32_storedqu256 (pchar,v32qi)
19395 void __builtin_ia32_storeupd256 (pdouble,v4df)
19396 void __builtin_ia32_storeups256 (pfloat,v8sf)
19397 v4df __builtin_ia32_subpd256 (v4df,v4df)
19398 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
19399 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
19400 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
19401 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
19402 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
19403 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
19404 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
19405 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
19406 v4sf __builtin_ia32_vbroadcastss (pcfloat)
19407 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
19408 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
19409 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
19410 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
19411 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
19412 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
19413 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
19414 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
19415 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
19416 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
19417 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
19418 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
19419 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
19420 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
19421 v2df __builtin_ia32_vpermilpd (v2df,int)
19422 v4df __builtin_ia32_vpermilpd256 (v4df,int)
19423 v4sf __builtin_ia32_vpermilps (v4sf,int)
19424 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
19425 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
19426 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
19427 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
19428 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
19429 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
19430 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
19431 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
19432 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
19433 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
19434 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
19435 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
19436 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
19437 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
19438 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
19439 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
19440 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
19441 void __builtin_ia32_vzeroall (void)
19442 void __builtin_ia32_vzeroupper (void)
19443 v4df __builtin_ia32_xorpd256 (v4df,v4df)
19444 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
19447 The following built-in functions are available when @option{-mavx2} is
19448 used. All of them generate the machine instruction that is part of the
19452 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
19453 v32qi __builtin_ia32_pabsb256 (v32qi)
19454 v16hi __builtin_ia32_pabsw256 (v16hi)
19455 v8si __builtin_ia32_pabsd256 (v8si)
19456 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
19457 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
19458 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
19459 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
19460 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
19461 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
19462 v8si __builtin_ia32_paddd256 (v8si,v8si)
19463 v4di __builtin_ia32_paddq256 (v4di,v4di)
19464 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
19465 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
19466 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
19467 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
19468 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
19469 v4di __builtin_ia32_andsi256 (v4di,v4di)
19470 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
19471 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
19472 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
19473 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
19474 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
19475 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
19476 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
19477 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
19478 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
19479 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
19480 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
19481 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
19482 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
19483 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
19484 v8si __builtin_ia32_phaddd256 (v8si,v8si)
19485 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
19486 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
19487 v8si __builtin_ia32_phsubd256 (v8si,v8si)
19488 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
19489 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
19490 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
19491 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
19492 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
19493 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
19494 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
19495 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
19496 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
19497 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
19498 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
19499 v8si __builtin_ia32_pminsd256 (v8si,v8si)
19500 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
19501 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
19502 v8si __builtin_ia32_pminud256 (v8si,v8si)
19503 int __builtin_ia32_pmovmskb256 (v32qi)
19504 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
19505 v8si __builtin_ia32_pmovsxbd256 (v16qi)
19506 v4di __builtin_ia32_pmovsxbq256 (v16qi)
19507 v8si __builtin_ia32_pmovsxwd256 (v8hi)
19508 v4di __builtin_ia32_pmovsxwq256 (v8hi)
19509 v4di __builtin_ia32_pmovsxdq256 (v4si)
19510 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
19511 v8si __builtin_ia32_pmovzxbd256 (v16qi)
19512 v4di __builtin_ia32_pmovzxbq256 (v16qi)
19513 v8si __builtin_ia32_pmovzxwd256 (v8hi)
19514 v4di __builtin_ia32_pmovzxwq256 (v8hi)
19515 v4di __builtin_ia32_pmovzxdq256 (v4si)
19516 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
19517 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
19518 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
19519 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
19520 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
19521 v8si __builtin_ia32_pmulld256 (v8si,v8si)
19522 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
19523 v4di __builtin_ia32_por256 (v4di,v4di)
19524 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
19525 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
19526 v8si __builtin_ia32_pshufd256 (v8si,int)
19527 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
19528 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
19529 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
19530 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
19531 v8si __builtin_ia32_psignd256 (v8si,v8si)
19532 v4di __builtin_ia32_pslldqi256 (v4di,int)
19533 v16hi __builtin_ia32_psllwi256 (16hi,int)
19534 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
19535 v8si __builtin_ia32_pslldi256 (v8si,int)
19536 v8si __builtin_ia32_pslld256(v8si,v4si)
19537 v4di __builtin_ia32_psllqi256 (v4di,int)
19538 v4di __builtin_ia32_psllq256(v4di,v2di)
19539 v16hi __builtin_ia32_psrawi256 (v16hi,int)
19540 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
19541 v8si __builtin_ia32_psradi256 (v8si,int)
19542 v8si __builtin_ia32_psrad256 (v8si,v4si)
19543 v4di __builtin_ia32_psrldqi256 (v4di, int)
19544 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
19545 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
19546 v8si __builtin_ia32_psrldi256 (v8si,int)
19547 v8si __builtin_ia32_psrld256 (v8si,v4si)
19548 v4di __builtin_ia32_psrlqi256 (v4di,int)
19549 v4di __builtin_ia32_psrlq256(v4di,v2di)
19550 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
19551 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
19552 v8si __builtin_ia32_psubd256 (v8si,v8si)
19553 v4di __builtin_ia32_psubq256 (v4di,v4di)
19554 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
19555 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
19556 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
19557 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
19558 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
19559 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
19560 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
19561 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
19562 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
19563 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
19564 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
19565 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
19566 v4di __builtin_ia32_pxor256 (v4di,v4di)
19567 v4di __builtin_ia32_movntdqa256 (pv4di)
19568 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
19569 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
19570 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
19571 v4di __builtin_ia32_vbroadcastsi256 (v2di)
19572 v4si __builtin_ia32_pblendd128 (v4si,v4si)
19573 v8si __builtin_ia32_pblendd256 (v8si,v8si)
19574 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
19575 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
19576 v8si __builtin_ia32_pbroadcastd256 (v4si)
19577 v4di __builtin_ia32_pbroadcastq256 (v2di)
19578 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
19579 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
19580 v4si __builtin_ia32_pbroadcastd128 (v4si)
19581 v2di __builtin_ia32_pbroadcastq128 (v2di)
19582 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
19583 v4df __builtin_ia32_permdf256 (v4df,int)
19584 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
19585 v4di __builtin_ia32_permdi256 (v4di,int)
19586 v4di __builtin_ia32_permti256 (v4di,v4di,int)
19587 v4di __builtin_ia32_extract128i256 (v4di,int)
19588 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
19589 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
19590 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
19591 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
19592 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
19593 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
19594 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
19595 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
19596 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
19597 v8si __builtin_ia32_psllv8si (v8si,v8si)
19598 v4si __builtin_ia32_psllv4si (v4si,v4si)
19599 v4di __builtin_ia32_psllv4di (v4di,v4di)
19600 v2di __builtin_ia32_psllv2di (v2di,v2di)
19601 v8si __builtin_ia32_psrav8si (v8si,v8si)
19602 v4si __builtin_ia32_psrav4si (v4si,v4si)
19603 v8si __builtin_ia32_psrlv8si (v8si,v8si)
19604 v4si __builtin_ia32_psrlv4si (v4si,v4si)
19605 v4di __builtin_ia32_psrlv4di (v4di,v4di)
19606 v2di __builtin_ia32_psrlv2di (v2di,v2di)
19607 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
19608 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
19609 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
19610 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
19611 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
19612 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
19613 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
19614 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
19615 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
19616 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
19617 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
19618 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
19619 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
19620 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
19621 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
19622 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
19625 The following built-in functions are available when @option{-maes} is
19626 used. All of them generate the machine instruction that is part of the
19630 v2di __builtin_ia32_aesenc128 (v2di, v2di)
19631 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
19632 v2di __builtin_ia32_aesdec128 (v2di, v2di)
19633 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
19634 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
19635 v2di __builtin_ia32_aesimc128 (v2di)
19638 The following built-in function is available when @option{-mpclmul} is
19642 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
19643 Generates the @code{pclmulqdq} machine instruction.
19646 The following built-in function is available when @option{-mfsgsbase} is
19647 used. All of them generate the machine instruction that is part of the
19651 unsigned int __builtin_ia32_rdfsbase32 (void)
19652 unsigned long long __builtin_ia32_rdfsbase64 (void)
19653 unsigned int __builtin_ia32_rdgsbase32 (void)
19654 unsigned long long __builtin_ia32_rdgsbase64 (void)
19655 void _writefsbase_u32 (unsigned int)
19656 void _writefsbase_u64 (unsigned long long)
19657 void _writegsbase_u32 (unsigned int)
19658 void _writegsbase_u64 (unsigned long long)
19661 The following built-in function is available when @option{-mrdrnd} is
19662 used. All of them generate the machine instruction that is part of the
19666 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
19667 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
19668 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
19671 The following built-in functions are available when @option{-msse4a} is used.
19672 All of them generate the machine instruction that is part of the name.
19675 void __builtin_ia32_movntsd (double *, v2df)
19676 void __builtin_ia32_movntss (float *, v4sf)
19677 v2di __builtin_ia32_extrq (v2di, v16qi)
19678 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
19679 v2di __builtin_ia32_insertq (v2di, v2di)
19680 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
19683 The following built-in functions are available when @option{-mxop} is used.
19685 v2df __builtin_ia32_vfrczpd (v2df)
19686 v4sf __builtin_ia32_vfrczps (v4sf)
19687 v2df __builtin_ia32_vfrczsd (v2df)
19688 v4sf __builtin_ia32_vfrczss (v4sf)
19689 v4df __builtin_ia32_vfrczpd256 (v4df)
19690 v8sf __builtin_ia32_vfrczps256 (v8sf)
19691 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
19692 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
19693 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
19694 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
19695 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
19696 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
19697 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
19698 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
19699 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
19700 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
19701 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
19702 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
19703 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
19704 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
19705 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
19706 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
19707 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
19708 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
19709 v4si __builtin_ia32_vpcomequd (v4si, v4si)
19710 v2di __builtin_ia32_vpcomequq (v2di, v2di)
19711 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
19712 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
19713 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
19714 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
19715 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
19716 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
19717 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
19718 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
19719 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
19720 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
19721 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
19722 v4si __builtin_ia32_vpcomged (v4si, v4si)
19723 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
19724 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
19725 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
19726 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
19727 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
19728 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
19729 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
19730 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
19731 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
19732 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
19733 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
19734 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
19735 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
19736 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
19737 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
19738 v4si __builtin_ia32_vpcomled (v4si, v4si)
19739 v2di __builtin_ia32_vpcomleq (v2di, v2di)
19740 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
19741 v4si __builtin_ia32_vpcomleud (v4si, v4si)
19742 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
19743 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
19744 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
19745 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
19746 v4si __builtin_ia32_vpcomltd (v4si, v4si)
19747 v2di __builtin_ia32_vpcomltq (v2di, v2di)
19748 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
19749 v4si __builtin_ia32_vpcomltud (v4si, v4si)
19750 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
19751 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
19752 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
19753 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
19754 v4si __builtin_ia32_vpcomned (v4si, v4si)
19755 v2di __builtin_ia32_vpcomneq (v2di, v2di)
19756 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
19757 v4si __builtin_ia32_vpcomneud (v4si, v4si)
19758 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
19759 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
19760 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
19761 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
19762 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
19763 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
19764 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
19765 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
19766 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
19767 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
19768 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
19769 v4si __builtin_ia32_vphaddbd (v16qi)
19770 v2di __builtin_ia32_vphaddbq (v16qi)
19771 v8hi __builtin_ia32_vphaddbw (v16qi)
19772 v2di __builtin_ia32_vphadddq (v4si)
19773 v4si __builtin_ia32_vphaddubd (v16qi)
19774 v2di __builtin_ia32_vphaddubq (v16qi)
19775 v8hi __builtin_ia32_vphaddubw (v16qi)
19776 v2di __builtin_ia32_vphaddudq (v4si)
19777 v4si __builtin_ia32_vphadduwd (v8hi)
19778 v2di __builtin_ia32_vphadduwq (v8hi)
19779 v4si __builtin_ia32_vphaddwd (v8hi)
19780 v2di __builtin_ia32_vphaddwq (v8hi)
19781 v8hi __builtin_ia32_vphsubbw (v16qi)
19782 v2di __builtin_ia32_vphsubdq (v4si)
19783 v4si __builtin_ia32_vphsubwd (v8hi)
19784 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
19785 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
19786 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
19787 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
19788 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
19789 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
19790 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
19791 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
19792 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
19793 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
19794 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
19795 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
19796 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
19797 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
19798 v4si __builtin_ia32_vprotd (v4si, v4si)
19799 v2di __builtin_ia32_vprotq (v2di, v2di)
19800 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
19801 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
19802 v4si __builtin_ia32_vpshad (v4si, v4si)
19803 v2di __builtin_ia32_vpshaq (v2di, v2di)
19804 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
19805 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
19806 v4si __builtin_ia32_vpshld (v4si, v4si)
19807 v2di __builtin_ia32_vpshlq (v2di, v2di)
19808 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
19811 The following built-in functions are available when @option{-mfma4} is used.
19812 All of them generate the machine instruction that is part of the name.
19815 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
19816 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
19817 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
19818 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
19819 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
19820 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
19821 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
19822 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
19823 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
19824 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
19825 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
19826 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
19827 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
19828 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
19829 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
19830 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
19831 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
19832 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
19833 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
19834 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
19835 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
19836 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
19837 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
19838 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
19839 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
19840 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
19841 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
19842 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
19843 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
19844 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
19845 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
19846 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
19850 The following built-in functions are available when @option{-mlwp} is used.
19853 void __builtin_ia32_llwpcb16 (void *);
19854 void __builtin_ia32_llwpcb32 (void *);
19855 void __builtin_ia32_llwpcb64 (void *);
19856 void * __builtin_ia32_llwpcb16 (void);
19857 void * __builtin_ia32_llwpcb32 (void);
19858 void * __builtin_ia32_llwpcb64 (void);
19859 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
19860 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
19861 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
19862 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
19863 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
19864 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
19867 The following built-in functions are available when @option{-mbmi} is used.
19868 All of them generate the machine instruction that is part of the name.
19870 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
19871 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
19874 The following built-in functions are available when @option{-mbmi2} is used.
19875 All of them generate the machine instruction that is part of the name.
19877 unsigned int _bzhi_u32 (unsigned int, unsigned int)
19878 unsigned int _pdep_u32 (unsigned int, unsigned int)
19879 unsigned int _pext_u32 (unsigned int, unsigned int)
19880 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
19881 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
19882 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
19885 The following built-in functions are available when @option{-mlzcnt} is used.
19886 All of them generate the machine instruction that is part of the name.
19888 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
19889 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
19890 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
19893 The following built-in functions are available when @option{-mfxsr} is used.
19894 All of them generate the machine instruction that is part of the name.
19896 void __builtin_ia32_fxsave (void *)
19897 void __builtin_ia32_fxrstor (void *)
19898 void __builtin_ia32_fxsave64 (void *)
19899 void __builtin_ia32_fxrstor64 (void *)
19902 The following built-in functions are available when @option{-mxsave} is used.
19903 All of them generate the machine instruction that is part of the name.
19905 void __builtin_ia32_xsave (void *, long long)
19906 void __builtin_ia32_xrstor (void *, long long)
19907 void __builtin_ia32_xsave64 (void *, long long)
19908 void __builtin_ia32_xrstor64 (void *, long long)
19911 The following built-in functions are available when @option{-mxsaveopt} is used.
19912 All of them generate the machine instruction that is part of the name.
19914 void __builtin_ia32_xsaveopt (void *, long long)
19915 void __builtin_ia32_xsaveopt64 (void *, long long)
19918 The following built-in functions are available when @option{-mtbm} is used.
19919 Both of them generate the immediate form of the bextr machine instruction.
19921 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
19922 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
19926 The following built-in functions are available when @option{-m3dnow} is used.
19927 All of them generate the machine instruction that is part of the name.
19930 void __builtin_ia32_femms (void)
19931 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
19932 v2si __builtin_ia32_pf2id (v2sf)
19933 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
19934 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
19935 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
19936 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
19937 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
19938 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
19939 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
19940 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
19941 v2sf __builtin_ia32_pfrcp (v2sf)
19942 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
19943 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
19944 v2sf __builtin_ia32_pfrsqrt (v2sf)
19945 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
19946 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
19947 v2sf __builtin_ia32_pi2fd (v2si)
19948 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
19951 The following built-in functions are available when both @option{-m3dnow}
19952 and @option{-march=athlon} are used. All of them generate the machine
19953 instruction that is part of the name.
19956 v2si __builtin_ia32_pf2iw (v2sf)
19957 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
19958 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
19959 v2sf __builtin_ia32_pi2fw (v2si)
19960 v2sf __builtin_ia32_pswapdsf (v2sf)
19961 v2si __builtin_ia32_pswapdsi (v2si)
19964 The following built-in functions are available when @option{-mrtm} is used
19965 They are used for restricted transactional memory. These are the internal
19966 low level functions. Normally the functions in
19967 @ref{x86 transactional memory intrinsics} should be used instead.
19970 int __builtin_ia32_xbegin ()
19971 void __builtin_ia32_xend ()
19972 void __builtin_ia32_xabort (status)
19973 int __builtin_ia32_xtest ()
19976 The following built-in functions are available when @option{-mmwaitx} is used.
19977 All of them generate the machine instruction that is part of the name.
19979 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
19980 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
19983 The following built-in functions are available when @option{-mclzero} is used.
19984 All of them generate the machine instruction that is part of the name.
19986 void __builtin_i32_clzero (void *)
19989 The following built-in functions are available when @option{-mpku} is used.
19990 They generate reads and writes to PKRU.
19992 void __builtin_ia32_wrpkru (unsigned int)
19993 unsigned int __builtin_ia32_rdpkru ()
19996 @node x86 transactional memory intrinsics
19997 @subsection x86 Transactional Memory Intrinsics
19999 These hardware transactional memory intrinsics for x86 allow you to use
20000 memory transactions with RTM (Restricted Transactional Memory).
20001 This support is enabled with the @option{-mrtm} option.
20002 For using HLE (Hardware Lock Elision) see
20003 @ref{x86 specific memory model extensions for transactional memory} instead.
20005 A memory transaction commits all changes to memory in an atomic way,
20006 as visible to other threads. If the transaction fails it is rolled back
20007 and all side effects discarded.
20009 Generally there is no guarantee that a memory transaction ever succeeds
20010 and suitable fallback code always needs to be supplied.
20012 @deftypefn {RTM Function} {unsigned} _xbegin ()
20013 Start a RTM (Restricted Transactional Memory) transaction.
20014 Returns @code{_XBEGIN_STARTED} when the transaction
20015 started successfully (note this is not 0, so the constant has to be
20016 explicitly tested).
20018 If the transaction aborts, all side-effects
20019 are undone and an abort code encoded as a bit mask is returned.
20020 The following macros are defined:
20023 @item _XABORT_EXPLICIT
20024 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
20025 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
20026 @item _XABORT_RETRY
20027 Transaction retry is possible.
20028 @item _XABORT_CONFLICT
20029 Transaction abort due to a memory conflict with another thread.
20030 @item _XABORT_CAPACITY
20031 Transaction abort due to the transaction using too much memory.
20032 @item _XABORT_DEBUG
20033 Transaction abort due to a debug trap.
20034 @item _XABORT_NESTED
20035 Transaction abort in an inner nested transaction.
20038 There is no guarantee
20039 any transaction ever succeeds, so there always needs to be a valid
20043 @deftypefn {RTM Function} {void} _xend ()
20044 Commit the current transaction. When no transaction is active this faults.
20045 All memory side-effects of the transaction become visible
20046 to other threads in an atomic manner.
20049 @deftypefn {RTM Function} {int} _xtest ()
20050 Return a nonzero value if a transaction is currently active, otherwise 0.
20053 @deftypefn {RTM Function} {void} _xabort (status)
20054 Abort the current transaction. When no transaction is active this is a no-op.
20055 The @var{status} is an 8-bit constant; its value is encoded in the return
20056 value from @code{_xbegin}.
20059 Here is an example showing handling for @code{_XABORT_RETRY}
20060 and a fallback path for other failures:
20063 #include <immintrin.h>
20065 int n_tries, max_tries;
20066 unsigned status = _XABORT_EXPLICIT;
20069 for (n_tries = 0; n_tries < max_tries; n_tries++)
20071 status = _xbegin ();
20072 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
20075 if (status == _XBEGIN_STARTED)
20077 ... transaction code...
20082 ... non-transactional fallback path...
20087 Note that, in most cases, the transactional and non-transactional code
20088 must synchronize together to ensure consistency.
20090 @node Target Format Checks
20091 @section Format Checks Specific to Particular Target Machines
20093 For some target machines, GCC supports additional options to the
20095 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
20098 * Solaris Format Checks::
20099 * Darwin Format Checks::
20102 @node Solaris Format Checks
20103 @subsection Solaris Format Checks
20105 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
20106 check. @code{cmn_err} accepts a subset of the standard @code{printf}
20107 conversions, and the two-argument @code{%b} conversion for displaying
20108 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
20110 @node Darwin Format Checks
20111 @subsection Darwin Format Checks
20113 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
20114 attribute context. Declarations made with such attribution are parsed for correct syntax
20115 and format argument types. However, parsing of the format string itself is currently undefined
20116 and is not carried out by this version of the compiler.
20118 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
20119 also be used as format arguments. Note that the relevant headers are only likely to be
20120 available on Darwin (OSX) installations. On such installations, the XCode and system
20121 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
20122 associated functions.
20125 @section Pragmas Accepted by GCC
20127 @cindex @code{#pragma}
20129 GCC supports several types of pragmas, primarily in order to compile
20130 code originally written for other compilers. Note that in general
20131 we do not recommend the use of pragmas; @xref{Function Attributes},
20132 for further explanation.
20135 * AArch64 Pragmas::
20139 * RS/6000 and PowerPC Pragmas::
20142 * Solaris Pragmas::
20143 * Symbol-Renaming Pragmas::
20144 * Structure-Layout Pragmas::
20146 * Diagnostic Pragmas::
20147 * Visibility Pragmas::
20148 * Push/Pop Macro Pragmas::
20149 * Function Specific Option Pragmas::
20150 * Loop-Specific Pragmas::
20153 @node AArch64 Pragmas
20154 @subsection AArch64 Pragmas
20156 The pragmas defined by the AArch64 target correspond to the AArch64
20157 target function attributes. They can be specified as below:
20159 #pragma GCC target("string")
20162 where @code{@var{string}} can be any string accepted as an AArch64 target
20163 attribute. @xref{AArch64 Function Attributes}, for more details
20164 on the permissible values of @code{string}.
20167 @subsection ARM Pragmas
20169 The ARM target defines pragmas for controlling the default addition of
20170 @code{long_call} and @code{short_call} attributes to functions.
20171 @xref{Function Attributes}, for information about the effects of these
20176 @cindex pragma, long_calls
20177 Set all subsequent functions to have the @code{long_call} attribute.
20179 @item no_long_calls
20180 @cindex pragma, no_long_calls
20181 Set all subsequent functions to have the @code{short_call} attribute.
20183 @item long_calls_off
20184 @cindex pragma, long_calls_off
20185 Do not affect the @code{long_call} or @code{short_call} attributes of
20186 subsequent functions.
20190 @subsection M32C Pragmas
20193 @item GCC memregs @var{number}
20194 @cindex pragma, memregs
20195 Overrides the command-line option @code{-memregs=} for the current
20196 file. Use with care! This pragma must be before any function in the
20197 file, and mixing different memregs values in different objects may
20198 make them incompatible. This pragma is useful when a
20199 performance-critical function uses a memreg for temporary values,
20200 as it may allow you to reduce the number of memregs used.
20202 @item ADDRESS @var{name} @var{address}
20203 @cindex pragma, address
20204 For any declared symbols matching @var{name}, this does three things
20205 to that symbol: it forces the symbol to be located at the given
20206 address (a number), it forces the symbol to be volatile, and it
20207 changes the symbol's scope to be static. This pragma exists for
20208 compatibility with other compilers, but note that the common
20209 @code{1234H} numeric syntax is not supported (use @code{0x1234}
20213 #pragma ADDRESS port3 0x103
20220 @subsection MeP Pragmas
20224 @item custom io_volatile (on|off)
20225 @cindex pragma, custom io_volatile
20226 Overrides the command-line option @code{-mio-volatile} for the current
20227 file. Note that for compatibility with future GCC releases, this
20228 option should only be used once before any @code{io} variables in each
20231 @item GCC coprocessor available @var{registers}
20232 @cindex pragma, coprocessor available
20233 Specifies which coprocessor registers are available to the register
20234 allocator. @var{registers} may be a single register, register range
20235 separated by ellipses, or comma-separated list of those. Example:
20238 #pragma GCC coprocessor available $c0...$c10, $c28
20241 @item GCC coprocessor call_saved @var{registers}
20242 @cindex pragma, coprocessor call_saved
20243 Specifies which coprocessor registers are to be saved and restored by
20244 any function using them. @var{registers} may be a single register,
20245 register range separated by ellipses, or comma-separated list of
20249 #pragma GCC coprocessor call_saved $c4...$c6, $c31
20252 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
20253 @cindex pragma, coprocessor subclass
20254 Creates and defines a register class. These register classes can be
20255 used by inline @code{asm} constructs. @var{registers} may be a single
20256 register, register range separated by ellipses, or comma-separated
20257 list of those. Example:
20260 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
20262 asm ("cpfoo %0" : "=B" (x));
20265 @item GCC disinterrupt @var{name} , @var{name} @dots{}
20266 @cindex pragma, disinterrupt
20267 For the named functions, the compiler adds code to disable interrupts
20268 for the duration of those functions. If any functions so named
20269 are not encountered in the source, a warning is emitted that the pragma is
20270 not used. Examples:
20273 #pragma disinterrupt foo
20274 #pragma disinterrupt bar, grill
20275 int foo () @{ @dots{} @}
20278 @item GCC call @var{name} , @var{name} @dots{}
20279 @cindex pragma, call
20280 For the named functions, the compiler always uses a register-indirect
20281 call model when calling the named functions. Examples:
20290 @node RS/6000 and PowerPC Pragmas
20291 @subsection RS/6000 and PowerPC Pragmas
20293 The RS/6000 and PowerPC targets define one pragma for controlling
20294 whether or not the @code{longcall} attribute is added to function
20295 declarations by default. This pragma overrides the @option{-mlongcall}
20296 option, but not the @code{longcall} and @code{shortcall} attributes.
20297 @xref{RS/6000 and PowerPC Options}, for more information about when long
20298 calls are and are not necessary.
20302 @cindex pragma, longcall
20303 Apply the @code{longcall} attribute to all subsequent function
20307 Do not apply the @code{longcall} attribute to subsequent function
20311 @c Describe h8300 pragmas here.
20312 @c Describe sh pragmas here.
20313 @c Describe v850 pragmas here.
20315 @node S/390 Pragmas
20316 @subsection S/390 Pragmas
20318 The pragmas defined by the S/390 target correspond to the S/390
20319 target function attributes and some the additional options:
20326 Note that options of the pragma, unlike options of the target
20327 attribute, do change the value of preprocessor macros like
20328 @code{__VEC__}. They can be specified as below:
20331 #pragma GCC target("string[,string]...")
20332 #pragma GCC target("string"[,"string"]...)
20335 @node Darwin Pragmas
20336 @subsection Darwin Pragmas
20338 The following pragmas are available for all architectures running the
20339 Darwin operating system. These are useful for compatibility with other
20343 @item mark @var{tokens}@dots{}
20344 @cindex pragma, mark
20345 This pragma is accepted, but has no effect.
20347 @item options align=@var{alignment}
20348 @cindex pragma, options align
20349 This pragma sets the alignment of fields in structures. The values of
20350 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
20351 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
20352 properly; to restore the previous setting, use @code{reset} for the
20355 @item segment @var{tokens}@dots{}
20356 @cindex pragma, segment
20357 This pragma is accepted, but has no effect.
20359 @item unused (@var{var} [, @var{var}]@dots{})
20360 @cindex pragma, unused
20361 This pragma declares variables to be possibly unused. GCC does not
20362 produce warnings for the listed variables. The effect is similar to
20363 that of the @code{unused} attribute, except that this pragma may appear
20364 anywhere within the variables' scopes.
20367 @node Solaris Pragmas
20368 @subsection Solaris Pragmas
20370 The Solaris target supports @code{#pragma redefine_extname}
20371 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
20372 @code{#pragma} directives for compatibility with the system compiler.
20375 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
20376 @cindex pragma, align
20378 Increase the minimum alignment of each @var{variable} to @var{alignment}.
20379 This is the same as GCC's @code{aligned} attribute @pxref{Variable
20380 Attributes}). Macro expansion occurs on the arguments to this pragma
20381 when compiling C and Objective-C@. It does not currently occur when
20382 compiling C++, but this is a bug which may be fixed in a future
20385 @item fini (@var{function} [, @var{function}]...)
20386 @cindex pragma, fini
20388 This pragma causes each listed @var{function} to be called after
20389 main, or during shared module unloading, by adding a call to the
20390 @code{.fini} section.
20392 @item init (@var{function} [, @var{function}]...)
20393 @cindex pragma, init
20395 This pragma causes each listed @var{function} to be called during
20396 initialization (before @code{main}) or during shared module loading, by
20397 adding a call to the @code{.init} section.
20401 @node Symbol-Renaming Pragmas
20402 @subsection Symbol-Renaming Pragmas
20404 GCC supports a @code{#pragma} directive that changes the name used in
20405 assembly for a given declaration. While this pragma is supported on all
20406 platforms, it is intended primarily to provide compatibility with the
20407 Solaris system headers. This effect can also be achieved using the asm
20408 labels extension (@pxref{Asm Labels}).
20411 @item redefine_extname @var{oldname} @var{newname}
20412 @cindex pragma, redefine_extname
20414 This pragma gives the C function @var{oldname} the assembly symbol
20415 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
20416 is defined if this pragma is available (currently on all platforms).
20419 This pragma and the asm labels extension interact in a complicated
20420 manner. Here are some corner cases you may want to be aware of:
20423 @item This pragma silently applies only to declarations with external
20424 linkage. Asm labels do not have this restriction.
20426 @item In C++, this pragma silently applies only to declarations with
20427 ``C'' linkage. Again, asm labels do not have this restriction.
20429 @item If either of the ways of changing the assembly name of a
20430 declaration are applied to a declaration whose assembly name has
20431 already been determined (either by a previous use of one of these
20432 features, or because the compiler needed the assembly name in order to
20433 generate code), and the new name is different, a warning issues and
20434 the name does not change.
20436 @item The @var{oldname} used by @code{#pragma redefine_extname} is
20437 always the C-language name.
20440 @node Structure-Layout Pragmas
20441 @subsection Structure-Layout Pragmas
20443 For compatibility with Microsoft Windows compilers, GCC supports a
20444 set of @code{#pragma} directives that change the maximum alignment of
20445 members of structures (other than zero-width bit-fields), unions, and
20446 classes subsequently defined. The @var{n} value below always is required
20447 to be a small power of two and specifies the new alignment in bytes.
20450 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
20451 @item @code{#pragma pack()} sets the alignment to the one that was in
20452 effect when compilation started (see also command-line option
20453 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
20454 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
20455 setting on an internal stack and then optionally sets the new alignment.
20456 @item @code{#pragma pack(pop)} restores the alignment setting to the one
20457 saved at the top of the internal stack (and removes that stack entry).
20458 Note that @code{#pragma pack([@var{n}])} does not influence this internal
20459 stack; thus it is possible to have @code{#pragma pack(push)} followed by
20460 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
20461 @code{#pragma pack(pop)}.
20464 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
20465 directive which lays out structures and unions subsequently defined as the
20466 documented @code{__attribute__ ((ms_struct))}.
20469 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
20470 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
20471 @item @code{#pragma ms_struct reset} goes back to the default layout.
20474 Most targets also support the @code{#pragma scalar_storage_order} directive
20475 which lays out structures and unions subsequently defined as the documented
20476 @code{__attribute__ ((scalar_storage_order))}.
20479 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
20480 of the scalar fields to big-endian.
20481 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
20482 of the scalar fields to little-endian.
20483 @item @code{#pragma scalar_storage_order default} goes back to the endianness
20484 that was in effect when compilation started (see also command-line option
20485 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
20489 @subsection Weak Pragmas
20491 For compatibility with SVR4, GCC supports a set of @code{#pragma}
20492 directives for declaring symbols to be weak, and defining weak
20496 @item #pragma weak @var{symbol}
20497 @cindex pragma, weak
20498 This pragma declares @var{symbol} to be weak, as if the declaration
20499 had the attribute of the same name. The pragma may appear before
20500 or after the declaration of @var{symbol}. It is not an error for
20501 @var{symbol} to never be defined at all.
20503 @item #pragma weak @var{symbol1} = @var{symbol2}
20504 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
20505 It is an error if @var{symbol2} is not defined in the current
20509 @node Diagnostic Pragmas
20510 @subsection Diagnostic Pragmas
20512 GCC allows the user to selectively enable or disable certain types of
20513 diagnostics, and change the kind of the diagnostic. For example, a
20514 project's policy might require that all sources compile with
20515 @option{-Werror} but certain files might have exceptions allowing
20516 specific types of warnings. Or, a project might selectively enable
20517 diagnostics and treat them as errors depending on which preprocessor
20518 macros are defined.
20521 @item #pragma GCC diagnostic @var{kind} @var{option}
20522 @cindex pragma, diagnostic
20524 Modifies the disposition of a diagnostic. Note that not all
20525 diagnostics are modifiable; at the moment only warnings (normally
20526 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
20527 Use @option{-fdiagnostics-show-option} to determine which diagnostics
20528 are controllable and which option controls them.
20530 @var{kind} is @samp{error} to treat this diagnostic as an error,
20531 @samp{warning} to treat it like a warning (even if @option{-Werror} is
20532 in effect), or @samp{ignored} if the diagnostic is to be ignored.
20533 @var{option} is a double quoted string that matches the command-line
20537 #pragma GCC diagnostic warning "-Wformat"
20538 #pragma GCC diagnostic error "-Wformat"
20539 #pragma GCC diagnostic ignored "-Wformat"
20542 Note that these pragmas override any command-line options. GCC keeps
20543 track of the location of each pragma, and issues diagnostics according
20544 to the state as of that point in the source file. Thus, pragmas occurring
20545 after a line do not affect diagnostics caused by that line.
20547 @item #pragma GCC diagnostic push
20548 @itemx #pragma GCC diagnostic pop
20550 Causes GCC to remember the state of the diagnostics as of each
20551 @code{push}, and restore to that point at each @code{pop}. If a
20552 @code{pop} has no matching @code{push}, the command-line options are
20556 #pragma GCC diagnostic error "-Wuninitialized"
20557 foo(a); /* error is given for this one */
20558 #pragma GCC diagnostic push
20559 #pragma GCC diagnostic ignored "-Wuninitialized"
20560 foo(b); /* no diagnostic for this one */
20561 #pragma GCC diagnostic pop
20562 foo(c); /* error is given for this one */
20563 #pragma GCC diagnostic pop
20564 foo(d); /* depends on command-line options */
20569 GCC also offers a simple mechanism for printing messages during
20573 @item #pragma message @var{string}
20574 @cindex pragma, diagnostic
20576 Prints @var{string} as a compiler message on compilation. The message
20577 is informational only, and is neither a compilation warning nor an error.
20580 #pragma message "Compiling " __FILE__ "..."
20583 @var{string} may be parenthesized, and is printed with location
20584 information. For example,
20587 #define DO_PRAGMA(x) _Pragma (#x)
20588 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
20590 TODO(Remember to fix this)
20594 prints @samp{/tmp/file.c:4: note: #pragma message:
20595 TODO - Remember to fix this}.
20599 @node Visibility Pragmas
20600 @subsection Visibility Pragmas
20603 @item #pragma GCC visibility push(@var{visibility})
20604 @itemx #pragma GCC visibility pop
20605 @cindex pragma, visibility
20607 This pragma allows the user to set the visibility for multiple
20608 declarations without having to give each a visibility attribute
20609 (@pxref{Function Attributes}).
20611 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
20612 declarations. Class members and template specializations are not
20613 affected; if you want to override the visibility for a particular
20614 member or instantiation, you must use an attribute.
20619 @node Push/Pop Macro Pragmas
20620 @subsection Push/Pop Macro Pragmas
20622 For compatibility with Microsoft Windows compilers, GCC supports
20623 @samp{#pragma push_macro(@var{"macro_name"})}
20624 and @samp{#pragma pop_macro(@var{"macro_name"})}.
20627 @item #pragma push_macro(@var{"macro_name"})
20628 @cindex pragma, push_macro
20629 This pragma saves the value of the macro named as @var{macro_name} to
20630 the top of the stack for this macro.
20632 @item #pragma pop_macro(@var{"macro_name"})
20633 @cindex pragma, pop_macro
20634 This pragma sets the value of the macro named as @var{macro_name} to
20635 the value on top of the stack for this macro. If the stack for
20636 @var{macro_name} is empty, the value of the macro remains unchanged.
20643 #pragma push_macro("X")
20646 #pragma pop_macro("X")
20651 In this example, the definition of X as 1 is saved by @code{#pragma
20652 push_macro} and restored by @code{#pragma pop_macro}.
20654 @node Function Specific Option Pragmas
20655 @subsection Function Specific Option Pragmas
20658 @item #pragma GCC target (@var{"string"}...)
20659 @cindex pragma GCC target
20661 This pragma allows you to set target specific options for functions
20662 defined later in the source file. One or more strings can be
20663 specified. Each function that is defined after this point is as
20664 if @code{attribute((target("STRING")))} was specified for that
20665 function. The parenthesis around the options is optional.
20666 @xref{Function Attributes}, for more information about the
20667 @code{target} attribute and the attribute syntax.
20669 The @code{#pragma GCC target} pragma is presently implemented for
20670 x86, PowerPC, and Nios II targets only.
20674 @item #pragma GCC optimize (@var{"string"}...)
20675 @cindex pragma GCC optimize
20677 This pragma allows you to set global optimization options for functions
20678 defined later in the source file. One or more strings can be
20679 specified. Each function that is defined after this point is as
20680 if @code{attribute((optimize("STRING")))} was specified for that
20681 function. The parenthesis around the options is optional.
20682 @xref{Function Attributes}, for more information about the
20683 @code{optimize} attribute and the attribute syntax.
20687 @item #pragma GCC push_options
20688 @itemx #pragma GCC pop_options
20689 @cindex pragma GCC push_options
20690 @cindex pragma GCC pop_options
20692 These pragmas maintain a stack of the current target and optimization
20693 options. It is intended for include files where you temporarily want
20694 to switch to using a different @samp{#pragma GCC target} or
20695 @samp{#pragma GCC optimize} and then to pop back to the previous
20700 @item #pragma GCC reset_options
20701 @cindex pragma GCC reset_options
20703 This pragma clears the current @code{#pragma GCC target} and
20704 @code{#pragma GCC optimize} to use the default switches as specified
20705 on the command line.
20708 @node Loop-Specific Pragmas
20709 @subsection Loop-Specific Pragmas
20712 @item #pragma GCC ivdep
20713 @cindex pragma GCC ivdep
20716 With this pragma, the programmer asserts that there are no loop-carried
20717 dependencies which would prevent consecutive iterations of
20718 the following loop from executing concurrently with SIMD
20719 (single instruction multiple data) instructions.
20721 For example, the compiler can only unconditionally vectorize the following
20722 loop with the pragma:
20725 void foo (int n, int *a, int *b, int *c)
20729 for (i = 0; i < n; ++i)
20730 a[i] = b[i] + c[i];
20735 In this example, using the @code{restrict} qualifier had the same
20736 effect. In the following example, that would not be possible. Assume
20737 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
20738 that it can unconditionally vectorize the following loop:
20741 void ignore_vec_dep (int *a, int k, int c, int m)
20744 for (int i = 0; i < m; i++)
20745 a[i] = a[i + k] * c;
20750 @node Unnamed Fields
20751 @section Unnamed Structure and Union Fields
20752 @cindex @code{struct}
20753 @cindex @code{union}
20755 As permitted by ISO C11 and for compatibility with other compilers,
20756 GCC allows you to define
20757 a structure or union that contains, as fields, structures and unions
20758 without names. For example:
20772 In this example, you are able to access members of the unnamed
20773 union with code like @samp{foo.b}. Note that only unnamed structs and
20774 unions are allowed, you may not have, for example, an unnamed
20777 You must never create such structures that cause ambiguous field definitions.
20778 For example, in this structure:
20790 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
20791 The compiler gives errors for such constructs.
20793 @opindex fms-extensions
20794 Unless @option{-fms-extensions} is used, the unnamed field must be a
20795 structure or union definition without a tag (for example, @samp{struct
20796 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
20797 also be a definition with a tag such as @samp{struct foo @{ int a;
20798 @};}, a reference to a previously defined structure or union such as
20799 @samp{struct foo;}, or a reference to a @code{typedef} name for a
20800 previously defined structure or union type.
20802 @opindex fplan9-extensions
20803 The option @option{-fplan9-extensions} enables
20804 @option{-fms-extensions} as well as two other extensions. First, a
20805 pointer to a structure is automatically converted to a pointer to an
20806 anonymous field for assignments and function calls. For example:
20809 struct s1 @{ int a; @};
20810 struct s2 @{ struct s1; @};
20811 extern void f1 (struct s1 *);
20812 void f2 (struct s2 *p) @{ f1 (p); @}
20816 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
20817 converted into a pointer to the anonymous field.
20819 Second, when the type of an anonymous field is a @code{typedef} for a
20820 @code{struct} or @code{union}, code may refer to the field using the
20821 name of the @code{typedef}.
20824 typedef struct @{ int a; @} s1;
20825 struct s2 @{ s1; @};
20826 s1 f1 (struct s2 *p) @{ return p->s1; @}
20829 These usages are only permitted when they are not ambiguous.
20832 @section Thread-Local Storage
20833 @cindex Thread-Local Storage
20834 @cindex @acronym{TLS}
20835 @cindex @code{__thread}
20837 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
20838 are allocated such that there is one instance of the variable per extant
20839 thread. The runtime model GCC uses to implement this originates
20840 in the IA-64 processor-specific ABI, but has since been migrated
20841 to other processors as well. It requires significant support from
20842 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
20843 system libraries (@file{libc.so} and @file{libpthread.so}), so it
20844 is not available everywhere.
20846 At the user level, the extension is visible with a new storage
20847 class keyword: @code{__thread}. For example:
20851 extern __thread struct state s;
20852 static __thread char *p;
20855 The @code{__thread} specifier may be used alone, with the @code{extern}
20856 or @code{static} specifiers, but with no other storage class specifier.
20857 When used with @code{extern} or @code{static}, @code{__thread} must appear
20858 immediately after the other storage class specifier.
20860 The @code{__thread} specifier may be applied to any global, file-scoped
20861 static, function-scoped static, or static data member of a class. It may
20862 not be applied to block-scoped automatic or non-static data member.
20864 When the address-of operator is applied to a thread-local variable, it is
20865 evaluated at run time and returns the address of the current thread's
20866 instance of that variable. An address so obtained may be used by any
20867 thread. When a thread terminates, any pointers to thread-local variables
20868 in that thread become invalid.
20870 No static initialization may refer to the address of a thread-local variable.
20872 In C++, if an initializer is present for a thread-local variable, it must
20873 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
20876 See @uref{http://www.akkadia.org/drepper/tls.pdf,
20877 ELF Handling For Thread-Local Storage} for a detailed explanation of
20878 the four thread-local storage addressing models, and how the runtime
20879 is expected to function.
20882 * C99 Thread-Local Edits::
20883 * C++98 Thread-Local Edits::
20886 @node C99 Thread-Local Edits
20887 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
20889 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
20890 that document the exact semantics of the language extension.
20894 @cite{5.1.2 Execution environments}
20896 Add new text after paragraph 1
20899 Within either execution environment, a @dfn{thread} is a flow of
20900 control within a program. It is implementation defined whether
20901 or not there may be more than one thread associated with a program.
20902 It is implementation defined how threads beyond the first are
20903 created, the name and type of the function called at thread
20904 startup, and how threads may be terminated. However, objects
20905 with thread storage duration shall be initialized before thread
20910 @cite{6.2.4 Storage durations of objects}
20912 Add new text before paragraph 3
20915 An object whose identifier is declared with the storage-class
20916 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
20917 Its lifetime is the entire execution of the thread, and its
20918 stored value is initialized only once, prior to thread startup.
20922 @cite{6.4.1 Keywords}
20924 Add @code{__thread}.
20927 @cite{6.7.1 Storage-class specifiers}
20929 Add @code{__thread} to the list of storage class specifiers in
20932 Change paragraph 2 to
20935 With the exception of @code{__thread}, at most one storage-class
20936 specifier may be given [@dots{}]. The @code{__thread} specifier may
20937 be used alone, or immediately following @code{extern} or
20941 Add new text after paragraph 6
20944 The declaration of an identifier for a variable that has
20945 block scope that specifies @code{__thread} shall also
20946 specify either @code{extern} or @code{static}.
20948 The @code{__thread} specifier shall be used only with
20953 @node C++98 Thread-Local Edits
20954 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
20956 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
20957 that document the exact semantics of the language extension.
20961 @b{[intro.execution]}
20963 New text after paragraph 4
20966 A @dfn{thread} is a flow of control within the abstract machine.
20967 It is implementation defined whether or not there may be more than
20971 New text after paragraph 7
20974 It is unspecified whether additional action must be taken to
20975 ensure when and whether side effects are visible to other threads.
20981 Add @code{__thread}.
20984 @b{[basic.start.main]}
20986 Add after paragraph 5
20989 The thread that begins execution at the @code{main} function is called
20990 the @dfn{main thread}. It is implementation defined how functions
20991 beginning threads other than the main thread are designated or typed.
20992 A function so designated, as well as the @code{main} function, is called
20993 a @dfn{thread startup function}. It is implementation defined what
20994 happens if a thread startup function returns. It is implementation
20995 defined what happens to other threads when any thread calls @code{exit}.
20999 @b{[basic.start.init]}
21001 Add after paragraph 4
21004 The storage for an object of thread storage duration shall be
21005 statically initialized before the first statement of the thread startup
21006 function. An object of thread storage duration shall not require
21007 dynamic initialization.
21011 @b{[basic.start.term]}
21013 Add after paragraph 3
21016 The type of an object with thread storage duration shall not have a
21017 non-trivial destructor, nor shall it be an array type whose elements
21018 (directly or indirectly) have non-trivial destructors.
21024 Add ``thread storage duration'' to the list in paragraph 1.
21029 Thread, static, and automatic storage durations are associated with
21030 objects introduced by declarations [@dots{}].
21033 Add @code{__thread} to the list of specifiers in paragraph 3.
21036 @b{[basic.stc.thread]}
21038 New section before @b{[basic.stc.static]}
21041 The keyword @code{__thread} applied to a non-local object gives the
21042 object thread storage duration.
21044 A local variable or class data member declared both @code{static}
21045 and @code{__thread} gives the variable or member thread storage
21050 @b{[basic.stc.static]}
21055 All objects that have neither thread storage duration, dynamic
21056 storage duration nor are local [@dots{}].
21062 Add @code{__thread} to the list in paragraph 1.
21067 With the exception of @code{__thread}, at most one
21068 @var{storage-class-specifier} shall appear in a given
21069 @var{decl-specifier-seq}. The @code{__thread} specifier may
21070 be used alone, or immediately following the @code{extern} or
21071 @code{static} specifiers. [@dots{}]
21074 Add after paragraph 5
21077 The @code{__thread} specifier can be applied only to the names of objects
21078 and to anonymous unions.
21084 Add after paragraph 6
21087 Non-@code{static} members shall not be @code{__thread}.
21091 @node Binary constants
21092 @section Binary Constants using the @samp{0b} Prefix
21093 @cindex Binary constants using the @samp{0b} prefix
21095 Integer constants can be written as binary constants, consisting of a
21096 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
21097 @samp{0B}. This is particularly useful in environments that operate a
21098 lot on the bit level (like microcontrollers).
21100 The following statements are identical:
21109 The type of these constants follows the same rules as for octal or
21110 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
21113 @node C++ Extensions
21114 @chapter Extensions to the C++ Language
21115 @cindex extensions, C++ language
21116 @cindex C++ language extensions
21118 The GNU compiler provides these extensions to the C++ language (and you
21119 can also use most of the C language extensions in your C++ programs). If you
21120 want to write code that checks whether these features are available, you can
21121 test for the GNU compiler the same way as for C programs: check for a
21122 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
21123 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
21124 Predefined Macros,cpp,The GNU C Preprocessor}).
21127 * C++ Volatiles:: What constitutes an access to a volatile object.
21128 * Restricted Pointers:: C99 restricted pointers and references.
21129 * Vague Linkage:: Where G++ puts inlines, vtables and such.
21130 * C++ Interface:: You can use a single C++ header file for both
21131 declarations and definitions.
21132 * Template Instantiation:: Methods for ensuring that exactly one copy of
21133 each needed template instantiation is emitted.
21134 * Bound member functions:: You can extract a function pointer to the
21135 method denoted by a @samp{->*} or @samp{.*} expression.
21136 * C++ Attributes:: Variable, function, and type attributes for C++ only.
21137 * Function Multiversioning:: Declaring multiple function versions.
21138 * Namespace Association:: Strong using-directives for namespace association.
21139 * Type Traits:: Compiler support for type traits.
21140 * C++ Concepts:: Improved support for generic programming.
21141 * Java Exceptions:: Tweaking exception handling to work with Java.
21142 * Deprecated Features:: Things will disappear from G++.
21143 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
21146 @node C++ Volatiles
21147 @section When is a Volatile C++ Object Accessed?
21148 @cindex accessing volatiles
21149 @cindex volatile read
21150 @cindex volatile write
21151 @cindex volatile access
21153 The C++ standard differs from the C standard in its treatment of
21154 volatile objects. It fails to specify what constitutes a volatile
21155 access, except to say that C++ should behave in a similar manner to C
21156 with respect to volatiles, where possible. However, the different
21157 lvalueness of expressions between C and C++ complicate the behavior.
21158 G++ behaves the same as GCC for volatile access, @xref{C
21159 Extensions,,Volatiles}, for a description of GCC's behavior.
21161 The C and C++ language specifications differ when an object is
21162 accessed in a void context:
21165 volatile int *src = @var{somevalue};
21169 The C++ standard specifies that such expressions do not undergo lvalue
21170 to rvalue conversion, and that the type of the dereferenced object may
21171 be incomplete. The C++ standard does not specify explicitly that it
21172 is lvalue to rvalue conversion that is responsible for causing an
21173 access. There is reason to believe that it is, because otherwise
21174 certain simple expressions become undefined. However, because it
21175 would surprise most programmers, G++ treats dereferencing a pointer to
21176 volatile object of complete type as GCC would do for an equivalent
21177 type in C@. When the object has incomplete type, G++ issues a
21178 warning; if you wish to force an error, you must force a conversion to
21179 rvalue with, for instance, a static cast.
21181 When using a reference to volatile, G++ does not treat equivalent
21182 expressions as accesses to volatiles, but instead issues a warning that
21183 no volatile is accessed. The rationale for this is that otherwise it
21184 becomes difficult to determine where volatile access occur, and not
21185 possible to ignore the return value from functions returning volatile
21186 references. Again, if you wish to force a read, cast the reference to
21189 G++ implements the same behavior as GCC does when assigning to a
21190 volatile object---there is no reread of the assigned-to object, the
21191 assigned rvalue is reused. Note that in C++ assignment expressions
21192 are lvalues, and if used as an lvalue, the volatile object is
21193 referred to. For instance, @var{vref} refers to @var{vobj}, as
21194 expected, in the following example:
21198 volatile int &vref = vobj = @var{something};
21201 @node Restricted Pointers
21202 @section Restricting Pointer Aliasing
21203 @cindex restricted pointers
21204 @cindex restricted references
21205 @cindex restricted this pointer
21207 As with the C front end, G++ understands the C99 feature of restricted pointers,
21208 specified with the @code{__restrict__}, or @code{__restrict} type
21209 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
21210 language flag, @code{restrict} is not a keyword in C++.
21212 In addition to allowing restricted pointers, you can specify restricted
21213 references, which indicate that the reference is not aliased in the local
21217 void fn (int *__restrict__ rptr, int &__restrict__ rref)
21224 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
21225 @var{rref} refers to a (different) unaliased integer.
21227 You may also specify whether a member function's @var{this} pointer is
21228 unaliased by using @code{__restrict__} as a member function qualifier.
21231 void T::fn () __restrict__
21238 Within the body of @code{T::fn}, @var{this} has the effective
21239 definition @code{T *__restrict__ const this}. Notice that the
21240 interpretation of a @code{__restrict__} member function qualifier is
21241 different to that of @code{const} or @code{volatile} qualifier, in that it
21242 is applied to the pointer rather than the object. This is consistent with
21243 other compilers that implement restricted pointers.
21245 As with all outermost parameter qualifiers, @code{__restrict__} is
21246 ignored in function definition matching. This means you only need to
21247 specify @code{__restrict__} in a function definition, rather than
21248 in a function prototype as well.
21250 @node Vague Linkage
21251 @section Vague Linkage
21252 @cindex vague linkage
21254 There are several constructs in C++ that require space in the object
21255 file but are not clearly tied to a single translation unit. We say that
21256 these constructs have ``vague linkage''. Typically such constructs are
21257 emitted wherever they are needed, though sometimes we can be more
21261 @item Inline Functions
21262 Inline functions are typically defined in a header file which can be
21263 included in many different compilations. Hopefully they can usually be
21264 inlined, but sometimes an out-of-line copy is necessary, if the address
21265 of the function is taken or if inlining fails. In general, we emit an
21266 out-of-line copy in all translation units where one is needed. As an
21267 exception, we only emit inline virtual functions with the vtable, since
21268 it always requires a copy.
21270 Local static variables and string constants used in an inline function
21271 are also considered to have vague linkage, since they must be shared
21272 between all inlined and out-of-line instances of the function.
21276 C++ virtual functions are implemented in most compilers using a lookup
21277 table, known as a vtable. The vtable contains pointers to the virtual
21278 functions provided by a class, and each object of the class contains a
21279 pointer to its vtable (or vtables, in some multiple-inheritance
21280 situations). If the class declares any non-inline, non-pure virtual
21281 functions, the first one is chosen as the ``key method'' for the class,
21282 and the vtable is only emitted in the translation unit where the key
21285 @emph{Note:} If the chosen key method is later defined as inline, the
21286 vtable is still emitted in every translation unit that defines it.
21287 Make sure that any inline virtuals are declared inline in the class
21288 body, even if they are not defined there.
21290 @item @code{type_info} objects
21291 @cindex @code{type_info}
21293 C++ requires information about types to be written out in order to
21294 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
21295 For polymorphic classes (classes with virtual functions), the @samp{type_info}
21296 object is written out along with the vtable so that @samp{dynamic_cast}
21297 can determine the dynamic type of a class object at run time. For all
21298 other types, we write out the @samp{type_info} object when it is used: when
21299 applying @samp{typeid} to an expression, throwing an object, or
21300 referring to a type in a catch clause or exception specification.
21302 @item Template Instantiations
21303 Most everything in this section also applies to template instantiations,
21304 but there are other options as well.
21305 @xref{Template Instantiation,,Where's the Template?}.
21309 When used with GNU ld version 2.8 or later on an ELF system such as
21310 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
21311 these constructs will be discarded at link time. This is known as
21314 On targets that don't support COMDAT, but do support weak symbols, GCC
21315 uses them. This way one copy overrides all the others, but
21316 the unused copies still take up space in the executable.
21318 For targets that do not support either COMDAT or weak symbols,
21319 most entities with vague linkage are emitted as local symbols to
21320 avoid duplicate definition errors from the linker. This does not happen
21321 for local statics in inlines, however, as having multiple copies
21322 almost certainly breaks things.
21324 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
21325 another way to control placement of these constructs.
21327 @node C++ Interface
21328 @section C++ Interface and Implementation Pragmas
21330 @cindex interface and implementation headers, C++
21331 @cindex C++ interface and implementation headers
21332 @cindex pragmas, interface and implementation
21334 @code{#pragma interface} and @code{#pragma implementation} provide the
21335 user with a way of explicitly directing the compiler to emit entities
21336 with vague linkage (and debugging information) in a particular
21339 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
21340 by COMDAT support and the ``key method'' heuristic
21341 mentioned in @ref{Vague Linkage}. Using them can actually cause your
21342 program to grow due to unnecessary out-of-line copies of inline
21346 @item #pragma interface
21347 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
21348 @kindex #pragma interface
21349 Use this directive in @emph{header files} that define object classes, to save
21350 space in most of the object files that use those classes. Normally,
21351 local copies of certain information (backup copies of inline member
21352 functions, debugging information, and the internal tables that implement
21353 virtual functions) must be kept in each object file that includes class
21354 definitions. You can use this pragma to avoid such duplication. When a
21355 header file containing @samp{#pragma interface} is included in a
21356 compilation, this auxiliary information is not generated (unless
21357 the main input source file itself uses @samp{#pragma implementation}).
21358 Instead, the object files contain references to be resolved at link
21361 The second form of this directive is useful for the case where you have
21362 multiple headers with the same name in different directories. If you
21363 use this form, you must specify the same string to @samp{#pragma
21366 @item #pragma implementation
21367 @itemx #pragma implementation "@var{objects}.h"
21368 @kindex #pragma implementation
21369 Use this pragma in a @emph{main input file}, when you want full output from
21370 included header files to be generated (and made globally visible). The
21371 included header file, in turn, should use @samp{#pragma interface}.
21372 Backup copies of inline member functions, debugging information, and the
21373 internal tables used to implement virtual functions are all generated in
21374 implementation files.
21376 @cindex implied @code{#pragma implementation}
21377 @cindex @code{#pragma implementation}, implied
21378 @cindex naming convention, implementation headers
21379 If you use @samp{#pragma implementation} with no argument, it applies to
21380 an include file with the same basename@footnote{A file's @dfn{basename}
21381 is the name stripped of all leading path information and of trailing
21382 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
21383 file. For example, in @file{allclass.cc}, giving just
21384 @samp{#pragma implementation}
21385 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
21387 Use the string argument if you want a single implementation file to
21388 include code from multiple header files. (You must also use
21389 @samp{#include} to include the header file; @samp{#pragma
21390 implementation} only specifies how to use the file---it doesn't actually
21393 There is no way to split up the contents of a single header file into
21394 multiple implementation files.
21397 @cindex inlining and C++ pragmas
21398 @cindex C++ pragmas, effect on inlining
21399 @cindex pragmas in C++, effect on inlining
21400 @samp{#pragma implementation} and @samp{#pragma interface} also have an
21401 effect on function inlining.
21403 If you define a class in a header file marked with @samp{#pragma
21404 interface}, the effect on an inline function defined in that class is
21405 similar to an explicit @code{extern} declaration---the compiler emits
21406 no code at all to define an independent version of the function. Its
21407 definition is used only for inlining with its callers.
21409 @opindex fno-implement-inlines
21410 Conversely, when you include the same header file in a main source file
21411 that declares it as @samp{#pragma implementation}, the compiler emits
21412 code for the function itself; this defines a version of the function
21413 that can be found via pointers (or by callers compiled without
21414 inlining). If all calls to the function can be inlined, you can avoid
21415 emitting the function by compiling with @option{-fno-implement-inlines}.
21416 If any calls are not inlined, you will get linker errors.
21418 @node Template Instantiation
21419 @section Where's the Template?
21420 @cindex template instantiation
21422 C++ templates were the first language feature to require more
21423 intelligence from the environment than was traditionally found on a UNIX
21424 system. Somehow the compiler and linker have to make sure that each
21425 template instance occurs exactly once in the executable if it is needed,
21426 and not at all otherwise. There are two basic approaches to this
21427 problem, which are referred to as the Borland model and the Cfront model.
21430 @item Borland model
21431 Borland C++ solved the template instantiation problem by adding the code
21432 equivalent of common blocks to their linker; the compiler emits template
21433 instances in each translation unit that uses them, and the linker
21434 collapses them together. The advantage of this model is that the linker
21435 only has to consider the object files themselves; there is no external
21436 complexity to worry about. The disadvantage is that compilation time
21437 is increased because the template code is being compiled repeatedly.
21438 Code written for this model tends to include definitions of all
21439 templates in the header file, since they must be seen to be
21443 The AT&T C++ translator, Cfront, solved the template instantiation
21444 problem by creating the notion of a template repository, an
21445 automatically maintained place where template instances are stored. A
21446 more modern version of the repository works as follows: As individual
21447 object files are built, the compiler places any template definitions and
21448 instantiations encountered in the repository. At link time, the link
21449 wrapper adds in the objects in the repository and compiles any needed
21450 instances that were not previously emitted. The advantages of this
21451 model are more optimal compilation speed and the ability to use the
21452 system linker; to implement the Borland model a compiler vendor also
21453 needs to replace the linker. The disadvantages are vastly increased
21454 complexity, and thus potential for error; for some code this can be
21455 just as transparent, but in practice it can been very difficult to build
21456 multiple programs in one directory and one program in multiple
21457 directories. Code written for this model tends to separate definitions
21458 of non-inline member templates into a separate file, which should be
21459 compiled separately.
21462 G++ implements the Borland model on targets where the linker supports it,
21463 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
21464 Otherwise G++ implements neither automatic model.
21466 You have the following options for dealing with template instantiations:
21470 Do nothing. Code written for the Borland model works fine, but
21471 each translation unit contains instances of each of the templates it
21472 uses. The duplicate instances will be discarded by the linker, but in
21473 a large program, this can lead to an unacceptable amount of code
21474 duplication in object files or shared libraries.
21476 Duplicate instances of a template can be avoided by defining an explicit
21477 instantiation in one object file, and preventing the compiler from doing
21478 implicit instantiations in any other object files by using an explicit
21479 instantiation declaration, using the @code{extern template} syntax:
21482 extern template int max (int, int);
21485 This syntax is defined in the C++ 2011 standard, but has been supported by
21486 G++ and other compilers since well before 2011.
21488 Explicit instantiations can be used for the largest or most frequently
21489 duplicated instances, without having to know exactly which other instances
21490 are used in the rest of the program. You can scatter the explicit
21491 instantiations throughout your program, perhaps putting them in the
21492 translation units where the instances are used or the translation units
21493 that define the templates themselves; you can put all of the explicit
21494 instantiations you need into one big file; or you can create small files
21501 template class Foo<int>;
21502 template ostream& operator <<
21503 (ostream&, const Foo<int>&);
21507 for each of the instances you need, and create a template instantiation
21508 library from those.
21510 This is the simplest option, but also offers flexibility and
21511 fine-grained control when necessary. It is also the most portable
21512 alternative and programs using this approach will work with most modern
21517 Compile your template-using code with @option{-frepo}. The compiler
21518 generates files with the extension @samp{.rpo} listing all of the
21519 template instantiations used in the corresponding object files that
21520 could be instantiated there; the link wrapper, @samp{collect2},
21521 then updates the @samp{.rpo} files to tell the compiler where to place
21522 those instantiations and rebuild any affected object files. The
21523 link-time overhead is negligible after the first pass, as the compiler
21524 continues to place the instantiations in the same files.
21526 This can be a suitable option for application code written for the Borland
21527 model, as it usually just works. Code written for the Cfront model
21528 needs to be modified so that the template definitions are available at
21529 one or more points of instantiation; usually this is as simple as adding
21530 @code{#include <tmethods.cc>} to the end of each template header.
21532 For library code, if you want the library to provide all of the template
21533 instantiations it needs, just try to link all of its object files
21534 together; the link will fail, but cause the instantiations to be
21535 generated as a side effect. Be warned, however, that this may cause
21536 conflicts if multiple libraries try to provide the same instantiations.
21537 For greater control, use explicit instantiation as described in the next
21541 @opindex fno-implicit-templates
21542 Compile your code with @option{-fno-implicit-templates} to disable the
21543 implicit generation of template instances, and explicitly instantiate
21544 all the ones you use. This approach requires more knowledge of exactly
21545 which instances you need than do the others, but it's less
21546 mysterious and allows greater control if you want to ensure that only
21547 the intended instances are used.
21549 If you are using Cfront-model code, you can probably get away with not
21550 using @option{-fno-implicit-templates} when compiling files that don't
21551 @samp{#include} the member template definitions.
21553 If you use one big file to do the instantiations, you may want to
21554 compile it without @option{-fno-implicit-templates} so you get all of the
21555 instances required by your explicit instantiations (but not by any
21556 other files) without having to specify them as well.
21558 In addition to forward declaration of explicit instantiations
21559 (with @code{extern}), G++ has extended the template instantiation
21560 syntax to support instantiation of the compiler support data for a
21561 template class (i.e.@: the vtable) without instantiating any of its
21562 members (with @code{inline}), and instantiation of only the static data
21563 members of a template class, without the support data or member
21564 functions (with @code{static}):
21567 inline template class Foo<int>;
21568 static template class Foo<int>;
21572 @node Bound member functions
21573 @section Extracting the Function Pointer from a Bound Pointer to Member Function
21575 @cindex pointer to member function
21576 @cindex bound pointer to member function
21578 In C++, pointer to member functions (PMFs) are implemented using a wide
21579 pointer of sorts to handle all the possible call mechanisms; the PMF
21580 needs to store information about how to adjust the @samp{this} pointer,
21581 and if the function pointed to is virtual, where to find the vtable, and
21582 where in the vtable to look for the member function. If you are using
21583 PMFs in an inner loop, you should really reconsider that decision. If
21584 that is not an option, you can extract the pointer to the function that
21585 would be called for a given object/PMF pair and call it directly inside
21586 the inner loop, to save a bit of time.
21588 Note that you still pay the penalty for the call through a
21589 function pointer; on most modern architectures, such a call defeats the
21590 branch prediction features of the CPU@. This is also true of normal
21591 virtual function calls.
21593 The syntax for this extension is
21597 extern int (A::*fp)();
21598 typedef int (*fptr)(A *);
21600 fptr p = (fptr)(a.*fp);
21603 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
21604 no object is needed to obtain the address of the function. They can be
21605 converted to function pointers directly:
21608 fptr p1 = (fptr)(&A::foo);
21611 @opindex Wno-pmf-conversions
21612 You must specify @option{-Wno-pmf-conversions} to use this extension.
21614 @node C++ Attributes
21615 @section C++-Specific Variable, Function, and Type Attributes
21617 Some attributes only make sense for C++ programs.
21620 @item abi_tag ("@var{tag}", ...)
21621 @cindex @code{abi_tag} function attribute
21622 @cindex @code{abi_tag} variable attribute
21623 @cindex @code{abi_tag} type attribute
21624 The @code{abi_tag} attribute can be applied to a function, variable, or class
21625 declaration. It modifies the mangled name of the entity to
21626 incorporate the tag name, in order to distinguish the function or
21627 class from an earlier version with a different ABI; perhaps the class
21628 has changed size, or the function has a different return type that is
21629 not encoded in the mangled name.
21631 The attribute can also be applied to an inline namespace, but does not
21632 affect the mangled name of the namespace; in this case it is only used
21633 for @option{-Wabi-tag} warnings and automatic tagging of functions and
21634 variables. Tagging inline namespaces is generally preferable to
21635 tagging individual declarations, but the latter is sometimes
21636 necessary, such as when only certain members of a class need to be
21639 The argument can be a list of strings of arbitrary length. The
21640 strings are sorted on output, so the order of the list is
21643 A redeclaration of an entity must not add new ABI tags,
21644 since doing so would change the mangled name.
21646 The ABI tags apply to a name, so all instantiations and
21647 specializations of a template have the same tags. The attribute will
21648 be ignored if applied to an explicit specialization or instantiation.
21650 The @option{-Wabi-tag} flag enables a warning about a class which does
21651 not have all the ABI tags used by its subobjects and virtual functions; for users with code
21652 that needs to coexist with an earlier ABI, using this option can help
21653 to find all affected types that need to be tagged.
21655 When a type involving an ABI tag is used as the type of a variable or
21656 return type of a function where that tag is not already present in the
21657 signature of the function, the tag is automatically applied to the
21658 variable or function. @option{-Wabi-tag} also warns about this
21659 situation; this warning can be avoided by explicitly tagging the
21660 variable or function or moving it into a tagged inline namespace.
21662 @item init_priority (@var{priority})
21663 @cindex @code{init_priority} variable attribute
21665 In Standard C++, objects defined at namespace scope are guaranteed to be
21666 initialized in an order in strict accordance with that of their definitions
21667 @emph{in a given translation unit}. No guarantee is made for initializations
21668 across translation units. However, GNU C++ allows users to control the
21669 order of initialization of objects defined at namespace scope with the
21670 @code{init_priority} attribute by specifying a relative @var{priority},
21671 a constant integral expression currently bounded between 101 and 65535
21672 inclusive. Lower numbers indicate a higher priority.
21674 In the following example, @code{A} would normally be created before
21675 @code{B}, but the @code{init_priority} attribute reverses that order:
21678 Some_Class A __attribute__ ((init_priority (2000)));
21679 Some_Class B __attribute__ ((init_priority (543)));
21683 Note that the particular values of @var{priority} do not matter; only their
21686 @item java_interface
21687 @cindex @code{java_interface} type attribute
21689 This type attribute informs C++ that the class is a Java interface. It may
21690 only be applied to classes declared within an @code{extern "Java"} block.
21691 Calls to methods declared in this interface are dispatched using GCJ's
21692 interface table mechanism, instead of regular virtual table dispatch.
21695 @cindex @code{warn_unused} type attribute
21697 For C++ types with non-trivial constructors and/or destructors it is
21698 impossible for the compiler to determine whether a variable of this
21699 type is truly unused if it is not referenced. This type attribute
21700 informs the compiler that variables of this type should be warned
21701 about if they appear to be unused, just like variables of fundamental
21704 This attribute is appropriate for types which just represent a value,
21705 such as @code{std::string}; it is not appropriate for types which
21706 control a resource, such as @code{std::lock_guard}.
21708 This attribute is also accepted in C, but it is unnecessary because C
21709 does not have constructors or destructors.
21713 See also @ref{Namespace Association}.
21715 @node Function Multiversioning
21716 @section Function Multiversioning
21717 @cindex function versions
21719 With the GNU C++ front end, for x86 targets, you may specify multiple
21720 versions of a function, where each function is specialized for a
21721 specific target feature. At runtime, the appropriate version of the
21722 function is automatically executed depending on the characteristics of
21723 the execution platform. Here is an example.
21726 __attribute__ ((target ("default")))
21729 // The default version of foo.
21733 __attribute__ ((target ("sse4.2")))
21736 // foo version for SSE4.2
21740 __attribute__ ((target ("arch=atom")))
21743 // foo version for the Intel ATOM processor
21747 __attribute__ ((target ("arch=amdfam10")))
21750 // foo version for the AMD Family 0x10 processors.
21757 assert ((*p) () == foo ());
21762 In the above example, four versions of function foo are created. The
21763 first version of foo with the target attribute "default" is the default
21764 version. This version gets executed when no other target specific
21765 version qualifies for execution on a particular platform. A new version
21766 of foo is created by using the same function signature but with a
21767 different target string. Function foo is called or a pointer to it is
21768 taken just like a regular function. GCC takes care of doing the
21769 dispatching to call the right version at runtime. Refer to the
21770 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
21771 Function Multiversioning} for more details.
21773 @node Namespace Association
21774 @section Namespace Association
21776 @strong{Caution:} The semantics of this extension are equivalent
21777 to C++ 2011 inline namespaces. Users should use inline namespaces
21778 instead as this extension will be removed in future versions of G++.
21780 A using-directive with @code{__attribute ((strong))} is stronger
21781 than a normal using-directive in two ways:
21785 Templates from the used namespace can be specialized and explicitly
21786 instantiated as though they were members of the using namespace.
21789 The using namespace is considered an associated namespace of all
21790 templates in the used namespace for purposes of argument-dependent
21794 The used namespace must be nested within the using namespace so that
21795 normal unqualified lookup works properly.
21797 This is useful for composing a namespace transparently from
21798 implementation namespaces. For example:
21803 template <class T> struct A @{ @};
21805 using namespace debug __attribute ((__strong__));
21806 template <> struct A<int> @{ @}; // @r{OK to specialize}
21808 template <class T> void f (A<T>);
21813 f (std::A<float>()); // @r{lookup finds} std::f
21819 @section Type Traits
21821 The C++ front end implements syntactic extensions that allow
21822 compile-time determination of
21823 various characteristics of a type (or of a
21827 @item __has_nothrow_assign (type)
21828 If @code{type} is const qualified or is a reference type then the trait is
21829 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
21830 is true, else if @code{type} is a cv class or union type with copy assignment
21831 operators that are known not to throw an exception then the trait is true,
21832 else it is false. Requires: @code{type} shall be a complete type,
21833 (possibly cv-qualified) @code{void}, or an array of unknown bound.
21835 @item __has_nothrow_copy (type)
21836 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
21837 @code{type} is a cv class or union type with copy constructors that
21838 are known not to throw an exception then the trait is true, else it is false.
21839 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
21840 @code{void}, or an array of unknown bound.
21842 @item __has_nothrow_constructor (type)
21843 If @code{__has_trivial_constructor (type)} is true then the trait is
21844 true, else if @code{type} is a cv class or union type (or array
21845 thereof) with a default constructor that is known not to throw an
21846 exception then the trait is true, else it is false. Requires:
21847 @code{type} shall be a complete type, (possibly cv-qualified)
21848 @code{void}, or an array of unknown bound.
21850 @item __has_trivial_assign (type)
21851 If @code{type} is const qualified or is a reference type then the trait is
21852 false. Otherwise if @code{__is_pod (type)} is true then the trait is
21853 true, else if @code{type} is a cv class or union type with a trivial
21854 copy assignment ([class.copy]) then the trait is true, else it is
21855 false. Requires: @code{type} shall be a complete type, (possibly
21856 cv-qualified) @code{void}, or an array of unknown bound.
21858 @item __has_trivial_copy (type)
21859 If @code{__is_pod (type)} is true or @code{type} is a reference type
21860 then the trait is true, else if @code{type} is a cv class or union type
21861 with a trivial copy constructor ([class.copy]) then the trait
21862 is true, else it is false. Requires: @code{type} shall be a complete
21863 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21865 @item __has_trivial_constructor (type)
21866 If @code{__is_pod (type)} is true then the trait is true, else if
21867 @code{type} is a cv class or union type (or array thereof) with a
21868 trivial default constructor ([class.ctor]) then the trait is true,
21869 else it is false. Requires: @code{type} shall be a complete
21870 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21872 @item __has_trivial_destructor (type)
21873 If @code{__is_pod (type)} is true or @code{type} is a reference type then
21874 the trait is true, else if @code{type} is a cv class or union type (or
21875 array thereof) with a trivial destructor ([class.dtor]) then the trait
21876 is true, else it is false. Requires: @code{type} shall be a complete
21877 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21879 @item __has_virtual_destructor (type)
21880 If @code{type} is a class type with a virtual destructor
21881 ([class.dtor]) then the trait is true, else it is false. Requires:
21882 @code{type} shall be a complete type, (possibly cv-qualified)
21883 @code{void}, or an array of unknown bound.
21885 @item __is_abstract (type)
21886 If @code{type} is an abstract class ([class.abstract]) then the trait
21887 is true, else it is false. Requires: @code{type} shall be a complete
21888 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21890 @item __is_base_of (base_type, derived_type)
21891 If @code{base_type} is a base class of @code{derived_type}
21892 ([class.derived]) then the trait is true, otherwise it is false.
21893 Top-level cv qualifications of @code{base_type} and
21894 @code{derived_type} are ignored. For the purposes of this trait, a
21895 class type is considered is own base. Requires: if @code{__is_class
21896 (base_type)} and @code{__is_class (derived_type)} are true and
21897 @code{base_type} and @code{derived_type} are not the same type
21898 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
21899 type. A diagnostic is produced if this requirement is not met.
21901 @item __is_class (type)
21902 If @code{type} is a cv class type, and not a union type
21903 ([basic.compound]) the trait is true, else it is false.
21905 @item __is_empty (type)
21906 If @code{__is_class (type)} is false then the trait is false.
21907 Otherwise @code{type} is considered empty if and only if: @code{type}
21908 has no non-static data members, or all non-static data members, if
21909 any, are bit-fields of length 0, and @code{type} has no virtual
21910 members, and @code{type} has no virtual base classes, and @code{type}
21911 has no base classes @code{base_type} for which
21912 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
21913 be a complete type, (possibly cv-qualified) @code{void}, or an array
21916 @item __is_enum (type)
21917 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
21918 true, else it is false.
21920 @item __is_literal_type (type)
21921 If @code{type} is a literal type ([basic.types]) the trait is
21922 true, else it is false. Requires: @code{type} shall be a complete type,
21923 (possibly cv-qualified) @code{void}, or an array of unknown bound.
21925 @item __is_pod (type)
21926 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
21927 else it is false. Requires: @code{type} shall be a complete type,
21928 (possibly cv-qualified) @code{void}, or an array of unknown bound.
21930 @item __is_polymorphic (type)
21931 If @code{type} is a polymorphic class ([class.virtual]) then the trait
21932 is true, else it is false. Requires: @code{type} shall be a complete
21933 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21935 @item __is_standard_layout (type)
21936 If @code{type} is a standard-layout type ([basic.types]) the trait is
21937 true, else it is false. Requires: @code{type} shall be a complete
21938 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21940 @item __is_trivial (type)
21941 If @code{type} is a trivial type ([basic.types]) the trait is
21942 true, else it is false. Requires: @code{type} shall be a complete
21943 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
21945 @item __is_union (type)
21946 If @code{type} is a cv union type ([basic.compound]) the trait is
21947 true, else it is false.
21949 @item __underlying_type (type)
21950 The underlying type of @code{type}. Requires: @code{type} shall be
21951 an enumeration type ([dcl.enum]).
21957 @section C++ Concepts
21959 C++ concepts provide much-improved support for generic programming. In
21960 particular, they allow the specification of constraints on template arguments.
21961 The constraints are used to extend the usual overloading and partial
21962 specialization capabilities of the language, allowing generic data structures
21963 and algorithms to be ``refined'' based on their properties rather than their
21966 The following keywords are reserved for concepts.
21970 States an expression as an assumption, and if possible, verifies that the
21971 assumption is valid. For example, @code{assume(n > 0)}.
21974 Introduces an axiom definition. Axioms introduce requirements on values.
21977 Introduces a universally quantified object in an axiom. For example,
21978 @code{forall (int n) n + 0 == n}).
21981 Introduces a concept definition. Concepts are sets of syntactic and semantic
21982 requirements on types and their values.
21985 Introduces constraints on template arguments or requirements for a member
21986 function of a class template.
21990 The front end also exposes a number of internal mechanism that can be used
21991 to simplify the writing of type traits. Note that some of these traits are
21992 likely to be removed in the future.
21995 @item __is_same (type1, type2)
21996 A binary type trait: true whenever the type arguments are the same.
22001 @node Java Exceptions
22002 @section Java Exceptions
22004 The Java language uses a slightly different exception handling model
22005 from C++. Normally, GNU C++ automatically detects when you are
22006 writing C++ code that uses Java exceptions, and handle them
22007 appropriately. However, if C++ code only needs to execute destructors
22008 when Java exceptions are thrown through it, GCC guesses incorrectly.
22009 Sample problematic code is:
22012 struct S @{ ~S(); @};
22013 extern void bar(); // @r{is written in Java, and may throw exceptions}
22022 The usual effect of an incorrect guess is a link failure, complaining of
22023 a missing routine called @samp{__gxx_personality_v0}.
22025 You can inform the compiler that Java exceptions are to be used in a
22026 translation unit, irrespective of what it might think, by writing
22027 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
22028 @samp{#pragma} must appear before any functions that throw or catch
22029 exceptions, or run destructors when exceptions are thrown through them.
22031 You cannot mix Java and C++ exceptions in the same translation unit. It
22032 is believed to be safe to throw a C++ exception from one file through
22033 another file compiled for the Java exception model, or vice versa, but
22034 there may be bugs in this area.
22036 @node Deprecated Features
22037 @section Deprecated Features
22039 In the past, the GNU C++ compiler was extended to experiment with new
22040 features, at a time when the C++ language was still evolving. Now that
22041 the C++ standard is complete, some of those features are superseded by
22042 superior alternatives. Using the old features might cause a warning in
22043 some cases that the feature will be dropped in the future. In other
22044 cases, the feature might be gone already.
22046 While the list below is not exhaustive, it documents some of the options
22047 that are now deprecated:
22050 @item -fexternal-templates
22051 @itemx -falt-external-templates
22052 These are two of the many ways for G++ to implement template
22053 instantiation. @xref{Template Instantiation}. The C++ standard clearly
22054 defines how template definitions have to be organized across
22055 implementation units. G++ has an implicit instantiation mechanism that
22056 should work just fine for standard-conforming code.
22058 @item -fstrict-prototype
22059 @itemx -fno-strict-prototype
22060 Previously it was possible to use an empty prototype parameter list to
22061 indicate an unspecified number of parameters (like C), rather than no
22062 parameters, as C++ demands. This feature has been removed, except where
22063 it is required for backwards compatibility. @xref{Backwards Compatibility}.
22066 G++ allows a virtual function returning @samp{void *} to be overridden
22067 by one returning a different pointer type. This extension to the
22068 covariant return type rules is now deprecated and will be removed from a
22071 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
22072 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
22073 and are now removed from G++. Code using these operators should be
22074 modified to use @code{std::min} and @code{std::max} instead.
22076 The named return value extension has been deprecated, and is now
22079 The use of initializer lists with new expressions has been deprecated,
22080 and is now removed from G++.
22082 Floating and complex non-type template parameters have been deprecated,
22083 and are now removed from G++.
22085 The implicit typename extension has been deprecated and is now
22088 The use of default arguments in function pointers, function typedefs
22089 and other places where they are not permitted by the standard is
22090 deprecated and will be removed from a future version of G++.
22092 G++ allows floating-point literals to appear in integral constant expressions,
22093 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
22094 This extension is deprecated and will be removed from a future version.
22096 G++ allows static data members of const floating-point type to be declared
22097 with an initializer in a class definition. The standard only allows
22098 initializers for static members of const integral types and const
22099 enumeration types so this extension has been deprecated and will be removed
22100 from a future version.
22102 @node Backwards Compatibility
22103 @section Backwards Compatibility
22104 @cindex Backwards Compatibility
22105 @cindex ARM [Annotated C++ Reference Manual]
22107 Now that there is a definitive ISO standard C++, G++ has a specification
22108 to adhere to. The C++ language evolved over time, and features that
22109 used to be acceptable in previous drafts of the standard, such as the ARM
22110 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
22111 compilation of C++ written to such drafts, G++ contains some backwards
22112 compatibilities. @emph{All such backwards compatibility features are
22113 liable to disappear in future versions of G++.} They should be considered
22114 deprecated. @xref{Deprecated Features}.
22118 If a variable is declared at for scope, it used to remain in scope until
22119 the end of the scope that contained the for statement (rather than just
22120 within the for scope). G++ retains this, but issues a warning, if such a
22121 variable is accessed outside the for scope.
22123 @item Implicit C language
22124 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
22125 scope to set the language. On such systems, all header files are
22126 implicitly scoped inside a C language scope. Also, an empty prototype
22127 @code{()} is treated as an unspecified number of arguments, rather
22128 than no arguments, as C++ demands.
22131 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
22132 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr