1 c Copyright (C) 1988-2018 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 * Statement Attributes:: Specifying attributes on statements.
64 * Attribute Syntax:: Formal syntax for attributes.
65 * Function Prototypes:: Prototype declarations and old-style definitions.
66 * C++ Comments:: C++ comments are recognized.
67 * Dollar Signs:: Dollar sign is allowed in identifiers.
68 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
69 * Alignment:: Inquiring about the alignment of a type or variable.
70 * Inline:: Defining inline functions (as fast as macros).
71 * Volatiles:: What constitutes an access to a volatile object.
72 * Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
73 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
74 * Incomplete Enums:: @code{enum foo;}, with details to follow.
75 * Function Names:: Printable strings which are the name of the current
77 * Return Address:: Getting the return or frame address of a function.
78 * Vector Extensions:: Using vector instructions through built-in functions.
79 * Offsetof:: Special syntax for implementing @code{offsetof}.
80 * __sync Builtins:: Legacy built-in functions for atomic memory access.
81 * __atomic Builtins:: Atomic built-in functions with memory model.
82 * Integer Overflow Builtins:: Built-in functions to perform arithmetics and
83 arithmetic overflow checking.
84 * x86 specific memory model extensions for transactional memory:: x86 memory models.
85 * Object Size Checking:: Built-in functions for limited buffer overflow
87 * Pointer Bounds Checker builtins:: Built-in functions for Pointer Bounds Checker.
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 The ISO C++14 library also defines the @samp{i} suffix, so C++14 code
901 that includes the @samp{<complex>} header cannot use @samp{i} for the
902 GNU extension. The @samp{j} suffix still has the GNU meaning.
904 @cindex @code{__real__} keyword
905 @cindex @code{__imag__} keyword
906 To extract the real part of a complex-valued expression @var{exp}, write
907 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
908 extract the imaginary part. This is a GNU extension; for values of
909 floating type, you should use the ISO C99 functions @code{crealf},
910 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
911 @code{cimagl}, declared in @code{<complex.h>} and also provided as
912 built-in functions by GCC@.
914 @cindex complex conjugation
915 The operator @samp{~} performs complex conjugation when used on a value
916 with a complex type. This is a GNU extension; for values of
917 floating type, you should use the ISO C99 functions @code{conjf},
918 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
919 provided as built-in functions by GCC@.
921 GCC can allocate complex automatic variables in a noncontiguous
922 fashion; it's even possible for the real part to be in a register while
923 the imaginary part is on the stack (or vice versa). Only the DWARF
924 debug info format can represent this, so use of DWARF is recommended.
925 If you are using the stabs debug info format, GCC describes a noncontiguous
926 complex variable as if it were two separate variables of noncomplex type.
927 If the variable's actual name is @code{foo}, the two fictitious
928 variables are named @code{foo$real} and @code{foo$imag}. You can
929 examine and set these two fictitious variables with your debugger.
932 @section Additional Floating Types
933 @cindex additional floating types
934 @cindex @code{_Float@var{n}} data types
935 @cindex @code{_Float@var{n}x} data types
936 @cindex @code{__float80} data type
937 @cindex @code{__float128} data type
938 @cindex @code{__ibm128} data type
939 @cindex @code{w} floating point suffix
940 @cindex @code{q} floating point suffix
941 @cindex @code{W} floating point suffix
942 @cindex @code{Q} floating point suffix
944 ISO/IEC TS 18661-3:2015 defines C support for additional floating
945 types @code{_Float@var{n}} and @code{_Float@var{n}x}, and GCC supports
946 these type names; the set of types supported depends on the target
947 architecture. These types are not supported when compiling C++.
948 Constants with these types use suffixes @code{f@var{n}} or
949 @code{F@var{n}} and @code{f@var{n}x} or @code{F@var{n}x}. These type
950 names can be used together with @code{_Complex} to declare complex
953 As an extension, GNU C and GNU C++ support additional floating
954 types, which are not supported by all targets.
956 @item @code{__float128} is available on i386, x86_64, IA-64, and
957 hppa HP-UX, as well as on PowerPC GNU/Linux targets that enable
958 the vector scalar (VSX) instruction set. @code{__float128} supports
959 the 128-bit floating type. On i386, x86_64, PowerPC, and IA-64
960 other than HP-UX, @code{__float128} is an alias for @code{_Float128}.
961 On hppa and IA-64 HP-UX, @code{__float128} is an alias for @code{long
964 @item @code{__float80} is available on the i386, x86_64, and IA-64
965 targets, and supports the 80-bit (@code{XFmode}) floating type. It is
966 an alias for the type name @code{_Float64x} on these targets.
968 @item @code{__ibm128} is available on PowerPC targets, and provides
969 access to the IBM extended double format which is the current format
970 used for @code{long double}. When @code{long double} transitions to
971 @code{__float128} on PowerPC in the future, @code{__ibm128} will remain
972 for use in conversions between the two types.
975 Support for these additional types includes the arithmetic operators:
976 add, subtract, multiply, divide; unary arithmetic operators;
977 relational operators; equality operators; and conversions to and from
978 integer and other floating types. Use a suffix @samp{w} or @samp{W}
979 in a literal constant of type @code{__float80} or type
980 @code{__ibm128}. Use a suffix @samp{q} or @samp{Q} for @code{_float128}.
982 In order to use @code{_Float128}, @code{__float128}, and @code{__ibm128}
983 on PowerPC Linux systems, you must use the @option{-mfloat128} option. It is
984 expected in future versions of GCC that @code{_Float128} and @code{__float128}
985 will be enabled automatically.
987 The @code{_Float128} type is supported on all systems where
988 @code{__float128} is supported or where @code{long double} has the
989 IEEE binary128 format. The @code{_Float64x} type is supported on all
990 systems where @code{__float128} is supported. The @code{_Float32}
991 type is supported on all systems supporting IEEE binary32; the
992 @code{_Float64} and @code{_Float32x} types are supported on all systems
993 supporting IEEE binary64. The @code{_Float16} type is supported on AArch64
994 systems by default, and on ARM systems when the IEEE format for 16-bit
995 floating-point types is selected with @option{-mfp16-format=ieee}.
996 GCC does not currently support @code{_Float128x} on any systems.
998 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
999 types using the corresponding internal complex type, @code{XCmode} for
1000 @code{__float80} type and @code{TCmode} for @code{__float128} type:
1003 typedef _Complex float __attribute__((mode(TC))) _Complex128;
1004 typedef _Complex float __attribute__((mode(XC))) _Complex80;
1007 On the PowerPC Linux VSX targets, you can declare complex types using
1008 the corresponding internal complex type, @code{KCmode} for
1009 @code{__float128} type and @code{ICmode} for @code{__ibm128} type:
1012 typedef _Complex float __attribute__((mode(KC))) _Complex_float128;
1013 typedef _Complex float __attribute__((mode(IC))) _Complex_ibm128;
1016 @node Half-Precision
1017 @section Half-Precision Floating Point
1018 @cindex half-precision floating point
1019 @cindex @code{__fp16} data type
1021 On ARM and AArch64 targets, GCC supports half-precision (16-bit) floating
1022 point via the @code{__fp16} type defined in the ARM C Language Extensions.
1023 On ARM systems, you must enable this type explicitly with the
1024 @option{-mfp16-format} command-line option in order to use it.
1026 ARM targets support two incompatible representations for half-precision
1027 floating-point values. You must choose one of the representations and
1028 use it consistently in your program.
1030 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
1031 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
1032 There are 11 bits of significand precision, approximately 3
1035 Specifying @option{-mfp16-format=alternative} selects the ARM
1036 alternative format. This representation is similar to the IEEE
1037 format, but does not support infinities or NaNs. Instead, the range
1038 of exponents is extended, so that this format can represent normalized
1039 values in the range of @math{2^{-14}} to 131008.
1041 The GCC port for AArch64 only supports the IEEE 754-2008 format, and does
1042 not require use of the @option{-mfp16-format} command-line option.
1044 The @code{__fp16} type may only be used as an argument to intrinsics defined
1045 in @code{<arm_fp16.h>}, or as a storage format. For purposes of
1046 arithmetic and other operations, @code{__fp16} values in C or C++
1047 expressions are automatically promoted to @code{float}.
1049 The ARM target provides hardware support for conversions between
1050 @code{__fp16} and @code{float} values
1051 as an extension to VFP and NEON (Advanced SIMD), and from ARMv8-A provides
1052 hardware support for conversions between @code{__fp16} and @code{double}
1053 values. GCC generates code using these hardware instructions if you
1054 compile with options to select an FPU that provides them;
1055 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
1056 in addition to the @option{-mfp16-format} option to select
1057 a half-precision format.
1059 Language-level support for the @code{__fp16} data type is
1060 independent of whether GCC generates code using hardware floating-point
1061 instructions. In cases where hardware support is not specified, GCC
1062 implements conversions between @code{__fp16} and other types as library
1065 It is recommended that portable code use the @code{_Float16} type defined
1066 by ISO/IEC TS 18661-3:2015. @xref{Floating Types}.
1069 @section Decimal Floating Types
1070 @cindex decimal floating types
1071 @cindex @code{_Decimal32} data type
1072 @cindex @code{_Decimal64} data type
1073 @cindex @code{_Decimal128} data type
1074 @cindex @code{df} integer suffix
1075 @cindex @code{dd} integer suffix
1076 @cindex @code{dl} integer suffix
1077 @cindex @code{DF} integer suffix
1078 @cindex @code{DD} integer suffix
1079 @cindex @code{DL} integer suffix
1081 As an extension, GNU C supports decimal floating types as
1082 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1083 floating types in GCC will evolve as the draft technical report changes.
1084 Calling conventions for any target might also change. Not all targets
1085 support decimal floating types.
1087 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1088 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1089 @code{float}, @code{double}, and @code{long double} whose radix is not
1090 specified by the C standard but is usually two.
1092 Support for decimal floating types includes the arithmetic operators
1093 add, subtract, multiply, divide; unary arithmetic operators;
1094 relational operators; equality operators; and conversions to and from
1095 integer and other floating types. Use a suffix @samp{df} or
1096 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1097 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1100 GCC support of decimal float as specified by the draft technical report
1105 When the value of a decimal floating type cannot be represented in the
1106 integer type to which it is being converted, the result is undefined
1107 rather than the result value specified by the draft technical report.
1110 GCC does not provide the C library functionality associated with
1111 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1112 @file{wchar.h}, which must come from a separate C library implementation.
1113 Because of this the GNU C compiler does not define macro
1114 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1115 the technical report.
1118 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1119 are supported by the DWARF debug information format.
1125 ISO C99 supports floating-point numbers written not only in the usual
1126 decimal notation, such as @code{1.55e1}, but also numbers such as
1127 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1128 supports this in C90 mode (except in some cases when strictly
1129 conforming) and in C++. In that format the
1130 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1131 mandatory. The exponent is a decimal number that indicates the power of
1132 2 by which the significant part is multiplied. Thus @samp{0x1.f} is
1139 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1140 is the same as @code{1.55e1}.
1142 Unlike for floating-point numbers in the decimal notation the exponent
1143 is always required in the hexadecimal notation. Otherwise the compiler
1144 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1145 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1146 extension for floating-point constants of type @code{float}.
1149 @section Fixed-Point Types
1150 @cindex fixed-point types
1151 @cindex @code{_Fract} data type
1152 @cindex @code{_Accum} data type
1153 @cindex @code{_Sat} data type
1154 @cindex @code{hr} fixed-suffix
1155 @cindex @code{r} fixed-suffix
1156 @cindex @code{lr} fixed-suffix
1157 @cindex @code{llr} fixed-suffix
1158 @cindex @code{uhr} fixed-suffix
1159 @cindex @code{ur} fixed-suffix
1160 @cindex @code{ulr} fixed-suffix
1161 @cindex @code{ullr} fixed-suffix
1162 @cindex @code{hk} fixed-suffix
1163 @cindex @code{k} fixed-suffix
1164 @cindex @code{lk} fixed-suffix
1165 @cindex @code{llk} fixed-suffix
1166 @cindex @code{uhk} fixed-suffix
1167 @cindex @code{uk} fixed-suffix
1168 @cindex @code{ulk} fixed-suffix
1169 @cindex @code{ullk} fixed-suffix
1170 @cindex @code{HR} fixed-suffix
1171 @cindex @code{R} fixed-suffix
1172 @cindex @code{LR} fixed-suffix
1173 @cindex @code{LLR} fixed-suffix
1174 @cindex @code{UHR} fixed-suffix
1175 @cindex @code{UR} fixed-suffix
1176 @cindex @code{ULR} fixed-suffix
1177 @cindex @code{ULLR} fixed-suffix
1178 @cindex @code{HK} fixed-suffix
1179 @cindex @code{K} fixed-suffix
1180 @cindex @code{LK} fixed-suffix
1181 @cindex @code{LLK} fixed-suffix
1182 @cindex @code{UHK} fixed-suffix
1183 @cindex @code{UK} fixed-suffix
1184 @cindex @code{ULK} fixed-suffix
1185 @cindex @code{ULLK} fixed-suffix
1187 As an extension, GNU C supports fixed-point types as
1188 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1189 types in GCC will evolve as the draft technical report changes.
1190 Calling conventions for any target might also change. Not all targets
1191 support fixed-point types.
1193 The fixed-point types are
1194 @code{short _Fract},
1197 @code{long long _Fract},
1198 @code{unsigned short _Fract},
1199 @code{unsigned _Fract},
1200 @code{unsigned long _Fract},
1201 @code{unsigned long long _Fract},
1202 @code{_Sat short _Fract},
1204 @code{_Sat long _Fract},
1205 @code{_Sat long long _Fract},
1206 @code{_Sat unsigned short _Fract},
1207 @code{_Sat unsigned _Fract},
1208 @code{_Sat unsigned long _Fract},
1209 @code{_Sat unsigned long long _Fract},
1210 @code{short _Accum},
1213 @code{long long _Accum},
1214 @code{unsigned short _Accum},
1215 @code{unsigned _Accum},
1216 @code{unsigned long _Accum},
1217 @code{unsigned long long _Accum},
1218 @code{_Sat short _Accum},
1220 @code{_Sat long _Accum},
1221 @code{_Sat long long _Accum},
1222 @code{_Sat unsigned short _Accum},
1223 @code{_Sat unsigned _Accum},
1224 @code{_Sat unsigned long _Accum},
1225 @code{_Sat unsigned long long _Accum}.
1227 Fixed-point data values contain fractional and optional integral parts.
1228 The format of fixed-point data varies and depends on the target machine.
1230 Support for fixed-point types includes:
1233 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1235 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1237 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1239 binary shift operators (@code{<<}, @code{>>})
1241 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1243 equality operators (@code{==}, @code{!=})
1245 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1246 @code{<<=}, @code{>>=})
1248 conversions to and from integer, floating-point, or fixed-point types
1251 Use a suffix in a fixed-point literal constant:
1253 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1254 @code{_Sat short _Fract}
1255 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1256 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1257 @code{_Sat long _Fract}
1258 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1259 @code{_Sat long long _Fract}
1260 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1261 @code{_Sat unsigned short _Fract}
1262 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1263 @code{_Sat unsigned _Fract}
1264 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1265 @code{_Sat unsigned long _Fract}
1266 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1267 and @code{_Sat unsigned long long _Fract}
1268 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1269 @code{_Sat short _Accum}
1270 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1271 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1272 @code{_Sat long _Accum}
1273 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1274 @code{_Sat long long _Accum}
1275 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1276 @code{_Sat unsigned short _Accum}
1277 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1278 @code{_Sat unsigned _Accum}
1279 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1280 @code{_Sat unsigned long _Accum}
1281 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1282 and @code{_Sat unsigned long long _Accum}
1285 GCC support of fixed-point types as specified by the draft technical report
1290 Pragmas to control overflow and rounding behaviors are not implemented.
1293 Fixed-point types are supported by the DWARF debug information format.
1295 @node Named Address Spaces
1296 @section Named Address Spaces
1297 @cindex Named Address Spaces
1299 As an extension, GNU C supports named address spaces as
1300 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1301 address spaces in GCC will evolve as the draft technical report
1302 changes. Calling conventions for any target might also change. At
1303 present, only the AVR, SPU, M32C, RL78, and x86 targets support
1304 address spaces other than the generic address space.
1306 Address space identifiers may be used exactly like any other C type
1307 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1308 document for more details.
1310 @anchor{AVR Named Address Spaces}
1311 @subsection AVR Named Address Spaces
1313 On the AVR target, there are several address spaces that can be used
1314 in order to put read-only data into the flash memory and access that
1315 data by means of the special instructions @code{LPM} or @code{ELPM}
1316 needed to read from flash.
1318 Devices belonging to @code{avrtiny} and @code{avrxmega3} can access
1319 flash memory by means of @code{LD*} instructions because the flash
1320 memory is mapped into the RAM address space. There is @emph{no need}
1321 for language extensions like @code{__flash} or attribute
1322 @ref{AVR Variable Attributes,,@code{progmem}}.
1323 The default linker description files for these devices cater for that
1324 feature and @code{.rodata} stays in flash: The compiler just generates
1325 @code{LD*} instructions, and the linker script adds core specific
1326 offsets to all @code{.rodata} symbols: @code{0x4000} in the case of
1327 @code{avrtiny} and @code{0x8000} in the case of @code{avrxmega3}.
1328 See @ref{AVR Options} for a list of respective devices.
1330 For devices not in @code{avrtiny} or @code{avrxmega3},
1331 any data including read-only data is located in RAM (the generic
1332 address space) because flash memory is not visible in the RAM address
1333 space. In order to locate read-only data in flash memory @emph{and}
1334 to generate the right instructions to access this data without
1335 using (inline) assembler code, special address spaces are needed.
1339 @cindex @code{__flash} AVR Named Address Spaces
1340 The @code{__flash} qualifier locates data in the
1341 @code{.progmem.data} section. Data is read using the @code{LPM}
1342 instruction. Pointers to this address space are 16 bits wide.
1349 @cindex @code{__flash1} AVR Named Address Spaces
1350 @cindex @code{__flash2} AVR Named Address Spaces
1351 @cindex @code{__flash3} AVR Named Address Spaces
1352 @cindex @code{__flash4} AVR Named Address Spaces
1353 @cindex @code{__flash5} AVR Named Address Spaces
1354 These are 16-bit address spaces locating data in section
1355 @code{.progmem@var{N}.data} where @var{N} refers to
1356 address space @code{__flash@var{N}}.
1357 The compiler sets the @code{RAMPZ} segment register appropriately
1358 before reading data by means of the @code{ELPM} instruction.
1361 @cindex @code{__memx} AVR Named Address Spaces
1362 This is a 24-bit address space that linearizes flash and RAM:
1363 If the high bit of the address is set, data is read from
1364 RAM using the lower two bytes as RAM address.
1365 If the high bit of the address is clear, data is read from flash
1366 with @code{RAMPZ} set according to the high byte of the address.
1367 @xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.
1369 Objects in this address space are located in @code{.progmemx.data}.
1375 char my_read (const __flash char ** p)
1377 /* p is a pointer to RAM that points to a pointer to flash.
1378 The first indirection of p reads that flash pointer
1379 from RAM and the second indirection reads a char from this
1385 /* Locate array[] in flash memory */
1386 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1392 /* Return 17 by reading from flash memory */
1393 return array[array[i]];
1398 For each named address space supported by avr-gcc there is an equally
1399 named but uppercase built-in macro defined.
1400 The purpose is to facilitate testing if respective address space
1401 support is available or not:
1405 const __flash int var = 1;
1412 #include <avr/pgmspace.h> /* From AVR-LibC */
1414 const int var PROGMEM = 1;
1418 return (int) pgm_read_word (&var);
1420 #endif /* __FLASH */
1424 Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
1425 locates data in flash but
1426 accesses to these data read from generic address space, i.e.@:
1428 so that you need special accessors like @code{pgm_read_byte}
1429 from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
1430 together with attribute @code{progmem}.
1433 @b{Limitations and caveats}
1437 Reading across the 64@tie{}KiB section boundary of
1438 the @code{__flash} or @code{__flash@var{N}} address spaces
1439 shows undefined behavior. The only address space that
1440 supports reading across the 64@tie{}KiB flash segment boundaries is
1444 If you use one of the @code{__flash@var{N}} address spaces
1445 you must arrange your linker script to locate the
1446 @code{.progmem@var{N}.data} sections according to your needs.
1449 Any data or pointers to the non-generic address spaces must
1450 be qualified as @code{const}, i.e.@: as read-only data.
1451 This still applies if the data in one of these address
1452 spaces like software version number or calibration lookup table are intended to
1453 be changed after load time by, say, a boot loader. In this case
1454 the right qualification is @code{const} @code{volatile} so that the compiler
1455 must not optimize away known values or insert them
1456 as immediates into operands of instructions.
1459 The following code initializes a variable @code{pfoo}
1460 located in static storage with a 24-bit address:
1462 extern const __memx char foo;
1463 const __memx void *pfoo = &foo;
1467 On the reduced Tiny devices like ATtiny40, no address spaces are supported.
1468 Just use vanilla C / C++ code without overhead as outlined above.
1469 Attribute @code{progmem} is supported but works differently,
1470 see @ref{AVR Variable Attributes}.
1474 @subsection M32C Named Address Spaces
1475 @cindex @code{__far} M32C Named Address Spaces
1477 On the M32C target, with the R8C and M16C CPU variants, variables
1478 qualified with @code{__far} are accessed using 32-bit addresses in
1479 order to access memory beyond the first 64@tie{}Ki bytes. If
1480 @code{__far} is used with the M32CM or M32C CPU variants, it has no
1483 @subsection RL78 Named Address Spaces
1484 @cindex @code{__far} RL78 Named Address Spaces
1486 On the RL78 target, variables qualified with @code{__far} are accessed
1487 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1488 addresses. Non-far variables are assumed to appear in the topmost
1489 64@tie{}KiB of the address space.
1491 @subsection SPU Named Address Spaces
1492 @cindex @code{__ea} SPU Named Address Spaces
1494 On the SPU target variables may be declared as
1495 belonging to another address space by qualifying the type with the
1496 @code{__ea} address space identifier:
1503 The compiler generates special code to access the variable @code{i}.
1504 It may use runtime library
1505 support, or generate special machine instructions to access that address
1508 @subsection x86 Named Address Spaces
1509 @cindex x86 named address spaces
1511 On the x86 target, variables may be declared as being relative
1512 to the @code{%fs} or @code{%gs} segments.
1517 @cindex @code{__seg_fs} x86 named address space
1518 @cindex @code{__seg_gs} x86 named address space
1519 The object is accessed with the respective segment override prefix.
1521 The respective segment base must be set via some method specific to
1522 the operating system. Rather than require an expensive system call
1523 to retrieve the segment base, these address spaces are not considered
1524 to be subspaces of the generic (flat) address space. This means that
1525 explicit casts are required to convert pointers between these address
1526 spaces and the generic address space. In practice the application
1527 should cast to @code{uintptr_t} and apply the segment base offset
1528 that it installed previously.
1530 The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
1531 defined when these address spaces are supported.
1535 @section Arrays of Length Zero
1536 @cindex arrays of length zero
1537 @cindex zero-length arrays
1538 @cindex length-zero arrays
1539 @cindex flexible array members
1541 Zero-length arrays are allowed in GNU C@. They are very useful as the
1542 last element of a structure that is really a header for a variable-length
1551 struct line *thisline = (struct line *)
1552 malloc (sizeof (struct line) + this_length);
1553 thisline->length = this_length;
1556 In ISO C90, you would have to give @code{contents} a length of 1, which
1557 means either you waste space or complicate the argument to @code{malloc}.
1559 In ISO C99, you would use a @dfn{flexible array member}, which is
1560 slightly different in syntax and semantics:
1564 Flexible array members are written as @code{contents[]} without
1568 Flexible array members have incomplete type, and so the @code{sizeof}
1569 operator may not be applied. As a quirk of the original implementation
1570 of zero-length arrays, @code{sizeof} evaluates to zero.
1573 Flexible array members may only appear as the last member of a
1574 @code{struct} that is otherwise non-empty.
1577 A structure containing a flexible array member, or a union containing
1578 such a structure (possibly recursively), may not be a member of a
1579 structure or an element of an array. (However, these uses are
1580 permitted by GCC as extensions.)
1583 Non-empty initialization of zero-length
1584 arrays is treated like any case where there are more initializer
1585 elements than the array holds, in that a suitable warning about ``excess
1586 elements in array'' is given, and the excess elements (all of them, in
1587 this case) are ignored.
1589 GCC allows static initialization of flexible array members.
1590 This is equivalent to defining a new structure containing the original
1591 structure followed by an array of sufficient size to contain the data.
1592 E.g.@: in the following, @code{f1} is constructed as if it were declared
1598 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1601 struct f1 f1; int data[3];
1602 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1606 The convenience of this extension is that @code{f1} has the desired
1607 type, eliminating the need to consistently refer to @code{f2.f1}.
1609 This has symmetry with normal static arrays, in that an array of
1610 unknown size is also written with @code{[]}.
1612 Of course, this extension only makes sense if the extra data comes at
1613 the end of a top-level object, as otherwise we would be overwriting
1614 data at subsequent offsets. To avoid undue complication and confusion
1615 with initialization of deeply nested arrays, we simply disallow any
1616 non-empty initialization except when the structure is the top-level
1617 object. For example:
1620 struct foo @{ int x; int y[]; @};
1621 struct bar @{ struct foo z; @};
1623 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1624 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1625 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1626 struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1629 @node Empty Structures
1630 @section Structures with No Members
1631 @cindex empty structures
1632 @cindex zero-size structures
1634 GCC permits a C structure to have no members:
1641 The structure has size zero. In C++, empty structures are part
1642 of the language. G++ treats empty structures as if they had a single
1643 member of type @code{char}.
1645 @node Variable Length
1646 @section Arrays of Variable Length
1647 @cindex variable-length arrays
1648 @cindex arrays of variable length
1651 Variable-length automatic arrays are allowed in ISO C99, and as an
1652 extension GCC accepts them in C90 mode and in C++. These arrays are
1653 declared like any other automatic arrays, but with a length that is not
1654 a constant expression. The storage is allocated at the point of
1655 declaration and deallocated when the block scope containing the declaration
1661 concat_fopen (char *s1, char *s2, char *mode)
1663 char str[strlen (s1) + strlen (s2) + 1];
1666 return fopen (str, mode);
1670 @cindex scope of a variable length array
1671 @cindex variable-length array scope
1672 @cindex deallocating variable length arrays
1673 Jumping or breaking out of the scope of the array name deallocates the
1674 storage. Jumping into the scope is not allowed; you get an error
1677 @cindex variable-length array in a structure
1678 As an extension, GCC accepts variable-length arrays as a member of
1679 a structure or a union. For example:
1685 struct S @{ int x[n]; @};
1689 @cindex @code{alloca} vs variable-length arrays
1690 You can use the function @code{alloca} to get an effect much like
1691 variable-length arrays. The function @code{alloca} is available in
1692 many other C implementations (but not in all). On the other hand,
1693 variable-length arrays are more elegant.
1695 There are other differences between these two methods. Space allocated
1696 with @code{alloca} exists until the containing @emph{function} returns.
1697 The space for a variable-length array is deallocated as soon as the array
1698 name's scope ends, unless you also use @code{alloca} in this scope.
1700 You can also use variable-length arrays as arguments to functions:
1704 tester (int len, char data[len][len])
1710 The length of an array is computed once when the storage is allocated
1711 and is remembered for the scope of the array in case you access it with
1714 If you want to pass the array first and the length afterward, you can
1715 use a forward declaration in the parameter list---another GNU extension.
1719 tester (int len; char data[len][len], int len)
1725 @cindex parameter forward declaration
1726 The @samp{int len} before the semicolon is a @dfn{parameter forward
1727 declaration}, and it serves the purpose of making the name @code{len}
1728 known when the declaration of @code{data} is parsed.
1730 You can write any number of such parameter forward declarations in the
1731 parameter list. They can be separated by commas or semicolons, but the
1732 last one must end with a semicolon, which is followed by the ``real''
1733 parameter declarations. Each forward declaration must match a ``real''
1734 declaration in parameter name and data type. ISO C99 does not support
1735 parameter forward declarations.
1737 @node Variadic Macros
1738 @section Macros with a Variable Number of Arguments.
1739 @cindex variable number of arguments
1740 @cindex macro with variable arguments
1741 @cindex rest argument (in macro)
1742 @cindex variadic macros
1744 In the ISO C standard of 1999, a macro can be declared to accept a
1745 variable number of arguments much as a function can. The syntax for
1746 defining the macro is similar to that of a function. Here is an
1750 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1754 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1755 such a macro, it represents the zero or more tokens until the closing
1756 parenthesis that ends the invocation, including any commas. This set of
1757 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1758 wherever it appears. See the CPP manual for more information.
1760 GCC has long supported variadic macros, and used a different syntax that
1761 allowed you to give a name to the variable arguments just like any other
1762 argument. Here is an example:
1765 #define debug(format, args...) fprintf (stderr, format, args)
1769 This is in all ways equivalent to the ISO C example above, but arguably
1770 more readable and descriptive.
1772 GNU CPP has two further variadic macro extensions, and permits them to
1773 be used with either of the above forms of macro definition.
1775 In standard C, you are not allowed to leave the variable argument out
1776 entirely; but you are allowed to pass an empty argument. For example,
1777 this invocation is invalid in ISO C, because there is no comma after
1784 GNU CPP permits you to completely omit the variable arguments in this
1785 way. In the above examples, the compiler would complain, though since
1786 the expansion of the macro still has the extra comma after the format
1789 To help solve this problem, CPP behaves specially for variable arguments
1790 used with the token paste operator, @samp{##}. If instead you write
1793 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1797 and if the variable arguments are omitted or empty, the @samp{##}
1798 operator causes the preprocessor to remove the comma before it. If you
1799 do provide some variable arguments in your macro invocation, GNU CPP
1800 does not complain about the paste operation and instead places the
1801 variable arguments after the comma. Just like any other pasted macro
1802 argument, these arguments are not macro expanded.
1804 @node Escaped Newlines
1805 @section Slightly Looser Rules for Escaped Newlines
1806 @cindex escaped newlines
1807 @cindex newlines (escaped)
1809 The preprocessor treatment of escaped newlines is more relaxed
1810 than that specified by the C90 standard, which requires the newline
1811 to immediately follow a backslash.
1812 GCC's implementation allows whitespace in the form
1813 of spaces, horizontal and vertical tabs, and form feeds between the
1814 backslash and the subsequent newline. The preprocessor issues a
1815 warning, but treats it as a valid escaped newline and combines the two
1816 lines to form a single logical line. This works within comments and
1817 tokens, as well as between tokens. Comments are @emph{not} treated as
1818 whitespace for the purposes of this relaxation, since they have not
1819 yet been replaced with spaces.
1822 @section Non-Lvalue Arrays May Have Subscripts
1823 @cindex subscripting
1824 @cindex arrays, non-lvalue
1826 @cindex subscripting and function values
1827 In ISO C99, arrays that are not lvalues still decay to pointers, and
1828 may be subscripted, although they may not be modified or used after
1829 the next sequence point and the unary @samp{&} operator may not be
1830 applied to them. As an extension, GNU C allows such arrays to be
1831 subscripted in C90 mode, though otherwise they do not decay to
1832 pointers outside C99 mode. For example,
1833 this is valid in GNU C though not valid in C90:
1837 struct foo @{int a[4];@};
1843 return f().a[index];
1849 @section Arithmetic on @code{void}- and Function-Pointers
1850 @cindex void pointers, arithmetic
1851 @cindex void, size of pointer to
1852 @cindex function pointers, arithmetic
1853 @cindex function, size of pointer to
1855 In GNU C, addition and subtraction operations are supported on pointers to
1856 @code{void} and on pointers to functions. This is done by treating the
1857 size of a @code{void} or of a function as 1.
1859 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1860 and on function types, and returns 1.
1862 @opindex Wpointer-arith
1863 The option @option{-Wpointer-arith} requests a warning if these extensions
1866 @node Pointers to Arrays
1867 @section Pointers to Arrays with Qualifiers Work as Expected
1868 @cindex pointers to arrays
1869 @cindex const qualifier
1871 In GNU C, pointers to arrays with qualifiers work similar to pointers
1872 to other qualified types. For example, a value of type @code{int (*)[5]}
1873 can be used to initialize a variable of type @code{const int (*)[5]}.
1874 These types are incompatible in ISO C because the @code{const} qualifier
1875 is formally attached to the element type of the array and not the
1880 transpose (int N, int M, double out[M][N], const double in[N][M]);
1884 transpose(3, 2, y, x);
1888 @section Non-Constant Initializers
1889 @cindex initializers, non-constant
1890 @cindex non-constant initializers
1892 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1893 automatic variable are not required to be constant expressions in GNU C@.
1894 Here is an example of an initializer with run-time varying elements:
1897 foo (float f, float g)
1899 float beat_freqs[2] = @{ f-g, f+g @};
1904 @node Compound Literals
1905 @section Compound Literals
1906 @cindex constructor expressions
1907 @cindex initializations in expressions
1908 @cindex structures, constructor expression
1909 @cindex expressions, constructor
1910 @cindex compound literals
1911 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1913 A compound literal looks like a cast of a brace-enclosed aggregate
1914 initializer list. Its value is an object of the type specified in
1915 the cast, containing the elements specified in the initializer.
1916 Unlike the result of a cast, a compound literal is an lvalue. ISO
1917 C99 and later support compound literals. As an extension, GCC
1918 supports compound literals also in C90 mode and in C++, although
1919 as explained below, the C++ semantics are somewhat different.
1921 Usually, the specified type of a compound literal is a structure. Assume
1922 that @code{struct foo} and @code{structure} are declared as shown:
1925 struct foo @{int a; char b[2];@} structure;
1929 Here is an example of constructing a @code{struct foo} with a compound literal:
1932 structure = ((struct foo) @{x + y, 'a', 0@});
1936 This is equivalent to writing the following:
1940 struct foo temp = @{x + y, 'a', 0@};
1945 You can also construct an array, though this is dangerous in C++, as
1946 explained below. If all the elements of the compound literal are
1947 (made up of) simple constant expressions suitable for use in
1948 initializers of objects of static storage duration, then the compound
1949 literal can be coerced to a pointer to its first element and used in
1950 such an initializer, as shown here:
1953 char **foo = (char *[]) @{ "x", "y", "z" @};
1956 Compound literals for scalar types and union types are also allowed. In
1957 the following example the variable @code{i} is initialized to the value
1958 @code{2}, the result of incrementing the unnamed object created by
1959 the compound literal.
1962 int i = ++(int) @{ 1 @};
1965 As a GNU extension, GCC allows initialization of objects with static storage
1966 duration by compound literals (which is not possible in ISO C99 because
1967 the initializer is not a constant).
1968 It is handled as if the object were initialized only with the brace-enclosed
1969 list if the types of the compound literal and the object match.
1970 The elements of the compound literal must be constant.
1971 If the object being initialized has array type of unknown size, the size is
1972 determined by the size of the compound literal.
1975 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1976 static int y[] = (int []) @{1, 2, 3@};
1977 static int z[] = (int [3]) @{1@};
1981 The above lines are equivalent to the following:
1983 static struct foo x = @{1, 'a', 'b'@};
1984 static int y[] = @{1, 2, 3@};
1985 static int z[] = @{1, 0, 0@};
1988 In C, a compound literal designates an unnamed object with static or
1989 automatic storage duration. In C++, a compound literal designates a
1990 temporary object that only lives until the end of its full-expression.
1991 As a result, well-defined C code that takes the address of a subobject
1992 of a compound literal can be undefined in C++, so G++ rejects
1993 the conversion of a temporary array to a pointer. For instance, if
1994 the array compound literal example above appeared inside a function,
1995 any subsequent use of @code{foo} in C++ would have undefined behavior
1996 because the lifetime of the array ends after the declaration of @code{foo}.
1998 As an optimization, G++ sometimes gives array compound literals longer
1999 lifetimes: when the array either appears outside a function or has
2000 a @code{const}-qualified type. If @code{foo} and its initializer had
2001 elements of type @code{char *const} rather than @code{char *}, or if
2002 @code{foo} were a global variable, the array would have static storage
2003 duration. But it is probably safest just to avoid the use of array
2004 compound literals in C++ code.
2006 @node Designated Inits
2007 @section Designated Initializers
2008 @cindex initializers with labeled elements
2009 @cindex labeled elements in initializers
2010 @cindex case labels in initializers
2011 @cindex designated initializers
2013 Standard C90 requires the elements of an initializer to appear in a fixed
2014 order, the same as the order of the elements in the array or structure
2017 In ISO C99 you can give the elements in any order, specifying the array
2018 indices or structure field names they apply to, and GNU C allows this as
2019 an extension in C90 mode as well. This extension is not
2020 implemented in GNU C++.
2022 To specify an array index, write
2023 @samp{[@var{index}] =} before the element value. For example,
2026 int a[6] = @{ [4] = 29, [2] = 15 @};
2033 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
2037 The index values must be constant expressions, even if the array being
2038 initialized is automatic.
2040 An alternative syntax for this that has been obsolete since GCC 2.5 but
2041 GCC still accepts is to write @samp{[@var{index}]} before the element
2042 value, with no @samp{=}.
2044 To initialize a range of elements to the same value, write
2045 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
2046 extension. For example,
2049 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
2053 If the value in it has side-effects, the side-effects happen only once,
2054 not for each initialized field by the range initializer.
2057 Note that the length of the array is the highest value specified
2060 In a structure initializer, specify the name of a field to initialize
2061 with @samp{.@var{fieldname} =} before the element value. For example,
2062 given the following structure,
2065 struct point @{ int x, y; @};
2069 the following initialization
2072 struct point p = @{ .y = yvalue, .x = xvalue @};
2079 struct point p = @{ xvalue, yvalue @};
2082 Another syntax that has the same meaning, obsolete since GCC 2.5, is
2083 @samp{@var{fieldname}:}, as shown here:
2086 struct point p = @{ y: yvalue, x: xvalue @};
2089 Omitted field members are implicitly initialized the same as objects
2090 that have static storage duration.
2093 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
2094 @dfn{designator}. You can also use a designator (or the obsolete colon
2095 syntax) when initializing a union, to specify which element of the union
2096 should be used. For example,
2099 union foo @{ int i; double d; @};
2101 union foo f = @{ .d = 4 @};
2105 converts 4 to a @code{double} to store it in the union using
2106 the second element. By contrast, casting 4 to type @code{union foo}
2107 stores it into the union as the integer @code{i}, since it is
2108 an integer. @xref{Cast to Union}.
2110 You can combine this technique of naming elements with ordinary C
2111 initialization of successive elements. Each initializer element that
2112 does not have a designator applies to the next consecutive element of the
2113 array or structure. For example,
2116 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
2123 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
2126 Labeling the elements of an array initializer is especially useful
2127 when the indices are characters or belong to an @code{enum} type.
2132 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
2133 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
2136 @cindex designator lists
2137 You can also write a series of @samp{.@var{fieldname}} and
2138 @samp{[@var{index}]} designators before an @samp{=} to specify a
2139 nested subobject to initialize; the list is taken relative to the
2140 subobject corresponding to the closest surrounding brace pair. For
2141 example, with the @samp{struct point} declaration above:
2144 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
2148 If the same field is initialized multiple times, it has the value from
2149 the last initialization. If any such overridden initialization has
2150 side-effect, it is unspecified whether the side-effect happens or not.
2151 Currently, GCC discards them and issues a warning.
2154 @section Case Ranges
2156 @cindex ranges in case statements
2158 You can specify a range of consecutive values in a single @code{case} label,
2162 case @var{low} ... @var{high}:
2166 This has the same effect as the proper number of individual @code{case}
2167 labels, one for each integer value from @var{low} to @var{high}, inclusive.
2169 This feature is especially useful for ranges of ASCII character codes:
2175 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
2176 it may be parsed wrong when you use it with integer values. For example,
2191 @section Cast to a Union Type
2192 @cindex cast to a union
2193 @cindex union, casting to a
2195 A cast to union type looks similar to other casts, except that the type
2196 specified is a union type. You can specify the type either with the
2197 @code{union} keyword or with a @code{typedef} name that refers to
2198 a union. A cast to a union actually creates a compound literal and
2199 yields an lvalue, not an rvalue like true casts do.
2200 @xref{Compound Literals}.
2202 The types that may be cast to the union type are those of the members
2203 of the union. Thus, given the following union and variables:
2206 union foo @{ int i; double d; @};
2212 both @code{x} and @code{y} can be cast to type @code{union foo}.
2214 Using the cast as the right-hand side of an assignment to a variable of
2215 union type is equivalent to storing in a member of the union:
2220 u = (union foo) x @equiv{} u.i = x
2221 u = (union foo) y @equiv{} u.d = y
2224 You can also use the union cast as a function argument:
2227 void hack (union foo);
2229 hack ((union foo) x);
2232 @node Mixed Declarations
2233 @section Mixed Declarations and Code
2234 @cindex mixed declarations and code
2235 @cindex declarations, mixed with code
2236 @cindex code, mixed with declarations
2238 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2239 within compound statements. As an extension, GNU C also allows this in
2240 C90 mode. For example, you could do:
2249 Each identifier is visible from where it is declared until the end of
2250 the enclosing block.
2252 @node Function Attributes
2253 @section Declaring Attributes of Functions
2254 @cindex function attributes
2255 @cindex declaring attributes of functions
2256 @cindex @code{volatile} applied to function
2257 @cindex @code{const} applied to function
2259 In GNU C, you can use function attributes to declare certain things
2260 about functions called in your program which help the compiler
2261 optimize calls and check your code more carefully. For example, you
2262 can use attributes to declare that a function never returns
2263 (@code{noreturn}), returns a value depending only on its arguments
2264 (@code{pure}), or has @code{printf}-style arguments (@code{format}).
2266 You can also use attributes to control memory placement, code
2267 generation options or call/return conventions within the function
2268 being annotated. Many of these attributes are target-specific. For
2269 example, many targets support attributes for defining interrupt
2270 handler functions, which typically must follow special register usage
2271 and return conventions.
2273 Function attributes are introduced by the @code{__attribute__} keyword
2274 on a declaration, followed by an attribute specification inside double
2275 parentheses. You can specify multiple attributes in a declaration by
2276 separating them by commas within the double parentheses or by
2277 immediately following an attribute declaration with another attribute
2278 declaration. @xref{Attribute Syntax}, for the exact rules on
2279 attribute syntax and placement.
2281 GCC also supports attributes on
2282 variable declarations (@pxref{Variable Attributes}),
2283 labels (@pxref{Label Attributes}),
2284 enumerators (@pxref{Enumerator Attributes}),
2285 statements (@pxref{Statement Attributes}),
2286 and types (@pxref{Type Attributes}).
2288 There is some overlap between the purposes of attributes and pragmas
2289 (@pxref{Pragmas,,Pragmas Accepted by GCC}). It has been
2290 found convenient to use @code{__attribute__} to achieve a natural
2291 attachment of attributes to their corresponding declarations, whereas
2292 @code{#pragma} is of use for compatibility with other compilers
2293 or constructs that do not naturally form part of the grammar.
2295 In addition to the attributes documented here,
2296 GCC plugins may provide their own attributes.
2299 * Common Function Attributes::
2300 * AArch64 Function Attributes::
2301 * ARC Function Attributes::
2302 * ARM Function Attributes::
2303 * AVR Function Attributes::
2304 * Blackfin Function Attributes::
2305 * CR16 Function Attributes::
2306 * Epiphany Function Attributes::
2307 * H8/300 Function Attributes::
2308 * IA-64 Function Attributes::
2309 * M32C Function Attributes::
2310 * M32R/D Function Attributes::
2311 * m68k Function Attributes::
2312 * MCORE Function Attributes::
2313 * MeP Function Attributes::
2314 * MicroBlaze Function Attributes::
2315 * Microsoft Windows Function Attributes::
2316 * MIPS Function Attributes::
2317 * MSP430 Function Attributes::
2318 * NDS32 Function Attributes::
2319 * Nios II Function Attributes::
2320 * Nvidia PTX Function Attributes::
2321 * PowerPC Function Attributes::
2322 * RISC-V Function Attributes::
2323 * RL78 Function Attributes::
2324 * RX Function Attributes::
2325 * S/390 Function Attributes::
2326 * SH Function Attributes::
2327 * SPU Function Attributes::
2328 * Symbian OS Function Attributes::
2329 * V850 Function Attributes::
2330 * Visium Function Attributes::
2331 * x86 Function Attributes::
2332 * Xstormy16 Function Attributes::
2335 @node Common Function Attributes
2336 @subsection Common Function Attributes
2338 The following attributes are supported on most targets.
2341 @c Keep this table alphabetized by attribute name. Treat _ as space.
2343 @item alias ("@var{target}")
2344 @cindex @code{alias} function attribute
2345 The @code{alias} attribute causes the declaration to be emitted as an
2346 alias for another symbol, which must be specified. For instance,
2349 void __f () @{ /* @r{Do something.} */; @}
2350 void f () __attribute__ ((weak, alias ("__f")));
2354 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2355 mangled name for the target must be used. It is an error if @samp{__f}
2356 is not defined in the same translation unit.
2358 This attribute requires assembler and object file support,
2359 and may not be available on all targets.
2361 @item aligned (@var{alignment})
2362 @cindex @code{aligned} function attribute
2363 This attribute specifies a minimum alignment for the function,
2366 You cannot use this attribute to decrease the alignment of a function,
2367 only to increase it. However, when you explicitly specify a function
2368 alignment this overrides the effect of the
2369 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2372 Note that the effectiveness of @code{aligned} attributes may be
2373 limited by inherent limitations in your linker. On many systems, the
2374 linker is only able to arrange for functions to be aligned up to a
2375 certain maximum alignment. (For some linkers, the maximum supported
2376 alignment may be very very small.) See your linker documentation for
2377 further information.
2379 The @code{aligned} attribute can also be used for variables and fields
2380 (@pxref{Variable Attributes}.)
2383 @cindex @code{alloc_align} function attribute
2384 The @code{alloc_align} attribute is used to tell the compiler that the
2385 function return value points to memory, where the returned pointer minimum
2386 alignment is given by one of the functions parameters. GCC uses this
2387 information to improve pointer alignment analysis.
2389 The function parameter denoting the allocated alignment is specified by
2390 one integer argument, whose number is the argument of the attribute.
2391 Argument numbering starts at one.
2396 void* my_memalign(size_t, size_t) __attribute__((alloc_align(1)))
2400 declares that @code{my_memalign} returns memory with minimum alignment
2401 given by parameter 1.
2404 @cindex @code{alloc_size} function attribute
2405 The @code{alloc_size} attribute is used to tell the compiler that the
2406 function return value points to memory, where the size is given by
2407 one or two of the functions parameters. GCC uses this
2408 information to improve the correctness of @code{__builtin_object_size}.
2410 The function parameter(s) denoting the allocated size are specified by
2411 one or two integer arguments supplied to the attribute. The allocated size
2412 is either the value of the single function argument specified or the product
2413 of the two function arguments specified. Argument numbering starts at
2419 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2420 void* my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2424 declares that @code{my_calloc} returns memory of the size given by
2425 the product of parameter 1 and 2 and that @code{my_realloc} returns memory
2426 of the size given by parameter 2.
2429 @cindex @code{always_inline} function attribute
2430 Generally, functions are not inlined unless optimization is specified.
2431 For functions declared inline, this attribute inlines the function
2432 independent of any restrictions that otherwise apply to inlining.
2433 Failure to inline such a function is diagnosed as an error.
2434 Note that if such a function is called indirectly the compiler may
2435 or may not inline it depending on optimization level and a failure
2436 to inline an indirect call may or may not be diagnosed.
2439 @cindex @code{artificial} function attribute
2440 This attribute is useful for small inline wrappers that if possible
2441 should appear during debugging as a unit. Depending on the debug
2442 info format it either means marking the function as artificial
2443 or using the caller location for all instructions within the inlined
2446 @item assume_aligned
2447 @cindex @code{assume_aligned} function attribute
2448 The @code{assume_aligned} attribute is used to tell the compiler that the
2449 function return value points to memory, where the returned pointer minimum
2450 alignment is given by the first argument.
2451 If the attribute has two arguments, the second argument is misalignment offset.
2456 void* my_alloc1(size_t) __attribute__((assume_aligned(16)))
2457 void* my_alloc2(size_t) __attribute__((assume_aligned(32, 8)))
2461 declares that @code{my_alloc1} returns 16-byte aligned pointer and
2462 that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
2465 @item bnd_instrument
2466 @cindex @code{bnd_instrument} function attribute
2467 The @code{bnd_instrument} attribute on functions is used to inform the
2468 compiler that the function should be instrumented when compiled
2469 with the @option{-fchkp-instrument-marked-only} option.
2472 @cindex @code{bnd_legacy} function attribute
2473 @cindex Pointer Bounds Checker attributes
2474 The @code{bnd_legacy} attribute on functions is used to inform the
2475 compiler that the function should not be instrumented when compiled
2476 with the @option{-fcheck-pointer-bounds} option.
2479 @cindex @code{cold} function attribute
2480 The @code{cold} attribute on functions is used to inform the compiler that
2481 the function is unlikely to be executed. The function is optimized for
2482 size rather than speed and on many targets it is placed into a special
2483 subsection of the text section so all cold functions appear close together,
2484 improving code locality of non-cold parts of program. The paths leading
2485 to calls of cold functions within code are marked as unlikely by the branch
2486 prediction mechanism. It is thus useful to mark functions used to handle
2487 unlikely conditions, such as @code{perror}, as cold to improve optimization
2488 of hot functions that do call marked functions in rare occasions.
2490 When profile feedback is available, via @option{-fprofile-use}, cold functions
2491 are automatically detected and this attribute is ignored.
2494 @cindex @code{const} function attribute
2495 @cindex functions that have no side effects
2496 Many functions do not examine any values except their arguments, and
2497 have no effects except to return a value. Calls to such functions lend
2498 themselves to optimization such as common subexpression elimination.
2499 The @code{const} attribute imposes greater restrictions on a function's
2500 definition than the similar @code{pure} attribute below because it prohibits
2501 the function from reading global variables. Consequently, the presence of
2502 the attribute on a function declarations allows GCC to emit more efficient
2503 code for some calls to the function. Decorating the same function with
2504 both the @code{const} and the @code{pure} attribute is diagnosed.
2506 @cindex pointer arguments
2507 Note that a function that has pointer arguments and examines the data
2508 pointed to must @emph{not} be declared @code{const}. Likewise, a
2509 function that calls a non-@code{const} function usually must not be
2510 @code{const}. It does not make sense for a @code{const} function to
2515 @itemx constructor (@var{priority})
2516 @itemx destructor (@var{priority})
2517 @cindex @code{constructor} function attribute
2518 @cindex @code{destructor} function attribute
2519 The @code{constructor} attribute causes the function to be called
2520 automatically before execution enters @code{main ()}. Similarly, the
2521 @code{destructor} attribute causes the function to be called
2522 automatically after @code{main ()} completes or @code{exit ()} is
2523 called. Functions with these attributes are useful for
2524 initializing data that is used implicitly during the execution of
2527 You may provide an optional integer priority to control the order in
2528 which constructor and destructor functions are run. A constructor
2529 with a smaller priority number runs before a constructor with a larger
2530 priority number; the opposite relationship holds for destructors. So,
2531 if you have a constructor that allocates a resource and a destructor
2532 that deallocates the same resource, both functions typically have the
2533 same priority. The priorities for constructor and destructor
2534 functions are the same as those specified for namespace-scope C++
2535 objects (@pxref{C++ Attributes}). However, at present, the order in which
2536 constructors for C++ objects with static storage duration and functions
2537 decorated with attribute @code{constructor} are invoked is unspecified.
2538 In mixed declarations, attribute @code{init_priority} can be used to
2539 impose a specific ordering.
2542 @itemx deprecated (@var{msg})
2543 @cindex @code{deprecated} function attribute
2544 The @code{deprecated} attribute results in a warning if the function
2545 is used anywhere in the source file. This is useful when identifying
2546 functions that are expected to be removed in a future version of a
2547 program. The warning also includes the location of the declaration
2548 of the deprecated function, to enable users to easily find further
2549 information about why the function is deprecated, or what they should
2550 do instead. Note that the warnings only occurs for uses:
2553 int old_fn () __attribute__ ((deprecated));
2555 int (*fn_ptr)() = old_fn;
2559 results in a warning on line 3 but not line 2. The optional @var{msg}
2560 argument, which must be a string, is printed in the warning if
2563 The @code{deprecated} attribute can also be used for variables and
2564 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2566 @item error ("@var{message}")
2567 @itemx warning ("@var{message}")
2568 @cindex @code{error} function attribute
2569 @cindex @code{warning} function attribute
2570 If the @code{error} or @code{warning} attribute
2571 is used on a function declaration and a call to such a function
2572 is not eliminated through dead code elimination or other optimizations,
2573 an error or warning (respectively) that includes @var{message} is diagnosed.
2575 for compile-time checking, especially together with @code{__builtin_constant_p}
2576 and inline functions where checking the inline function arguments is not
2577 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2579 While it is possible to leave the function undefined and thus invoke
2580 a link failure (to define the function with
2581 a message in @code{.gnu.warning*} section),
2582 when using these attributes the problem is diagnosed
2583 earlier and with exact location of the call even in presence of inline
2584 functions or when not emitting debugging information.
2586 @item externally_visible
2587 @cindex @code{externally_visible} function attribute
2588 This attribute, attached to a global variable or function, nullifies
2589 the effect of the @option{-fwhole-program} command-line option, so the
2590 object remains visible outside the current compilation unit.
2592 If @option{-fwhole-program} is used together with @option{-flto} and
2593 @command{gold} is used as the linker plugin,
2594 @code{externally_visible} attributes are automatically added to functions
2595 (not variable yet due to a current @command{gold} issue)
2596 that are accessed outside of LTO objects according to resolution file
2597 produced by @command{gold}.
2598 For other linkers that cannot generate resolution file,
2599 explicit @code{externally_visible} attributes are still necessary.
2602 @cindex @code{flatten} function attribute
2603 Generally, inlining into a function is limited. For a function marked with
2604 this attribute, every call inside this function is inlined, if possible.
2605 Whether the function itself is considered for inlining depends on its size and
2606 the current inlining parameters.
2608 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2609 @cindex @code{format} function attribute
2610 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2612 The @code{format} attribute specifies that a function takes @code{printf},
2613 @code{scanf}, @code{strftime} or @code{strfmon} style arguments that
2614 should be type-checked against a format string. For example, the
2619 my_printf (void *my_object, const char *my_format, ...)
2620 __attribute__ ((format (printf, 2, 3)));
2624 causes the compiler to check the arguments in calls to @code{my_printf}
2625 for consistency with the @code{printf} style format string argument
2628 The parameter @var{archetype} determines how the format string is
2629 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2630 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2631 @code{strfmon}. (You can also use @code{__printf__},
2632 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2633 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2634 @code{ms_strftime} are also present.
2635 @var{archetype} values such as @code{printf} refer to the formats accepted
2636 by the system's C runtime library,
2637 while values prefixed with @samp{gnu_} always refer
2638 to the formats accepted by the GNU C Library. On Microsoft Windows
2639 targets, values prefixed with @samp{ms_} refer to the formats accepted by the
2640 @file{msvcrt.dll} library.
2641 The parameter @var{string-index}
2642 specifies which argument is the format string argument (starting
2643 from 1), while @var{first-to-check} is the number of the first
2644 argument to check against the format string. For functions
2645 where the arguments are not available to be checked (such as
2646 @code{vprintf}), specify the third parameter as zero. In this case the
2647 compiler only checks the format string for consistency. For
2648 @code{strftime} formats, the third parameter is required to be zero.
2649 Since non-static C++ methods have an implicit @code{this} argument, the
2650 arguments of such methods should be counted from two, not one, when
2651 giving values for @var{string-index} and @var{first-to-check}.
2653 In the example above, the format string (@code{my_format}) is the second
2654 argument of the function @code{my_print}, and the arguments to check
2655 start with the third argument, so the correct parameters for the format
2656 attribute are 2 and 3.
2658 @opindex ffreestanding
2659 @opindex fno-builtin
2660 The @code{format} attribute allows you to identify your own functions
2661 that take format strings as arguments, so that GCC can check the
2662 calls to these functions for errors. The compiler always (unless
2663 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2664 for the standard library functions @code{printf}, @code{fprintf},
2665 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2666 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2667 warnings are requested (using @option{-Wformat}), so there is no need to
2668 modify the header file @file{stdio.h}. In C99 mode, the functions
2669 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2670 @code{vsscanf} are also checked. Except in strictly conforming C
2671 standard modes, the X/Open function @code{strfmon} is also checked as
2672 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2673 @xref{C Dialect Options,,Options Controlling C Dialect}.
2675 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2676 recognized in the same context. Declarations including these format attributes
2677 are parsed for correct syntax, however the result of checking of such format
2678 strings is not yet defined, and is not carried out by this version of the
2681 The target may also provide additional types of format checks.
2682 @xref{Target Format Checks,,Format Checks Specific to Particular
2685 @item format_arg (@var{string-index})
2686 @cindex @code{format_arg} function attribute
2687 @opindex Wformat-nonliteral
2688 The @code{format_arg} attribute specifies that a function takes a format
2689 string for a @code{printf}, @code{scanf}, @code{strftime} or
2690 @code{strfmon} style function and modifies it (for example, to translate
2691 it into another language), so the result can be passed to a
2692 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2693 function (with the remaining arguments to the format function the same
2694 as they would have been for the unmodified string). For example, the
2699 my_dgettext (char *my_domain, const char *my_format)
2700 __attribute__ ((format_arg (2)));
2704 causes the compiler to check the arguments in calls to a @code{printf},
2705 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2706 format string argument is a call to the @code{my_dgettext} function, for
2707 consistency with the format string argument @code{my_format}. If the
2708 @code{format_arg} attribute had not been specified, all the compiler
2709 could tell in such calls to format functions would be that the format
2710 string argument is not constant; this would generate a warning when
2711 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2712 without the attribute.
2714 The parameter @var{string-index} specifies which argument is the format
2715 string argument (starting from one). Since non-static C++ methods have
2716 an implicit @code{this} argument, the arguments of such methods should
2717 be counted from two.
2719 The @code{format_arg} attribute allows you to identify your own
2720 functions that modify format strings, so that GCC can check the
2721 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2722 type function whose operands are a call to one of your own function.
2723 The compiler always treats @code{gettext}, @code{dgettext}, and
2724 @code{dcgettext} in this manner except when strict ISO C support is
2725 requested by @option{-ansi} or an appropriate @option{-std} option, or
2726 @option{-ffreestanding} or @option{-fno-builtin}
2727 is used. @xref{C Dialect Options,,Options
2728 Controlling C Dialect}.
2730 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2731 @code{NSString} reference for compatibility with the @code{format} attribute
2734 The target may also allow additional types in @code{format-arg} attributes.
2735 @xref{Target Format Checks,,Format Checks Specific to Particular
2739 @cindex @code{gnu_inline} function attribute
2740 This attribute should be used with a function that is also declared
2741 with the @code{inline} keyword. It directs GCC to treat the function
2742 as if it were defined in gnu90 mode even when compiling in C99 or
2745 If the function is declared @code{extern}, then this definition of the
2746 function is used only for inlining. In no case is the function
2747 compiled as a standalone function, not even if you take its address
2748 explicitly. Such an address becomes an external reference, as if you
2749 had only declared the function, and had not defined it. This has
2750 almost the effect of a macro. The way to use this is to put a
2751 function definition in a header file with this attribute, and put
2752 another copy of the function, without @code{extern}, in a library
2753 file. The definition in the header file causes most calls to the
2754 function to be inlined. If any uses of the function remain, they
2755 refer to the single copy in the library. Note that the two
2756 definitions of the functions need not be precisely the same, although
2757 if they do not have the same effect your program may behave oddly.
2759 In C, if the function is neither @code{extern} nor @code{static}, then
2760 the function is compiled as a standalone function, as well as being
2761 inlined where possible.
2763 This is how GCC traditionally handled functions declared
2764 @code{inline}. Since ISO C99 specifies a different semantics for
2765 @code{inline}, this function attribute is provided as a transition
2766 measure and as a useful feature in its own right. This attribute is
2767 available in GCC 4.1.3 and later. It is available if either of the
2768 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2769 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2770 Function is As Fast As a Macro}.
2772 In C++, this attribute does not depend on @code{extern} in any way,
2773 but it still requires the @code{inline} keyword to enable its special
2777 @cindex @code{hot} function attribute
2778 The @code{hot} attribute on a function is used to inform the compiler that
2779 the function is a hot spot of the compiled program. The function is
2780 optimized more aggressively and on many targets it is placed into a special
2781 subsection of the text section so all hot functions appear close together,
2784 When profile feedback is available, via @option{-fprofile-use}, hot functions
2785 are automatically detected and this attribute is ignored.
2787 @item ifunc ("@var{resolver}")
2788 @cindex @code{ifunc} function attribute
2789 @cindex indirect functions
2790 @cindex functions that are dynamically resolved
2791 The @code{ifunc} attribute is used to mark a function as an indirect
2792 function using the STT_GNU_IFUNC symbol type extension to the ELF
2793 standard. This allows the resolution of the symbol value to be
2794 determined dynamically at load time, and an optimized version of the
2795 routine to be selected for the particular processor or other system
2796 characteristics determined then. To use this attribute, first define
2797 the implementation functions available, and a resolver function that
2798 returns a pointer to the selected implementation function. The
2799 implementation functions' declarations must match the API of the
2800 function being implemented. The resolver should be declared to
2801 be a function taking no arguments and returning a pointer to
2802 a function of the same type as the implementation. For example:
2805 void *my_memcpy (void *dst, const void *src, size_t len)
2811 static void * (*resolve_memcpy (void))(void *, const void *, size_t)
2813 return my_memcpy; // we will just always select this routine
2818 The exported header file declaring the function the user calls would
2822 extern void *memcpy (void *, const void *, size_t);
2826 allowing the user to call @code{memcpy} as a regular function, unaware of
2827 the actual implementation. Finally, the indirect function needs to be
2828 defined in the same translation unit as the resolver function:
2831 void *memcpy (void *, const void *, size_t)
2832 __attribute__ ((ifunc ("resolve_memcpy")));
2835 In C++, the @code{ifunc} attribute takes a string that is the mangled name
2836 of the resolver function. A C++ resolver for a non-static member function
2837 of class @code{C} should be declared to return a pointer to a non-member
2838 function taking pointer to @code{C} as the first argument, followed by
2839 the same arguments as of the implementation function. G++ checks
2840 the signatures of the two functions and issues
2841 a @option{-Wattribute-alias} warning for mismatches. To suppress a warning
2842 for the necessary cast from a pointer to the implementation member function
2843 to the type of the corresponding non-member function use
2844 the @option{-Wno-pmf-conversions} option. For example:
2850 int debug_impl (int);
2851 int optimized_impl (int);
2853 typedef int Func (S*, int);
2855 static Func* resolver ();
2858 int interface (int);
2861 int S::debug_impl (int) @{ /* @r{@dots{}} */ @}
2862 int S::optimized_impl (int) @{ /* @r{@dots{}} */ @}
2864 S::Func* S::resolver ()
2866 int (S::*pimpl) (int)
2867 = getenv ("DEBUG") ? &S::debug_impl : &S::optimized_impl;
2869 // Cast triggers -Wno-pmf-conversions.
2870 return reinterpret_cast<Func*>(pimpl);
2873 int S::interface (int) __attribute__ ((ifunc ("_ZN1S8resolverEv")));
2876 Indirect functions cannot be weak. Binutils version 2.20.1 or higher
2877 and GNU C Library version 2.11.1 are required to use this feature.
2880 @itemx interrupt_handler
2881 Many GCC back ends support attributes to indicate that a function is
2882 an interrupt handler, which tells the compiler to generate function
2883 entry and exit sequences that differ from those from regular
2884 functions. The exact syntax and behavior are target-specific;
2885 refer to the following subsections for details.
2888 @cindex @code{leaf} function attribute
2889 Calls to external functions with this attribute must return to the
2890 current compilation unit only by return or by exception handling. In
2891 particular, a leaf function is not allowed to invoke callback functions
2892 passed to it from the current compilation unit, directly call functions
2893 exported by the unit, or @code{longjmp} into the unit. Leaf functions
2894 might still call functions from other compilation units and thus they
2895 are not necessarily leaf in the sense that they contain no function
2898 The attribute is intended for library functions to improve dataflow
2899 analysis. The compiler takes the hint that any data not escaping the
2900 current compilation unit cannot be used or modified by the leaf
2901 function. For example, the @code{sin} function is a leaf function, but
2902 @code{qsort} is not.
2904 Note that leaf functions might indirectly run a signal handler defined
2905 in the current compilation unit that uses static variables. Similarly,
2906 when lazy symbol resolution is in effect, leaf functions might invoke
2907 indirect functions whose resolver function or implementation function is
2908 defined in the current compilation unit and uses static variables. There
2909 is no standard-compliant way to write such a signal handler, resolver
2910 function, or implementation function, and the best that you can do is to
2911 remove the @code{leaf} attribute or mark all such static variables
2912 @code{volatile}. Lastly, for ELF-based systems that support symbol
2913 interposition, care should be taken that functions defined in the
2914 current compilation unit do not unexpectedly interpose other symbols
2915 based on the defined standards mode and defined feature test macros;
2916 otherwise an inadvertent callback would be added.
2918 The attribute has no effect on functions defined within the current
2919 compilation unit. This is to allow easy merging of multiple compilation
2920 units into one, for example, by using the link-time optimization. For
2921 this reason the attribute is not allowed on types to annotate indirect
2925 @cindex @code{malloc} function attribute
2926 @cindex functions that behave like malloc
2927 This tells the compiler that a function is @code{malloc}-like, i.e.,
2928 that the pointer @var{P} returned by the function cannot alias any
2929 other pointer valid when the function returns, and moreover no
2930 pointers to valid objects occur in any storage addressed by @var{P}.
2932 Using this attribute can improve optimization. Functions like
2933 @code{malloc} and @code{calloc} have this property because they return
2934 a pointer to uninitialized or zeroed-out storage. However, functions
2935 like @code{realloc} do not have this property, as they can return a
2936 pointer to storage containing pointers.
2939 @cindex @code{no_icf} function attribute
2940 This function attribute prevents a functions from being merged with another
2941 semantically equivalent function.
2943 @item no_instrument_function
2944 @cindex @code{no_instrument_function} function attribute
2945 @opindex finstrument-functions
2946 If @option{-finstrument-functions} is given, profiling function calls are
2947 generated at entry and exit of most user-compiled functions.
2948 Functions with this attribute are not so instrumented.
2950 @item no_profile_instrument_function
2951 @cindex @code{no_profile_instrument_function} function attribute
2952 The @code{no_profile_instrument_function} attribute on functions is used
2953 to inform the compiler that it should not process any profile feedback based
2954 optimization code instrumentation.
2957 @cindex @code{no_reorder} function attribute
2958 Do not reorder functions or variables marked @code{no_reorder}
2959 against each other or top level assembler statements the executable.
2960 The actual order in the program will depend on the linker command
2961 line. Static variables marked like this are also not removed.
2962 This has a similar effect
2963 as the @option{-fno-toplevel-reorder} option, but only applies to the
2966 @item no_sanitize ("@var{sanitize_option}")
2967 @cindex @code{no_sanitize} function attribute
2968 The @code{no_sanitize} attribute on functions is used
2969 to inform the compiler that it should not do sanitization of all options
2970 mentioned in @var{sanitize_option}. A list of values acceptable by
2971 @option{-fsanitize} option can be provided.
2974 void __attribute__ ((no_sanitize ("alignment", "object-size")))
2975 f () @{ /* @r{Do something.} */; @}
2978 @item no_sanitize_address
2979 @itemx no_address_safety_analysis
2980 @cindex @code{no_sanitize_address} function attribute
2981 The @code{no_sanitize_address} attribute on functions is used
2982 to inform the compiler that it should not instrument memory accesses
2983 in the function when compiling with the @option{-fsanitize=address} option.
2984 The @code{no_address_safety_analysis} is a deprecated alias of the
2985 @code{no_sanitize_address} attribute, new code should use
2986 @code{no_sanitize_address}.
2988 @item no_sanitize_thread
2989 @cindex @code{no_sanitize_thread} function attribute
2990 The @code{no_sanitize_thread} attribute on functions is used
2991 to inform the compiler that it should not instrument memory accesses
2992 in the function when compiling with the @option{-fsanitize=thread} option.
2994 @item no_sanitize_undefined
2995 @cindex @code{no_sanitize_undefined} function attribute
2996 The @code{no_sanitize_undefined} attribute on functions is used
2997 to inform the compiler that it should not check for undefined behavior
2998 in the function when compiling with the @option{-fsanitize=undefined} option.
3000 @item no_split_stack
3001 @cindex @code{no_split_stack} function attribute
3002 @opindex fsplit-stack
3003 If @option{-fsplit-stack} is given, functions have a small
3004 prologue which decides whether to split the stack. Functions with the
3005 @code{no_split_stack} attribute do not have that prologue, and thus
3006 may run with only a small amount of stack space available.
3008 @item no_stack_limit
3009 @cindex @code{no_stack_limit} function attribute
3010 This attribute locally overrides the @option{-fstack-limit-register}
3011 and @option{-fstack-limit-symbol} command-line options; it has the effect
3012 of disabling stack limit checking in the function it applies to.
3015 @cindex @code{noclone} function attribute
3016 This function attribute prevents a function from being considered for
3017 cloning---a mechanism that produces specialized copies of functions
3018 and which is (currently) performed by interprocedural constant
3022 @cindex @code{noinline} function attribute
3023 This function attribute prevents a function from being considered for
3025 @c Don't enumerate the optimizations by name here; we try to be
3026 @c future-compatible with this mechanism.
3027 If the function does not have side-effects, there are optimizations
3028 other than inlining that cause function calls to be optimized away,
3029 although the function call is live. To keep such calls from being
3036 (@pxref{Extended Asm}) in the called function, to serve as a special
3040 @cindex @code{noipa} function attribute
3041 Disable interprocedural optimizations between the function with this
3042 attribute and its callers, as if the body of the function is not available
3043 when optimizing callers and the callers are unavailable when optimizing
3044 the body. This attribute implies @code{noinline}, @code{noclone} and
3045 @code{no_icf} attributes. However, this attribute is not equivalent
3046 to a combination of other attributes, because its purpose is to suppress
3047 existing and future optimizations employing interprocedural analysis,
3048 including those that do not have an attribute suitable for disabling
3049 them individually. This attribute is supported mainly for the purpose
3050 of testing the compiler.
3052 @item nonnull (@var{arg-index}, @dots{})
3053 @cindex @code{nonnull} function attribute
3054 @cindex functions with non-null pointer arguments
3055 The @code{nonnull} attribute specifies that some function parameters should
3056 be non-null pointers. For instance, the declaration:
3060 my_memcpy (void *dest, const void *src, size_t len)
3061 __attribute__((nonnull (1, 2)));
3065 causes the compiler to check that, in calls to @code{my_memcpy},
3066 arguments @var{dest} and @var{src} are non-null. If the compiler
3067 determines that a null pointer is passed in an argument slot marked
3068 as non-null, and the @option{-Wnonnull} option is enabled, a warning
3069 is issued. The compiler may also choose to make optimizations based
3070 on the knowledge that certain function arguments will never be null.
3072 If no argument index list is given to the @code{nonnull} attribute,
3073 all pointer arguments are marked as non-null. To illustrate, the
3074 following declaration is equivalent to the previous example:
3078 my_memcpy (void *dest, const void *src, size_t len)
3079 __attribute__((nonnull));
3083 @cindex @code{noplt} function attribute
3084 The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
3085 Calls to functions marked with this attribute in position-independent code
3090 /* Externally defined function foo. */
3091 int foo () __attribute__ ((noplt));
3094 main (/* @r{@dots{}} */)
3103 The @code{noplt} attribute on function @code{foo}
3104 tells the compiler to assume that
3105 the function @code{foo} is externally defined and that the call to
3106 @code{foo} must avoid the PLT
3107 in position-independent code.
3109 In position-dependent code, a few targets also convert calls to
3110 functions that are marked to not use the PLT to use the GOT instead.
3113 @cindex @code{noreturn} function attribute
3114 @cindex functions that never return
3115 A few standard library functions, such as @code{abort} and @code{exit},
3116 cannot return. GCC knows this automatically. Some programs define
3117 their own functions that never return. You can declare them
3118 @code{noreturn} to tell the compiler this fact. For example,
3122 void fatal () __attribute__ ((noreturn));
3125 fatal (/* @r{@dots{}} */)
3127 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3133 The @code{noreturn} keyword tells the compiler to assume that
3134 @code{fatal} cannot return. It can then optimize without regard to what
3135 would happen if @code{fatal} ever did return. This makes slightly
3136 better code. More importantly, it helps avoid spurious warnings of
3137 uninitialized variables.
3139 The @code{noreturn} keyword does not affect the exceptional path when that
3140 applies: a @code{noreturn}-marked function may still return to the caller
3141 by throwing an exception or calling @code{longjmp}.
3143 Do not assume that registers saved by the calling function are
3144 restored before calling the @code{noreturn} function.
3146 It does not make sense for a @code{noreturn} function to have a return
3147 type other than @code{void}.
3150 @cindex @code{nothrow} function attribute
3151 The @code{nothrow} attribute is used to inform the compiler that a
3152 function cannot throw an exception. For example, most functions in
3153 the standard C library can be guaranteed not to throw an exception
3154 with the notable exceptions of @code{qsort} and @code{bsearch} that
3155 take function pointer arguments.
3158 @cindex @code{optimize} function attribute
3159 The @code{optimize} attribute is used to specify that a function is to
3160 be compiled with different optimization options than specified on the
3161 command line. Arguments can either be numbers or strings. Numbers
3162 are assumed to be an optimization level. Strings that begin with
3163 @code{O} are assumed to be an optimization option, while other options
3164 are assumed to be used with a @code{-f} prefix. You can also use the
3165 @samp{#pragma GCC optimize} pragma to set the optimization options
3166 that affect more than one function.
3167 @xref{Function Specific Option Pragmas}, for details about the
3168 @samp{#pragma GCC optimize} pragma.
3170 This attribute should be used for debugging purposes only. It is not
3171 suitable in production code.
3173 @item patchable_function_entry
3174 @cindex @code{patchable_function_entry} function attribute
3175 @cindex extra NOP instructions at the function entry point
3176 In case the target's text segment can be made writable at run time by
3177 any means, padding the function entry with a number of NOPs can be
3178 used to provide a universal tool for instrumentation.
3180 The @code{patchable_function_entry} function attribute can be used to
3181 change the number of NOPs to any desired value. The two-value syntax
3182 is the same as for the command-line switch
3183 @option{-fpatchable-function-entry=N,M}, generating @var{N} NOPs, with
3184 the function entry point before the @var{M}th NOP instruction.
3185 @var{M} defaults to 0 if omitted e.g. function entry point is before
3188 If patchable function entries are enabled globally using the command-line
3189 option @option{-fpatchable-function-entry=N,M}, then you must disable
3190 instrumentation on all functions that are part of the instrumentation
3191 framework with the attribute @code{patchable_function_entry (0)}
3192 to prevent recursion.
3195 @cindex @code{pure} function attribute
3196 @cindex functions that have no side effects
3197 Many functions have no effects except the return value and their
3198 return value depends only on the parameters and/or global variables.
3199 Calls to such functions can be subject
3200 to common subexpression elimination and loop optimization just as an
3201 arithmetic operator would be. These functions should be declared
3202 with the attribute @code{pure}. For example,
3205 int square (int) __attribute__ ((pure));
3209 says that the hypothetical function @code{square} is safe to call
3210 fewer times than the program says.
3212 Some common examples of pure functions are @code{strlen} or @code{memcmp}.
3213 Interesting non-pure functions are functions with infinite loops or those
3214 depending on volatile memory or other system resource, that may change between
3215 two consecutive calls (such as @code{feof} in a multithreading environment).
3217 The @code{pure} attribute imposes similar but looser restrictions on
3218 a function's defintion than the @code{const} attribute: it allows the
3219 function to read global variables. Decorating the same function with
3220 both the @code{pure} and the @code{const} attribute is diagnosed.
3222 @item returns_nonnull
3223 @cindex @code{returns_nonnull} function attribute
3224 The @code{returns_nonnull} attribute specifies that the function
3225 return value should be a non-null pointer. For instance, the declaration:
3229 mymalloc (size_t len) __attribute__((returns_nonnull));
3233 lets the compiler optimize callers based on the knowledge
3234 that the return value will never be null.
3237 @cindex @code{returns_twice} function attribute
3238 @cindex functions that return more than once
3239 The @code{returns_twice} attribute tells the compiler that a function may
3240 return more than one time. The compiler ensures that all registers
3241 are dead before calling such a function and emits a warning about
3242 the variables that may be clobbered after the second return from the
3243 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3244 The @code{longjmp}-like counterpart of such function, if any, might need
3245 to be marked with the @code{noreturn} attribute.
3247 @item section ("@var{section-name}")
3248 @cindex @code{section} function attribute
3249 @cindex functions in arbitrary sections
3250 Normally, the compiler places the code it generates in the @code{text} section.
3251 Sometimes, however, you need additional sections, or you need certain
3252 particular functions to appear in special sections. The @code{section}
3253 attribute specifies that a function lives in a particular section.
3254 For example, the declaration:
3257 extern void foobar (void) __attribute__ ((section ("bar")));
3261 puts the function @code{foobar} in the @code{bar} section.
3263 Some file formats do not support arbitrary sections so the @code{section}
3264 attribute is not available on all platforms.
3265 If you need to map the entire contents of a module to a particular
3266 section, consider using the facilities of the linker instead.
3269 @cindex @code{sentinel} function attribute
3270 This function attribute ensures that a parameter in a function call is
3271 an explicit @code{NULL}. The attribute is only valid on variadic
3272 functions. By default, the sentinel is located at position zero, the
3273 last parameter of the function call. If an optional integer position
3274 argument P is supplied to the attribute, the sentinel must be located at
3275 position P counting backwards from the end of the argument list.
3278 __attribute__ ((sentinel))
3280 __attribute__ ((sentinel(0)))
3283 The attribute is automatically set with a position of 0 for the built-in
3284 functions @code{execl} and @code{execlp}. The built-in function
3285 @code{execle} has the attribute set with a position of 1.
3287 A valid @code{NULL} in this context is defined as zero with any pointer
3288 type. If your system defines the @code{NULL} macro with an integer type
3289 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3290 with a copy that redefines NULL appropriately.
3292 The warnings for missing or incorrect sentinels are enabled with
3296 @itemx simd("@var{mask}")
3297 @cindex @code{simd} function attribute
3298 This attribute enables creation of one or more function versions that
3299 can process multiple arguments using SIMD instructions from a
3300 single invocation. Specifying this attribute allows compiler to
3301 assume that such versions are available at link time (provided
3302 in the same or another translation unit). Generated versions are
3303 target-dependent and described in the corresponding Vector ABI document. For
3304 x86_64 target this document can be found
3305 @w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.
3307 The optional argument @var{mask} may have the value
3308 @code{notinbranch} or @code{inbranch},
3309 and instructs the compiler to generate non-masked or masked
3310 clones correspondingly. By default, all clones are generated.
3312 If the attribute is specified and @code{#pragma omp declare simd} is
3313 present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
3314 switch is specified, then the attribute is ignored.
3317 @cindex @code{stack_protect} function attribute
3318 This attribute adds stack protection code to the function if
3319 flags @option{-fstack-protector}, @option{-fstack-protector-strong}
3320 or @option{-fstack-protector-explicit} are set.
3322 @item target (@var{options})
3323 @cindex @code{target} function attribute
3324 Multiple target back ends implement the @code{target} attribute
3325 to specify that a function is to
3326 be compiled with different target options than specified on the
3327 command line. This can be used for instance to have functions
3328 compiled with a different ISA (instruction set architecture) than the
3329 default. You can also use the @samp{#pragma GCC target} pragma to set
3330 more than one function to be compiled with specific target options.
3331 @xref{Function Specific Option Pragmas}, for details about the
3332 @samp{#pragma GCC target} pragma.
3334 For instance, on an x86, you could declare one function with the
3335 @code{target("sse4.1,arch=core2")} attribute and another with
3336 @code{target("sse4a,arch=amdfam10")}. This is equivalent to
3337 compiling the first function with @option{-msse4.1} and
3338 @option{-march=core2} options, and the second function with
3339 @option{-msse4a} and @option{-march=amdfam10} options. It is up to you
3340 to make sure that a function is only invoked on a machine that
3341 supports the particular ISA it is compiled for (for example by using
3342 @code{cpuid} on x86 to determine what feature bits and architecture
3346 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3347 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3350 You can either use multiple
3351 strings separated by commas to specify multiple options,
3352 or separate the options with a comma (@samp{,}) within a single string.
3354 The options supported are specific to each target; refer to @ref{x86
3355 Function Attributes}, @ref{PowerPC Function Attributes},
3356 @ref{ARM Function Attributes}, @ref{AArch64 Function Attributes},
3357 @ref{Nios II Function Attributes}, and @ref{S/390 Function Attributes}
3360 @item target_clones (@var{options})
3361 @cindex @code{target_clones} function attribute
3362 The @code{target_clones} attribute is used to specify that a function
3363 be cloned into multiple versions compiled with different target options
3364 than specified on the command line. The supported options and restrictions
3365 are the same as for @code{target} attribute.
3367 For instance, on an x86, you could compile a function with
3368 @code{target_clones("sse4.1,avx")}. GCC creates two function clones,
3369 one compiled with @option{-msse4.1} and another with @option{-mavx}.
3371 On a PowerPC, you can compile a function with
3372 @code{target_clones("cpu=power9,default")}. GCC will create two
3373 function clones, one compiled with @option{-mcpu=power9} and another
3374 with the default options. GCC must be configured to use GLIBC 2.23 or
3375 newer in order to use the @code{target_clones} attribute.
3377 It also creates a resolver function (see
3378 the @code{ifunc} attribute above) that dynamically selects a clone
3379 suitable for current architecture. The resolver is created only if there
3380 is a usage of a function with @code{target_clones} attribute.
3383 @cindex @code{unused} function attribute
3384 This attribute, attached to a function, means that the function is meant
3385 to be possibly unused. GCC does not produce a warning for this
3389 @cindex @code{used} function attribute
3390 This attribute, attached to a function, means that code must be emitted
3391 for the function even if it appears that the function is not referenced.
3392 This is useful, for example, when the function is referenced only in
3395 When applied to a member function of a C++ class template, the
3396 attribute also means that the function is instantiated if the
3397 class itself is instantiated.
3399 @item visibility ("@var{visibility_type}")
3400 @cindex @code{visibility} function attribute
3401 This attribute affects the linkage of the declaration to which it is attached.
3402 It can be applied to variables (@pxref{Common Variable Attributes}) and types
3403 (@pxref{Common Type Attributes}) as well as functions.
3405 There are four supported @var{visibility_type} values: default,
3406 hidden, protected or internal visibility.
3409 void __attribute__ ((visibility ("protected")))
3410 f () @{ /* @r{Do something.} */; @}
3411 int i __attribute__ ((visibility ("hidden")));
3414 The possible values of @var{visibility_type} correspond to the
3415 visibility settings in the ELF gABI.
3418 @c keep this list of visibilities in alphabetical order.
3421 Default visibility is the normal case for the object file format.
3422 This value is available for the visibility attribute to override other
3423 options that may change the assumed visibility of entities.
3425 On ELF, default visibility means that the declaration is visible to other
3426 modules and, in shared libraries, means that the declared entity may be
3429 On Darwin, default visibility means that the declaration is visible to
3432 Default visibility corresponds to ``external linkage'' in the language.
3435 Hidden visibility indicates that the entity declared has a new
3436 form of linkage, which we call ``hidden linkage''. Two
3437 declarations of an object with hidden linkage refer to the same object
3438 if they are in the same shared object.
3441 Internal visibility is like hidden visibility, but with additional
3442 processor specific semantics. Unless otherwise specified by the
3443 psABI, GCC defines internal visibility to mean that a function is
3444 @emph{never} called from another module. Compare this with hidden
3445 functions which, while they cannot be referenced directly by other
3446 modules, can be referenced indirectly via function pointers. By
3447 indicating that a function cannot be called from outside the module,
3448 GCC may for instance omit the load of a PIC register since it is known
3449 that the calling function loaded the correct value.
3452 Protected visibility is like default visibility except that it
3453 indicates that references within the defining module bind to the
3454 definition in that module. That is, the declared entity cannot be
3455 overridden by another module.
3459 All visibilities are supported on many, but not all, ELF targets
3460 (supported when the assembler supports the @samp{.visibility}
3461 pseudo-op). Default visibility is supported everywhere. Hidden
3462 visibility is supported on Darwin targets.
3464 The visibility attribute should be applied only to declarations that
3465 would otherwise have external linkage. The attribute should be applied
3466 consistently, so that the same entity should not be declared with
3467 different settings of the attribute.
3469 In C++, the visibility attribute applies to types as well as functions
3470 and objects, because in C++ types have linkage. A class must not have
3471 greater visibility than its non-static data member types and bases,
3472 and class members default to the visibility of their class. Also, a
3473 declaration without explicit visibility is limited to the visibility
3476 In C++, you can mark member functions and static member variables of a
3477 class with the visibility attribute. This is useful if you know a
3478 particular method or static member variable should only be used from
3479 one shared object; then you can mark it hidden while the rest of the
3480 class has default visibility. Care must be taken to avoid breaking
3481 the One Definition Rule; for example, it is usually not useful to mark
3482 an inline method as hidden without marking the whole class as hidden.
3484 A C++ namespace declaration can also have the visibility attribute.
3487 namespace nspace1 __attribute__ ((visibility ("protected")))
3488 @{ /* @r{Do something.} */; @}
3491 This attribute applies only to the particular namespace body, not to
3492 other definitions of the same namespace; it is equivalent to using
3493 @samp{#pragma GCC visibility} before and after the namespace
3494 definition (@pxref{Visibility Pragmas}).
3496 In C++, if a template argument has limited visibility, this
3497 restriction is implicitly propagated to the template instantiation.
3498 Otherwise, template instantiations and specializations default to the
3499 visibility of their template.
3501 If both the template and enclosing class have explicit visibility, the
3502 visibility from the template is used.
3504 @item warn_unused_result
3505 @cindex @code{warn_unused_result} function attribute
3506 The @code{warn_unused_result} attribute causes a warning to be emitted
3507 if a caller of the function with this attribute does not use its
3508 return value. This is useful for functions where not checking
3509 the result is either a security problem or always a bug, such as
3513 int fn () __attribute__ ((warn_unused_result));
3516 if (fn () < 0) return -1;
3523 results in warning on line 5.
3526 @cindex @code{weak} function attribute
3527 The @code{weak} attribute causes the declaration to be emitted as a weak
3528 symbol rather than a global. This is primarily useful in defining
3529 library functions that can be overridden in user code, though it can
3530 also be used with non-function declarations. Weak symbols are supported
3531 for ELF targets, and also for a.out targets when using the GNU assembler
3535 @itemx weakref ("@var{target}")
3536 @cindex @code{weakref} function attribute
3537 The @code{weakref} attribute marks a declaration as a weak reference.
3538 Without arguments, it should be accompanied by an @code{alias} attribute
3539 naming the target symbol. Optionally, the @var{target} may be given as
3540 an argument to @code{weakref} itself. In either case, @code{weakref}
3541 implicitly marks the declaration as @code{weak}. Without a
3542 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3543 @code{weakref} is equivalent to @code{weak}.
3546 static int x() __attribute__ ((weakref ("y")));
3547 /* is equivalent to... */
3548 static int x() __attribute__ ((weak, weakref, alias ("y")));
3550 static int x() __attribute__ ((weakref));
3551 static int x() __attribute__ ((alias ("y")));
3554 A weak reference is an alias that does not by itself require a
3555 definition to be given for the target symbol. If the target symbol is
3556 only referenced through weak references, then it becomes a @code{weak}
3557 undefined symbol. If it is directly referenced, however, then such
3558 strong references prevail, and a definition is required for the
3559 symbol, not necessarily in the same translation unit.
3561 The effect is equivalent to moving all references to the alias to a
3562 separate translation unit, renaming the alias to the aliased symbol,
3563 declaring it as weak, compiling the two separate translation units and
3564 performing a reloadable link on them.
3566 At present, a declaration to which @code{weakref} is attached can
3567 only be @code{static}.
3572 @c This is the end of the target-independent attribute table
3574 @node AArch64 Function Attributes
3575 @subsection AArch64 Function Attributes
3577 The following target-specific function attributes are available for the
3578 AArch64 target. For the most part, these options mirror the behavior of
3579 similar command-line options (@pxref{AArch64 Options}), but on a
3583 @item general-regs-only
3584 @cindex @code{general-regs-only} function attribute, AArch64
3585 Indicates that no floating-point or Advanced SIMD registers should be
3586 used when generating code for this function. If the function explicitly
3587 uses floating-point code, then the compiler gives an error. This is
3588 the same behavior as that of the command-line option
3589 @option{-mgeneral-regs-only}.
3591 @item fix-cortex-a53-835769
3592 @cindex @code{fix-cortex-a53-835769} function attribute, AArch64
3593 Indicates that the workaround for the Cortex-A53 erratum 835769 should be
3594 applied to this function. To explicitly disable the workaround for this
3595 function specify the negated form: @code{no-fix-cortex-a53-835769}.
3596 This corresponds to the behavior of the command line options
3597 @option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.
3600 @cindex @code{cmodel=} function attribute, AArch64
3601 Indicates that code should be generated for a particular code model for
3602 this function. The behavior and permissible arguments are the same as
3603 for the command line option @option{-mcmodel=}.
3606 @cindex @code{strict-align} function attribute, AArch64
3607 Indicates that the compiler should not assume that unaligned memory references
3608 are handled by the system. The behavior is the same as for the command-line
3609 option @option{-mstrict-align}.
3611 @item omit-leaf-frame-pointer
3612 @cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
3613 Indicates that the frame pointer should be omitted for a leaf function call.
3614 To keep the frame pointer, the inverse attribute
3615 @code{no-omit-leaf-frame-pointer} can be specified. These attributes have
3616 the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
3617 and @option{-mno-omit-leaf-frame-pointer}.
3620 @cindex @code{tls-dialect=} function attribute, AArch64
3621 Specifies the TLS dialect to use for this function. The behavior and
3622 permissible arguments are the same as for the command-line option
3623 @option{-mtls-dialect=}.
3626 @cindex @code{arch=} function attribute, AArch64
3627 Specifies the architecture version and architectural extensions to use
3628 for this function. The behavior and permissible arguments are the same as
3629 for the @option{-march=} command-line option.
3632 @cindex @code{tune=} function attribute, AArch64
3633 Specifies the core for which to tune the performance of this function.
3634 The behavior and permissible arguments are the same as for the @option{-mtune=}
3635 command-line option.
3638 @cindex @code{cpu=} function attribute, AArch64
3639 Specifies the core for which to tune the performance of this function and also
3640 whose architectural features to use. The behavior and valid arguments are the
3641 same as for the @option{-mcpu=} command-line option.
3643 @item sign-return-address
3644 @cindex @code{sign-return-address} function attribute, AArch64
3645 Select the function scope on which return address signing will be applied. The
3646 behavior and permissible arguments are the same as for the command-line option
3647 @option{-msign-return-address=}. The default value is @code{none}.
3651 The above target attributes can be specified as follows:
3654 __attribute__((target("@var{attr-string}")))
3662 where @code{@var{attr-string}} is one of the attribute strings specified above.
3664 Additionally, the architectural extension string may be specified on its
3665 own. This can be used to turn on and off particular architectural extensions
3666 without having to specify a particular architecture version or core. Example:
3669 __attribute__((target("+crc+nocrypto")))
3677 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3678 extension and disables the @code{crypto} extension for the function @code{foo}
3679 without modifying an existing @option{-march=} or @option{-mcpu} option.
3681 Multiple target function attributes can be specified by separating them with
3682 a comma. For example:
3684 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
3692 is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
3693 and @code{crypto} extensions and tunes it for @code{cortex-a53}.
3695 @subsubsection Inlining rules
3696 Specifying target attributes on individual functions or performing link-time
3697 optimization across translation units compiled with different target options
3698 can affect function inlining rules:
3700 In particular, a caller function can inline a callee function only if the
3701 architectural features available to the callee are a subset of the features
3702 available to the caller.
3703 For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
3704 or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
3705 can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
3706 because the all the architectural features that function @code{bar} requires
3707 are available to function @code{foo}. Conversely, function @code{bar} cannot
3708 inline function @code{foo}.
3710 Additionally inlining a function compiled with @option{-mstrict-align} into a
3711 function compiled without @code{-mstrict-align} is not allowed.
3712 However, inlining a function compiled without @option{-mstrict-align} into a
3713 function compiled with @option{-mstrict-align} is allowed.
3715 Note that CPU tuning options and attributes such as the @option{-mcpu=},
3716 @option{-mtune=} do not inhibit inlining unless the CPU specified by the
3717 @option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
3718 architectural feature rules specified above.
3720 @node ARC Function Attributes
3721 @subsection ARC Function Attributes
3723 These function attributes are supported by the ARC back end:
3727 @cindex @code{interrupt} function attribute, ARC
3728 Use this attribute to indicate
3729 that the specified function is an interrupt handler. The compiler generates
3730 function entry and exit sequences suitable for use in an interrupt handler
3731 when this attribute is present.
3733 On the ARC, you must specify the kind of interrupt to be handled
3734 in a parameter to the interrupt attribute like this:
3737 void f () __attribute__ ((interrupt ("ilink1")));
3740 Permissible values for this parameter are: @w{@code{ilink1}} and
3746 @cindex @code{long_call} function attribute, ARC
3747 @cindex @code{medium_call} function attribute, ARC
3748 @cindex @code{short_call} function attribute, ARC
3749 @cindex indirect calls, ARC
3750 These attributes specify how a particular function is called.
3751 These attributes override the
3752 @option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
3753 command-line switches and @code{#pragma long_calls} settings.
3755 For ARC, a function marked with the @code{long_call} attribute is
3756 always called using register-indirect jump-and-link instructions,
3757 thereby enabling the called function to be placed anywhere within the
3758 32-bit address space. A function marked with the @code{medium_call}
3759 attribute will always be close enough to be called with an unconditional
3760 branch-and-link instruction, which has a 25-bit offset from
3761 the call site. A function marked with the @code{short_call}
3762 attribute will always be close enough to be called with a conditional
3763 branch-and-link instruction, which has a 21-bit offset from
3767 @cindex @code{jli_always} function attribute, ARC
3768 Forces a particular function to be called using @code{jli}
3769 instruction. The @code{jli} instruction makes use of a table stored
3770 into @code{.jlitab} section, which holds the location of the functions
3771 which are addressed using this instruction.
3774 @cindex @code{jli_fixed} function attribute, ARC
3775 Identical like the above one, but the location of the function in the
3776 @code{jli} table is known and given as an attribute parameter.
3779 @cindex @code{secure_call} function attribute, ARC
3780 This attribute allows one to mark secure-code functions that are
3781 callable from normal mode. The location of the secure call function
3782 into the @code{sjli} table needs to be passed as argument.
3786 @node ARM Function Attributes
3787 @subsection ARM Function Attributes
3789 These function attributes are supported for ARM targets:
3793 @cindex @code{interrupt} function attribute, ARM
3794 Use this attribute to indicate
3795 that the specified function is an interrupt handler. The compiler generates
3796 function entry and exit sequences suitable for use in an interrupt handler
3797 when this attribute is present.
3799 You can specify the kind of interrupt to be handled by
3800 adding an optional parameter to the interrupt attribute like this:
3803 void f () __attribute__ ((interrupt ("IRQ")));
3807 Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
3808 @code{SWI}, @code{ABORT} and @code{UNDEF}.
3810 On ARMv7-M the interrupt type is ignored, and the attribute means the function
3811 may be called with a word-aligned stack pointer.
3814 @cindex @code{isr} function attribute, ARM
3815 Use this attribute on ARM to write Interrupt Service Routines. This is an
3816 alias to the @code{interrupt} attribute above.
3820 @cindex @code{long_call} function attribute, ARM
3821 @cindex @code{short_call} function attribute, ARM
3822 @cindex indirect calls, ARM
3823 These attributes specify how a particular function is called.
3824 These attributes override the
3825 @option{-mlong-calls} (@pxref{ARM Options})
3826 command-line switch and @code{#pragma long_calls} settings. For ARM, the
3827 @code{long_call} attribute indicates that the function might be far
3828 away from the call site and require a different (more expensive)
3829 calling sequence. The @code{short_call} attribute always places
3830 the offset to the function from the call site into the @samp{BL}
3831 instruction directly.
3834 @cindex @code{naked} function attribute, ARM
3835 This attribute allows the compiler to construct the
3836 requisite function declaration, while allowing the body of the
3837 function to be assembly code. The specified function will not have
3838 prologue/epilogue sequences generated by the compiler. Only basic
3839 @code{asm} statements can safely be included in naked functions
3840 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3841 basic @code{asm} and C code may appear to work, they cannot be
3842 depended upon to work reliably and are not supported.
3845 @cindex @code{pcs} function attribute, ARM
3847 The @code{pcs} attribute can be used to control the calling convention
3848 used for a function on ARM. The attribute takes an argument that specifies
3849 the calling convention to use.
3851 When compiling using the AAPCS ABI (or a variant of it) then valid
3852 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3853 order to use a variant other than @code{"aapcs"} then the compiler must
3854 be permitted to use the appropriate co-processor registers (i.e., the
3855 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3859 /* Argument passed in r0, and result returned in r0+r1. */
3860 double f2d (float) __attribute__((pcs("aapcs")));
3863 Variadic functions always use the @code{"aapcs"} calling convention and
3864 the compiler rejects attempts to specify an alternative.
3866 @item target (@var{options})
3867 @cindex @code{target} function attribute
3868 As discussed in @ref{Common Function Attributes}, this attribute
3869 allows specification of target-specific compilation options.
3871 On ARM, the following options are allowed:
3875 @cindex @code{target("thumb")} function attribute, ARM
3876 Force code generation in the Thumb (T16/T32) ISA, depending on the
3880 @cindex @code{target("arm")} function attribute, ARM
3881 Force code generation in the ARM (A32) ISA.
3883 Functions from different modes can be inlined in the caller's mode.
3886 @cindex @code{target("fpu=")} function attribute, ARM
3887 Specifies the fpu for which to tune the performance of this function.
3888 The behavior and permissible arguments are the same as for the @option{-mfpu=}
3889 command-line option.
3892 @cindex @code{arch=} function attribute, ARM
3893 Specifies the architecture version and architectural extensions to use
3894 for this function. The behavior and permissible arguments are the same as
3895 for the @option{-march=} command-line option.
3897 The above target attributes can be specified as follows:
3900 __attribute__((target("arch=armv8-a+crc")))
3908 Additionally, the architectural extension string may be specified on its
3909 own. This can be used to turn on and off particular architectural extensions
3910 without having to specify a particular architecture version or core. Example:
3913 __attribute__((target("+crc+nocrypto")))
3921 In this example @code{target("+crc+nocrypto")} enables the @code{crc}
3922 extension and disables the @code{crypto} extension for the function @code{foo}
3923 without modifying an existing @option{-march=} or @option{-mcpu} option.
3929 @node AVR Function Attributes
3930 @subsection AVR Function Attributes
3932 These function attributes are supported by the AVR back end:
3936 @cindex @code{interrupt} function attribute, AVR
3937 Use this attribute to indicate
3938 that the specified function is an interrupt handler. The compiler generates
3939 function entry and exit sequences suitable for use in an interrupt handler
3940 when this attribute is present.
3942 On the AVR, the hardware globally disables interrupts when an
3943 interrupt is executed. The first instruction of an interrupt handler
3944 declared with this attribute is a @code{SEI} instruction to
3945 re-enable interrupts. See also the @code{signal} function attribute
3946 that does not insert a @code{SEI} instruction. If both @code{signal} and
3947 @code{interrupt} are specified for the same function, @code{signal}
3948 is silently ignored.
3951 @cindex @code{naked} function attribute, AVR
3952 This attribute allows the compiler to construct the
3953 requisite function declaration, while allowing the body of the
3954 function to be assembly code. The specified function will not have
3955 prologue/epilogue sequences generated by the compiler. Only basic
3956 @code{asm} statements can safely be included in naked functions
3957 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
3958 basic @code{asm} and C code may appear to work, they cannot be
3959 depended upon to work reliably and are not supported.
3962 @cindex @code{no_gccisr} function attribute, AVR
3963 Do not use @code{__gcc_isr} pseudo instructions in a function with
3964 the @code{interrupt} or @code{signal} attribute aka. interrupt
3965 service routine (ISR).
3966 Use this attribute if the preamble of the ISR prologue should always read
3970 in __tmp_reg__, __SREG__
3974 and accordingly for the postamble of the epilogue --- no matter whether
3975 the mentioned registers are actually used in the ISR or not.
3976 Situations where you might want to use this attribute include:
3979 Code that (effectively) clobbers bits of @code{SREG} other than the
3980 @code{I}-flag by writing to the memory location of @code{SREG}.
3982 Code that uses inline assembler to jump to a different function which
3983 expects (parts of) the prologue code as outlined above to be present.
3985 To disable @code{__gcc_isr} generation for the whole compilation unit,
3986 there is option @option{-mno-gas-isr-prologues}, @pxref{AVR Options}.
3990 @cindex @code{OS_main} function attribute, AVR
3991 @cindex @code{OS_task} function attribute, AVR
3992 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3993 do not save/restore any call-saved register in their prologue/epilogue.
3995 The @code{OS_main} attribute can be used when there @emph{is
3996 guarantee} that interrupts are disabled at the time when the function
3997 is entered. This saves resources when the stack pointer has to be
3998 changed to set up a frame for local variables.
4000 The @code{OS_task} attribute can be used when there is @emph{no
4001 guarantee} that interrupts are disabled at that time when the function
4002 is entered like for, e@.g@. task functions in a multi-threading operating
4003 system. In that case, changing the stack pointer register is
4004 guarded by save/clear/restore of the global interrupt enable flag.
4006 The differences to the @code{naked} function attribute are:
4008 @item @code{naked} functions do not have a return instruction whereas
4009 @code{OS_main} and @code{OS_task} functions have a @code{RET} or
4010 @code{RETI} return instruction.
4011 @item @code{naked} functions do not set up a frame for local variables
4012 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
4017 @cindex @code{signal} function attribute, AVR
4018 Use this attribute on the AVR to indicate that the specified
4019 function is an interrupt handler. The compiler generates function
4020 entry and exit sequences suitable for use in an interrupt handler when this
4021 attribute is present.
4023 See also the @code{interrupt} function attribute.
4025 The AVR hardware globally disables interrupts when an interrupt is executed.
4026 Interrupt handler functions defined with the @code{signal} attribute
4027 do not re-enable interrupts. It is save to enable interrupts in a
4028 @code{signal} handler. This ``save'' only applies to the code
4029 generated by the compiler and not to the IRQ layout of the
4030 application which is responsibility of the application.
4032 If both @code{signal} and @code{interrupt} are specified for the same
4033 function, @code{signal} is silently ignored.
4036 @node Blackfin Function Attributes
4037 @subsection Blackfin Function Attributes
4039 These function attributes are supported by the Blackfin back end:
4043 @item exception_handler
4044 @cindex @code{exception_handler} function attribute
4045 @cindex exception handler functions, Blackfin
4046 Use this attribute on the Blackfin to indicate that the specified function
4047 is an exception handler. The compiler generates function entry and
4048 exit sequences suitable for use in an exception handler when this
4049 attribute is present.
4051 @item interrupt_handler
4052 @cindex @code{interrupt_handler} function attribute, Blackfin
4053 Use this attribute to
4054 indicate that the specified function is an interrupt handler. The compiler
4055 generates function entry and exit sequences suitable for use in an
4056 interrupt handler when this attribute is present.
4059 @cindex @code{kspisusp} function attribute, Blackfin
4060 @cindex User stack pointer in interrupts on the Blackfin
4061 When used together with @code{interrupt_handler}, @code{exception_handler}
4062 or @code{nmi_handler}, code is generated to load the stack pointer
4063 from the USP register in the function prologue.
4066 @cindex @code{l1_text} function attribute, Blackfin
4067 This attribute specifies a function to be placed into L1 Instruction
4068 SRAM@. The function is put into a specific section named @code{.l1.text}.
4069 With @option{-mfdpic}, function calls with a such function as the callee
4070 or caller uses inlined PLT.
4073 @cindex @code{l2} function attribute, Blackfin
4074 This attribute specifies a function to be placed into L2
4075 SRAM. The function is put into a specific section named
4076 @code{.l2.text}. With @option{-mfdpic}, callers of such functions use
4081 @cindex indirect calls, Blackfin
4082 @cindex @code{longcall} function attribute, Blackfin
4083 @cindex @code{shortcall} function attribute, Blackfin
4084 The @code{longcall} attribute
4085 indicates that the function might be far away from the call site and
4086 require a different (more expensive) calling sequence. The
4087 @code{shortcall} attribute indicates that the function is always close
4088 enough for the shorter calling sequence to be used. These attributes
4089 override the @option{-mlongcall} switch.
4092 @cindex @code{nesting} function attribute, Blackfin
4093 @cindex Allow nesting in an interrupt handler on the Blackfin processor
4094 Use this attribute together with @code{interrupt_handler},
4095 @code{exception_handler} or @code{nmi_handler} to indicate that the function
4096 entry code should enable nested interrupts or exceptions.
4099 @cindex @code{nmi_handler} function attribute, Blackfin
4100 @cindex NMI handler functions on the Blackfin processor
4101 Use this attribute on the Blackfin to indicate that the specified function
4102 is an NMI handler. The compiler generates function entry and
4103 exit sequences suitable for use in an NMI handler when this
4104 attribute is present.
4107 @cindex @code{saveall} function attribute, Blackfin
4108 @cindex save all registers on the Blackfin
4109 Use this attribute to indicate that
4110 all registers except the stack pointer should be saved in the prologue
4111 regardless of whether they are used or not.
4114 @node CR16 Function Attributes
4115 @subsection CR16 Function Attributes
4117 These function attributes are supported by the CR16 back end:
4121 @cindex @code{interrupt} function attribute, CR16
4122 Use this attribute to indicate
4123 that the specified function is an interrupt handler. The compiler generates
4124 function entry and exit sequences suitable for use in an interrupt handler
4125 when this attribute is present.
4128 @node Epiphany Function Attributes
4129 @subsection Epiphany Function Attributes
4131 These function attributes are supported by the Epiphany back end:
4135 @cindex @code{disinterrupt} function attribute, Epiphany
4136 This attribute causes the compiler to emit
4137 instructions to disable interrupts for the duration of the given
4140 @item forwarder_section
4141 @cindex @code{forwarder_section} function attribute, Epiphany
4142 This attribute modifies the behavior of an interrupt handler.
4143 The interrupt handler may be in external memory which cannot be
4144 reached by a branch instruction, so generate a local memory trampoline
4145 to transfer control. The single parameter identifies the section where
4146 the trampoline is placed.
4149 @cindex @code{interrupt} function attribute, Epiphany
4150 Use this attribute to indicate
4151 that the specified function is an interrupt handler. The compiler generates
4152 function entry and exit sequences suitable for use in an interrupt handler
4153 when this attribute is present. It may also generate
4154 a special section with code to initialize the interrupt vector table.
4156 On Epiphany targets one or more optional parameters can be added like this:
4159 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
4162 Permissible values for these parameters are: @w{@code{reset}},
4163 @w{@code{software_exception}}, @w{@code{page_miss}},
4164 @w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
4165 @w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
4166 Multiple parameters indicate that multiple entries in the interrupt
4167 vector table should be initialized for this function, i.e.@: for each
4168 parameter @w{@var{name}}, a jump to the function is emitted in
4169 the section @w{ivt_entry_@var{name}}. The parameter(s) may be omitted
4170 entirely, in which case no interrupt vector table entry is provided.
4172 Note that interrupts are enabled inside the function
4173 unless the @code{disinterrupt} attribute is also specified.
4175 The following examples are all valid uses of these attributes on
4178 void __attribute__ ((interrupt)) universal_handler ();
4179 void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
4180 void __attribute__ ((interrupt ("dma0, dma1")))
4181 universal_dma_handler ();
4182 void __attribute__ ((interrupt ("timer0"), disinterrupt))
4183 fast_timer_handler ();
4184 void __attribute__ ((interrupt ("dma0, dma1"),
4185 forwarder_section ("tramp")))
4186 external_dma_handler ();
4191 @cindex @code{long_call} function attribute, Epiphany
4192 @cindex @code{short_call} function attribute, Epiphany
4193 @cindex indirect calls, Epiphany
4194 These attributes specify how a particular function is called.
4195 These attributes override the
4196 @option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
4197 command-line switch and @code{#pragma long_calls} settings.
4201 @node H8/300 Function Attributes
4202 @subsection H8/300 Function Attributes
4204 These function attributes are available for H8/300 targets:
4207 @item function_vector
4208 @cindex @code{function_vector} function attribute, H8/300
4209 Use this attribute on the H8/300, H8/300H, and H8S to indicate
4210 that the specified function should be called through the function vector.
4211 Calling a function through the function vector reduces code size; however,
4212 the function vector has a limited size (maximum 128 entries on the H8/300
4213 and 64 entries on the H8/300H and H8S)
4214 and shares space with the interrupt vector.
4216 @item interrupt_handler
4217 @cindex @code{interrupt_handler} function attribute, H8/300
4218 Use this attribute on the H8/300, H8/300H, and H8S to
4219 indicate that the specified function is an interrupt handler. The compiler
4220 generates function entry and exit sequences suitable for use in an
4221 interrupt handler when this attribute is present.
4224 @cindex @code{saveall} function attribute, H8/300
4225 @cindex save all registers on the H8/300, H8/300H, and H8S
4226 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
4227 all registers except the stack pointer should be saved in the prologue
4228 regardless of whether they are used or not.
4231 @node IA-64 Function Attributes
4232 @subsection IA-64 Function Attributes
4234 These function attributes are supported on IA-64 targets:
4237 @item syscall_linkage
4238 @cindex @code{syscall_linkage} function attribute, IA-64
4239 This attribute is used to modify the IA-64 calling convention by marking
4240 all input registers as live at all function exits. This makes it possible
4241 to restart a system call after an interrupt without having to save/restore
4242 the input registers. This also prevents kernel data from leaking into
4246 @cindex @code{version_id} function attribute, IA-64
4247 This IA-64 HP-UX attribute, attached to a global variable or function, renames a
4248 symbol to contain a version string, thus allowing for function level
4249 versioning. HP-UX system header files may use function level versioning
4250 for some system calls.
4253 extern int foo () __attribute__((version_id ("20040821")));
4257 Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
4260 @node M32C Function Attributes
4261 @subsection M32C Function Attributes
4263 These function attributes are supported by the M32C back end:
4267 @cindex @code{bank_switch} function attribute, M32C
4268 When added to an interrupt handler with the M32C port, causes the
4269 prologue and epilogue to use bank switching to preserve the registers
4270 rather than saving them on the stack.
4272 @item fast_interrupt
4273 @cindex @code{fast_interrupt} function attribute, M32C
4274 Use this attribute on the M32C port to indicate that the specified
4275 function is a fast interrupt handler. This is just like the
4276 @code{interrupt} attribute, except that @code{freit} is used to return
4277 instead of @code{reit}.
4279 @item function_vector
4280 @cindex @code{function_vector} function attribute, M16C/M32C
4281 On M16C/M32C targets, the @code{function_vector} attribute declares a
4282 special page subroutine call function. Use of this attribute reduces
4283 the code size by 2 bytes for each call generated to the
4284 subroutine. The argument to the attribute is the vector number entry
4285 from the special page vector table which contains the 16 low-order
4286 bits of the subroutine's entry address. Each vector table has special
4287 page number (18 to 255) that is used in @code{jsrs} instructions.
4288 Jump addresses of the routines are generated by adding 0x0F0000 (in
4289 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
4290 2-byte addresses set in the vector table. Therefore you need to ensure
4291 that all the special page vector routines should get mapped within the
4292 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
4295 In the following example 2 bytes are saved for each call to
4296 function @code{foo}.
4299 void foo (void) __attribute__((function_vector(0x18)));
4310 If functions are defined in one file and are called in another file,
4311 then be sure to write this declaration in both files.
4313 This attribute is ignored for R8C target.
4316 @cindex @code{interrupt} function attribute, M32C
4317 Use this attribute to indicate
4318 that the specified function is an interrupt handler. The compiler generates
4319 function entry and exit sequences suitable for use in an interrupt handler
4320 when this attribute is present.
4323 @node M32R/D Function Attributes
4324 @subsection M32R/D Function Attributes
4326 These function attributes are supported by the M32R/D back end:
4330 @cindex @code{interrupt} function attribute, M32R/D
4331 Use this attribute to indicate
4332 that the specified function is an interrupt handler. The compiler generates
4333 function entry and exit sequences suitable for use in an interrupt handler
4334 when this attribute is present.
4336 @item model (@var{model-name})
4337 @cindex @code{model} function attribute, M32R/D
4338 @cindex function addressability on the M32R/D
4340 On the M32R/D, use this attribute to set the addressability of an
4341 object, and of the code generated for a function. The identifier
4342 @var{model-name} is one of @code{small}, @code{medium}, or
4343 @code{large}, representing each of the code models.
4345 Small model objects live in the lower 16MB of memory (so that their
4346 addresses can be loaded with the @code{ld24} instruction), and are
4347 callable with the @code{bl} instruction.
4349 Medium model objects may live anywhere in the 32-bit address space (the
4350 compiler generates @code{seth/add3} instructions to load their addresses),
4351 and are callable with the @code{bl} instruction.
4353 Large model objects may live anywhere in the 32-bit address space (the
4354 compiler generates @code{seth/add3} instructions to load their addresses),
4355 and may not be reachable with the @code{bl} instruction (the compiler
4356 generates the much slower @code{seth/add3/jl} instruction sequence).
4359 @node m68k Function Attributes
4360 @subsection m68k Function Attributes
4362 These function attributes are supported by the m68k back end:
4366 @itemx interrupt_handler
4367 @cindex @code{interrupt} function attribute, m68k
4368 @cindex @code{interrupt_handler} function attribute, m68k
4369 Use this attribute to
4370 indicate that the specified function is an interrupt handler. The compiler
4371 generates function entry and exit sequences suitable for use in an
4372 interrupt handler when this attribute is present. Either name may be used.
4374 @item interrupt_thread
4375 @cindex @code{interrupt_thread} function attribute, fido
4376 Use this attribute on fido, a subarchitecture of the m68k, to indicate
4377 that the specified function is an interrupt handler that is designed
4378 to run as a thread. The compiler omits generate prologue/epilogue
4379 sequences and replaces the return instruction with a @code{sleep}
4380 instruction. This attribute is available only on fido.
4383 @node MCORE Function Attributes
4384 @subsection MCORE Function Attributes
4386 These function attributes are supported by the MCORE back end:
4390 @cindex @code{naked} function attribute, MCORE
4391 This attribute allows the compiler to construct the
4392 requisite function declaration, while allowing the body of the
4393 function to be assembly code. The specified function will not have
4394 prologue/epilogue sequences generated by the compiler. Only basic
4395 @code{asm} statements can safely be included in naked functions
4396 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4397 basic @code{asm} and C code may appear to work, they cannot be
4398 depended upon to work reliably and are not supported.
4401 @node MeP Function Attributes
4402 @subsection MeP Function Attributes
4404 These function attributes are supported by the MeP back end:
4408 @cindex @code{disinterrupt} function attribute, MeP
4409 On MeP targets, this attribute causes the compiler to emit
4410 instructions to disable interrupts for the duration of the given
4414 @cindex @code{interrupt} function attribute, MeP
4415 Use this attribute to indicate
4416 that the specified function is an interrupt handler. The compiler generates
4417 function entry and exit sequences suitable for use in an interrupt handler
4418 when this attribute is present.
4421 @cindex @code{near} function attribute, MeP
4422 This attribute causes the compiler to assume the called
4423 function is close enough to use the normal calling convention,
4424 overriding the @option{-mtf} command-line option.
4427 @cindex @code{far} function attribute, MeP
4428 On MeP targets this causes the compiler to use a calling convention
4429 that assumes the called function is too far away for the built-in
4433 @cindex @code{vliw} function attribute, MeP
4434 The @code{vliw} attribute tells the compiler to emit
4435 instructions in VLIW mode instead of core mode. Note that this
4436 attribute is not allowed unless a VLIW coprocessor has been configured
4437 and enabled through command-line options.
4440 @node MicroBlaze Function Attributes
4441 @subsection MicroBlaze Function Attributes
4443 These function attributes are supported on MicroBlaze targets:
4446 @item save_volatiles
4447 @cindex @code{save_volatiles} function attribute, MicroBlaze
4448 Use this attribute to indicate that the function is
4449 an interrupt handler. All volatile registers (in addition to non-volatile
4450 registers) are saved in the function prologue. If the function is a leaf
4451 function, only volatiles used by the function are saved. A normal function
4452 return is generated instead of a return from interrupt.
4455 @cindex @code{break_handler} function attribute, MicroBlaze
4456 @cindex break handler functions
4457 Use this attribute to indicate that
4458 the specified function is a break handler. The compiler generates function
4459 entry and exit sequences suitable for use in an break handler when this
4460 attribute is present. The return from @code{break_handler} is done through
4461 the @code{rtbd} instead of @code{rtsd}.
4464 void f () __attribute__ ((break_handler));
4467 @item interrupt_handler
4468 @itemx fast_interrupt
4469 @cindex @code{interrupt_handler} function attribute, MicroBlaze
4470 @cindex @code{fast_interrupt} function attribute, MicroBlaze
4471 These attributes indicate that the specified function is an interrupt
4472 handler. Use the @code{fast_interrupt} attribute to indicate handlers
4473 used in low-latency interrupt mode, and @code{interrupt_handler} for
4474 interrupts that do not use low-latency handlers. In both cases, GCC
4475 emits appropriate prologue code and generates a return from the handler
4476 using @code{rtid} instead of @code{rtsd}.
4479 @node Microsoft Windows Function Attributes
4480 @subsection Microsoft Windows Function Attributes
4482 The following attributes are available on Microsoft Windows and Symbian OS
4487 @cindex @code{dllexport} function attribute
4488 @cindex @code{__declspec(dllexport)}
4489 On Microsoft Windows targets and Symbian OS targets the
4490 @code{dllexport} attribute causes the compiler to provide a global
4491 pointer to a pointer in a DLL, so that it can be referenced with the
4492 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
4493 name is formed by combining @code{_imp__} and the function or variable
4496 You can use @code{__declspec(dllexport)} as a synonym for
4497 @code{__attribute__ ((dllexport))} for compatibility with other
4500 On systems that support the @code{visibility} attribute, this
4501 attribute also implies ``default'' visibility. It is an error to
4502 explicitly specify any other visibility.
4504 GCC's default behavior is to emit all inline functions with the
4505 @code{dllexport} attribute. Since this can cause object file-size bloat,
4506 you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
4507 ignore the attribute for inlined functions unless the
4508 @option{-fkeep-inline-functions} flag is used instead.
4510 The attribute is ignored for undefined symbols.
4512 When applied to C++ classes, the attribute marks defined non-inlined
4513 member functions and static data members as exports. Static consts
4514 initialized in-class are not marked unless they are also defined
4517 For Microsoft Windows targets there are alternative methods for
4518 including the symbol in the DLL's export table such as using a
4519 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
4520 the @option{--export-all} linker flag.
4523 @cindex @code{dllimport} function attribute
4524 @cindex @code{__declspec(dllimport)}
4525 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
4526 attribute causes the compiler to reference a function or variable via
4527 a global pointer to a pointer that is set up by the DLL exporting the
4528 symbol. The attribute implies @code{extern}. On Microsoft Windows
4529 targets, the pointer name is formed by combining @code{_imp__} and the
4530 function or variable name.
4532 You can use @code{__declspec(dllimport)} as a synonym for
4533 @code{__attribute__ ((dllimport))} for compatibility with other
4536 On systems that support the @code{visibility} attribute, this
4537 attribute also implies ``default'' visibility. It is an error to
4538 explicitly specify any other visibility.
4540 Currently, the attribute is ignored for inlined functions. If the
4541 attribute is applied to a symbol @emph{definition}, an error is reported.
4542 If a symbol previously declared @code{dllimport} is later defined, the
4543 attribute is ignored in subsequent references, and a warning is emitted.
4544 The attribute is also overridden by a subsequent declaration as
4547 When applied to C++ classes, the attribute marks non-inlined
4548 member functions and static data members as imports. However, the
4549 attribute is ignored for virtual methods to allow creation of vtables
4552 On the SH Symbian OS target the @code{dllimport} attribute also has
4553 another affect---it can cause the vtable and run-time type information
4554 for a class to be exported. This happens when the class has a
4555 dllimported constructor or a non-inline, non-pure virtual function
4556 and, for either of those two conditions, the class also has an inline
4557 constructor or destructor and has a key function that is defined in
4558 the current translation unit.
4560 For Microsoft Windows targets the use of the @code{dllimport}
4561 attribute on functions is not necessary, but provides a small
4562 performance benefit by eliminating a thunk in the DLL@. The use of the
4563 @code{dllimport} attribute on imported variables can be avoided by passing the
4564 @option{--enable-auto-import} switch to the GNU linker. As with
4565 functions, using the attribute for a variable eliminates a thunk in
4568 One drawback to using this attribute is that a pointer to a
4569 @emph{variable} marked as @code{dllimport} cannot be used as a constant
4570 address. However, a pointer to a @emph{function} with the
4571 @code{dllimport} attribute can be used as a constant initializer; in
4572 this case, the address of a stub function in the import lib is
4573 referenced. On Microsoft Windows targets, the attribute can be disabled
4574 for functions by setting the @option{-mnop-fun-dllimport} flag.
4577 @node MIPS Function Attributes
4578 @subsection MIPS Function Attributes
4580 These function attributes are supported by the MIPS back end:
4584 @cindex @code{interrupt} function attribute, MIPS
4585 Use this attribute to indicate that the specified function is an interrupt
4586 handler. The compiler generates function entry and exit sequences suitable
4587 for use in an interrupt handler when this attribute is present.
4588 An optional argument is supported for the interrupt attribute which allows
4589 the interrupt mode to be described. By default GCC assumes the external
4590 interrupt controller (EIC) mode is in use, this can be explicitly set using
4591 @code{eic}. When interrupts are non-masked then the requested Interrupt
4592 Priority Level (IPL) is copied to the current IPL which has the effect of only
4593 enabling higher priority interrupts. To use vectored interrupt mode use
4594 the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
4595 the behavior of the non-masked interrupt support and GCC will arrange to mask
4596 all interrupts from sw0 up to and including the specified interrupt vector.
4598 You can use the following attributes to modify the behavior
4599 of an interrupt handler:
4601 @item use_shadow_register_set
4602 @cindex @code{use_shadow_register_set} function attribute, MIPS
4603 Assume that the handler uses a shadow register set, instead of
4604 the main general-purpose registers. An optional argument @code{intstack} is
4605 supported to indicate that the shadow register set contains a valid stack
4608 @item keep_interrupts_masked
4609 @cindex @code{keep_interrupts_masked} function attribute, MIPS
4610 Keep interrupts masked for the whole function. Without this attribute,
4611 GCC tries to reenable interrupts for as much of the function as it can.
4613 @item use_debug_exception_return
4614 @cindex @code{use_debug_exception_return} function attribute, MIPS
4615 Return using the @code{deret} instruction. Interrupt handlers that don't
4616 have this attribute return using @code{eret} instead.
4619 You can use any combination of these attributes, as shown below:
4621 void __attribute__ ((interrupt)) v0 ();
4622 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
4623 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
4624 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
4625 void __attribute__ ((interrupt, use_shadow_register_set,
4626 keep_interrupts_masked)) v4 ();
4627 void __attribute__ ((interrupt, use_shadow_register_set,
4628 use_debug_exception_return)) v5 ();
4629 void __attribute__ ((interrupt, keep_interrupts_masked,
4630 use_debug_exception_return)) v6 ();
4631 void __attribute__ ((interrupt, use_shadow_register_set,
4632 keep_interrupts_masked,
4633 use_debug_exception_return)) v7 ();
4634 void __attribute__ ((interrupt("eic"))) v8 ();
4635 void __attribute__ ((interrupt("vector=hw3"))) v9 ();
4642 @cindex indirect calls, MIPS
4643 @cindex @code{long_call} function attribute, MIPS
4644 @cindex @code{short_call} function attribute, MIPS
4645 @cindex @code{near} function attribute, MIPS
4646 @cindex @code{far} function attribute, MIPS
4647 These attributes specify how a particular function is called on MIPS@.
4648 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
4649 command-line switch. The @code{long_call} and @code{far} attributes are
4650 synonyms, and cause the compiler to always call
4651 the function by first loading its address into a register, and then using
4652 the contents of that register. The @code{short_call} and @code{near}
4653 attributes are synonyms, and have the opposite
4654 effect; they specify that non-PIC calls should be made using the more
4655 efficient @code{jal} instruction.
4659 @cindex @code{mips16} function attribute, MIPS
4660 @cindex @code{nomips16} function attribute, MIPS
4662 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
4663 function attributes to locally select or turn off MIPS16 code generation.
4664 A function with the @code{mips16} attribute is emitted as MIPS16 code,
4665 while MIPS16 code generation is disabled for functions with the
4666 @code{nomips16} attribute. These attributes override the
4667 @option{-mips16} and @option{-mno-mips16} options on the command line
4668 (@pxref{MIPS Options}).
4670 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
4671 preprocessor symbol @code{__mips16} reflects the setting on the command line,
4672 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
4673 may interact badly with some GCC extensions such as @code{__builtin_apply}
4674 (@pxref{Constructing Calls}).
4676 @item micromips, MIPS
4677 @itemx nomicromips, MIPS
4678 @cindex @code{micromips} function attribute
4679 @cindex @code{nomicromips} function attribute
4681 On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
4682 function attributes to locally select or turn off microMIPS code generation.
4683 A function with the @code{micromips} attribute is emitted as microMIPS code,
4684 while microMIPS code generation is disabled for functions with the
4685 @code{nomicromips} attribute. These attributes override the
4686 @option{-mmicromips} and @option{-mno-micromips} options on the command line
4687 (@pxref{MIPS Options}).
4689 When compiling files containing mixed microMIPS and non-microMIPS code, the
4690 preprocessor symbol @code{__mips_micromips} reflects the setting on the
4692 not that within individual functions. Mixed microMIPS and non-microMIPS code
4693 may interact badly with some GCC extensions such as @code{__builtin_apply}
4694 (@pxref{Constructing Calls}).
4697 @cindex @code{nocompression} function attribute, MIPS
4698 On MIPS targets, you can use the @code{nocompression} function attribute
4699 to locally turn off MIPS16 and microMIPS code generation. This attribute
4700 overrides the @option{-mips16} and @option{-mmicromips} options on the
4701 command line (@pxref{MIPS Options}).
4704 @node MSP430 Function Attributes
4705 @subsection MSP430 Function Attributes
4707 These function attributes are supported by the MSP430 back end:
4711 @cindex @code{critical} function attribute, MSP430
4712 Critical functions disable interrupts upon entry and restore the
4713 previous interrupt state upon exit. Critical functions cannot also
4714 have the @code{naked} or @code{reentrant} attributes. They can have
4715 the @code{interrupt} attribute.
4718 @cindex @code{interrupt} function attribute, MSP430
4719 Use this attribute to indicate
4720 that the specified function is an interrupt handler. The compiler generates
4721 function entry and exit sequences suitable for use in an interrupt handler
4722 when this attribute is present.
4724 You can provide an argument to the interrupt
4725 attribute which specifies a name or number. If the argument is a
4726 number it indicates the slot in the interrupt vector table (0 - 31) to
4727 which this handler should be assigned. If the argument is a name it
4728 is treated as a symbolic name for the vector slot. These names should
4729 match up with appropriate entries in the linker script. By default
4730 the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
4731 @code{reset} for vector 31 are recognized.
4734 @cindex @code{naked} function attribute, MSP430
4735 This attribute allows the compiler to construct the
4736 requisite function declaration, while allowing the body of the
4737 function to be assembly code. The specified function will not have
4738 prologue/epilogue sequences generated by the compiler. Only basic
4739 @code{asm} statements can safely be included in naked functions
4740 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4741 basic @code{asm} and C code may appear to work, they cannot be
4742 depended upon to work reliably and are not supported.
4745 @cindex @code{reentrant} function attribute, MSP430
4746 Reentrant functions disable interrupts upon entry and enable them
4747 upon exit. Reentrant functions cannot also have the @code{naked}
4748 or @code{critical} attributes. They can have the @code{interrupt}
4752 @cindex @code{wakeup} function attribute, MSP430
4753 This attribute only applies to interrupt functions. It is silently
4754 ignored if applied to a non-interrupt function. A wakeup interrupt
4755 function will rouse the processor from any low-power state that it
4756 might be in when the function exits.
4761 @cindex @code{lower} function attribute, MSP430
4762 @cindex @code{upper} function attribute, MSP430
4763 @cindex @code{either} function attribute, MSP430
4764 On the MSP430 target these attributes can be used to specify whether
4765 the function or variable should be placed into low memory, high
4766 memory, or the placement should be left to the linker to decide. The
4767 attributes are only significant if compiling for the MSP430X
4770 The attributes work in conjunction with a linker script that has been
4771 augmented to specify where to place sections with a @code{.lower} and
4772 a @code{.upper} prefix. So, for example, as well as placing the
4773 @code{.data} section, the script also specifies the placement of a
4774 @code{.lower.data} and a @code{.upper.data} section. The intention
4775 is that @code{lower} sections are placed into a small but easier to
4776 access memory region and the upper sections are placed into a larger, but
4777 slower to access, region.
4779 The @code{either} attribute is special. It tells the linker to place
4780 the object into the corresponding @code{lower} section if there is
4781 room for it. If there is insufficient room then the object is placed
4782 into the corresponding @code{upper} section instead. Note that the
4783 placement algorithm is not very sophisticated. It does not attempt to
4784 find an optimal packing of the @code{lower} sections. It just makes
4785 one pass over the objects and does the best that it can. Using the
4786 @option{-ffunction-sections} and @option{-fdata-sections} command-line
4787 options can help the packing, however, since they produce smaller,
4788 easier to pack regions.
4791 @node NDS32 Function Attributes
4792 @subsection NDS32 Function Attributes
4794 These function attributes are supported by the NDS32 back end:
4798 @cindex @code{exception} function attribute
4799 @cindex exception handler functions, NDS32
4800 Use this attribute on the NDS32 target to indicate that the specified function
4801 is an exception handler. The compiler will generate corresponding sections
4802 for use in an exception handler.
4805 @cindex @code{interrupt} function attribute, NDS32
4806 On NDS32 target, this attribute indicates that the specified function
4807 is an interrupt handler. The compiler generates corresponding sections
4808 for use in an interrupt handler. You can use the following attributes
4809 to modify the behavior:
4812 @cindex @code{nested} function attribute, NDS32
4813 This interrupt service routine is interruptible.
4815 @cindex @code{not_nested} function attribute, NDS32
4816 This interrupt service routine is not interruptible.
4818 @cindex @code{nested_ready} function attribute, NDS32
4819 This interrupt service routine is interruptible after @code{PSW.GIE}
4820 (global interrupt enable) is set. This allows interrupt service routine to
4821 finish some short critical code before enabling interrupts.
4823 @cindex @code{save_all} function attribute, NDS32
4824 The system will help save all registers into stack before entering
4827 @cindex @code{partial_save} function attribute, NDS32
4828 The system will help save caller registers into stack before entering
4833 @cindex @code{naked} function attribute, NDS32
4834 This attribute allows the compiler to construct the
4835 requisite function declaration, while allowing the body of the
4836 function to be assembly code. The specified function will not have
4837 prologue/epilogue sequences generated by the compiler. Only basic
4838 @code{asm} statements can safely be included in naked functions
4839 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
4840 basic @code{asm} and C code may appear to work, they cannot be
4841 depended upon to work reliably and are not supported.
4844 @cindex @code{reset} function attribute, NDS32
4845 @cindex reset handler functions
4846 Use this attribute on the NDS32 target to indicate that the specified function
4847 is a reset handler. The compiler will generate corresponding sections
4848 for use in a reset handler. You can use the following attributes
4849 to provide extra exception handling:
4852 @cindex @code{nmi} function attribute, NDS32
4853 Provide a user-defined function to handle NMI exception.
4855 @cindex @code{warm} function attribute, NDS32
4856 Provide a user-defined function to handle warm reset exception.
4860 @node Nios II Function Attributes
4861 @subsection Nios II Function Attributes
4863 These function attributes are supported by the Nios II back end:
4866 @item target (@var{options})
4867 @cindex @code{target} function attribute
4868 As discussed in @ref{Common Function Attributes}, this attribute
4869 allows specification of target-specific compilation options.
4871 When compiling for Nios II, the following options are allowed:
4874 @item custom-@var{insn}=@var{N}
4875 @itemx no-custom-@var{insn}
4876 @cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
4877 @cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
4878 Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
4879 custom instruction with encoding @var{N} when generating code that uses
4880 @var{insn}. Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
4881 the custom instruction @var{insn}.
4882 These target attributes correspond to the
4883 @option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
4884 command-line options, and support the same set of @var{insn} keywords.
4885 @xref{Nios II Options}, for more information.
4887 @item custom-fpu-cfg=@var{name}
4888 @cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
4889 This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
4890 command-line option, to select a predefined set of custom instructions
4892 @xref{Nios II Options}, for more information.
4896 @node Nvidia PTX Function Attributes
4897 @subsection Nvidia PTX Function Attributes
4899 These function attributes are supported by the Nvidia PTX back end:
4903 @cindex @code{kernel} attribute, Nvidia PTX
4904 This attribute indicates that the corresponding function should be compiled
4905 as a kernel function, which can be invoked from the host via the CUDA RT
4907 By default functions are only callable only from other PTX functions.
4909 Kernel functions must have @code{void} return type.
4912 @node PowerPC Function Attributes
4913 @subsection PowerPC Function Attributes
4915 These function attributes are supported by the PowerPC back end:
4920 @cindex indirect calls, PowerPC
4921 @cindex @code{longcall} function attribute, PowerPC
4922 @cindex @code{shortcall} function attribute, PowerPC
4923 The @code{longcall} attribute
4924 indicates that the function might be far away from the call site and
4925 require a different (more expensive) calling sequence. The
4926 @code{shortcall} attribute indicates that the function is always close
4927 enough for the shorter calling sequence to be used. These attributes
4928 override both the @option{-mlongcall} switch and
4929 the @code{#pragma longcall} setting.
4931 @xref{RS/6000 and PowerPC Options}, for more information on whether long
4932 calls are necessary.
4934 @item target (@var{options})
4935 @cindex @code{target} function attribute
4936 As discussed in @ref{Common Function Attributes}, this attribute
4937 allows specification of target-specific compilation options.
4939 On the PowerPC, the following options are allowed:
4944 @cindex @code{target("altivec")} function attribute, PowerPC
4945 Generate code that uses (does not use) AltiVec instructions. In
4946 32-bit code, you cannot enable AltiVec instructions unless
4947 @option{-mabi=altivec} is used on the command line.
4951 @cindex @code{target("cmpb")} function attribute, PowerPC
4952 Generate code that uses (does not use) the compare bytes instruction
4953 implemented on the POWER6 processor and other processors that support
4954 the PowerPC V2.05 architecture.
4958 @cindex @code{target("dlmzb")} function attribute, PowerPC
4959 Generate code that uses (does not use) the string-search @samp{dlmzb}
4960 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
4961 generated by default when targeting those processors.
4965 @cindex @code{target("fprnd")} function attribute, PowerPC
4966 Generate code that uses (does not use) the FP round to integer
4967 instructions implemented on the POWER5+ processor and other processors
4968 that support the PowerPC V2.03 architecture.
4972 @cindex @code{target("hard-dfp")} function attribute, PowerPC
4973 Generate code that uses (does not use) the decimal floating-point
4974 instructions implemented on some POWER processors.
4978 @cindex @code{target("isel")} function attribute, PowerPC
4979 Generate code that uses (does not use) ISEL instruction.
4983 @cindex @code{target("mfcrf")} function attribute, PowerPC
4984 Generate code that uses (does not use) the move from condition
4985 register field instruction implemented on the POWER4 processor and
4986 other processors that support the PowerPC V2.01 architecture.
4990 @cindex @code{target("mfpgpr")} function attribute, PowerPC
4991 Generate code that uses (does not use) the FP move to/from general
4992 purpose register instructions implemented on the POWER6X processor and
4993 other processors that support the extended PowerPC V2.05 architecture.
4997 @cindex @code{target("mulhw")} function attribute, PowerPC
4998 Generate code that uses (does not use) the half-word multiply and
4999 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
5000 These instructions are generated by default when targeting those
5005 @cindex @code{target("multiple")} function attribute, PowerPC
5006 Generate code that uses (does not use) the load multiple word
5007 instructions and the store multiple word instructions.
5011 @cindex @code{target("update")} function attribute, PowerPC
5012 Generate code that uses (does not use) the load or store instructions
5013 that update the base register to the address of the calculated memory
5018 @cindex @code{target("popcntb")} function attribute, PowerPC
5019 Generate code that uses (does not use) the popcount and double-precision
5020 FP reciprocal estimate instruction implemented on the POWER5
5021 processor and other processors that support the PowerPC V2.02
5026 @cindex @code{target("popcntd")} function attribute, PowerPC
5027 Generate code that uses (does not use) the popcount instruction
5028 implemented on the POWER7 processor and other processors that support
5029 the PowerPC V2.06 architecture.
5031 @item powerpc-gfxopt
5032 @itemx no-powerpc-gfxopt
5033 @cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
5034 Generate code that uses (does not use) the optional PowerPC
5035 architecture instructions in the Graphics group, including
5036 floating-point select.
5039 @itemx no-powerpc-gpopt
5040 @cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
5041 Generate code that uses (does not use) the optional PowerPC
5042 architecture instructions in the General Purpose group, including
5043 floating-point square root.
5045 @item recip-precision
5046 @itemx no-recip-precision
5047 @cindex @code{target("recip-precision")} function attribute, PowerPC
5048 Assume (do not assume) that the reciprocal estimate instructions
5049 provide higher-precision estimates than is mandated by the PowerPC
5054 @cindex @code{target("string")} function attribute, PowerPC
5055 Generate code that uses (does not use) the load string instructions
5056 and the store string word instructions to save multiple registers and
5057 do small block moves.
5061 @cindex @code{target("vsx")} function attribute, PowerPC
5062 Generate code that uses (does not use) vector/scalar (VSX)
5063 instructions, and also enable the use of built-in functions that allow
5064 more direct access to the VSX instruction set. In 32-bit code, you
5065 cannot enable VSX or AltiVec instructions unless
5066 @option{-mabi=altivec} is used on the command line.
5070 @cindex @code{target("friz")} function attribute, PowerPC
5071 Generate (do not generate) the @code{friz} instruction when the
5072 @option{-funsafe-math-optimizations} option is used to optimize
5073 rounding a floating-point value to 64-bit integer and back to floating
5074 point. The @code{friz} instruction does not return the same value if
5075 the floating-point number is too large to fit in an integer.
5077 @item avoid-indexed-addresses
5078 @itemx no-avoid-indexed-addresses
5079 @cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
5080 Generate code that tries to avoid (not avoid) the use of indexed load
5081 or store instructions.
5085 @cindex @code{target("paired")} function attribute, PowerPC
5086 Generate code that uses (does not use) the generation of PAIRED simd
5091 @cindex @code{target("longcall")} function attribute, PowerPC
5092 Generate code that assumes (does not assume) that all calls are far
5093 away so that a longer more expensive calling sequence is required.
5096 @cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
5097 Specify the architecture to generate code for when compiling the
5098 function. If you select the @code{target("cpu=power7")} attribute when
5099 generating 32-bit code, VSX and AltiVec instructions are not generated
5100 unless you use the @option{-mabi=altivec} option on the command line.
5102 @item tune=@var{TUNE}
5103 @cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
5104 Specify the architecture to tune for when compiling the function. If
5105 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
5106 you do specify the @code{target("cpu=@var{CPU}")} attribute,
5107 compilation tunes for the @var{CPU} architecture, and not the
5108 default tuning specified on the command line.
5111 On the PowerPC, the inliner does not inline a
5112 function that has different target options than the caller, unless the
5113 callee has a subset of the target options of the caller.
5116 @node RISC-V Function Attributes
5117 @subsection RISC-V Function Attributes
5119 These function attributes are supported by the RISC-V back end:
5123 @cindex @code{naked} function attribute, RISC-V
5124 This attribute allows the compiler to construct the
5125 requisite function declaration, while allowing the body of the
5126 function to be assembly code. The specified function will not have
5127 prologue/epilogue sequences generated by the compiler. Only basic
5128 @code{asm} statements can safely be included in naked functions
5129 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5130 basic @code{asm} and C code may appear to work, they cannot be
5131 depended upon to work reliably and are not supported.
5134 @node RL78 Function Attributes
5135 @subsection RL78 Function Attributes
5137 These function attributes are supported by the RL78 back end:
5141 @itemx brk_interrupt
5142 @cindex @code{interrupt} function attribute, RL78
5143 @cindex @code{brk_interrupt} function attribute, RL78
5144 These attributes indicate
5145 that the specified function is an interrupt handler. The compiler generates
5146 function entry and exit sequences suitable for use in an interrupt handler
5147 when this attribute is present.
5149 Use @code{brk_interrupt} instead of @code{interrupt} for
5150 handlers intended to be used with the @code{BRK} opcode (i.e.@: those
5151 that must end with @code{RETB} instead of @code{RETI}).
5154 @cindex @code{naked} function attribute, RL78
5155 This attribute allows the compiler to construct the
5156 requisite function declaration, while allowing the body of the
5157 function to be assembly code. The specified function will not have
5158 prologue/epilogue sequences generated by the compiler. Only basic
5159 @code{asm} statements can safely be included in naked functions
5160 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5161 basic @code{asm} and C code may appear to work, they cannot be
5162 depended upon to work reliably and are not supported.
5165 @node RX Function Attributes
5166 @subsection RX Function Attributes
5168 These function attributes are supported by the RX back end:
5171 @item fast_interrupt
5172 @cindex @code{fast_interrupt} function attribute, RX
5173 Use this attribute on the RX port to indicate that the specified
5174 function is a fast interrupt handler. This is just like the
5175 @code{interrupt} attribute, except that @code{freit} is used to return
5176 instead of @code{reit}.
5179 @cindex @code{interrupt} function attribute, RX
5180 Use this attribute to indicate
5181 that the specified function is an interrupt handler. The compiler generates
5182 function entry and exit sequences suitable for use in an interrupt handler
5183 when this attribute is present.
5185 On RX targets, you may specify one or more vector numbers as arguments
5186 to the attribute, as well as naming an alternate table name.
5187 Parameters are handled sequentially, so one handler can be assigned to
5188 multiple entries in multiple tables. One may also pass the magic
5189 string @code{"$default"} which causes the function to be used for any
5190 unfilled slots in the current table.
5192 This example shows a simple assignment of a function to one vector in
5193 the default table (note that preprocessor macros may be used for
5194 chip-specific symbolic vector names):
5196 void __attribute__ ((interrupt (5))) txd1_handler ();
5199 This example assigns a function to two slots in the default table
5200 (using preprocessor macros defined elsewhere) and makes it the default
5201 for the @code{dct} table:
5203 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
5208 @cindex @code{naked} function attribute, RX
5209 This attribute allows the compiler to construct the
5210 requisite function declaration, while allowing the body of the
5211 function to be assembly code. The specified function will not have
5212 prologue/epilogue sequences generated by the compiler. Only basic
5213 @code{asm} statements can safely be included in naked functions
5214 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5215 basic @code{asm} and C code may appear to work, they cannot be
5216 depended upon to work reliably and are not supported.
5219 @cindex @code{vector} function attribute, RX
5220 This RX attribute is similar to the @code{interrupt} attribute, including its
5221 parameters, but does not make the function an interrupt-handler type
5222 function (i.e. it retains the normal C function calling ABI). See the
5223 @code{interrupt} attribute for a description of its arguments.
5226 @node S/390 Function Attributes
5227 @subsection S/390 Function Attributes
5229 These function attributes are supported on the S/390:
5232 @item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
5233 @cindex @code{hotpatch} function attribute, S/390
5235 On S/390 System z targets, you can use this function attribute to
5236 make GCC generate a ``hot-patching'' function prologue. If the
5237 @option{-mhotpatch=} command-line option is used at the same time,
5238 the @code{hotpatch} attribute takes precedence. The first of the
5239 two arguments specifies the number of halfwords to be added before
5240 the function label. A second argument can be used to specify the
5241 number of halfwords to be added after the function label. For
5242 both arguments the maximum allowed value is 1000000.
5244 If both arguments are zero, hotpatching is disabled.
5246 @item target (@var{options})
5247 @cindex @code{target} function attribute
5248 As discussed in @ref{Common Function Attributes}, this attribute
5249 allows specification of target-specific compilation options.
5251 On S/390, the following options are supported:
5259 @item warn-framesize=
5271 @itemx no-packed-stack
5273 @itemx no-small-exec
5276 @item warn-dynamicstack
5277 @itemx no-warn-dynamicstack
5280 The options work exactly like the S/390 specific command line
5281 options (without the prefix @option{-m}) except that they do not
5282 change any feature macros. For example,
5285 @code{target("no-vx")}
5288 does not undefine the @code{__VEC__} macro.
5291 @node SH Function Attributes
5292 @subsection SH Function Attributes
5294 These function attributes are supported on the SH family of processors:
5297 @item function_vector
5298 @cindex @code{function_vector} function attribute, SH
5299 @cindex calling functions through the function vector on SH2A
5300 On SH2A targets, this attribute declares a function to be called using the
5301 TBR relative addressing mode. The argument to this attribute is the entry
5302 number of the same function in a vector table containing all the TBR
5303 relative addressable functions. For correct operation the TBR must be setup
5304 accordingly to point to the start of the vector table before any functions with
5305 this attribute are invoked. Usually a good place to do the initialization is
5306 the startup routine. The TBR relative vector table can have at max 256 function
5307 entries. The jumps to these functions are generated using a SH2A specific,
5308 non delayed branch instruction JSR/N @@(disp8,TBR). You must use GAS and GLD
5309 from GNU binutils version 2.7 or later for this attribute to work correctly.
5311 In an application, for a function being called once, this attribute
5312 saves at least 8 bytes of code; and if other successive calls are being
5313 made to the same function, it saves 2 bytes of code per each of these
5316 @item interrupt_handler
5317 @cindex @code{interrupt_handler} function attribute, SH
5318 Use this attribute to
5319 indicate that the specified function is an interrupt handler. The compiler
5320 generates function entry and exit sequences suitable for use in an
5321 interrupt handler when this attribute is present.
5323 @item nosave_low_regs
5324 @cindex @code{nosave_low_regs} function attribute, SH
5325 Use this attribute on SH targets to indicate that an @code{interrupt_handler}
5326 function should not save and restore registers R0..R7. This can be used on SH3*
5327 and SH4* targets that have a second R0..R7 register bank for non-reentrant
5331 @cindex @code{renesas} function attribute, SH
5332 On SH targets this attribute specifies that the function or struct follows the
5336 @cindex @code{resbank} function attribute, SH
5337 On the SH2A target, this attribute enables the high-speed register
5338 saving and restoration using a register bank for @code{interrupt_handler}
5339 routines. Saving to the bank is performed automatically after the CPU
5340 accepts an interrupt that uses a register bank.
5342 The nineteen 32-bit registers comprising general register R0 to R14,
5343 control register GBR, and system registers MACH, MACL, and PR and the
5344 vector table address offset are saved into a register bank. Register
5345 banks are stacked in first-in last-out (FILO) sequence. Restoration
5346 from the bank is executed by issuing a RESBANK instruction.
5349 @cindex @code{sp_switch} function attribute, SH
5350 Use this attribute on the SH to indicate an @code{interrupt_handler}
5351 function should switch to an alternate stack. It expects a string
5352 argument that names a global variable holding the address of the
5357 void f () __attribute__ ((interrupt_handler,
5358 sp_switch ("alt_stack")));
5362 @cindex @code{trap_exit} function attribute, SH
5363 Use this attribute on the SH for an @code{interrupt_handler} to return using
5364 @code{trapa} instead of @code{rte}. This attribute expects an integer
5365 argument specifying the trap number to be used.
5368 @cindex @code{trapa_handler} function attribute, SH
5369 On SH targets this function attribute is similar to @code{interrupt_handler}
5370 but it does not save and restore all registers.
5373 @node SPU Function Attributes
5374 @subsection SPU Function Attributes
5376 These function attributes are supported by the SPU back end:
5380 @cindex @code{naked} function attribute, SPU
5381 This attribute allows the compiler to construct the
5382 requisite function declaration, while allowing the body of the
5383 function to be assembly code. The specified function will not have
5384 prologue/epilogue sequences generated by the compiler. Only basic
5385 @code{asm} statements can safely be included in naked functions
5386 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5387 basic @code{asm} and C code may appear to work, they cannot be
5388 depended upon to work reliably and are not supported.
5391 @node Symbian OS Function Attributes
5392 @subsection Symbian OS Function Attributes
5394 @xref{Microsoft Windows Function Attributes}, for discussion of the
5395 @code{dllexport} and @code{dllimport} attributes.
5397 @node V850 Function Attributes
5398 @subsection V850 Function Attributes
5400 The V850 back end supports these function attributes:
5404 @itemx interrupt_handler
5405 @cindex @code{interrupt} function attribute, V850
5406 @cindex @code{interrupt_handler} function attribute, V850
5407 Use these attributes to indicate
5408 that the specified function is an interrupt handler. The compiler generates
5409 function entry and exit sequences suitable for use in an interrupt handler
5410 when either attribute is present.
5413 @node Visium Function Attributes
5414 @subsection Visium Function Attributes
5416 These function attributes are supported by the Visium back end:
5420 @cindex @code{interrupt} function attribute, Visium
5421 Use this attribute to indicate
5422 that the specified function is an interrupt handler. The compiler generates
5423 function entry and exit sequences suitable for use in an interrupt handler
5424 when this attribute is present.
5427 @node x86 Function Attributes
5428 @subsection x86 Function Attributes
5430 These function attributes are supported by the x86 back end:
5434 @cindex @code{cdecl} function attribute, x86-32
5435 @cindex functions that pop the argument stack on x86-32
5437 On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
5438 assume that the calling function pops off the stack space used to
5439 pass arguments. This is
5440 useful to override the effects of the @option{-mrtd} switch.
5443 @cindex @code{fastcall} function attribute, x86-32
5444 @cindex functions that pop the argument stack on x86-32
5445 On x86-32 targets, the @code{fastcall} attribute causes the compiler to
5446 pass the first argument (if of integral type) in the register ECX and
5447 the second argument (if of integral type) in the register EDX@. Subsequent
5448 and other typed arguments are passed on the stack. The called function
5449 pops the arguments off the stack. If the number of arguments is variable all
5450 arguments are pushed on the stack.
5453 @cindex @code{thiscall} function attribute, x86-32
5454 @cindex functions that pop the argument stack on x86-32
5455 On x86-32 targets, the @code{thiscall} attribute causes the compiler to
5456 pass the first argument (if of integral type) in the register ECX.
5457 Subsequent and other typed arguments are passed on the stack. The called
5458 function pops the arguments off the stack.
5459 If the number of arguments is variable all arguments are pushed on the
5461 The @code{thiscall} attribute is intended for C++ non-static member functions.
5462 As a GCC extension, this calling convention can be used for C functions
5463 and for static member methods.
5467 @cindex @code{ms_abi} function attribute, x86
5468 @cindex @code{sysv_abi} function attribute, x86
5470 On 32-bit and 64-bit x86 targets, you can use an ABI attribute
5471 to indicate which calling convention should be used for a function. The
5472 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
5473 while the @code{sysv_abi} attribute tells the compiler to use the ABI
5474 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
5475 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
5477 Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
5478 requires the @option{-maccumulate-outgoing-args} option.
5480 @item callee_pop_aggregate_return (@var{number})
5481 @cindex @code{callee_pop_aggregate_return} function attribute, x86
5483 On x86-32 targets, you can use this attribute to control how
5484 aggregates are returned in memory. If the caller is responsible for
5485 popping the hidden pointer together with the rest of the arguments, specify
5486 @var{number} equal to zero. If callee is responsible for popping the
5487 hidden pointer, specify @var{number} equal to one.
5489 The default x86-32 ABI assumes that the callee pops the
5490 stack for hidden pointer. However, on x86-32 Microsoft Windows targets,
5491 the compiler assumes that the
5492 caller pops the stack for hidden pointer.
5494 @item ms_hook_prologue
5495 @cindex @code{ms_hook_prologue} function attribute, x86
5497 On 32-bit and 64-bit x86 targets, you can use
5498 this function attribute to make GCC generate the ``hot-patching'' function
5499 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
5503 @cindex @code{naked} function attribute, x86
5504 This attribute allows the compiler to construct the
5505 requisite function declaration, while allowing the body of the
5506 function to be assembly code. The specified function will not have
5507 prologue/epilogue sequences generated by the compiler. Only basic
5508 @code{asm} statements can safely be included in naked functions
5509 (@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
5510 basic @code{asm} and C code may appear to work, they cannot be
5511 depended upon to work reliably and are not supported.
5513 @item regparm (@var{number})
5514 @cindex @code{regparm} function attribute, x86
5515 @cindex functions that are passed arguments in registers on x86-32
5516 On x86-32 targets, the @code{regparm} attribute causes the compiler to
5517 pass arguments number one to @var{number} if they are of integral type
5518 in registers EAX, EDX, and ECX instead of on the stack. Functions that
5519 take a variable number of arguments continue to be passed all of their
5520 arguments on the stack.
5522 Beware that on some ELF systems this attribute is unsuitable for
5523 global functions in shared libraries with lazy binding (which is the
5524 default). Lazy binding sends the first call via resolving code in
5525 the loader, which might assume EAX, EDX and ECX can be clobbered, as
5526 per the standard calling conventions. Solaris 8 is affected by this.
5527 Systems with the GNU C Library version 2.1 or higher
5528 and FreeBSD are believed to be
5529 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
5530 disabled with the linker or the loader if desired, to avoid the
5534 @cindex @code{sseregparm} function attribute, x86
5535 On x86-32 targets with SSE support, the @code{sseregparm} attribute
5536 causes the compiler to pass up to 3 floating-point arguments in
5537 SSE registers instead of on the stack. Functions that take a
5538 variable number of arguments continue to pass all of their
5539 floating-point arguments on the stack.
5541 @item force_align_arg_pointer
5542 @cindex @code{force_align_arg_pointer} function attribute, x86
5543 On x86 targets, the @code{force_align_arg_pointer} attribute may be
5544 applied to individual function definitions, generating an alternate
5545 prologue and epilogue that realigns the run-time stack if necessary.
5546 This supports mixing legacy codes that run with a 4-byte aligned stack
5547 with modern codes that keep a 16-byte stack for SSE compatibility.
5550 @cindex @code{stdcall} function attribute, x86-32
5551 @cindex functions that pop the argument stack on x86-32
5552 On x86-32 targets, the @code{stdcall} attribute causes the compiler to
5553 assume that the called function pops off the stack space used to
5554 pass arguments, unless it takes a variable number of arguments.
5556 @item no_caller_saved_registers
5557 @cindex @code{no_caller_saved_registers} function attribute, x86
5558 Use this attribute to indicate that the specified function has no
5559 caller-saved registers. That is, all registers are callee-saved. For
5560 example, this attribute can be used for a function called from an
5561 interrupt handler. The compiler generates proper function entry and
5562 exit sequences to save and restore any modified registers, except for
5563 the EFLAGS register. Since GCC doesn't preserve MPX, SSE, MMX nor x87
5564 states, the GCC option @option{-mgeneral-regs-only} should be used to
5565 compile functions with @code{no_caller_saved_registers} attribute.
5568 @cindex @code{interrupt} function attribute, x86
5569 Use this attribute to indicate that the specified function is an
5570 interrupt handler or an exception handler (depending on parameters passed
5571 to the function, explained further). The compiler generates function
5572 entry and exit sequences suitable for use in an interrupt handler when
5573 this attribute is present. The @code{IRET} instruction, instead of the
5574 @code{RET} instruction, is used to return from interrupt handlers. All
5575 registers, except for the EFLAGS register which is restored by the
5576 @code{IRET} instruction, are preserved by the compiler. Since GCC
5577 doesn't preserve MPX, SSE, MMX nor x87 states, the GCC option
5578 @option{-mgeneral-regs-only} should be used to compile interrupt and
5581 Any interruptible-without-stack-switch code must be compiled with
5582 @option{-mno-red-zone} since interrupt handlers can and will, because
5583 of the hardware design, touch the red zone.
5585 An interrupt handler must be declared with a mandatory pointer
5589 struct interrupt_frame;
5591 __attribute__ ((interrupt))
5593 f (struct interrupt_frame *frame)
5599 and you must define @code{struct interrupt_frame} as described in the
5602 Exception handlers differ from interrupt handlers because the system
5603 pushes an error code on the stack. An exception handler declaration is
5604 similar to that for an interrupt handler, but with a different mandatory
5605 function signature. The compiler arranges to pop the error code off the
5606 stack before the @code{IRET} instruction.
5610 typedef unsigned long long int uword_t;
5612 typedef unsigned int uword_t;
5615 struct interrupt_frame;
5617 __attribute__ ((interrupt))
5619 f (struct interrupt_frame *frame, uword_t error_code)
5625 Exception handlers should only be used for exceptions that push an error
5626 code; you should use an interrupt handler in other cases. The system
5627 will crash if the wrong kind of handler is used.
5629 @item target (@var{options})
5630 @cindex @code{target} function attribute
5631 As discussed in @ref{Common Function Attributes}, this attribute
5632 allows specification of target-specific compilation options.
5634 On the x86, the following options are allowed:
5638 @cindex @code{target("abm")} function attribute, x86
5639 Enable/disable the generation of the advanced bit instructions.
5643 @cindex @code{target("aes")} function attribute, x86
5644 Enable/disable the generation of the AES instructions.
5647 @cindex @code{target("default")} function attribute, x86
5648 @xref{Function Multiversioning}, where it is used to specify the
5649 default function version.
5653 @cindex @code{target("mmx")} function attribute, x86
5654 Enable/disable the generation of the MMX instructions.
5658 @cindex @code{target("pclmul")} function attribute, x86
5659 Enable/disable the generation of the PCLMUL instructions.
5663 @cindex @code{target("popcnt")} function attribute, x86
5664 Enable/disable the generation of the POPCNT instruction.
5668 @cindex @code{target("sse")} function attribute, x86
5669 Enable/disable the generation of the SSE instructions.
5673 @cindex @code{target("sse2")} function attribute, x86
5674 Enable/disable the generation of the SSE2 instructions.
5678 @cindex @code{target("sse3")} function attribute, x86
5679 Enable/disable the generation of the SSE3 instructions.
5683 @cindex @code{target("sse4")} function attribute, x86
5684 Enable/disable the generation of the SSE4 instructions (both SSE4.1
5689 @cindex @code{target("sse4.1")} function attribute, x86
5690 Enable/disable the generation of the sse4.1 instructions.
5694 @cindex @code{target("sse4.2")} function attribute, x86
5695 Enable/disable the generation of the sse4.2 instructions.
5699 @cindex @code{target("sse4a")} function attribute, x86
5700 Enable/disable the generation of the SSE4A instructions.
5704 @cindex @code{target("fma4")} function attribute, x86
5705 Enable/disable the generation of the FMA4 instructions.
5709 @cindex @code{target("xop")} function attribute, x86
5710 Enable/disable the generation of the XOP instructions.
5714 @cindex @code{target("lwp")} function attribute, x86
5715 Enable/disable the generation of the LWP instructions.
5719 @cindex @code{target("ssse3")} function attribute, x86
5720 Enable/disable the generation of the SSSE3 instructions.
5724 @cindex @code{target("cld")} function attribute, x86
5725 Enable/disable the generation of the CLD before string moves.
5727 @item fancy-math-387
5728 @itemx no-fancy-math-387
5729 @cindex @code{target("fancy-math-387")} function attribute, x86
5730 Enable/disable the generation of the @code{sin}, @code{cos}, and
5731 @code{sqrt} instructions on the 387 floating-point unit.
5735 @cindex @code{target("ieee-fp")} function attribute, x86
5736 Enable/disable the generation of floating point that depends on IEEE arithmetic.
5738 @item inline-all-stringops
5739 @itemx no-inline-all-stringops
5740 @cindex @code{target("inline-all-stringops")} function attribute, x86
5741 Enable/disable inlining of string operations.
5743 @item inline-stringops-dynamically
5744 @itemx no-inline-stringops-dynamically
5745 @cindex @code{target("inline-stringops-dynamically")} function attribute, x86
5746 Enable/disable the generation of the inline code to do small string
5747 operations and calling the library routines for large operations.
5749 @item align-stringops
5750 @itemx no-align-stringops
5751 @cindex @code{target("align-stringops")} function attribute, x86
5752 Do/do not align destination of inlined string operations.
5756 @cindex @code{target("recip")} function attribute, x86
5757 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
5758 instructions followed an additional Newton-Raphson step instead of
5759 doing a floating-point division.
5761 @item arch=@var{ARCH}
5762 @cindex @code{target("arch=@var{ARCH}")} function attribute, x86
5763 Specify the architecture to generate code for in compiling the function.
5765 @item tune=@var{TUNE}
5766 @cindex @code{target("tune=@var{TUNE}")} function attribute, x86
5767 Specify the architecture to tune for in compiling the function.
5769 @item fpmath=@var{FPMATH}
5770 @cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
5771 Specify which floating-point unit to use. You must specify the
5772 @code{target("fpmath=sse,387")} option as
5773 @code{target("fpmath=sse+387")} because the comma would separate
5776 @item indirect_branch("@var{choice}")
5777 @cindex @code{indirect_branch} function attribute, x86
5778 On x86 targets, the @code{indirect_branch} attribute causes the compiler
5779 to convert indirect call and jump with @var{choice}. @samp{keep}
5780 keeps indirect call and jump unmodified. @samp{thunk} converts indirect
5781 call and jump to call and return thunk. @samp{thunk-inline} converts
5782 indirect call and jump to inlined call and return thunk.
5783 @samp{thunk-extern} converts indirect call and jump to external call
5784 and return thunk provided in a separate object file.
5786 @item function_return("@var{choice}")
5787 @cindex @code{function_return} function attribute, x86
5788 On x86 targets, the @code{function_return} attribute causes the compiler
5789 to convert function return with @var{choice}. @samp{keep} keeps function
5790 return unmodified. @samp{thunk} converts function return to call and
5791 return thunk. @samp{thunk-inline} converts function return to inlined
5792 call and return thunk. @samp{thunk-extern} converts function return to
5793 external call and return thunk provided in a separate object file.
5796 @cindex @code{nocf_check} function attribute
5797 The @code{nocf_check} attribute on a function is used to inform the
5798 compiler that the function's prologue should not be instrumented when
5799 compiled with the @option{-fcf-protection=branch} option. The
5800 compiler assumes that the function's address is a valid target for a
5801 control-flow transfer.
5803 The @code{nocf_check} attribute on a type of pointer to function is
5804 used to inform the compiler that a call through the pointer should
5805 not be instrumented when compiled with the
5806 @option{-fcf-protection=branch} option. The compiler assumes
5807 that the function's address from the pointer is a valid target for
5808 a control-flow transfer. A direct function call through a function
5809 name is assumed to be a safe call thus direct calls are not
5810 instrumented by the compiler.
5812 The @code{nocf_check} attribute is applied to an object's type.
5813 In case of assignment of a function address or a function pointer to
5814 another pointer, the attribute is not carried over from the right-hand
5815 object's type; the type of left-hand object stays unchanged. The
5816 compiler checks for @code{nocf_check} attribute mismatch and reports
5817 a warning in case of mismatch.
5821 int foo (void) __attribute__(nocf_check);
5822 void (*foo1)(void) __attribute__(nocf_check);
5825 /* foo's address is assumed to be valid. */
5829 /* This call site is not checked for control-flow
5833 /* A warning is issued about attribute mismatch. */
5836 /* This call site is still not checked. */
5839 /* This call site is checked. */
5842 /* A warning is issued about attribute mismatch. */
5845 /* This call site is still checked. */
5854 On the x86, the inliner does not inline a
5855 function that has different target options than the caller, unless the
5856 callee has a subset of the target options of the caller. For example
5857 a function declared with @code{target("sse3")} can inline a function
5858 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
5861 @node Xstormy16 Function Attributes
5862 @subsection Xstormy16 Function Attributes
5864 These function attributes are supported by the Xstormy16 back end:
5868 @cindex @code{interrupt} function attribute, Xstormy16
5869 Use this attribute to indicate
5870 that the specified function is an interrupt handler. The compiler generates
5871 function entry and exit sequences suitable for use in an interrupt handler
5872 when this attribute is present.
5875 @node Variable Attributes
5876 @section Specifying Attributes of Variables
5877 @cindex attribute of variables
5878 @cindex variable attributes
5880 The keyword @code{__attribute__} allows you to specify special
5881 attributes of variables or structure fields. This keyword is followed
5882 by an attribute specification inside double parentheses. Some
5883 attributes are currently defined generically for variables.
5884 Other attributes are defined for variables on particular target
5885 systems. Other attributes are available for functions
5886 (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
5887 enumerators (@pxref{Enumerator Attributes}), statements
5888 (@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).
5889 Other front ends might define more attributes
5890 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
5892 @xref{Attribute Syntax}, for details of the exact syntax for using
5896 * Common Variable Attributes::
5897 * ARC Variable Attributes::
5898 * AVR Variable Attributes::
5899 * Blackfin Variable Attributes::
5900 * H8/300 Variable Attributes::
5901 * IA-64 Variable Attributes::
5902 * M32R/D Variable Attributes::
5903 * MeP Variable Attributes::
5904 * Microsoft Windows Variable Attributes::
5905 * MSP430 Variable Attributes::
5906 * Nvidia PTX Variable Attributes::
5907 * PowerPC Variable Attributes::
5908 * RL78 Variable Attributes::
5909 * SPU Variable Attributes::
5910 * V850 Variable Attributes::
5911 * x86 Variable Attributes::
5912 * Xstormy16 Variable Attributes::
5915 @node Common Variable Attributes
5916 @subsection Common Variable Attributes
5918 The following attributes are supported on most targets.
5921 @cindex @code{aligned} variable attribute
5922 @item aligned (@var{alignment})
5923 This attribute specifies a minimum alignment for the variable or
5924 structure field, measured in bytes. For example, the declaration:
5927 int x __attribute__ ((aligned (16))) = 0;
5931 causes the compiler to allocate the global variable @code{x} on a
5932 16-byte boundary. On a 68040, this could be used in conjunction with
5933 an @code{asm} expression to access the @code{move16} instruction which
5934 requires 16-byte aligned operands.
5936 You can also specify the alignment of structure fields. For example, to
5937 create a double-word aligned @code{int} pair, you could write:
5940 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
5944 This is an alternative to creating a union with a @code{double} member,
5945 which forces the union to be double-word aligned.
5947 As in the preceding examples, you can explicitly specify the alignment
5948 (in bytes) that you wish the compiler to use for a given variable or
5949 structure field. Alternatively, you can leave out the alignment factor
5950 and just ask the compiler to align a variable or field to the
5951 default alignment for the target architecture you are compiling for.
5952 The default alignment is sufficient for all scalar types, but may not be
5953 enough for all vector types on a target that supports vector operations.
5954 The default alignment is fixed for a particular target ABI.
5956 GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
5957 which is the largest alignment ever used for any data type on the
5958 target machine you are compiling for. For example, you could write:
5961 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
5964 The compiler automatically sets the alignment for the declared
5965 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
5966 often make copy operations more efficient, because the compiler can
5967 use whatever instructions copy the biggest chunks of memory when
5968 performing copies to or from the variables or fields that you have
5969 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
5970 may change depending on command-line options.
5972 When used on a struct, or struct member, the @code{aligned} attribute can
5973 only increase the alignment; in order to decrease it, the @code{packed}
5974 attribute must be specified as well. When used as part of a typedef, the
5975 @code{aligned} attribute can both increase and decrease alignment, and
5976 specifying the @code{packed} attribute generates a warning.
5978 Note that the effectiveness of @code{aligned} attributes may be limited
5979 by inherent limitations in your linker. On many systems, the linker is
5980 only able to arrange for variables to be aligned up to a certain maximum
5981 alignment. (For some linkers, the maximum supported alignment may
5982 be very very small.) If your linker is only able to align variables
5983 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
5984 in an @code{__attribute__} still only provides you with 8-byte
5985 alignment. See your linker documentation for further information.
5987 The @code{aligned} attribute can also be used for functions
5988 (@pxref{Common Function Attributes}.)
5990 @cindex @code{warn_if_not_aligned} variable attribute
5991 @item warn_if_not_aligned (@var{alignment})
5992 This attribute specifies a threshold for the structure field, measured
5993 in bytes. If the structure field is aligned below the threshold, a
5994 warning will be issued. For example, the declaration:
6001 unsigned long long x __attribute__((warn_if_not_aligned(16)));
6006 causes the compiler to issue an warning on @code{struct foo}, like
6007 @samp{warning: alignment 8 of 'struct foo' is less than 16}.
6008 The compiler also issues a warning, like @samp{warning: 'x' offset
6009 8 in 'struct foo' isn't aligned to 16}, when the structure field has
6010 the misaligned offset:
6017 unsigned long long x __attribute__((warn_if_not_aligned(16)));
6018 @} __attribute__((aligned(16)));
6021 This warning can be disabled by @option{-Wno-if-not-aligned}.
6022 The @code{warn_if_not_aligned} attribute can also be used for types
6023 (@pxref{Common Type Attributes}.)
6025 @item cleanup (@var{cleanup_function})
6026 @cindex @code{cleanup} variable attribute
6027 The @code{cleanup} attribute runs a function when the variable goes
6028 out of scope. This attribute can only be applied to auto function
6029 scope variables; it may not be applied to parameters or variables
6030 with static storage duration. The function must take one parameter,
6031 a pointer to a type compatible with the variable. The return value
6032 of the function (if any) is ignored.
6034 If @option{-fexceptions} is enabled, then @var{cleanup_function}
6035 is run during the stack unwinding that happens during the
6036 processing of the exception. Note that the @code{cleanup} attribute
6037 does not allow the exception to be caught, only to perform an action.
6038 It is undefined what happens if @var{cleanup_function} does not
6043 @cindex @code{common} variable attribute
6044 @cindex @code{nocommon} variable attribute
6047 The @code{common} attribute requests GCC to place a variable in
6048 ``common'' storage. The @code{nocommon} attribute requests the
6049 opposite---to allocate space for it directly.
6051 These attributes override the default chosen by the
6052 @option{-fno-common} and @option{-fcommon} flags respectively.
6055 @itemx deprecated (@var{msg})
6056 @cindex @code{deprecated} variable attribute
6057 The @code{deprecated} attribute results in a warning if the variable
6058 is used anywhere in the source file. This is useful when identifying
6059 variables that are expected to be removed in a future version of a
6060 program. The warning also includes the location of the declaration
6061 of the deprecated variable, to enable users to easily find further
6062 information about why the variable is deprecated, or what they should
6063 do instead. Note that the warning only occurs for uses:
6066 extern int old_var __attribute__ ((deprecated));
6068 int new_fn () @{ return old_var; @}
6072 results in a warning on line 3 but not line 2. The optional @var{msg}
6073 argument, which must be a string, is printed in the warning if
6076 The @code{deprecated} attribute can also be used for functions and
6077 types (@pxref{Common Function Attributes},
6078 @pxref{Common Type Attributes}).
6081 @cindex @code{nonstring} variable attribute
6082 The @code{nonstring} variable attribute specifies that an object or member
6083 declaration with type array of @code{char} or pointer to @code{char} is
6084 intended to store character arrays that do not necessarily contain
6085 a terminating @code{NUL} character. This is useful in detecting uses
6086 of such arrays or pointers with functions that expect NUL-terminated
6087 strings, and to avoid warnings when such an array or pointer is used
6088 as an argument to a bounded string manipulation function such as
6089 @code{strncpy}. For example, without the attribute, GCC will issue
6090 a warning for the @code{strncpy} call below because it may truncate
6091 the copy without appending the terminating @code{NUL} character. Using
6092 the attribute makes it possible to suppress the warning. However, when
6093 the array is declared with the attribute the call to @code{strlen} is
6094 diagnosed because when the array doesn't contain a @code{NUL}-terminated
6095 string the call is undefined. To copy, compare, of search non-string
6096 character arrays use the @code{memcpy}, @code{memcmp}, @code{memchr},
6097 and other functions that operate on arrays of bytes. In addition,
6098 calling @code{strnlen} and @code{strndup} with such arrays is safe
6099 provided a suitable bound is specified, and not diagnosed.
6104 char name [32] __attribute__ ((nonstring));
6107 int f (struct Data *pd, const char *s)
6109 strncpy (pd->name, s, sizeof pd->name);
6111 return strlen (pd->name); // unsafe, gets a warning
6115 @item mode (@var{mode})
6116 @cindex @code{mode} variable attribute
6117 This attribute specifies the data type for the declaration---whichever
6118 type corresponds to the mode @var{mode}. This in effect lets you
6119 request an integer or floating-point type according to its width.
6121 @xref{Machine Modes,,, gccint, GNU Compiler Collection (GCC) Internals},
6122 for a list of the possible keywords for @var{mode}.
6123 You may also specify a mode of @code{byte} or @code{__byte__} to
6124 indicate the mode corresponding to a one-byte integer, @code{word} or
6125 @code{__word__} for the mode of a one-word integer, and @code{pointer}
6126 or @code{__pointer__} for the mode used to represent pointers.
6129 @cindex @code{packed} variable attribute
6130 The @code{packed} attribute specifies that a variable or structure field
6131 should have the smallest possible alignment---one byte for a variable,
6132 and one bit for a field, unless you specify a larger value with the
6133 @code{aligned} attribute.
6135 Here is a structure in which the field @code{x} is packed, so that it
6136 immediately follows @code{a}:
6142 int x[2] __attribute__ ((packed));
6146 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
6147 @code{packed} attribute on bit-fields of type @code{char}. This has
6148 been fixed in GCC 4.4 but the change can lead to differences in the
6149 structure layout. See the documentation of
6150 @option{-Wpacked-bitfield-compat} for more information.
6152 @item section ("@var{section-name}")
6153 @cindex @code{section} variable attribute
6154 Normally, the compiler places the objects it generates in sections like
6155 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
6156 or you need certain particular variables to appear in special sections,
6157 for example to map to special hardware. The @code{section}
6158 attribute specifies that a variable (or function) lives in a particular
6159 section. For example, this small program uses several specific section names:
6162 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
6163 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
6164 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
6165 int init_data __attribute__ ((section ("INITDATA")));
6169 /* @r{Initialize stack pointer} */
6170 init_sp (stack + sizeof (stack));
6172 /* @r{Initialize initialized data} */
6173 memcpy (&init_data, &data, &edata - &data);
6175 /* @r{Turn on the serial ports} */
6182 Use the @code{section} attribute with
6183 @emph{global} variables and not @emph{local} variables,
6184 as shown in the example.
6186 You may use the @code{section} attribute with initialized or
6187 uninitialized global variables but the linker requires
6188 each object be defined once, with the exception that uninitialized
6189 variables tentatively go in the @code{common} (or @code{bss}) section
6190 and can be multiply ``defined''. Using the @code{section} attribute
6191 changes what section the variable goes into and may cause the
6192 linker to issue an error if an uninitialized variable has multiple
6193 definitions. You can force a variable to be initialized with the
6194 @option{-fno-common} flag or the @code{nocommon} attribute.
6196 Some file formats do not support arbitrary sections so the @code{section}
6197 attribute is not available on all platforms.
6198 If you need to map the entire contents of a module to a particular
6199 section, consider using the facilities of the linker instead.
6201 @item tls_model ("@var{tls_model}")
6202 @cindex @code{tls_model} variable attribute
6203 The @code{tls_model} attribute sets thread-local storage model
6204 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
6205 overriding @option{-ftls-model=} command-line switch on a per-variable
6207 The @var{tls_model} argument should be one of @code{global-dynamic},
6208 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
6210 Not all targets support this attribute.
6213 @cindex @code{unused} variable attribute
6214 This attribute, attached to a variable, means that the variable is meant
6215 to be possibly unused. GCC does not produce a warning for this
6219 @cindex @code{used} variable attribute
6220 This attribute, attached to a variable with static storage, means that
6221 the variable must be emitted even if it appears that the variable is not
6224 When applied to a static data member of a C++ class template, the
6225 attribute also means that the member is instantiated if the
6226 class itself is instantiated.
6228 @item vector_size (@var{bytes})
6229 @cindex @code{vector_size} variable attribute
6230 This attribute specifies the vector size for the variable, measured in
6231 bytes. For example, the declaration:
6234 int foo __attribute__ ((vector_size (16)));
6238 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
6239 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
6240 4 units of 4 bytes), the corresponding mode of @code{foo} is V4SI@.
6242 This attribute is only applicable to integral and float scalars,
6243 although arrays, pointers, and function return values are allowed in
6244 conjunction with this construct.
6246 Aggregates with this attribute are invalid, even if they are of the same
6247 size as a corresponding scalar. For example, the declaration:
6250 struct S @{ int a; @};
6251 struct S __attribute__ ((vector_size (16))) foo;
6255 is invalid even if the size of the structure is the same as the size of
6258 @item visibility ("@var{visibility_type}")
6259 @cindex @code{visibility} variable attribute
6260 This attribute affects the linkage of the declaration to which it is attached.
6261 The @code{visibility} attribute is described in
6262 @ref{Common Function Attributes}.
6265 @cindex @code{weak} variable attribute
6266 The @code{weak} attribute is described in
6267 @ref{Common Function Attributes}.
6271 @node ARC Variable Attributes
6272 @subsection ARC Variable Attributes
6276 @cindex @code{aux} variable attribute, ARC
6277 The @code{aux} attribute is used to directly access the ARC's
6278 auxiliary register space from C. The auxilirary register number is
6279 given via attribute argument.
6283 @node AVR Variable Attributes
6284 @subsection AVR Variable Attributes
6288 @cindex @code{progmem} variable attribute, AVR
6289 The @code{progmem} attribute is used on the AVR to place read-only
6290 data in the non-volatile program memory (flash). The @code{progmem}
6291 attribute accomplishes this by putting respective variables into a
6292 section whose name starts with @code{.progmem}.
6294 This attribute works similar to the @code{section} attribute
6295 but adds additional checking.
6298 @item @bullet{}@tie{} Ordinary AVR cores with 32 general purpose registers:
6299 @code{progmem} affects the location
6300 of the data but not how this data is accessed.
6301 In order to read data located with the @code{progmem} attribute
6302 (inline) assembler must be used.
6304 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
6305 #include <avr/pgmspace.h>
6307 /* Locate var in flash memory */
6308 const int var[2] PROGMEM = @{ 1, 2 @};
6310 int read_var (int i)
6312 /* Access var[] by accessor macro from avr/pgmspace.h */
6313 return (int) pgm_read_word (& var[i]);
6317 AVR is a Harvard architecture processor and data and read-only data
6318 normally resides in the data memory (RAM).
6320 See also the @ref{AVR Named Address Spaces} section for
6321 an alternate way to locate and access data in flash memory.
6323 @item @bullet{}@tie{} AVR cores with flash memory visible in the RAM address range:
6324 On such devices, there is no need for attribute @code{progmem} or
6325 @ref{AVR Named Address Spaces,,@code{__flash}} qualifier at all.
6326 Just use standard C / C++. The compiler will generate @code{LD*}
6327 instructions. As flash memory is visible in the RAM address range,
6328 and the default linker script does @emph{not} locate @code{.rodata} in
6329 RAM, no special features are needed in order not to waste RAM for
6330 read-only data or to read from flash. You might even get slightly better
6332 avoiding @code{progmem} and @code{__flash}. This applies to devices from
6333 families @code{avrtiny} and @code{avrxmega3}, see @ref{AVR Options} for
6336 @item @bullet{}@tie{}Reduced AVR Tiny cores like ATtiny40:
6337 The compiler adds @code{0x4000}
6338 to the addresses of objects and declarations in @code{progmem} and locates
6339 the objects in flash memory, namely in section @code{.progmem.data}.
6340 The offset is needed because the flash memory is visible in the RAM
6341 address space starting at address @code{0x4000}.
6343 Data in @code{progmem} can be accessed by means of ordinary C@tie{}code,
6344 no special functions or macros are needed.
6347 /* var is located in flash memory */
6348 extern const int var[2] __attribute__((progmem));
6350 int read_var (int i)
6356 Please notice that on these devices, there is no need for @code{progmem}
6362 @itemx io (@var{addr})
6363 @cindex @code{io} variable attribute, AVR
6364 Variables with the @code{io} attribute are used to address
6365 memory-mapped peripherals in the io address range.
6366 If an address is specified, the variable
6367 is assigned that address, and the value is interpreted as an
6368 address in the data address space.
6372 volatile int porta __attribute__((io (0x22)));
6375 The address specified in the address in the data address range.
6377 Otherwise, the variable it is not assigned an address, but the
6378 compiler will still use in/out instructions where applicable,
6379 assuming some other module assigns an address in the io address range.
6383 extern volatile int porta __attribute__((io));
6387 @itemx io_low (@var{addr})
6388 @cindex @code{io_low} variable attribute, AVR
6389 This is like the @code{io} attribute, but additionally it informs the
6390 compiler that the object lies in the lower half of the I/O area,
6391 allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
6395 @itemx address (@var{addr})
6396 @cindex @code{address} variable attribute, AVR
6397 Variables with the @code{address} attribute are used to address
6398 memory-mapped peripherals that may lie outside the io address range.
6401 volatile int porta __attribute__((address (0x600)));
6405 @cindex @code{absdata} variable attribute, AVR
6406 Variables in static storage and with the @code{absdata} attribute can
6407 be accessed by the @code{LDS} and @code{STS} instructions which take
6412 This attribute is only supported for the reduced AVR Tiny core
6416 You must make sure that respective data is located in the
6417 address range @code{0x40}@dots{}@code{0xbf} accessible by
6418 @code{LDS} and @code{STS}. One way to achieve this as an
6419 appropriate linker description file.
6422 If the location does not fit the address range of @code{LDS}
6423 and @code{STS}, there is currently (Binutils 2.26) just an unspecific
6426 @code{module.c:(.text+0x1c): warning: internal error: out of range error}
6431 See also the @option{-mabsdata} @ref{AVR Options,command-line option}.
6435 @node Blackfin Variable Attributes
6436 @subsection Blackfin Variable Attributes
6438 Three attributes are currently defined for the Blackfin.
6444 @cindex @code{l1_data} variable attribute, Blackfin
6445 @cindex @code{l1_data_A} variable attribute, Blackfin
6446 @cindex @code{l1_data_B} variable attribute, Blackfin
6447 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
6448 Variables with @code{l1_data} attribute are put into the specific section
6449 named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
6450 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
6451 attribute are put into the specific section named @code{.l1.data.B}.
6454 @cindex @code{l2} variable attribute, Blackfin
6455 Use this attribute on the Blackfin to place the variable into L2 SRAM.
6456 Variables with @code{l2} attribute are put into the specific section
6457 named @code{.l2.data}.
6460 @node H8/300 Variable Attributes
6461 @subsection H8/300 Variable Attributes
6463 These variable attributes are available for H8/300 targets:
6467 @cindex @code{eightbit_data} variable attribute, H8/300
6468 @cindex eight-bit data on the H8/300, H8/300H, and H8S
6469 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
6470 variable should be placed into the eight-bit data section.
6471 The compiler generates more efficient code for certain operations
6472 on data in the eight-bit data area. Note the eight-bit data area is limited to
6475 You must use GAS and GLD from GNU binutils version 2.7 or later for
6476 this attribute to work correctly.
6479 @cindex @code{tiny_data} variable attribute, H8/300
6480 @cindex tiny data section on the H8/300H and H8S
6481 Use this attribute on the H8/300H and H8S to indicate that the specified
6482 variable should be placed into the tiny data section.
6483 The compiler generates more efficient code for loads and stores
6484 on data in the tiny data section. Note the tiny data area is limited to
6485 slightly under 32KB of data.
6489 @node IA-64 Variable Attributes
6490 @subsection IA-64 Variable Attributes
6492 The IA-64 back end supports the following variable attribute:
6495 @item model (@var{model-name})
6496 @cindex @code{model} variable attribute, IA-64
6498 On IA-64, use this attribute to set the addressability of an object.
6499 At present, the only supported identifier for @var{model-name} is
6500 @code{small}, indicating addressability via ``small'' (22-bit)
6501 addresses (so that their addresses can be loaded with the @code{addl}
6502 instruction). Caveat: such addressing is by definition not position
6503 independent and hence this attribute must not be used for objects
6504 defined by shared libraries.
6508 @node M32R/D Variable Attributes
6509 @subsection M32R/D Variable Attributes
6511 One attribute is currently defined for the M32R/D@.
6514 @item model (@var{model-name})
6515 @cindex @code{model-name} variable attribute, M32R/D
6516 @cindex variable addressability on the M32R/D
6517 Use this attribute on the M32R/D to set the addressability of an object.
6518 The identifier @var{model-name} is one of @code{small}, @code{medium},
6519 or @code{large}, representing each of the code models.
6521 Small model objects live in the lower 16MB of memory (so that their
6522 addresses can be loaded with the @code{ld24} instruction).
6524 Medium and large model objects may live anywhere in the 32-bit address space
6525 (the compiler generates @code{seth/add3} instructions to load their
6529 @node MeP Variable Attributes
6530 @subsection MeP Variable Attributes
6532 The MeP target has a number of addressing modes and busses. The
6533 @code{near} space spans the standard memory space's first 16 megabytes
6534 (24 bits). The @code{far} space spans the entire 32-bit memory space.
6535 The @code{based} space is a 128-byte region in the memory space that
6536 is addressed relative to the @code{$tp} register. The @code{tiny}
6537 space is a 65536-byte region relative to the @code{$gp} register. In
6538 addition to these memory regions, the MeP target has a separate 16-bit
6539 control bus which is specified with @code{cb} attributes.
6544 @cindex @code{based} variable attribute, MeP
6545 Any variable with the @code{based} attribute is assigned to the
6546 @code{.based} section, and is accessed with relative to the
6547 @code{$tp} register.
6550 @cindex @code{tiny} variable attribute, MeP
6551 Likewise, the @code{tiny} attribute assigned variables to the
6552 @code{.tiny} section, relative to the @code{$gp} register.
6555 @cindex @code{near} variable attribute, MeP
6556 Variables with the @code{near} attribute are assumed to have addresses
6557 that fit in a 24-bit addressing mode. This is the default for large
6558 variables (@code{-mtiny=4} is the default) but this attribute can
6559 override @code{-mtiny=} for small variables, or override @code{-ml}.
6562 @cindex @code{far} variable attribute, MeP
6563 Variables with the @code{far} attribute are addressed using a full
6564 32-bit address. Since this covers the entire memory space, this
6565 allows modules to make no assumptions about where variables might be
6569 @cindex @code{io} variable attribute, MeP
6570 @itemx io (@var{addr})
6571 Variables with the @code{io} attribute are used to address
6572 memory-mapped peripherals. If an address is specified, the variable
6573 is assigned that address, else it is not assigned an address (it is
6574 assumed some other module assigns an address). Example:
6577 int timer_count __attribute__((io(0x123)));
6581 @itemx cb (@var{addr})
6582 @cindex @code{cb} variable attribute, MeP
6583 Variables with the @code{cb} attribute are used to access the control
6584 bus, using special instructions. @code{addr} indicates the control bus
6588 int cpu_clock __attribute__((cb(0x123)));
6593 @node Microsoft Windows Variable Attributes
6594 @subsection Microsoft Windows Variable Attributes
6596 You can use these attributes on Microsoft Windows targets.
6597 @ref{x86 Variable Attributes} for additional Windows compatibility
6598 attributes available on all x86 targets.
6603 @cindex @code{dllimport} variable attribute
6604 @cindex @code{dllexport} variable attribute
6605 The @code{dllimport} and @code{dllexport} attributes are described in
6606 @ref{Microsoft Windows Function Attributes}.
6609 @cindex @code{selectany} variable attribute
6610 The @code{selectany} attribute causes an initialized global variable to
6611 have link-once semantics. When multiple definitions of the variable are
6612 encountered by the linker, the first is selected and the remainder are
6613 discarded. Following usage by the Microsoft compiler, the linker is told
6614 @emph{not} to warn about size or content differences of the multiple
6617 Although the primary usage of this attribute is for POD types, the
6618 attribute can also be applied to global C++ objects that are initialized
6619 by a constructor. In this case, the static initialization and destruction
6620 code for the object is emitted in each translation defining the object,
6621 but the calls to the constructor and destructor are protected by a
6622 link-once guard variable.
6624 The @code{selectany} attribute is only available on Microsoft Windows
6625 targets. You can use @code{__declspec (selectany)} as a synonym for
6626 @code{__attribute__ ((selectany))} for compatibility with other
6630 @cindex @code{shared} variable attribute
6631 On Microsoft Windows, in addition to putting variable definitions in a named
6632 section, the section can also be shared among all running copies of an
6633 executable or DLL@. For example, this small program defines shared data
6634 by putting it in a named section @code{shared} and marking the section
6638 int foo __attribute__((section ("shared"), shared)) = 0;
6643 /* @r{Read and write foo. All running
6644 copies see the same value.} */
6650 You may only use the @code{shared} attribute along with @code{section}
6651 attribute with a fully-initialized global definition because of the way
6652 linkers work. See @code{section} attribute for more information.
6654 The @code{shared} attribute is only available on Microsoft Windows@.
6658 @node MSP430 Variable Attributes
6659 @subsection MSP430 Variable Attributes
6663 @cindex @code{noinit} variable attribute, MSP430
6664 Any data with the @code{noinit} attribute will not be initialised by
6665 the C runtime startup code, or the program loader. Not initialising
6666 data in this way can reduce program startup times.
6669 @cindex @code{persistent} variable attribute, MSP430
6670 Any variable with the @code{persistent} attribute will not be
6671 initialised by the C runtime startup code. Instead its value will be
6672 set once, when the application is loaded, and then never initialised
6673 again, even if the processor is reset or the program restarts.
6674 Persistent data is intended to be placed into FLASH RAM, where its
6675 value will be retained across resets. The linker script being used to
6676 create the application should ensure that persistent data is correctly
6682 @cindex @code{lower} variable attribute, MSP430
6683 @cindex @code{upper} variable attribute, MSP430
6684 @cindex @code{either} variable attribute, MSP430
6685 These attributes are the same as the MSP430 function attributes of the
6686 same name (@pxref{MSP430 Function Attributes}).
6687 These attributes can be applied to both functions and variables.
6690 @node Nvidia PTX Variable Attributes
6691 @subsection Nvidia PTX Variable Attributes
6693 These variable attributes are supported by the Nvidia PTX back end:
6697 @cindex @code{shared} attribute, Nvidia PTX
6698 Use this attribute to place a variable in the @code{.shared} memory space.
6699 This memory space is private to each cooperative thread array; only threads
6700 within one thread block refer to the same instance of the variable.
6701 The runtime does not initialize variables in this memory space.
6704 @node PowerPC Variable Attributes
6705 @subsection PowerPC Variable Attributes
6707 Three attributes currently are defined for PowerPC configurations:
6708 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
6710 @cindex @code{ms_struct} variable attribute, PowerPC
6711 @cindex @code{gcc_struct} variable attribute, PowerPC
6712 For full documentation of the struct attributes please see the
6713 documentation in @ref{x86 Variable Attributes}.
6715 @cindex @code{altivec} variable attribute, PowerPC
6716 For documentation of @code{altivec} attribute please see the
6717 documentation in @ref{PowerPC Type Attributes}.
6719 @node RL78 Variable Attributes
6720 @subsection RL78 Variable Attributes
6722 @cindex @code{saddr} variable attribute, RL78
6723 The RL78 back end supports the @code{saddr} variable attribute. This
6724 specifies placement of the corresponding variable in the SADDR area,
6725 which can be accessed more efficiently than the default memory region.
6727 @node SPU Variable Attributes
6728 @subsection SPU Variable Attributes
6730 @cindex @code{spu_vector} variable attribute, SPU
6731 The SPU supports the @code{spu_vector} attribute for variables. For
6732 documentation of this attribute please see the documentation in
6733 @ref{SPU Type Attributes}.
6735 @node V850 Variable Attributes
6736 @subsection V850 Variable Attributes
6738 These variable attributes are supported by the V850 back end:
6743 @cindex @code{sda} variable attribute, V850
6744 Use this attribute to explicitly place a variable in the small data area,
6745 which can hold up to 64 kilobytes.
6748 @cindex @code{tda} variable attribute, V850
6749 Use this attribute to explicitly place a variable in the tiny data area,
6750 which can hold up to 256 bytes in total.
6753 @cindex @code{zda} variable attribute, V850
6754 Use this attribute to explicitly place a variable in the first 32 kilobytes
6758 @node x86 Variable Attributes
6759 @subsection x86 Variable Attributes
6761 Two attributes are currently defined for x86 configurations:
6762 @code{ms_struct} and @code{gcc_struct}.
6767 @cindex @code{ms_struct} variable attribute, x86
6768 @cindex @code{gcc_struct} variable attribute, x86
6770 If @code{packed} is used on a structure, or if bit-fields are used,
6771 it may be that the Microsoft ABI lays out the structure differently
6772 than the way GCC normally does. Particularly when moving packed
6773 data between functions compiled with GCC and the native Microsoft compiler
6774 (either via function call or as data in a file), it may be necessary to access
6777 The @code{ms_struct} and @code{gcc_struct} attributes correspond
6778 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
6779 command-line options, respectively;
6780 see @ref{x86 Options}, for details of how structure layout is affected.
6781 @xref{x86 Type Attributes}, for information about the corresponding
6782 attributes on types.
6786 @node Xstormy16 Variable Attributes
6787 @subsection Xstormy16 Variable Attributes
6789 One attribute is currently defined for xstormy16 configurations:
6794 @cindex @code{below100} variable attribute, Xstormy16
6796 If a variable has the @code{below100} attribute (@code{BELOW100} is
6797 allowed also), GCC places the variable in the first 0x100 bytes of
6798 memory and use special opcodes to access it. Such variables are
6799 placed in either the @code{.bss_below100} section or the
6800 @code{.data_below100} section.
6804 @node Type Attributes
6805 @section Specifying Attributes of Types
6806 @cindex attribute of types
6807 @cindex type attributes
6809 The keyword @code{__attribute__} allows you to specify special
6810 attributes of types. Some type attributes apply only to @code{struct}
6811 and @code{union} types, while others can apply to any type defined
6812 via a @code{typedef} declaration. Other attributes are defined for
6813 functions (@pxref{Function Attributes}), labels (@pxref{Label
6814 Attributes}), enumerators (@pxref{Enumerator Attributes}),
6815 statements (@pxref{Statement Attributes}), and for
6816 variables (@pxref{Variable Attributes}).
6818 The @code{__attribute__} keyword is followed by an attribute specification
6819 inside double parentheses.
6821 You may specify type attributes in an enum, struct or union type
6822 declaration or definition by placing them immediately after the
6823 @code{struct}, @code{union} or @code{enum} keyword. A less preferred
6824 syntax is to place them just past the closing curly brace of the
6827 You can also include type attributes in a @code{typedef} declaration.
6828 @xref{Attribute Syntax}, for details of the exact syntax for using
6832 * Common Type Attributes::
6833 * ARC Type Attributes::
6834 * ARM Type Attributes::
6835 * MeP Type Attributes::
6836 * PowerPC Type Attributes::
6837 * SPU Type Attributes::
6838 * x86 Type Attributes::
6841 @node Common Type Attributes
6842 @subsection Common Type Attributes
6844 The following type attributes are supported on most targets.
6847 @cindex @code{aligned} type attribute
6848 @item aligned (@var{alignment})
6849 This attribute specifies a minimum alignment (in bytes) for variables
6850 of the specified type. For example, the declarations:
6853 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
6854 typedef int more_aligned_int __attribute__ ((aligned (8)));
6858 force the compiler to ensure (as far as it can) that each variable whose
6859 type is @code{struct S} or @code{more_aligned_int} is allocated and
6860 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
6861 variables of type @code{struct S} aligned to 8-byte boundaries allows
6862 the compiler to use the @code{ldd} and @code{std} (doubleword load and
6863 store) instructions when copying one variable of type @code{struct S} to
6864 another, thus improving run-time efficiency.
6866 Note that the alignment of any given @code{struct} or @code{union} type
6867 is required by the ISO C standard to be at least a perfect multiple of
6868 the lowest common multiple of the alignments of all of the members of
6869 the @code{struct} or @code{union} in question. This means that you @emph{can}
6870 effectively adjust the alignment of a @code{struct} or @code{union}
6871 type by attaching an @code{aligned} attribute to any one of the members
6872 of such a type, but the notation illustrated in the example above is a
6873 more obvious, intuitive, and readable way to request the compiler to
6874 adjust the alignment of an entire @code{struct} or @code{union} type.
6876 As in the preceding example, you can explicitly specify the alignment
6877 (in bytes) that you wish the compiler to use for a given @code{struct}
6878 or @code{union} type. Alternatively, you can leave out the alignment factor
6879 and just ask the compiler to align a type to the maximum
6880 useful alignment for the target machine you are compiling for. For
6881 example, you could write:
6884 struct S @{ short f[3]; @} __attribute__ ((aligned));
6887 Whenever you leave out the alignment factor in an @code{aligned}
6888 attribute specification, the compiler automatically sets the alignment
6889 for the type to the largest alignment that is ever used for any data
6890 type on the target machine you are compiling for. Doing this can often
6891 make copy operations more efficient, because the compiler can use
6892 whatever instructions copy the biggest chunks of memory when performing
6893 copies to or from the variables that have types that you have aligned
6896 In the example above, if the size of each @code{short} is 2 bytes, then
6897 the size of the entire @code{struct S} type is 6 bytes. The smallest
6898 power of two that is greater than or equal to that is 8, so the
6899 compiler sets the alignment for the entire @code{struct S} type to 8
6902 Note that although you can ask the compiler to select a time-efficient
6903 alignment for a given type and then declare only individual stand-alone
6904 objects of that type, the compiler's ability to select a time-efficient
6905 alignment is primarily useful only when you plan to create arrays of
6906 variables having the relevant (efficiently aligned) type. If you
6907 declare or use arrays of variables of an efficiently-aligned type, then
6908 it is likely that your program also does pointer arithmetic (or
6909 subscripting, which amounts to the same thing) on pointers to the
6910 relevant type, and the code that the compiler generates for these
6911 pointer arithmetic operations is often more efficient for
6912 efficiently-aligned types than for other types.
6914 Note that the effectiveness of @code{aligned} attributes may be limited
6915 by inherent limitations in your linker. On many systems, the linker is
6916 only able to arrange for variables to be aligned up to a certain maximum
6917 alignment. (For some linkers, the maximum supported alignment may
6918 be very very small.) If your linker is only able to align variables
6919 up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
6920 in an @code{__attribute__} still only provides you with 8-byte
6921 alignment. See your linker documentation for further information.
6923 The @code{aligned} attribute can only increase alignment. Alignment
6924 can be decreased by specifying the @code{packed} attribute. See below.
6926 @cindex @code{warn_if_not_aligned} type attribute
6927 @item warn_if_not_aligned (@var{alignment})
6928 This attribute specifies a threshold for the structure field, measured
6929 in bytes. If the structure field is aligned below the threshold, a
6930 warning will be issued. For example, the declaration:
6933 typedef unsigned long long __u64
6934 __attribute__((aligned(4),warn_if_not_aligned(8)));
6945 causes the compiler to issue an warning on @code{struct foo}, like
6946 @samp{warning: alignment 4 of 'struct foo' is less than 8}.
6947 It is used to define @code{struct foo} in such a way that
6948 @code{struct foo} has the same layout and the structure field @code{x}
6949 has the same alignment when @code{__u64} is aligned at either 4 or
6950 8 bytes. Align @code{struct foo} to 8 bytes:
6958 @} __attribute__((aligned(8)));
6962 silences the warning. The compiler also issues a warning, like
6963 @samp{warning: 'x' offset 12 in 'struct foo' isn't aligned to 8},
6964 when the structure field has the misaligned offset:
6973 @} __attribute__((aligned(8)));
6976 This warning can be disabled by @option{-Wno-if-not-aligned}.
6978 @item bnd_variable_size
6979 @cindex @code{bnd_variable_size} type attribute
6980 @cindex Pointer Bounds Checker attributes
6981 When applied to a structure field, this attribute tells Pointer
6982 Bounds Checker that the size of this field should not be computed
6983 using static type information. It may be used to mark variably-sized
6984 static array fields placed at the end of a structure.
6992 S *p = (S *)malloc (sizeof(S) + 100);
6993 p->data[10] = 0; //Bounds violation
6997 By using an attribute for the field we may avoid unwanted bound
7004 char data[1] __attribute__((bnd_variable_size));
7006 S *p = (S *)malloc (sizeof(S) + 100);
7007 p->data[10] = 0; //OK
7011 @itemx deprecated (@var{msg})
7012 @cindex @code{deprecated} type attribute
7013 The @code{deprecated} attribute results in a warning if the type
7014 is used anywhere in the source file. This is useful when identifying
7015 types that are expected to be removed in a future version of a program.
7016 If possible, the warning also includes the location of the declaration
7017 of the deprecated type, to enable users to easily find further
7018 information about why the type is deprecated, or what they should do
7019 instead. Note that the warnings only occur for uses and then only
7020 if the type is being applied to an identifier that itself is not being
7021 declared as deprecated.
7024 typedef int T1 __attribute__ ((deprecated));
7028 typedef T1 T3 __attribute__ ((deprecated));
7029 T3 z __attribute__ ((deprecated));
7033 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
7034 warning is issued for line 4 because T2 is not explicitly
7035 deprecated. Line 5 has no warning because T3 is explicitly
7036 deprecated. Similarly for line 6. The optional @var{msg}
7037 argument, which must be a string, is printed in the warning if
7040 The @code{deprecated} attribute can also be used for functions and
7041 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
7043 @item designated_init
7044 @cindex @code{designated_init} type attribute
7045 This attribute may only be applied to structure types. It indicates
7046 that any initialization of an object of this type must use designated
7047 initializers rather than positional initializers. The intent of this
7048 attribute is to allow the programmer to indicate that a structure's
7049 layout may change, and that therefore relying on positional
7050 initialization will result in future breakage.
7052 GCC emits warnings based on this attribute by default; use
7053 @option{-Wno-designated-init} to suppress them.
7056 @cindex @code{may_alias} type attribute
7057 Accesses through pointers to types with this attribute are not subject
7058 to type-based alias analysis, but are instead assumed to be able to alias
7059 any other type of objects.
7060 In the context of section 6.5 paragraph 7 of the C99 standard,
7061 an lvalue expression
7062 dereferencing such a pointer is treated like having a character type.
7063 See @option{-fstrict-aliasing} for more information on aliasing issues.
7064 This extension exists to support some vector APIs, in which pointers to
7065 one vector type are permitted to alias pointers to a different vector type.
7067 Note that an object of a type with this attribute does not have any
7073 typedef short __attribute__((__may_alias__)) short_a;
7079 short_a *b = (short_a *) &a;
7083 if (a == 0x12345678)
7091 If you replaced @code{short_a} with @code{short} in the variable
7092 declaration, the above program would abort when compiled with
7093 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
7097 @cindex @code{packed} type attribute
7098 This attribute, attached to @code{struct} or @code{union} type
7099 definition, specifies that each member (other than zero-width bit-fields)
7100 of the structure or union is placed to minimize the memory required. When
7101 attached to an @code{enum} definition, it indicates that the smallest
7102 integral type should be used.
7104 @opindex fshort-enums
7105 Specifying the @code{packed} attribute for @code{struct} and @code{union}
7106 types is equivalent to specifying the @code{packed} attribute on each
7107 of the structure or union members. Specifying the @option{-fshort-enums}
7108 flag on the command line is equivalent to specifying the @code{packed}
7109 attribute on all @code{enum} definitions.
7111 In the following example @code{struct my_packed_struct}'s members are
7112 packed closely together, but the internal layout of its @code{s} member
7113 is not packed---to do that, @code{struct my_unpacked_struct} needs to
7117 struct my_unpacked_struct
7123 struct __attribute__ ((__packed__)) my_packed_struct
7127 struct my_unpacked_struct s;
7131 You may only specify the @code{packed} attribute attribute on the definition
7132 of an @code{enum}, @code{struct} or @code{union}, not on a @code{typedef}
7133 that does not also define the enumerated type, structure or union.
7135 @item scalar_storage_order ("@var{endianness}")
7136 @cindex @code{scalar_storage_order} type attribute
7137 When attached to a @code{union} or a @code{struct}, this attribute sets
7138 the storage order, aka endianness, of the scalar fields of the type, as
7139 well as the array fields whose component is scalar. The supported
7140 endiannesses are @code{big-endian} and @code{little-endian}. The attribute
7141 has no effects on fields which are themselves a @code{union}, a @code{struct}
7142 or an array whose component is a @code{union} or a @code{struct}, and it is
7143 possible for these fields to have a different scalar storage order than the
7146 This attribute is supported only for targets that use a uniform default
7147 scalar storage order (fortunately, most of them), i.e. targets that store
7148 the scalars either all in big-endian or all in little-endian.
7150 Additional restrictions are enforced for types with the reverse scalar
7151 storage order with regard to the scalar storage order of the target:
7154 @item Taking the address of a scalar field of a @code{union} or a
7155 @code{struct} with reverse scalar storage order is not permitted and yields
7157 @item Taking the address of an array field, whose component is scalar, of
7158 a @code{union} or a @code{struct} with reverse scalar storage order is
7159 permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
7161 @item Taking the address of a @code{union} or a @code{struct} with reverse
7162 scalar storage order is permitted.
7165 These restrictions exist because the storage order attribute is lost when
7166 the address of a scalar or the address of an array with scalar component is
7167 taken, so storing indirectly through this address generally does not work.
7168 The second case is nevertheless allowed to be able to perform a block copy
7169 from or to the array.
7171 Moreover, the use of type punning or aliasing to toggle the storage order
7172 is not supported; that is to say, a given scalar object cannot be accessed
7173 through distinct types that assign a different storage order to it.
7175 @item transparent_union
7176 @cindex @code{transparent_union} type attribute
7178 This attribute, attached to a @code{union} type definition, indicates
7179 that any function parameter having that union type causes calls to that
7180 function to be treated in a special way.
7182 First, the argument corresponding to a transparent union type can be of
7183 any type in the union; no cast is required. Also, if the union contains
7184 a pointer type, the corresponding argument can be a null pointer
7185 constant or a void pointer expression; and if the union contains a void
7186 pointer type, the corresponding argument can be any pointer expression.
7187 If the union member type is a pointer, qualifiers like @code{const} on
7188 the referenced type must be respected, just as with normal pointer
7191 Second, the argument is passed to the function using the calling
7192 conventions of the first member of the transparent union, not the calling
7193 conventions of the union itself. All members of the union must have the
7194 same machine representation; this is necessary for this argument passing
7197 Transparent unions are designed for library functions that have multiple
7198 interfaces for compatibility reasons. For example, suppose the
7199 @code{wait} function must accept either a value of type @code{int *} to
7200 comply with POSIX, or a value of type @code{union wait *} to comply with
7201 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
7202 @code{wait} would accept both kinds of arguments, but it would also
7203 accept any other pointer type and this would make argument type checking
7204 less useful. Instead, @code{<sys/wait.h>} might define the interface
7208 typedef union __attribute__ ((__transparent_union__))
7212 @} wait_status_ptr_t;
7214 pid_t wait (wait_status_ptr_t);
7218 This interface allows either @code{int *} or @code{union wait *}
7219 arguments to be passed, using the @code{int *} calling convention.
7220 The program can call @code{wait} with arguments of either type:
7223 int w1 () @{ int w; return wait (&w); @}
7224 int w2 () @{ union wait w; return wait (&w); @}
7228 With this interface, @code{wait}'s implementation might look like this:
7231 pid_t wait (wait_status_ptr_t p)
7233 return waitpid (-1, p.__ip, 0);
7238 @cindex @code{unused} type attribute
7239 When attached to a type (including a @code{union} or a @code{struct}),
7240 this attribute means that variables of that type are meant to appear
7241 possibly unused. GCC does not produce a warning for any variables of
7242 that type, even if the variable appears to do nothing. This is often
7243 the case with lock or thread classes, which are usually defined and then
7244 not referenced, but contain constructors and destructors that have
7245 nontrivial bookkeeping functions.
7248 @cindex @code{visibility} type attribute
7249 In C++, attribute visibility (@pxref{Function Attributes}) can also be
7250 applied to class, struct, union and enum types. Unlike other type
7251 attributes, the attribute must appear between the initial keyword and
7252 the name of the type; it cannot appear after the body of the type.
7254 Note that the type visibility is applied to vague linkage entities
7255 associated with the class (vtable, typeinfo node, etc.). In
7256 particular, if a class is thrown as an exception in one shared object
7257 and caught in another, the class must have default visibility.
7258 Otherwise the two shared objects are unable to use the same
7259 typeinfo node and exception handling will break.
7263 To specify multiple attributes, separate them by commas within the
7264 double parentheses: for example, @samp{__attribute__ ((aligned (16),
7267 @node ARC Type Attributes
7268 @subsection ARC Type Attributes
7270 @cindex @code{uncached} type attribute, ARC
7271 Declaring objects with @code{uncached} allows you to exclude
7272 data-cache participation in load and store operations on those objects
7273 without involving the additional semantic implications of
7274 @code{volatile}. The @code{.di} instruction suffix is used for all
7275 loads and stores of data declared @code{uncached}.
7277 @node ARM Type Attributes
7278 @subsection ARM Type Attributes
7280 @cindex @code{notshared} type attribute, ARM
7281 On those ARM targets that support @code{dllimport} (such as Symbian
7282 OS), you can use the @code{notshared} attribute to indicate that the
7283 virtual table and other similar data for a class should not be
7284 exported from a DLL@. For example:
7287 class __declspec(notshared) C @{
7289 __declspec(dllimport) C();
7293 __declspec(dllexport)
7298 In this code, @code{C::C} is exported from the current DLL, but the
7299 virtual table for @code{C} is not exported. (You can use
7300 @code{__attribute__} instead of @code{__declspec} if you prefer, but
7301 most Symbian OS code uses @code{__declspec}.)
7303 @node MeP Type Attributes
7304 @subsection MeP Type Attributes
7306 @cindex @code{based} type attribute, MeP
7307 @cindex @code{tiny} type attribute, MeP
7308 @cindex @code{near} type attribute, MeP
7309 @cindex @code{far} type attribute, MeP
7310 Many of the MeP variable attributes may be applied to types as well.
7311 Specifically, the @code{based}, @code{tiny}, @code{near}, and
7312 @code{far} attributes may be applied to either. The @code{io} and
7313 @code{cb} attributes may not be applied to types.
7315 @node PowerPC Type Attributes
7316 @subsection PowerPC Type Attributes
7318 Three attributes currently are defined for PowerPC configurations:
7319 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
7321 @cindex @code{ms_struct} type attribute, PowerPC
7322 @cindex @code{gcc_struct} type attribute, PowerPC
7323 For full documentation of the @code{ms_struct} and @code{gcc_struct}
7324 attributes please see the documentation in @ref{x86 Type Attributes}.
7326 @cindex @code{altivec} type attribute, PowerPC
7327 The @code{altivec} attribute allows one to declare AltiVec vector data
7328 types supported by the AltiVec Programming Interface Manual. The
7329 attribute requires an argument to specify one of three vector types:
7330 @code{vector__}, @code{pixel__} (always followed by unsigned short),
7331 and @code{bool__} (always followed by unsigned).
7334 __attribute__((altivec(vector__)))
7335 __attribute__((altivec(pixel__))) unsigned short
7336 __attribute__((altivec(bool__))) unsigned
7339 These attributes mainly are intended to support the @code{__vector},
7340 @code{__pixel}, and @code{__bool} AltiVec keywords.
7342 @node SPU Type Attributes
7343 @subsection SPU Type Attributes
7345 @cindex @code{spu_vector} type attribute, SPU
7346 The SPU supports the @code{spu_vector} attribute for types. This attribute
7347 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
7348 Language Extensions Specification. It is intended to support the
7349 @code{__vector} keyword.
7351 @node x86 Type Attributes
7352 @subsection x86 Type Attributes
7354 Two attributes are currently defined for x86 configurations:
7355 @code{ms_struct} and @code{gcc_struct}.
7361 @cindex @code{ms_struct} type attribute, x86
7362 @cindex @code{gcc_struct} type attribute, x86
7364 If @code{packed} is used on a structure, or if bit-fields are used
7365 it may be that the Microsoft ABI packs them differently
7366 than GCC normally packs them. Particularly when moving packed
7367 data between functions compiled with GCC and the native Microsoft compiler
7368 (either via function call or as data in a file), it may be necessary to access
7371 The @code{ms_struct} and @code{gcc_struct} attributes correspond
7372 to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
7373 command-line options, respectively;
7374 see @ref{x86 Options}, for details of how structure layout is affected.
7375 @xref{x86 Variable Attributes}, for information about the corresponding
7376 attributes on variables.
7380 @node Label Attributes
7381 @section Label Attributes
7382 @cindex Label Attributes
7384 GCC allows attributes to be set on C labels. @xref{Attribute Syntax}, for
7385 details of the exact syntax for using attributes. Other attributes are
7386 available for functions (@pxref{Function Attributes}), variables
7387 (@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
7388 statements (@pxref{Statement Attributes}), and for types
7389 (@pxref{Type Attributes}).
7391 This example uses the @code{cold} label attribute to indicate the
7392 @code{ErrorHandling} branch is unlikely to be taken and that the
7393 @code{ErrorHandling} label is unused:
7397 asm goto ("some asm" : : : : NoError);
7399 /* This branch (the fall-through from the asm) is less commonly used */
7401 __attribute__((cold, unused)); /* Semi-colon is required here */
7406 printf("no error\n");
7412 @cindex @code{unused} label attribute
7413 This feature is intended for program-generated code that may contain
7414 unused labels, but which is compiled with @option{-Wall}. It is
7415 not normally appropriate to use in it human-written code, though it
7416 could be useful in cases where the code that jumps to the label is
7417 contained within an @code{#ifdef} conditional.
7420 @cindex @code{hot} label attribute
7421 The @code{hot} attribute on a label is used to inform the compiler that
7422 the path following the label is more likely than paths that are not so
7423 annotated. This attribute is used in cases where @code{__builtin_expect}
7424 cannot be used, for instance with computed goto or @code{asm goto}.
7427 @cindex @code{cold} label attribute
7428 The @code{cold} attribute on labels is used to inform the compiler that
7429 the path following the label is unlikely to be executed. This attribute
7430 is used in cases where @code{__builtin_expect} cannot be used, for instance
7431 with computed goto or @code{asm goto}.
7435 @node Enumerator Attributes
7436 @section Enumerator Attributes
7437 @cindex Enumerator Attributes
7439 GCC allows attributes to be set on enumerators. @xref{Attribute Syntax}, for
7440 details of the exact syntax for using attributes. Other attributes are
7441 available for functions (@pxref{Function Attributes}), variables
7442 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), statements
7443 (@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).
7445 This example uses the @code{deprecated} enumerator attribute to indicate the
7446 @code{oldval} enumerator is deprecated:
7450 oldval __attribute__((deprecated)),
7463 @cindex @code{deprecated} enumerator attribute
7464 The @code{deprecated} attribute results in a warning if the enumerator
7465 is used anywhere in the source file. This is useful when identifying
7466 enumerators that are expected to be removed in a future version of a
7467 program. The warning also includes the location of the declaration
7468 of the deprecated enumerator, to enable users to easily find further
7469 information about why the enumerator is deprecated, or what they should
7470 do instead. Note that the warnings only occurs for uses.
7474 @node Statement Attributes
7475 @section Statement Attributes
7476 @cindex Statement Attributes
7478 GCC allows attributes to be set on null statements. @xref{Attribute Syntax},
7479 for details of the exact syntax for using attributes. Other attributes are
7480 available for functions (@pxref{Function Attributes}), variables
7481 (@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), enumerators
7482 (@pxref{Enumerator Attributes}), and for types (@pxref{Type Attributes}).
7484 This example uses the @code{fallthrough} statement attribute to indicate that
7485 the @option{-Wimplicit-fallthrough} warning should not be emitted:
7492 __attribute__((fallthrough));
7500 @cindex @code{fallthrough} statement attribute
7501 The @code{fallthrough} attribute with a null statement serves as a
7502 fallthrough statement. It hints to the compiler that a statement
7503 that falls through to another case label, or user-defined label
7504 in a switch statement is intentional and thus the
7505 @option{-Wimplicit-fallthrough} warning must not trigger. The
7506 fallthrough attribute may appear at most once in each attribute
7507 list, and may not be mixed with other attributes. It can only
7508 be used in a switch statement (the compiler will issue an error
7509 otherwise), after a preceding statement and before a logically
7510 succeeding case label, or user-defined label.
7514 @node Attribute Syntax
7515 @section Attribute Syntax
7516 @cindex attribute syntax
7518 This section describes the syntax with which @code{__attribute__} may be
7519 used, and the constructs to which attribute specifiers bind, for the C
7520 language. Some details may vary for C++ and Objective-C@. Because of
7521 infelicities in the grammar for attributes, some forms described here
7522 may not be successfully parsed in all cases.
7524 There are some problems with the semantics of attributes in C++. For
7525 example, there are no manglings for attributes, although they may affect
7526 code generation, so problems may arise when attributed types are used in
7527 conjunction with templates or overloading. Similarly, @code{typeid}
7528 does not distinguish between types with different attributes. Support
7529 for attributes in C++ may be restricted in future to attributes on
7530 declarations only, but not on nested declarators.
7532 @xref{Function Attributes}, for details of the semantics of attributes
7533 applying to functions. @xref{Variable Attributes}, for details of the
7534 semantics of attributes applying to variables. @xref{Type Attributes},
7535 for details of the semantics of attributes applying to structure, union
7536 and enumerated types.
7537 @xref{Label Attributes}, for details of the semantics of attributes
7539 @xref{Enumerator Attributes}, for details of the semantics of attributes
7540 applying to enumerators.
7541 @xref{Statement Attributes}, for details of the semantics of attributes
7542 applying to statements.
7544 An @dfn{attribute specifier} is of the form
7545 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
7546 is a possibly empty comma-separated sequence of @dfn{attributes}, where
7547 each attribute is one of the following:
7551 Empty. Empty attributes are ignored.
7555 (which may be an identifier such as @code{unused}, or a reserved
7556 word such as @code{const}).
7559 An attribute name followed by a parenthesized list of
7560 parameters for the attribute.
7561 These parameters take one of the following forms:
7565 An identifier. For example, @code{mode} attributes use this form.
7568 An identifier followed by a comma and a non-empty comma-separated list
7569 of expressions. For example, @code{format} attributes use this form.
7572 A possibly empty comma-separated list of expressions. For example,
7573 @code{format_arg} attributes use this form with the list being a single
7574 integer constant expression, and @code{alias} attributes use this form
7575 with the list being a single string constant.
7579 An @dfn{attribute specifier list} is a sequence of one or more attribute
7580 specifiers, not separated by any other tokens.
7582 You may optionally specify attribute names with @samp{__}
7583 preceding and following the name.
7584 This allows you to use them in header files without
7585 being concerned about a possible macro of the same name. For example,
7586 you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.
7589 @subsubheading Label Attributes
7591 In GNU C, an attribute specifier list may appear after the colon following a
7592 label, other than a @code{case} or @code{default} label. GNU C++ only permits
7593 attributes on labels if the attribute specifier is immediately
7594 followed by a semicolon (i.e., the label applies to an empty
7595 statement). If the semicolon is missing, C++ label attributes are
7596 ambiguous, as it is permissible for a declaration, which could begin
7597 with an attribute list, to be labelled in C++. Declarations cannot be
7598 labelled in C90 or C99, so the ambiguity does not arise there.
7600 @subsubheading Enumerator Attributes
7602 In GNU C, an attribute specifier list may appear as part of an enumerator.
7603 The attribute goes after the enumeration constant, before @code{=}, if
7604 present. The optional attribute in the enumerator appertains to the
7605 enumeration constant. It is not possible to place the attribute after
7606 the constant expression, if present.
7608 @subsubheading Statement Attributes
7609 In GNU C, an attribute specifier list may appear as part of a null
7610 statement. The attribute goes before the semicolon.
7612 @subsubheading Type Attributes
7614 An attribute specifier list may appear as part of a @code{struct},
7615 @code{union} or @code{enum} specifier. It may go either immediately
7616 after the @code{struct}, @code{union} or @code{enum} keyword, or after
7617 the closing brace. The former syntax is preferred.
7618 Where attribute specifiers follow the closing brace, they are considered
7619 to relate to the structure, union or enumerated type defined, not to any
7620 enclosing declaration the type specifier appears in, and the type
7621 defined is not complete until after the attribute specifiers.
7622 @c Otherwise, there would be the following problems: a shift/reduce
7623 @c conflict between attributes binding the struct/union/enum and
7624 @c binding to the list of specifiers/qualifiers; and "aligned"
7625 @c attributes could use sizeof for the structure, but the size could be
7626 @c changed later by "packed" attributes.
7629 @subsubheading All other attributes
7631 Otherwise, an attribute specifier appears as part of a declaration,
7632 counting declarations of unnamed parameters and type names, and relates
7633 to that declaration (which may be nested in another declaration, for
7634 example in the case of a parameter declaration), or to a particular declarator
7635 within a declaration. Where an
7636 attribute specifier is applied to a parameter declared as a function or
7637 an array, it should apply to the function or array rather than the
7638 pointer to which the parameter is implicitly converted, but this is not
7639 yet correctly implemented.
7641 Any list of specifiers and qualifiers at the start of a declaration may
7642 contain attribute specifiers, whether or not such a list may in that
7643 context contain storage class specifiers. (Some attributes, however,
7644 are essentially in the nature of storage class specifiers, and only make
7645 sense where storage class specifiers may be used; for example,
7646 @code{section}.) There is one necessary limitation to this syntax: the
7647 first old-style parameter declaration in a function definition cannot
7648 begin with an attribute specifier, because such an attribute applies to
7649 the function instead by syntax described below (which, however, is not
7650 yet implemented in this case). In some other cases, attribute
7651 specifiers are permitted by this grammar but not yet supported by the
7652 compiler. All attribute specifiers in this place relate to the
7653 declaration as a whole. In the obsolescent usage where a type of
7654 @code{int} is implied by the absence of type specifiers, such a list of
7655 specifiers and qualifiers may be an attribute specifier list with no
7656 other specifiers or qualifiers.
7658 At present, the first parameter in a function prototype must have some
7659 type specifier that is not an attribute specifier; this resolves an
7660 ambiguity in the interpretation of @code{void f(int
7661 (__attribute__((foo)) x))}, but is subject to change. At present, if
7662 the parentheses of a function declarator contain only attributes then
7663 those attributes are ignored, rather than yielding an error or warning
7664 or implying a single parameter of type int, but this is subject to
7667 An attribute specifier list may appear immediately before a declarator
7668 (other than the first) in a comma-separated list of declarators in a
7669 declaration of more than one identifier using a single list of
7670 specifiers and qualifiers. Such attribute specifiers apply
7671 only to the identifier before whose declarator they appear. For
7675 __attribute__((noreturn)) void d0 (void),
7676 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
7681 the @code{noreturn} attribute applies to all the functions
7682 declared; the @code{format} attribute only applies to @code{d1}.
7684 An attribute specifier list may appear immediately before the comma,
7685 @code{=} or semicolon terminating the declaration of an identifier other
7686 than a function definition. Such attribute specifiers apply
7687 to the declared object or function. Where an
7688 assembler name for an object or function is specified (@pxref{Asm
7689 Labels}), the attribute must follow the @code{asm}
7692 An attribute specifier list may, in future, be permitted to appear after
7693 the declarator in a function definition (before any old-style parameter
7694 declarations or the function body).
7696 Attribute specifiers may be mixed with type qualifiers appearing inside
7697 the @code{[]} of a parameter array declarator, in the C99 construct by
7698 which such qualifiers are applied to the pointer to which the array is
7699 implicitly converted. Such attribute specifiers apply to the pointer,
7700 not to the array, but at present this is not implemented and they are
7703 An attribute specifier list may appear at the start of a nested
7704 declarator. At present, there are some limitations in this usage: the
7705 attributes correctly apply to the declarator, but for most individual
7706 attributes the semantics this implies are not implemented.
7707 When attribute specifiers follow the @code{*} of a pointer
7708 declarator, they may be mixed with any type qualifiers present.
7709 The following describes the formal semantics of this syntax. It makes the
7710 most sense if you are familiar with the formal specification of
7711 declarators in the ISO C standard.
7713 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
7714 D1}, where @code{T} contains declaration specifiers that specify a type
7715 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
7716 contains an identifier @var{ident}. The type specified for @var{ident}
7717 for derived declarators whose type does not include an attribute
7718 specifier is as in the ISO C standard.
7720 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
7721 and the declaration @code{T D} specifies the type
7722 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7723 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7724 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
7726 If @code{D1} has the form @code{*
7727 @var{type-qualifier-and-attribute-specifier-list} D}, and the
7728 declaration @code{T D} specifies the type
7729 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
7730 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
7731 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
7737 void (__attribute__((noreturn)) ****f) (void);
7741 specifies the type ``pointer to pointer to pointer to pointer to
7742 non-returning function returning @code{void}''. As another example,
7745 char *__attribute__((aligned(8))) *f;
7749 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
7750 Note again that this does not work with most attributes; for example,
7751 the usage of @samp{aligned} and @samp{noreturn} attributes given above
7752 is not yet supported.
7754 For compatibility with existing code written for compiler versions that
7755 did not implement attributes on nested declarators, some laxity is
7756 allowed in the placing of attributes. If an attribute that only applies
7757 to types is applied to a declaration, it is treated as applying to
7758 the type of that declaration. If an attribute that only applies to
7759 declarations is applied to the type of a declaration, it is treated
7760 as applying to that declaration; and, for compatibility with code
7761 placing the attributes immediately before the identifier declared, such
7762 an attribute applied to a function return type is treated as
7763 applying to the function type, and such an attribute applied to an array
7764 element type is treated as applying to the array type. If an
7765 attribute that only applies to function types is applied to a
7766 pointer-to-function type, it is treated as applying to the pointer
7767 target type; if such an attribute is applied to a function return type
7768 that is not a pointer-to-function type, it is treated as applying
7769 to the function type.
7771 @node Function Prototypes
7772 @section Prototypes and Old-Style Function Definitions
7773 @cindex function prototype declarations
7774 @cindex old-style function definitions
7775 @cindex promotion of formal parameters
7777 GNU C extends ISO C to allow a function prototype to override a later
7778 old-style non-prototype definition. Consider the following example:
7781 /* @r{Use prototypes unless the compiler is old-fashioned.} */
7788 /* @r{Prototype function declaration.} */
7789 int isroot P((uid_t));
7791 /* @r{Old-style function definition.} */
7793 isroot (x) /* @r{??? lossage here ???} */
7800 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
7801 not allow this example, because subword arguments in old-style
7802 non-prototype definitions are promoted. Therefore in this example the
7803 function definition's argument is really an @code{int}, which does not
7804 match the prototype argument type of @code{short}.
7806 This restriction of ISO C makes it hard to write code that is portable
7807 to traditional C compilers, because the programmer does not know
7808 whether the @code{uid_t} type is @code{short}, @code{int}, or
7809 @code{long}. Therefore, in cases like these GNU C allows a prototype
7810 to override a later old-style definition. More precisely, in GNU C, a
7811 function prototype argument type overrides the argument type specified
7812 by a later old-style definition if the former type is the same as the
7813 latter type before promotion. Thus in GNU C the above example is
7814 equivalent to the following:
7827 GNU C++ does not support old-style function definitions, so this
7828 extension is irrelevant.
7831 @section C++ Style Comments
7833 @cindex C++ comments
7834 @cindex comments, C++ style
7836 In GNU C, you may use C++ style comments, which start with @samp{//} and
7837 continue until the end of the line. Many other C implementations allow
7838 such comments, and they are included in the 1999 C standard. However,
7839 C++ style comments are not recognized if you specify an @option{-std}
7840 option specifying a version of ISO C before C99, or @option{-ansi}
7841 (equivalent to @option{-std=c90}).
7844 @section Dollar Signs in Identifier Names
7846 @cindex dollar signs in identifier names
7847 @cindex identifier names, dollar signs in
7849 In GNU C, you may normally use dollar signs in identifier names.
7850 This is because many traditional C implementations allow such identifiers.
7851 However, dollar signs in identifiers are not supported on a few target
7852 machines, typically because the target assembler does not allow them.
7854 @node Character Escapes
7855 @section The Character @key{ESC} in Constants
7857 You can use the sequence @samp{\e} in a string or character constant to
7858 stand for the ASCII character @key{ESC}.
7861 @section Inquiring on Alignment of Types or Variables
7863 @cindex type alignment
7864 @cindex variable alignment
7866 The keyword @code{__alignof__} allows you to inquire about how an object
7867 is aligned, or the minimum alignment usually required by a type. Its
7868 syntax is just like @code{sizeof}.
7870 For example, if the target machine requires a @code{double} value to be
7871 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
7872 This is true on many RISC machines. On more traditional machine
7873 designs, @code{__alignof__ (double)} is 4 or even 2.
7875 Some machines never actually require alignment; they allow reference to any
7876 data type even at an odd address. For these machines, @code{__alignof__}
7877 reports the smallest alignment that GCC gives the data type, usually as
7878 mandated by the target ABI.
7880 If the operand of @code{__alignof__} is an lvalue rather than a type,
7881 its value is the required alignment for its type, taking into account
7882 any minimum alignment specified with GCC's @code{__attribute__}
7883 extension (@pxref{Variable Attributes}). For example, after this
7887 struct foo @{ int x; char y; @} foo1;
7891 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
7892 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
7894 It is an error to ask for the alignment of an incomplete type.
7898 @section An Inline Function is As Fast As a Macro
7899 @cindex inline functions
7900 @cindex integrating function code
7902 @cindex macros, inline alternative
7904 By declaring a function inline, you can direct GCC to make
7905 calls to that function faster. One way GCC can achieve this is to
7906 integrate that function's code into the code for its callers. This
7907 makes execution faster by eliminating the function-call overhead; in
7908 addition, if any of the actual argument values are constant, their
7909 known values may permit simplifications at compile time so that not
7910 all of the inline function's code needs to be included. The effect on
7911 code size is less predictable; object code may be larger or smaller
7912 with function inlining, depending on the particular case. You can
7913 also direct GCC to try to integrate all ``simple enough'' functions
7914 into their callers with the option @option{-finline-functions}.
7916 GCC implements three different semantics of declaring a function
7917 inline. One is available with @option{-std=gnu89} or
7918 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
7919 on all inline declarations, another when
7921 @option{-std=gnu99} or an option for a later C version is used
7922 (without @option{-fgnu89-inline}), and the third
7923 is used when compiling C++.
7925 To declare a function inline, use the @code{inline} keyword in its
7926 declaration, like this:
7936 If you are writing a header file to be included in ISO C90 programs, write
7937 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
7939 The three types of inlining behave similarly in two important cases:
7940 when the @code{inline} keyword is used on a @code{static} function,
7941 like the example above, and when a function is first declared without
7942 using the @code{inline} keyword and then is defined with
7943 @code{inline}, like this:
7946 extern int inc (int *a);
7954 In both of these common cases, the program behaves the same as if you
7955 had not used the @code{inline} keyword, except for its speed.
7957 @cindex inline functions, omission of
7958 @opindex fkeep-inline-functions
7959 When a function is both inline and @code{static}, if all calls to the
7960 function are integrated into the caller, and the function's address is
7961 never used, then the function's own assembler code is never referenced.
7962 In this case, GCC does not actually output assembler code for the
7963 function, unless you specify the option @option{-fkeep-inline-functions}.
7964 If there is a nonintegrated call, then the function is compiled to
7965 assembler code as usual. The function must also be compiled as usual if
7966 the program refers to its address, because that cannot be inlined.
7969 Note that certain usages in a function definition can make it unsuitable
7970 for inline substitution. Among these usages are: variadic functions,
7971 use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
7972 use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
7973 of @code{__builtin_longjmp} and use of @code{__builtin_return} or
7974 @code{__builtin_apply_args}. Using @option{-Winline} warns when a
7975 function marked @code{inline} could not be substituted, and gives the
7976 reason for the failure.
7978 @cindex automatic @code{inline} for C++ member fns
7979 @cindex @code{inline} automatic for C++ member fns
7980 @cindex member fns, automatically @code{inline}
7981 @cindex C++ member fns, automatically @code{inline}
7982 @opindex fno-default-inline
7983 As required by ISO C++, GCC considers member functions defined within
7984 the body of a class to be marked inline even if they are
7985 not explicitly declared with the @code{inline} keyword. You can
7986 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
7987 Options,,Options Controlling C++ Dialect}.
7989 GCC does not inline any functions when not optimizing unless you specify
7990 the @samp{always_inline} attribute for the function, like this:
7993 /* @r{Prototype.} */
7994 inline void foo (const char) __attribute__((always_inline));
7997 The remainder of this section is specific to GNU C90 inlining.
7999 @cindex non-static inline function
8000 When an inline function is not @code{static}, then the compiler must assume
8001 that there may be calls from other source files; since a global symbol can
8002 be defined only once in any program, the function must not be defined in
8003 the other source files, so the calls therein cannot be integrated.
8004 Therefore, a non-@code{static} inline function is always compiled on its
8005 own in the usual fashion.
8007 If you specify both @code{inline} and @code{extern} in the function
8008 definition, then the definition is used only for inlining. In no case
8009 is the function compiled on its own, not even if you refer to its
8010 address explicitly. Such an address becomes an external reference, as
8011 if you had only declared the function, and had not defined it.
8013 This combination of @code{inline} and @code{extern} has almost the
8014 effect of a macro. The way to use it is to put a function definition in
8015 a header file with these keywords, and put another copy of the
8016 definition (lacking @code{inline} and @code{extern}) in a library file.
8017 The definition in the header file causes most calls to the function
8018 to be inlined. If any uses of the function remain, they refer to
8019 the single copy in the library.
8022 @section When is a Volatile Object Accessed?
8023 @cindex accessing volatiles
8024 @cindex volatile read
8025 @cindex volatile write
8026 @cindex volatile access
8028 C has the concept of volatile objects. These are normally accessed by
8029 pointers and used for accessing hardware or inter-thread
8030 communication. The standard encourages compilers to refrain from
8031 optimizations concerning accesses to volatile objects, but leaves it
8032 implementation defined as to what constitutes a volatile access. The
8033 minimum requirement is that at a sequence point all previous accesses
8034 to volatile objects have stabilized and no subsequent accesses have
8035 occurred. Thus an implementation is free to reorder and combine
8036 volatile accesses that occur between sequence points, but cannot do
8037 so for accesses across a sequence point. The use of volatile does
8038 not allow you to violate the restriction on updating objects multiple
8039 times between two sequence points.
8041 Accesses to non-volatile objects are not ordered with respect to
8042 volatile accesses. You cannot use a volatile object as a memory
8043 barrier to order a sequence of writes to non-volatile memory. For
8047 int *ptr = @var{something};
8049 *ptr = @var{something};
8054 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
8055 that the write to @var{*ptr} occurs by the time the update
8056 of @var{vobj} happens. If you need this guarantee, you must use
8057 a stronger memory barrier such as:
8060 int *ptr = @var{something};
8062 *ptr = @var{something};
8063 asm volatile ("" : : : "memory");
8067 A scalar volatile object is read when it is accessed in a void context:
8070 volatile int *src = @var{somevalue};
8074 Such expressions are rvalues, and GCC implements this as a
8075 read of the volatile object being pointed to.
8077 Assignments are also expressions and have an rvalue. However when
8078 assigning to a scalar volatile, the volatile object is not reread,
8079 regardless of whether the assignment expression's rvalue is used or
8080 not. If the assignment's rvalue is used, the value is that assigned
8081 to the volatile object. For instance, there is no read of @var{vobj}
8082 in all the following cases:
8087 vobj = @var{something};
8088 obj = vobj = @var{something};
8089 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
8090 obj = (@var{something}, vobj = @var{anotherthing});
8093 If you need to read the volatile object after an assignment has
8094 occurred, you must use a separate expression with an intervening
8097 As bit-fields are not individually addressable, volatile bit-fields may
8098 be implicitly read when written to, or when adjacent bit-fields are
8099 accessed. Bit-field operations may be optimized such that adjacent
8100 bit-fields are only partially accessed, if they straddle a storage unit
8101 boundary. For these reasons it is unwise to use volatile bit-fields to
8104 @node Using Assembly Language with C
8105 @section How to Use Inline Assembly Language in C Code
8106 @cindex @code{asm} keyword
8107 @cindex assembly language in C
8108 @cindex inline assembly language
8109 @cindex mixing assembly language and C
8111 The @code{asm} keyword allows you to embed assembler instructions
8112 within C code. GCC provides two forms of inline @code{asm}
8113 statements. A @dfn{basic @code{asm}} statement is one with no
8114 operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
8115 statement (@pxref{Extended Asm}) includes one or more operands.
8116 The extended form is preferred for mixing C and assembly language
8117 within a function, but to include assembly language at
8118 top level you must use basic @code{asm}.
8120 You can also use the @code{asm} keyword to override the assembler name
8121 for a C symbol, or to place a C variable in a specific register.
8124 * Basic Asm:: Inline assembler without operands.
8125 * Extended Asm:: Inline assembler with operands.
8126 * Constraints:: Constraints for @code{asm} operands
8127 * Asm Labels:: Specifying the assembler name to use for a C symbol.
8128 * Explicit Register Variables:: Defining variables residing in specified
8130 * Size of an asm:: How GCC calculates the size of an @code{asm} block.
8134 @subsection Basic Asm --- Assembler Instructions Without Operands
8135 @cindex basic @code{asm}
8136 @cindex assembly language in C, basic
8138 A basic @code{asm} statement has the following syntax:
8141 asm @r{[} volatile @r{]} ( @var{AssemblerInstructions} )
8144 The @code{asm} keyword is a GNU extension.
8145 When writing code that can be compiled with @option{-ansi} and the
8146 various @option{-std} options, use @code{__asm__} instead of
8147 @code{asm} (@pxref{Alternate Keywords}).
8149 @subsubheading Qualifiers
8152 The optional @code{volatile} qualifier has no effect.
8153 All basic @code{asm} blocks are implicitly volatile.
8156 @subsubheading Parameters
8159 @item AssemblerInstructions
8160 This is a literal string that specifies the assembler code. The string can
8161 contain any instructions recognized by the assembler, including directives.
8162 GCC does not parse the assembler instructions themselves and
8163 does not know what they mean or even whether they are valid assembler input.
8165 You may place multiple assembler instructions together in a single @code{asm}
8166 string, separated by the characters normally used in assembly code for the
8167 system. A combination that works in most places is a newline to break the
8168 line, plus a tab character (written as @samp{\n\t}).
8169 Some assemblers allow semicolons as a line separator. However,
8170 note that some assembler dialects use semicolons to start a comment.
8173 @subsubheading Remarks
8174 Using extended @code{asm} (@pxref{Extended Asm}) typically produces
8175 smaller, safer, and more efficient code, and in most cases it is a
8176 better solution than basic @code{asm}. However, there are two
8177 situations where only basic @code{asm} can be used:
8181 Extended @code{asm} statements have to be inside a C
8182 function, so to write inline assembly language at file scope (``top-level''),
8183 outside of C functions, you must use basic @code{asm}.
8184 You can use this technique to emit assembler directives,
8185 define assembly language macros that can be invoked elsewhere in the file,
8186 or write entire functions in assembly language.
8190 with the @code{naked} attribute also require basic @code{asm}
8191 (@pxref{Function Attributes}).
8194 Safely accessing C data and calling functions from basic @code{asm} is more
8195 complex than it may appear. To access C data, it is better to use extended
8198 Do not expect a sequence of @code{asm} statements to remain perfectly
8199 consecutive after compilation. If certain instructions need to remain
8200 consecutive in the output, put them in a single multi-instruction @code{asm}
8201 statement. Note that GCC's optimizers can move @code{asm} statements
8202 relative to other code, including across jumps.
8204 @code{asm} statements may not perform jumps into other @code{asm} statements.
8205 GCC does not know about these jumps, and therefore cannot take
8206 account of them when deciding how to optimize. Jumps from @code{asm} to C
8207 labels are only supported in extended @code{asm}.
8209 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
8210 assembly code when optimizing. This can lead to unexpected duplicate
8211 symbol errors during compilation if your assembly code defines symbols or
8214 @strong{Warning:} The C standards do not specify semantics for @code{asm},
8215 making it a potential source of incompatibilities between compilers. These
8216 incompatibilities may not produce compiler warnings/errors.
8218 GCC does not parse basic @code{asm}'s @var{AssemblerInstructions}, which
8219 means there is no way to communicate to the compiler what is happening
8220 inside them. GCC has no visibility of symbols in the @code{asm} and may
8221 discard them as unreferenced. It also does not know about side effects of
8222 the assembler code, such as modifications to memory or registers. Unlike
8223 some compilers, GCC assumes that no changes to general purpose registers
8224 occur. This assumption may change in a future release.
8226 To avoid complications from future changes to the semantics and the
8227 compatibility issues between compilers, consider replacing basic @code{asm}
8228 with extended @code{asm}. See
8229 @uref{https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended, How to convert
8230 from basic asm to extended asm} for information about how to perform this
8233 The compiler copies the assembler instructions in a basic @code{asm}
8234 verbatim to the assembly language output file, without
8235 processing dialects or any of the @samp{%} operators that are available with
8236 extended @code{asm}. This results in minor differences between basic
8237 @code{asm} strings and extended @code{asm} templates. For example, to refer to
8238 registers you might use @samp{%eax} in basic @code{asm} and
8239 @samp{%%eax} in extended @code{asm}.
8241 On targets such as x86 that support multiple assembler dialects,
8242 all basic @code{asm} blocks use the assembler dialect specified by the
8243 @option{-masm} command-line option (@pxref{x86 Options}).
8244 Basic @code{asm} provides no
8245 mechanism to provide different assembler strings for different dialects.
8247 For basic @code{asm} with non-empty assembler string GCC assumes
8248 the assembler block does not change any general purpose registers,
8249 but it may read or write any globally accessible variable.
8251 Here is an example of basic @code{asm} for i386:
8254 /* Note that this code will not compile with -masm=intel */
8255 #define DebugBreak() asm("int $3")
8259 @subsection Extended Asm - Assembler Instructions with C Expression Operands
8260 @cindex extended @code{asm}
8261 @cindex assembly language in C, extended
8263 With extended @code{asm} you can read and write C variables from
8264 assembler and perform jumps from assembler code to C labels.
8265 Extended @code{asm} syntax uses colons (@samp{:}) to delimit
8266 the operand parameters after the assembler template:
8269 asm @r{[}volatile@r{]} ( @var{AssemblerTemplate}
8270 : @var{OutputOperands}
8271 @r{[} : @var{InputOperands}
8272 @r{[} : @var{Clobbers} @r{]} @r{]})
8274 asm @r{[}volatile@r{]} goto ( @var{AssemblerTemplate}
8276 : @var{InputOperands}
8281 The @code{asm} keyword is a GNU extension.
8282 When writing code that can be compiled with @option{-ansi} and the
8283 various @option{-std} options, use @code{__asm__} instead of
8284 @code{asm} (@pxref{Alternate Keywords}).
8286 @subsubheading Qualifiers
8290 The typical use of extended @code{asm} statements is to manipulate input
8291 values to produce output values. However, your @code{asm} statements may
8292 also produce side effects. If so, you may need to use the @code{volatile}
8293 qualifier to disable certain optimizations. @xref{Volatile}.
8296 This qualifier informs the compiler that the @code{asm} statement may
8297 perform a jump to one of the labels listed in the @var{GotoLabels}.
8301 @subsubheading Parameters
8303 @item AssemblerTemplate
8304 This is a literal string that is the template for the assembler code. It is a
8305 combination of fixed text and tokens that refer to the input, output,
8306 and goto parameters. @xref{AssemblerTemplate}.
8308 @item OutputOperands
8309 A comma-separated list of the C variables modified by the instructions in the
8310 @var{AssemblerTemplate}. An empty list is permitted. @xref{OutputOperands}.
8313 A comma-separated list of C expressions read by the instructions in the
8314 @var{AssemblerTemplate}. An empty list is permitted. @xref{InputOperands}.
8317 A comma-separated list of registers or other values changed by the
8318 @var{AssemblerTemplate}, beyond those listed as outputs.
8319 An empty list is permitted. @xref{Clobbers and Scratch Registers}.
8322 When you are using the @code{goto} form of @code{asm}, this section contains
8323 the list of all C labels to which the code in the
8324 @var{AssemblerTemplate} may jump.
8327 @code{asm} statements may not perform jumps into other @code{asm} statements,
8328 only to the listed @var{GotoLabels}.
8329 GCC's optimizers do not know about other jumps; therefore they cannot take
8330 account of them when deciding how to optimize.
8333 The total number of input + output + goto operands is limited to 30.
8335 @subsubheading Remarks
8336 The @code{asm} statement allows you to include assembly instructions directly
8337 within C code. This may help you to maximize performance in time-sensitive
8338 code or to access assembly instructions that are not readily available to C
8341 Note that extended @code{asm} statements must be inside a function. Only
8342 basic @code{asm} may be outside functions (@pxref{Basic Asm}).
8343 Functions declared with the @code{naked} attribute also require basic
8344 @code{asm} (@pxref{Function Attributes}).
8346 While the uses of @code{asm} are many and varied, it may help to think of an
8347 @code{asm} statement as a series of low-level instructions that convert input
8348 parameters to output parameters. So a simple (if not particularly useful)
8349 example for i386 using @code{asm} might look like this:
8355 asm ("mov %1, %0\n\t"
8360 printf("%d\n", dst);
8363 This code copies @code{src} to @code{dst} and add 1 to @code{dst}.
8366 @subsubsection Volatile
8367 @cindex volatile @code{asm}
8368 @cindex @code{asm} volatile
8370 GCC's optimizers sometimes discard @code{asm} statements if they determine
8371 there is no need for the output variables. Also, the optimizers may move
8372 code out of loops if they believe that the code will always return the same
8373 result (i.e. none of its input values change between calls). Using the
8374 @code{volatile} qualifier disables these optimizations. @code{asm} statements
8375 that have no output operands, including @code{asm goto} statements,
8376 are implicitly volatile.
8378 This i386 code demonstrates a case that does not use (or require) the
8379 @code{volatile} qualifier. If it is performing assertion checking, this code
8380 uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is
8381 unreferenced by any code. As a result, the optimizers can discard the
8382 @code{asm} statement, which in turn removes the need for the entire
8383 @code{DoCheck} routine. By omitting the @code{volatile} qualifier when it
8384 isn't needed you allow the optimizers to produce the most efficient code
8388 void DoCheck(uint32_t dwSomeValue)
8392 // Assumes dwSomeValue is not zero.
8402 The next example shows a case where the optimizers can recognize that the input
8403 (@code{dwSomeValue}) never changes during the execution of the function and can
8404 therefore move the @code{asm} outside the loop to produce more efficient code.
8405 Again, using @code{volatile} disables this type of optimization.
8408 void do_print(uint32_t dwSomeValue)
8412 for (uint32_t x=0; x < 5; x++)
8414 // Assumes dwSomeValue is not zero.
8420 printf("%u: %u %u\n", x, dwSomeValue, dwRes);
8425 The following example demonstrates a case where you need to use the
8426 @code{volatile} qualifier.
8427 It uses the x86 @code{rdtsc} instruction, which reads
8428 the computer's time-stamp counter. Without the @code{volatile} qualifier,
8429 the optimizers might assume that the @code{asm} block will always return the
8430 same value and therefore optimize away the second call.
8435 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
8436 "shl $32, %%rdx\n\t" // Shift the upper bits left.
8437 "or %%rdx, %0" // 'Or' in the lower bits.
8442 printf("msr: %llx\n", msr);
8446 // Reprint the timestamp
8447 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX.
8448 "shl $32, %%rdx\n\t" // Shift the upper bits left.
8449 "or %%rdx, %0" // 'Or' in the lower bits.
8454 printf("msr: %llx\n", msr);
8457 GCC's optimizers do not treat this code like the non-volatile code in the
8458 earlier examples. They do not move it out of loops or omit it on the
8459 assumption that the result from a previous call is still valid.
8461 Note that the compiler can move even volatile @code{asm} instructions relative
8462 to other code, including across jump instructions. For example, on many
8463 targets there is a system register that controls the rounding mode of
8464 floating-point operations. Setting it with a volatile @code{asm}, as in the
8465 following PowerPC example, does not work reliably.
8468 asm volatile("mtfsf 255, %0" : : "f" (fpenv));
8472 The compiler may move the addition back before the volatile @code{asm}. To
8473 make it work as expected, add an artificial dependency to the @code{asm} by
8474 referencing a variable in the subsequent code, for example:
8477 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
8481 Under certain circumstances, GCC may duplicate (or remove duplicates of) your
8482 assembly code when optimizing. This can lead to unexpected duplicate symbol
8483 errors during compilation if your asm code defines symbols or labels.
8485 (@pxref{AssemblerTemplate}) may help resolve this problem.
8487 @anchor{AssemblerTemplate}
8488 @subsubsection Assembler Template
8489 @cindex @code{asm} assembler template
8491 An assembler template is a literal string containing assembler instructions.
8492 The compiler replaces tokens in the template that refer
8493 to inputs, outputs, and goto labels,
8494 and then outputs the resulting string to the assembler. The
8495 string can contain any instructions recognized by the assembler, including
8496 directives. GCC does not parse the assembler instructions
8497 themselves and does not know what they mean or even whether they are valid
8498 assembler input. However, it does count the statements
8499 (@pxref{Size of an asm}).
8501 You may place multiple assembler instructions together in a single @code{asm}
8502 string, separated by the characters normally used in assembly code for the
8503 system. A combination that works in most places is a newline to break the
8504 line, plus a tab character to move to the instruction field (written as
8506 Some assemblers allow semicolons as a line separator. However, note
8507 that some assembler dialects use semicolons to start a comment.
8509 Do not expect a sequence of @code{asm} statements to remain perfectly
8510 consecutive after compilation, even when you are using the @code{volatile}
8511 qualifier. If certain instructions need to remain consecutive in the output,
8512 put them in a single multi-instruction asm statement.
8514 Accessing data from C programs without using input/output operands (such as
8515 by using global symbols directly from the assembler template) may not work as
8516 expected. Similarly, calling functions directly from an assembler template
8517 requires a detailed understanding of the target assembler and ABI.
8519 Since GCC does not parse the assembler template,
8520 it has no visibility of any
8521 symbols it references. This may result in GCC discarding those symbols as
8522 unreferenced unless they are also listed as input, output, or goto operands.
8524 @subsubheading Special format strings
8526 In addition to the tokens described by the input, output, and goto operands,
8527 these tokens have special meanings in the assembler template:
8531 Outputs a single @samp{%} into the assembler code.
8534 Outputs a number that is unique to each instance of the @code{asm}
8535 statement in the entire compilation. This option is useful when creating local
8536 labels and referring to them multiple times in a single template that
8537 generates multiple assembler instructions.
8542 Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
8543 into the assembler code. When unescaped, these characters have special
8544 meaning to indicate multiple assembler dialects, as described below.
8547 @subsubheading Multiple assembler dialects in @code{asm} templates
8549 On targets such as x86, GCC supports multiple assembler dialects.
8550 The @option{-masm} option controls which dialect GCC uses as its
8551 default for inline assembler. The target-specific documentation for the
8552 @option{-masm} option contains the list of supported dialects, as well as the
8553 default dialect if the option is not specified. This information may be
8554 important to understand, since assembler code that works correctly when
8555 compiled using one dialect will likely fail if compiled using another.
8558 If your code needs to support multiple assembler dialects (for example, if
8559 you are writing public headers that need to support a variety of compilation
8560 options), use constructs of this form:
8563 @{ dialect0 | dialect1 | dialect2... @}
8566 This construct outputs @code{dialect0}
8567 when using dialect #0 to compile the code,
8568 @code{dialect1} for dialect #1, etc. If there are fewer alternatives within the
8569 braces than the number of dialects the compiler supports, the construct
8572 For example, if an x86 compiler supports two dialects
8573 (@samp{att}, @samp{intel}), an
8574 assembler template such as this:
8577 "bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
8581 is equivalent to one of
8584 "btl %[Offset],%[Base] ; jc %l2" @r{/* att dialect */}
8585 "bt %[Base],%[Offset]; jc %l2" @r{/* intel dialect */}
8588 Using that same compiler, this code:
8591 "xchg@{l@}\t@{%%@}ebx, %1"
8595 corresponds to either
8598 "xchgl\t%%ebx, %1" @r{/* att dialect */}
8599 "xchg\tebx, %1" @r{/* intel dialect */}
8602 There is no support for nesting dialect alternatives.
8604 @anchor{OutputOperands}
8605 @subsubsection Output Operands
8606 @cindex @code{asm} output operands
8608 An @code{asm} statement has zero or more output operands indicating the names
8609 of C variables modified by the assembler code.
8611 In this i386 example, @code{old} (referred to in the template string as
8612 @code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset}
8613 (@code{%2}) is an input:
8618 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
8619 "sbb %0,%0" // Use the CF to calculate old.
8620 : "=r" (old), "+rm" (*Base)
8627 Operands are separated by commas. Each operand has this format:
8630 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
8634 @item asmSymbolicName
8635 Specifies a symbolic name for the operand.
8636 Reference the name in the assembler template
8637 by enclosing it in square brackets
8638 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8639 that contains the definition. Any valid C variable name is acceptable,
8640 including names already defined in the surrounding code. No two operands
8641 within the same @code{asm} statement can use the same symbolic name.
8643 When not using an @var{asmSymbolicName}, use the (zero-based) position
8645 in the list of operands in the assembler template. For example if there are
8646 three output operands, use @samp{%0} in the template to refer to the first,
8647 @samp{%1} for the second, and @samp{%2} for the third.
8650 A string constant specifying constraints on the placement of the operand;
8651 @xref{Constraints}, for details.
8653 Output constraints must begin with either @samp{=} (a variable overwriting an
8654 existing value) or @samp{+} (when reading and writing). When using
8655 @samp{=}, do not assume the location contains the existing value
8656 on entry to the @code{asm}, except
8657 when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.
8659 After the prefix, there must be one or more additional constraints
8660 (@pxref{Constraints}) that describe where the value resides. Common
8661 constraints include @samp{r} for register and @samp{m} for memory.
8662 When you list more than one possible location (for example, @code{"=rm"}),
8663 the compiler chooses the most efficient one based on the current context.
8664 If you list as many alternates as the @code{asm} statement allows, you permit
8665 the optimizers to produce the best possible code.
8666 If you must use a specific register, but your Machine Constraints do not
8667 provide sufficient control to select the specific register you want,
8668 local register variables may provide a solution (@pxref{Local Register
8672 Specifies a C lvalue expression to hold the output, typically a variable name.
8673 The enclosing parentheses are a required part of the syntax.
8677 When the compiler selects the registers to use to
8678 represent the output operands, it does not use any of the clobbered registers
8679 (@pxref{Clobbers and Scratch Registers}).
8681 Output operand expressions must be lvalues. The compiler cannot check whether
8682 the operands have data types that are reasonable for the instruction being
8683 executed. For output expressions that are not directly addressable (for
8684 example a bit-field), the constraint must allow a register. In that case, GCC
8685 uses the register as the output of the @code{asm}, and then stores that
8686 register into the output.
8688 Operands using the @samp{+} constraint modifier count as two operands
8689 (that is, both as input and output) towards the total maximum of 30 operands
8690 per @code{asm} statement.
8692 Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
8693 operands that must not overlap an input. Otherwise,
8694 GCC may allocate the output operand in the same register as an unrelated
8695 input operand, on the assumption that the assembler code consumes its
8696 inputs before producing outputs. This assumption may be false if the assembler
8697 code actually consists of more than one instruction.
8699 The same problem can occur if one output parameter (@var{a}) allows a register
8700 constraint and another output parameter (@var{b}) allows a memory constraint.
8701 The code generated by GCC to access the memory address in @var{b} can contain
8702 registers which @emph{might} be shared by @var{a}, and GCC considers those
8703 registers to be inputs to the asm. As above, GCC assumes that such input
8704 registers are consumed before any outputs are written. This assumption may
8705 result in incorrect behavior if the asm writes to @var{a} before using
8706 @var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
8707 ensures that modifying @var{a} does not affect the address referenced by
8708 @var{b}. Otherwise, the location of @var{b}
8709 is undefined if @var{a} is modified before using @var{b}.
8711 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8712 instead of simply @samp{%2}). Typically these qualifiers are hardware
8713 dependent. The list of supported modifiers for x86 is found at
8714 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8716 If the C code that follows the @code{asm} makes no use of any of the output
8717 operands, use @code{volatile} for the @code{asm} statement to prevent the
8718 optimizers from discarding the @code{asm} statement as unneeded
8719 (see @ref{Volatile}).
8721 This code makes no use of the optional @var{asmSymbolicName}. Therefore it
8722 references the first output operand as @code{%0} (were there a second, it
8723 would be @code{%1}, etc). The number of the first input operand is one greater
8724 than that of the last output operand. In this i386 example, that makes
8725 @code{Mask} referenced as @code{%1}:
8728 uint32_t Mask = 1234;
8737 That code overwrites the variable @code{Index} (@samp{=}),
8738 placing the value in a register (@samp{r}).
8739 Using the generic @samp{r} constraint instead of a constraint for a specific
8740 register allows the compiler to pick the register to use, which can result
8741 in more efficient code. This may not be possible if an assembler instruction
8742 requires a specific register.
8744 The following i386 example uses the @var{asmSymbolicName} syntax.
8746 same result as the code above, but some may consider it more readable or more
8747 maintainable since reordering index numbers is not necessary when adding or
8748 removing operands. The names @code{aIndex} and @code{aMask}
8749 are only used in this example to emphasize which
8750 names get used where.
8751 It is acceptable to reuse the names @code{Index} and @code{Mask}.
8754 uint32_t Mask = 1234;
8757 asm ("bsfl %[aMask], %[aIndex]"
8758 : [aIndex] "=r" (Index)
8759 : [aMask] "r" (Mask)
8763 Here are some more examples of output operands.
8770 asm ("mov %[e], %[d]"
8775 Here, @code{d} may either be in a register or in memory. Since the compiler
8776 might already have the current value of the @code{uint32_t} location
8777 pointed to by @code{e}
8778 in a register, you can enable it to choose the best location
8779 for @code{d} by specifying both constraints.
8781 @anchor{FlagOutputOperands}
8782 @subsubsection Flag Output Operands
8783 @cindex @code{asm} flag output operands
8785 Some targets have a special register that holds the ``flags'' for the
8786 result of an operation or comparison. Normally, the contents of that
8787 register are either unmodifed by the asm, or the asm is considered to
8788 clobber the contents.
8790 On some targets, a special form of output operand exists by which
8791 conditions in the flags register may be outputs of the asm. The set of
8792 conditions supported are target specific, but the general rule is that
8793 the output variable must be a scalar integer, and the value is boolean.
8794 When supported, the target defines the preprocessor symbol
8795 @code{__GCC_ASM_FLAG_OUTPUTS__}.
8797 Because of the special nature of the flag output operands, the constraint
8798 may not include alternatives.
8800 Most often, the target has only one flags register, and thus is an implied
8801 operand of many instructions. In this case, the operand should not be
8802 referenced within the assembler template via @code{%0} etc, as there's
8803 no corresponding text in the assembly language.
8807 The flag output constraints for the x86 family are of the form
8808 @samp{=@@cc@var{cond}} where @var{cond} is one of the standard
8809 conditions defined in the ISA manual for @code{j@var{cc}} or
8814 ``above'' or unsigned greater than
8816 ``above or equal'' or unsigned greater than or equal
8818 ``below'' or unsigned less than
8820 ``below or equal'' or unsigned less than or equal
8825 ``equal'' or zero flag set
8829 signed greater than or equal
8833 signed less than or equal
8854 ``not'' @var{flag}, or inverted versions of those above
8859 @anchor{InputOperands}
8860 @subsubsection Input Operands
8861 @cindex @code{asm} input operands
8862 @cindex @code{asm} expressions
8864 Input operands make values from C variables and expressions available to the
8867 Operands are separated by commas. Each operand has this format:
8870 @r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
8874 @item asmSymbolicName
8875 Specifies a symbolic name for the operand.
8876 Reference the name in the assembler template
8877 by enclosing it in square brackets
8878 (i.e. @samp{%[Value]}). The scope of the name is the @code{asm} statement
8879 that contains the definition. Any valid C variable name is acceptable,
8880 including names already defined in the surrounding code. No two operands
8881 within the same @code{asm} statement can use the same symbolic name.
8883 When not using an @var{asmSymbolicName}, use the (zero-based) position
8885 in the list of operands in the assembler template. For example if there are
8886 two output operands and three inputs,
8887 use @samp{%2} in the template to refer to the first input operand,
8888 @samp{%3} for the second, and @samp{%4} for the third.
8891 A string constant specifying constraints on the placement of the operand;
8892 @xref{Constraints}, for details.
8894 Input constraint strings may not begin with either @samp{=} or @samp{+}.
8895 When you list more than one possible location (for example, @samp{"irm"}),
8896 the compiler chooses the most efficient one based on the current context.
8897 If you must use a specific register, but your Machine Constraints do not
8898 provide sufficient control to select the specific register you want,
8899 local register variables may provide a solution (@pxref{Local Register
8902 Input constraints can also be digits (for example, @code{"0"}). This indicates
8903 that the specified input must be in the same place as the output constraint
8904 at the (zero-based) index in the output constraint list.
8905 When using @var{asmSymbolicName} syntax for the output operands,
8906 you may use these names (enclosed in brackets @samp{[]}) instead of digits.
8909 This is the C variable or expression being passed to the @code{asm} statement
8910 as input. The enclosing parentheses are a required part of the syntax.
8914 When the compiler selects the registers to use to represent the input
8915 operands, it does not use any of the clobbered registers
8916 (@pxref{Clobbers and Scratch Registers}).
8918 If there are no output operands but there are input operands, place two
8919 consecutive colons where the output operands would go:
8922 __asm__ ("some instructions"
8924 : "r" (Offset / 8));
8927 @strong{Warning:} Do @emph{not} modify the contents of input-only operands
8928 (except for inputs tied to outputs). The compiler assumes that on exit from
8929 the @code{asm} statement these operands contain the same values as they
8930 had before executing the statement.
8931 It is @emph{not} possible to use clobbers
8932 to inform the compiler that the values in these inputs are changing. One
8933 common work-around is to tie the changing input variable to an output variable
8934 that never gets used. Note, however, that if the code that follows the
8935 @code{asm} statement makes no use of any of the output operands, the GCC
8936 optimizers may discard the @code{asm} statement as unneeded
8937 (see @ref{Volatile}).
8939 @code{asm} supports operand modifiers on operands (for example @samp{%k2}
8940 instead of simply @samp{%2}). Typically these qualifiers are hardware
8941 dependent. The list of supported modifiers for x86 is found at
8942 @ref{x86Operandmodifiers,x86 Operand modifiers}.
8944 In this example using the fictitious @code{combine} instruction, the
8945 constraint @code{"0"} for input operand 1 says that it must occupy the same
8946 location as output operand 0. Only input operands may use numbers in
8947 constraints, and they must each refer to an output operand. Only a number (or
8948 the symbolic assembler name) in the constraint can guarantee that one operand
8949 is in the same place as another. The mere fact that @code{foo} is the value of
8950 both operands is not enough to guarantee that they are in the same place in
8951 the generated assembler code.
8954 asm ("combine %2, %0"
8956 : "0" (foo), "g" (bar));
8959 Here is an example using symbolic names.
8962 asm ("cmoveq %1, %2, %[result]"
8963 : [result] "=r"(result)
8964 : "r" (test), "r" (new), "[result]" (old));
8967 @anchor{Clobbers and Scratch Registers}
8968 @subsubsection Clobbers and Scratch Registers
8969 @cindex @code{asm} clobbers
8970 @cindex @code{asm} scratch registers
8972 While the compiler is aware of changes to entries listed in the output
8973 operands, the inline @code{asm} code may modify more than just the outputs. For
8974 example, calculations may require additional registers, or the processor may
8975 overwrite a register as a side effect of a particular assembler instruction.
8976 In order to inform the compiler of these changes, list them in the clobber
8977 list. Clobber list items are either register names or the special clobbers
8978 (listed below). Each clobber list item is a string constant
8979 enclosed in double quotes and separated by commas.
8981 Clobber descriptions may not in any way overlap with an input or output
8982 operand. For example, you may not have an operand describing a register class
8983 with one member when listing that register in the clobber list. Variables
8984 declared to live in specific registers (@pxref{Explicit Register
8985 Variables}) and used
8986 as @code{asm} input or output operands must have no part mentioned in the
8987 clobber description. In particular, there is no way to specify that input
8988 operands get modified without also specifying them as output operands.
8990 When the compiler selects which registers to use to represent input and output
8991 operands, it does not use any of the clobbered registers. As a result,
8992 clobbered registers are available for any use in the assembler code.
8994 Here is a realistic example for the VAX showing the use of clobbered
8998 asm volatile ("movc3 %0, %1, %2"
9000 : "g" (from), "g" (to), "g" (count)
9001 : "r0", "r1", "r2", "r3", "r4", "r5", "memory");
9004 Also, there are two special clobber arguments:
9008 The @code{"cc"} clobber indicates that the assembler code modifies the flags
9009 register. On some machines, GCC represents the condition codes as a specific
9010 hardware register; @code{"cc"} serves to name this register.
9011 On other machines, condition code handling is different,
9012 and specifying @code{"cc"} has no effect. But
9013 it is valid no matter what the target.
9016 The @code{"memory"} clobber tells the compiler that the assembly code
9018 reads or writes to items other than those listed in the input and output
9019 operands (for example, accessing the memory pointed to by one of the input
9020 parameters). To ensure memory contains correct values, GCC may need to flush
9021 specific register values to memory before executing the @code{asm}. Further,
9022 the compiler does not assume that any values read from memory before an
9023 @code{asm} remain unchanged after that @code{asm}; it reloads them as
9025 Using the @code{"memory"} clobber effectively forms a read/write
9026 memory barrier for the compiler.
9028 Note that this clobber does not prevent the @emph{processor} from doing
9029 speculative reads past the @code{asm} statement. To prevent that, you need
9030 processor-specific fence instructions.
9034 Flushing registers to memory has performance implications and may be
9035 an issue for time-sensitive code. You can provide better information
9036 to GCC to avoid this, as shown in the following examples. At a
9037 minimum, aliasing rules allow GCC to know what memory @emph{doesn't}
9040 Here is a fictitious sum of squares instruction, that takes two
9041 pointers to floating point values in memory and produces a floating
9042 point register output.
9043 Notice that @code{x}, and @code{y} both appear twice in the @code{asm}
9044 parameters, once to specify memory accessed, and once to specify a
9045 base register used by the @code{asm}. You won't normally be wasting a
9046 register by doing this as GCC can use the same register for both
9047 purposes. However, it would be foolish to use both @code{%1} and
9048 @code{%3} for @code{x} in this @code{asm} and expect them to be the
9049 same. In fact, @code{%3} may well not be a register. It might be a
9050 symbolic memory reference to the object pointed to by @code{x}.
9053 asm ("sumsq %0, %1, %2"
9055 : "r" (x), "r" (y), "m" (*x), "m" (*y));
9058 Here is a fictitious @code{*z++ = *x++ * *y++} instruction.
9059 Notice that the @code{x}, @code{y} and @code{z} pointer registers
9060 must be specified as input/output because the @code{asm} modifies
9064 asm ("vecmul %0, %1, %2"
9065 : "+r" (z), "+r" (x), "+r" (y), "=m" (*z)
9066 : "m" (*x), "m" (*y));
9069 An x86 example where the string memory argument is of unknown length.
9073 : "=c" (count), "+D" (p)
9074 : "m" (*(const char (*)[]) p), "0" (-1), "a" (0));
9077 If you know the above will only be reading a ten byte array then you
9078 could instead use a memory input like:
9079 @code{"m" (*(const char (*)[10]) p)}.
9081 Here is an example of a PowerPC vector scale implemented in assembly,
9082 complete with vector and condition code clobbers, and some initialized
9083 offset registers that are unchanged by the @code{asm}.
9087 dscal (size_t n, double *x, double alpha)
9089 asm ("/* lots of asm here */"
9090 : "+m" (*(double (*)[n]) x), "+&r" (n), "+b" (x)
9091 : "d" (alpha), "b" (32), "b" (48), "b" (64),
9092 "b" (80), "b" (96), "b" (112)
9094 "vs32","vs33","vs34","vs35","vs36","vs37","vs38","vs39",
9095 "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47");
9099 Rather than allocating fixed registers via clobbers to provide scratch
9100 registers for an @code{asm} statement, an alternative is to define a
9101 variable and make it an early-clobber output as with @code{a2} and
9102 @code{a3} in the example below. This gives the compiler register
9103 allocator more freedom. You can also define a variable and make it an
9104 output tied to an input as with @code{a0} and @code{a1}, tied
9105 respectively to @code{ap} and @code{lda}. Of course, with tied
9106 outputs your @code{asm} can't use the input value after modifying the
9107 output register since they are one and the same register. What's
9108 more, if you omit the early-clobber on the output, it is possible that
9109 GCC might allocate the same register to another of the inputs if GCC
9110 could prove they had the same value on entry to the @code{asm}. This
9111 is why @code{a1} has an early-clobber. Its tied input, @code{lda}
9112 might conceivably be known to have the value 16 and without an
9113 early-clobber share the same register as @code{%11}. On the other
9114 hand, @code{ap} can't be the same as any of the other inputs, so an
9115 early-clobber on @code{a0} is not needed. It is also not desirable in
9116 this case. An early-clobber on @code{a0} would cause GCC to allocate
9117 a separate register for the @code{"m" (*(const double (*)[]) ap)}
9118 input. Note that tying an input to an output is the way to set up an
9119 initialized temporary register modified by an @code{asm} statement.
9120 An input not tied to an output is assumed by GCC to be unchanged, for
9121 example @code{"b" (16)} below sets up @code{%11} to 16, and GCC might
9122 use that register in following code if the value 16 happened to be
9123 needed. You can even use a normal @code{asm} output for a scratch if
9124 all inputs that might share the same register are consumed before the
9125 scratch is used. The VSX registers clobbered by the @code{asm}
9126 statement could have used this technique except for GCC's limit on the
9127 number of @code{asm} parameters.
9131 dgemv_kernel_4x4 (long n, const double *ap, long lda,
9132 const double *x, double *y, double alpha)
9141 /* lots of asm here */
9142 "#n=%1 ap=%8=%12 lda=%13 x=%7=%10 y=%0=%2 alpha=%9 o16=%11\n"
9143 "#a0=%3 a1=%4 a2=%5 a3=%6"
9145 "+m" (*(double (*)[n]) y),
9153 "m" (*(const double (*)[n]) x),
9154 "m" (*(const double (*)[]) ap),
9162 "vs32","vs33","vs34","vs35","vs36","vs37",
9163 "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47"
9169 @subsubsection Goto Labels
9170 @cindex @code{asm} goto labels
9172 @code{asm goto} allows assembly code to jump to one or more C labels. The
9173 @var{GotoLabels} section in an @code{asm goto} statement contains
9175 list of all C labels to which the assembler code may jump. GCC assumes that
9176 @code{asm} execution falls through to the next statement (if this is not the
9177 case, consider using the @code{__builtin_unreachable} intrinsic after the
9178 @code{asm} statement). Optimization of @code{asm goto} may be improved by
9179 using the @code{hot} and @code{cold} label attributes (@pxref{Label
9182 An @code{asm goto} statement cannot have outputs.
9183 This is due to an internal restriction of
9184 the compiler: control transfer instructions cannot have outputs.
9185 If the assembler code does modify anything, use the @code{"memory"} clobber
9187 optimizers to flush all register values to memory and reload them if
9188 necessary after the @code{asm} statement.
9190 Also note that an @code{asm goto} statement is always implicitly
9191 considered volatile.
9193 To reference a label in the assembler template,
9194 prefix it with @samp{%l} (lowercase @samp{L}) followed
9195 by its (zero-based) position in @var{GotoLabels} plus the number of input
9196 operands. For example, if the @code{asm} has three inputs and references two
9197 labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).
9199 Alternately, you can reference labels using the actual C label name enclosed
9200 in brackets. For example, to reference a label named @code{carry}, you can
9201 use @samp{%l[carry]}. The label must still be listed in the @var{GotoLabels}
9202 section when using this approach.
9204 Here is an example of @code{asm goto} for i386:
9211 : "r" (p1), "r" (p2)
9221 The following example shows an @code{asm goto} that uses a memory clobber.
9227 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
9238 @anchor{x86Operandmodifiers}
9239 @subsubsection x86 Operand Modifiers
9241 References to input, output, and goto operands in the assembler template
9242 of extended @code{asm} statements can use
9243 modifiers to affect the way the operands are formatted in
9244 the code output to the assembler. For example, the
9245 following code uses the @samp{h} and @samp{b} modifiers for x86:
9249 asm volatile ("xchg %h0, %b0" : "+a" (num) );
9253 These modifiers generate this assembler code:
9259 The rest of this discussion uses the following code for illustrative purposes.
9268 asm volatile goto ("some assembler instructions here"
9270 : "q" (iInt), "X" (sizeof(unsigned char) + 1)
9271 : /* No clobbers. */
9276 With no modifiers, this is what the output from the operands would be for the
9277 @samp{att} and @samp{intel} dialects of assembler:
9279 @multitable {Operand} {$.L2} {OFFSET FLAT:.L2}
9280 @headitem Operand @tab @samp{att} @tab @samp{intel}
9289 @tab @code{OFFSET FLAT:.L2}
9292 The table below shows the list of supported modifiers and their effects.
9294 @multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {@samp{att}} {@samp{intel}}
9295 @headitem Modifier @tab Description @tab Operand @tab @samp{att} @tab @samp{intel}
9297 @tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
9302 @tab Print the QImode name of the register.
9307 @tab Print the QImode name for a ``high'' register.
9312 @tab Print the HImode name of the register.
9317 @tab Print the SImode name of the register.
9322 @tab Print the DImode name of the register.
9327 @tab Print the label name with no punctuation.
9332 @tab Require a constant operand and print the constant expression with no punctuation.
9338 @code{V} is a special modifier which prints the name of the full integer
9339 register without @code{%}.
9341 @anchor{x86floatingpointasmoperands}
9342 @subsubsection x86 Floating-Point @code{asm} Operands
9344 On x86 targets, there are several rules on the usage of stack-like registers
9345 in the operands of an @code{asm}. These rules apply only to the operands
9346 that are stack-like registers:
9350 Given a set of input registers that die in an @code{asm}, it is
9351 necessary to know which are implicitly popped by the @code{asm}, and
9352 which must be explicitly popped by GCC@.
9354 An input register that is implicitly popped by the @code{asm} must be
9355 explicitly clobbered, unless it is constrained to match an
9359 For any input register that is implicitly popped by an @code{asm}, it is
9360 necessary to know how to adjust the stack to compensate for the pop.
9361 If any non-popped input is closer to the top of the reg-stack than
9362 the implicitly popped register, it would not be possible to know what the
9363 stack looked like---it's not clear how the rest of the stack ``slides
9366 All implicitly popped input registers must be closer to the top of
9367 the reg-stack than any input that is not implicitly popped.
9369 It is possible that if an input dies in an @code{asm}, the compiler might
9370 use the input register for an output reload. Consider this example:
9373 asm ("foo" : "=t" (a) : "f" (b));
9377 This code says that input @code{b} is not popped by the @code{asm}, and that
9378 the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
9379 deeper after the @code{asm} than it was before. But, it is possible that
9380 reload may think that it can use the same register for both the input and
9383 To prevent this from happening,
9384 if any input operand uses the @samp{f} constraint, all output register
9385 constraints must use the @samp{&} early-clobber modifier.
9387 The example above is correctly written as:
9390 asm ("foo" : "=&t" (a) : "f" (b));
9394 Some operands need to be in particular places on the stack. All
9395 output operands fall in this category---GCC has no other way to
9396 know which registers the outputs appear in unless you indicate
9397 this in the constraints.
9399 Output operands must specifically indicate which register an output
9400 appears in after an @code{asm}. @samp{=f} is not allowed: the operand
9401 constraints must select a class with a single register.
9404 Output operands may not be ``inserted'' between existing stack registers.
9405 Since no 387 opcode uses a read/write operand, all output operands
9406 are dead before the @code{asm}, and are pushed by the @code{asm}.
9407 It makes no sense to push anywhere but the top of the reg-stack.
9409 Output operands must start at the top of the reg-stack: output
9410 operands may not ``skip'' a register.
9413 Some @code{asm} statements may need extra stack space for internal
9414 calculations. This can be guaranteed by clobbering stack registers
9415 unrelated to the inputs and outputs.
9420 takes one input, which is internally popped, and produces two outputs.
9423 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
9427 This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
9428 and replaces them with one output. The @code{st(1)} clobber is necessary
9429 for the compiler to know that @code{fyl2xp1} pops both inputs.
9432 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
9440 @subsection Controlling Names Used in Assembler Code
9441 @cindex assembler names for identifiers
9442 @cindex names used in assembler code
9443 @cindex identifiers, names in assembler code
9445 You can specify the name to be used in the assembler code for a C
9446 function or variable by writing the @code{asm} (or @code{__asm__})
9447 keyword after the declarator.
9448 It is up to you to make sure that the assembler names you choose do not
9449 conflict with any other assembler symbols, or reference registers.
9451 @subsubheading Assembler names for data:
9453 This sample shows how to specify the assembler name for data:
9456 int foo asm ("myfoo") = 2;
9460 This specifies that the name to be used for the variable @code{foo} in
9461 the assembler code should be @samp{myfoo} rather than the usual
9464 On systems where an underscore is normally prepended to the name of a C
9465 variable, this feature allows you to define names for the
9466 linker that do not start with an underscore.
9468 GCC does not support using this feature with a non-static local variable
9469 since such variables do not have assembler names. If you are
9470 trying to put the variable in a particular register, see
9471 @ref{Explicit Register Variables}.
9473 @subsubheading Assembler names for functions:
9475 To specify the assembler name for functions, write a declaration for the
9476 function before its definition and put @code{asm} there, like this:
9479 int func (int x, int y) asm ("MYFUNC");
9481 int func (int x, int y)
9487 This specifies that the name to be used for the function @code{func} in
9488 the assembler code should be @code{MYFUNC}.
9490 @node Explicit Register Variables
9491 @subsection Variables in Specified Registers
9492 @anchor{Explicit Reg Vars}
9493 @cindex explicit register variables
9494 @cindex variables in specified registers
9495 @cindex specified registers
9497 GNU C allows you to associate specific hardware registers with C
9498 variables. In almost all cases, allowing the compiler to assign
9499 registers produces the best code. However under certain unusual
9500 circumstances, more precise control over the variable storage is
9503 Both global and local variables can be associated with a register. The
9504 consequences of performing this association are very different between
9505 the two, as explained in the sections below.
9508 * Global Register Variables:: Variables declared at global scope.
9509 * Local Register Variables:: Variables declared within a function.
9512 @node Global Register Variables
9513 @subsubsection Defining Global Register Variables
9514 @anchor{Global Reg Vars}
9515 @cindex global register variables
9516 @cindex registers, global variables in
9517 @cindex registers, global allocation
9519 You can define a global register variable and associate it with a specified
9523 register int *foo asm ("r12");
9527 Here @code{r12} is the name of the register that should be used. Note that
9528 this is the same syntax used for defining local register variables, but for
9529 a global variable the declaration appears outside a function. The
9530 @code{register} keyword is required, and cannot be combined with
9531 @code{static}. The register name must be a valid register name for the
9534 Registers are a scarce resource on most systems and allowing the
9535 compiler to manage their usage usually results in the best code. However,
9536 under special circumstances it can make sense to reserve some globally.
9537 For example this may be useful in programs such as programming language
9538 interpreters that have a couple of global variables that are accessed
9541 After defining a global register variable, for the current compilation
9545 @item The register is reserved entirely for this use, and will not be
9546 allocated for any other purpose.
9547 @item The register is not saved and restored by any functions.
9548 @item Stores into this register are never deleted even if they appear to be
9549 dead, but references may be deleted, moved or simplified.
9552 Note that these points @emph{only} apply to code that is compiled with the
9553 definition. The behavior of code that is merely linked in (for example
9554 code from libraries) is not affected.
9556 If you want to recompile source files that do not actually use your global
9557 register variable so they do not use the specified register for any other
9558 purpose, you need not actually add the global register declaration to
9559 their source code. It suffices to specify the compiler option
9560 @option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the
9563 @subsubheading Declaring the variable
9565 Global register variables can not have initial values, because an
9566 executable file has no means to supply initial contents for a register.
9568 When selecting a register, choose one that is normally saved and
9569 restored by function calls on your machine. This ensures that code
9570 which is unaware of this reservation (such as library routines) will
9571 restore it before returning.
9573 On machines with register windows, be sure to choose a global
9574 register that is not affected magically by the function call mechanism.
9576 @subsubheading Using the variable
9578 @cindex @code{qsort}, and global register variables
9579 When calling routines that are not aware of the reservation, be
9580 cautious if those routines call back into code which uses them. As an
9581 example, if you call the system library version of @code{qsort}, it may
9582 clobber your registers during execution, but (if you have selected
9583 appropriate registers) it will restore them before returning. However
9584 it will @emph{not} restore them before calling @code{qsort}'s comparison
9585 function. As a result, global values will not reliably be available to
9586 the comparison function unless the @code{qsort} function itself is rebuilt.
9588 Similarly, it is not safe to access the global register variables from signal
9589 handlers or from more than one thread of control. Unless you recompile
9590 them specially for the task at hand, the system library routines may
9591 temporarily use the register for other things.
9593 @cindex register variable after @code{longjmp}
9594 @cindex global register after @code{longjmp}
9595 @cindex value after @code{longjmp}
9598 On most machines, @code{longjmp} restores to each global register
9599 variable the value it had at the time of the @code{setjmp}. On some
9600 machines, however, @code{longjmp} does not change the value of global
9601 register variables. To be portable, the function that called @code{setjmp}
9602 should make other arrangements to save the values of the global register
9603 variables, and to restore them in a @code{longjmp}. This way, the same
9604 thing happens regardless of what @code{longjmp} does.
9606 Eventually there may be a way of asking the compiler to choose a register
9607 automatically, but first we need to figure out how it should choose and
9608 how to enable you to guide the choice. No solution is evident.
9610 @node Local Register Variables
9611 @subsubsection Specifying Registers for Local Variables
9612 @anchor{Local Reg Vars}
9613 @cindex local variables, specifying registers
9614 @cindex specifying registers for local variables
9615 @cindex registers for local variables
9617 You can define a local register variable and associate it with a specified
9621 register int *foo asm ("r12");
9625 Here @code{r12} is the name of the register that should be used. Note
9626 that this is the same syntax used for defining global register variables,
9627 but for a local variable the declaration appears within a function. The
9628 @code{register} keyword is required, and cannot be combined with
9629 @code{static}. The register name must be a valid register name for the
9632 As with global register variables, it is recommended that you choose
9633 a register that is normally saved and restored by function calls on your
9634 machine, so that calls to library routines will not clobber it.
9636 The only supported use for this feature is to specify registers
9637 for input and output operands when calling Extended @code{asm}
9638 (@pxref{Extended Asm}). This may be necessary if the constraints for a
9639 particular machine don't provide sufficient control to select the desired
9640 register. To force an operand into a register, create a local variable
9641 and specify the register name after the variable's declaration. Then use
9642 the local variable for the @code{asm} operand and specify any constraint
9643 letter that matches the register:
9646 register int *p1 asm ("r0") = @dots{};
9647 register int *p2 asm ("r1") = @dots{};
9648 register int *result asm ("r0");
9649 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
9652 @emph{Warning:} In the above example, be aware that a register (for example
9653 @code{r0}) can be call-clobbered by subsequent code, including function
9654 calls and library calls for arithmetic operators on other variables (for
9655 example the initialization of @code{p2}). In this case, use temporary
9656 variables for expressions between the register assignments:
9660 register int *p1 asm ("r0") = @dots{};
9661 register int *p2 asm ("r1") = t1;
9662 register int *result asm ("r0");
9663 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
9666 Defining a register variable does not reserve the register. Other than
9667 when invoking the Extended @code{asm}, the contents of the specified
9668 register are not guaranteed. For this reason, the following uses
9669 are explicitly @emph{not} supported. If they appear to work, it is only
9670 happenstance, and may stop working as intended due to (seemingly)
9671 unrelated changes in surrounding code, or even minor changes in the
9672 optimization of a future version of gcc:
9675 @item Passing parameters to or from Basic @code{asm}
9676 @item Passing parameters to or from Extended @code{asm} without using input
9678 @item Passing parameters to or from routines written in assembler (or
9679 other languages) using non-standard calling conventions.
9682 Some developers use Local Register Variables in an attempt to improve
9683 gcc's allocation of registers, especially in large functions. In this
9684 case the register name is essentially a hint to the register allocator.
9685 While in some instances this can generate better code, improvements are
9686 subject to the whims of the allocator/optimizers. Since there are no
9687 guarantees that your improvements won't be lost, this usage of Local
9688 Register Variables is discouraged.
9690 On the MIPS platform, there is related use for local register variables
9691 with slightly different characteristics (@pxref{MIPS Coprocessors,,
9692 Defining coprocessor specifics for MIPS targets, gccint,
9693 GNU Compiler Collection (GCC) Internals}).
9695 @node Size of an asm
9696 @subsection Size of an @code{asm}
9698 Some targets require that GCC track the size of each instruction used
9699 in order to generate correct code. Because the final length of the
9700 code produced by an @code{asm} statement is only known by the
9701 assembler, GCC must make an estimate as to how big it will be. It
9702 does this by counting the number of instructions in the pattern of the
9703 @code{asm} and multiplying that by the length of the longest
9704 instruction supported by that processor. (When working out the number
9705 of instructions, it assumes that any occurrence of a newline or of
9706 whatever statement separator character is supported by the assembler --
9707 typically @samp{;} --- indicates the end of an instruction.)
9709 Normally, GCC's estimate is adequate to ensure that correct
9710 code is generated, but it is possible to confuse the compiler if you use
9711 pseudo instructions or assembler macros that expand into multiple real
9712 instructions, or if you use assembler directives that expand to more
9713 space in the object file than is needed for a single instruction.
9714 If this happens then the assembler may produce a diagnostic saying that
9715 a label is unreachable.
9717 @node Alternate Keywords
9718 @section Alternate Keywords
9719 @cindex alternate keywords
9720 @cindex keywords, alternate
9722 @option{-ansi} and the various @option{-std} options disable certain
9723 keywords. This causes trouble when you want to use GNU C extensions, or
9724 a general-purpose header file that should be usable by all programs,
9725 including ISO C programs. The keywords @code{asm}, @code{typeof} and
9726 @code{inline} are not available in programs compiled with
9727 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
9728 program compiled with @option{-std=c99} or @option{-std=c11}). The
9730 @code{restrict} is only available when @option{-std=gnu99} (which will
9731 eventually be the default) or @option{-std=c99} (or the equivalent
9732 @option{-std=iso9899:1999}), or an option for a later standard
9735 The way to solve these problems is to put @samp{__} at the beginning and
9736 end of each problematical keyword. For example, use @code{__asm__}
9737 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
9739 Other C compilers won't accept these alternative keywords; if you want to
9740 compile with another compiler, you can define the alternate keywords as
9741 macros to replace them with the customary keywords. It looks like this:
9749 @findex __extension__
9751 @option{-pedantic} and other options cause warnings for many GNU C extensions.
9753 prevent such warnings within one expression by writing
9754 @code{__extension__} before the expression. @code{__extension__} has no
9755 effect aside from this.
9757 @node Incomplete Enums
9758 @section Incomplete @code{enum} Types
9760 You can define an @code{enum} tag without specifying its possible values.
9761 This results in an incomplete type, much like what you get if you write
9762 @code{struct foo} without describing the elements. A later declaration
9763 that does specify the possible values completes the type.
9765 You cannot allocate variables or storage using the type while it is
9766 incomplete. However, you can work with pointers to that type.
9768 This extension may not be very useful, but it makes the handling of
9769 @code{enum} more consistent with the way @code{struct} and @code{union}
9772 This extension is not supported by GNU C++.
9774 @node Function Names
9775 @section Function Names as Strings
9776 @cindex @code{__func__} identifier
9777 @cindex @code{__FUNCTION__} identifier
9778 @cindex @code{__PRETTY_FUNCTION__} identifier
9780 GCC provides three magic constants that hold the name of the current
9781 function as a string. In C++11 and later modes, all three are treated
9782 as constant expressions and can be used in @code{constexpr} constexts.
9783 The first of these constants is @code{__func__}, which is part of
9786 The identifier @code{__func__} is implicitly declared by the translator
9787 as if, immediately following the opening brace of each function
9788 definition, the declaration
9791 static const char __func__[] = "function-name";
9795 appeared, where function-name is the name of the lexically-enclosing
9796 function. This name is the unadorned name of the function. As an
9797 extension, at file (or, in C++, namespace scope), @code{__func__}
9798 evaluates to the empty string.
9800 @code{__FUNCTION__} is another name for @code{__func__}, provided for
9801 backward compatibility with old versions of GCC.
9803 In C, @code{__PRETTY_FUNCTION__} is yet another name for
9804 @code{__func__}, except that at file (or, in C++, namespace scope),
9805 it evaluates to the string @code{"top level"}. In addition, in C++,
9806 @code{__PRETTY_FUNCTION__} contains the signature of the function as
9807 well as its bare name. For example, this program:
9810 extern "C" int printf (const char *, ...);
9816 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
9817 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
9835 __PRETTY_FUNCTION__ = void a::sub(int)
9838 These identifiers are variables, not preprocessor macros, and may not
9839 be used to initialize @code{char} arrays or be concatenated with string
9842 @node Return Address
9843 @section Getting the Return or Frame Address of a Function
9845 These functions may be used to get information about the callers of a
9848 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
9849 This function returns the return address of the current function, or of
9850 one of its callers. The @var{level} argument is number of frames to
9851 scan up the call stack. A value of @code{0} yields the return address
9852 of the current function, a value of @code{1} yields the return address
9853 of the caller of the current function, and so forth. When inlining
9854 the expected behavior is that the function returns the address of
9855 the function that is returned to. To work around this behavior use
9856 the @code{noinline} function attribute.
9858 The @var{level} argument must be a constant integer.
9860 On some machines it may be impossible to determine the return address of
9861 any function other than the current one; in such cases, or when the top
9862 of the stack has been reached, this function returns @code{0} or a
9863 random value. In addition, @code{__builtin_frame_address} may be used
9864 to determine if the top of the stack has been reached.
9866 Additional post-processing of the returned value may be needed, see
9867 @code{__builtin_extract_return_addr}.
9869 Calling this function with a nonzero argument can have unpredictable
9870 effects, including crashing the calling program. As a result, calls
9871 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9872 option is in effect. Such calls should only be made in debugging
9876 @deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
9877 The address as returned by @code{__builtin_return_address} may have to be fed
9878 through this function to get the actual encoded address. For example, on the
9879 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
9880 platforms an offset has to be added for the true next instruction to be
9883 If no fixup is needed, this function simply passes through @var{addr}.
9886 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
9887 This function does the reverse of @code{__builtin_extract_return_addr}.
9890 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
9891 This function is similar to @code{__builtin_return_address}, but it
9892 returns the address of the function frame rather than the return address
9893 of the function. Calling @code{__builtin_frame_address} with a value of
9894 @code{0} yields the frame address of the current function, a value of
9895 @code{1} yields the frame address of the caller of the current function,
9898 The frame is the area on the stack that holds local variables and saved
9899 registers. The frame address is normally the address of the first word
9900 pushed on to the stack by the function. However, the exact definition
9901 depends upon the processor and the calling convention. If the processor
9902 has a dedicated frame pointer register, and the function has a frame,
9903 then @code{__builtin_frame_address} returns the value of the frame
9906 On some machines it may be impossible to determine the frame address of
9907 any function other than the current one; in such cases, or when the top
9908 of the stack has been reached, this function returns @code{0} if
9909 the first frame pointer is properly initialized by the startup code.
9911 Calling this function with a nonzero argument can have unpredictable
9912 effects, including crashing the calling program. As a result, calls
9913 that are considered unsafe are diagnosed when the @option{-Wframe-address}
9914 option is in effect. Such calls should only be made in debugging
9918 @node Vector Extensions
9919 @section Using Vector Instructions through Built-in Functions
9921 On some targets, the instruction set contains SIMD vector instructions which
9922 operate on multiple values contained in one large register at the same time.
9923 For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
9926 The first step in using these extensions is to provide the necessary data
9927 types. This should be done using an appropriate @code{typedef}:
9930 typedef int v4si __attribute__ ((vector_size (16)));
9934 The @code{int} type specifies the base type, while the attribute specifies
9935 the vector size for the variable, measured in bytes. For example, the
9936 declaration above causes the compiler to set the mode for the @code{v4si}
9937 type to be 16 bytes wide and divided into @code{int} sized units. For
9938 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
9939 corresponding mode of @code{foo} is @acronym{V4SI}.
9941 The @code{vector_size} attribute is only applicable to integral and
9942 float scalars, although arrays, pointers, and function return values
9943 are allowed in conjunction with this construct. Only sizes that are
9944 a power of two are currently allowed.
9946 All the basic integer types can be used as base types, both as signed
9947 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
9948 @code{long long}. In addition, @code{float} and @code{double} can be
9949 used to build floating-point vector types.
9951 Specifying a combination that is not valid for the current architecture
9952 causes GCC to synthesize the instructions using a narrower mode.
9953 For example, if you specify a variable of type @code{V4SI} and your
9954 architecture does not allow for this specific SIMD type, GCC
9955 produces code that uses 4 @code{SIs}.
9957 The types defined in this manner can be used with a subset of normal C
9958 operations. Currently, GCC allows using the following operators
9959 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
9961 The operations behave like C++ @code{valarrays}. Addition is defined as
9962 the addition of the corresponding elements of the operands. For
9963 example, in the code below, each of the 4 elements in @var{a} is
9964 added to the corresponding 4 elements in @var{b} and the resulting
9965 vector is stored in @var{c}.
9968 typedef int v4si __attribute__ ((vector_size (16)));
9975 Subtraction, multiplication, division, and the logical operations
9976 operate in a similar manner. Likewise, the result of using the unary
9977 minus or complement operators on a vector type is a vector whose
9978 elements are the negative or complemented values of the corresponding
9979 elements in the operand.
9981 It is possible to use shifting operators @code{<<}, @code{>>} on
9982 integer-type vectors. The operation is defined as following: @code{@{a0,
9983 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
9984 @dots{}, an >> bn@}}@. Vector operands must have the same number of
9987 For convenience, it is allowed to use a binary vector operation
9988 where one operand is a scalar. In that case the compiler transforms
9989 the scalar operand into a vector where each element is the scalar from
9990 the operation. The transformation happens only if the scalar could be
9991 safely converted to the vector-element type.
9992 Consider the following code.
9995 typedef int v4si __attribute__ ((vector_size (16)));
10000 a = b + 1; /* a = b + @{1,1,1,1@}; */
10001 a = 2 * b; /* a = @{2,2,2,2@} * b; */
10003 a = l + a; /* Error, cannot convert long to int. */
10006 Vectors can be subscripted as if the vector were an array with
10007 the same number of elements and base type. Out of bound accesses
10008 invoke undefined behavior at run time. Warnings for out of bound
10009 accesses for vector subscription can be enabled with
10010 @option{-Warray-bounds}.
10012 Vector comparison is supported with standard comparison
10013 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
10014 vector expressions of integer-type or real-type. Comparison between
10015 integer-type vectors and real-type vectors are not supported. The
10016 result of the comparison is a vector of the same width and number of
10017 elements as the comparison operands with a signed integral element
10020 Vectors are compared element-wise producing 0 when comparison is false
10021 and -1 (constant of the appropriate type where all bits are set)
10022 otherwise. Consider the following example.
10025 typedef int v4si __attribute__ ((vector_size (16)));
10027 v4si a = @{1,2,3,4@};
10028 v4si b = @{3,2,1,4@};
10031 c = a > b; /* The result would be @{0, 0,-1, 0@} */
10032 c = a == b; /* The result would be @{0,-1, 0,-1@} */
10035 In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
10036 @code{b} and @code{c} are vectors of the same type and @code{a} is an
10037 integer vector with the same number of elements of the same size as @code{b}
10038 and @code{c}, computes all three arguments and creates a vector
10039 @code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}. Note that unlike in
10040 OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
10041 As in the case of binary operations, this syntax is also accepted when
10042 one of @code{b} or @code{c} is a scalar that is then transformed into a
10043 vector. If both @code{b} and @code{c} are scalars and the type of
10044 @code{true?b:c} has the same size as the element type of @code{a}, then
10045 @code{b} and @code{c} are converted to a vector type whose elements have
10046 this type and with the same number of elements as @code{a}.
10048 In C++, the logic operators @code{!, &&, ||} are available for vectors.
10049 @code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
10050 @code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
10051 For mixed operations between a scalar @code{s} and a vector @code{v},
10052 @code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
10053 short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.
10055 @findex __builtin_shuffle
10056 Vector shuffling is available using functions
10057 @code{__builtin_shuffle (vec, mask)} and
10058 @code{__builtin_shuffle (vec0, vec1, mask)}.
10059 Both functions construct a permutation of elements from one or two
10060 vectors and return a vector of the same type as the input vector(s).
10061 The @var{mask} is an integral vector with the same width (@var{W})
10062 and element count (@var{N}) as the output vector.
10064 The elements of the input vectors are numbered in memory ordering of
10065 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
10066 elements of @var{mask} are considered modulo @var{N} in the single-operand
10067 case and modulo @math{2*@var{N}} in the two-operand case.
10069 Consider the following example,
10072 typedef int v4si __attribute__ ((vector_size (16)));
10074 v4si a = @{1,2,3,4@};
10075 v4si b = @{5,6,7,8@};
10076 v4si mask1 = @{0,1,1,3@};
10077 v4si mask2 = @{0,4,2,5@};
10080 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
10081 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
10084 Note that @code{__builtin_shuffle} is intentionally semantically
10085 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
10087 You can declare variables and use them in function calls and returns, as
10088 well as in assignments and some casts. You can specify a vector type as
10089 a return type for a function. Vector types can also be used as function
10090 arguments. It is possible to cast from one vector type to another,
10091 provided they are of the same size (in fact, you can also cast vectors
10092 to and from other datatypes of the same size).
10094 You cannot operate between vectors of different lengths or different
10095 signedness without a cast.
10098 @section Support for @code{offsetof}
10099 @findex __builtin_offsetof
10101 GCC implements for both C and C++ a syntactic extension to implement
10102 the @code{offsetof} macro.
10106 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
10108 offsetof_member_designator:
10110 | offsetof_member_designator "." @code{identifier}
10111 | offsetof_member_designator "[" @code{expr} "]"
10114 This extension is sufficient such that
10117 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
10121 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
10122 may be dependent. In either case, @var{member} may consist of a single
10123 identifier, or a sequence of member accesses and array references.
10125 @node __sync Builtins
10126 @section Legacy @code{__sync} Built-in Functions for Atomic Memory Access
10128 The following built-in functions
10129 are intended to be compatible with those described
10130 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
10131 section 7.4. As such, they depart from normal GCC practice by not using
10132 the @samp{__builtin_} prefix and also by being overloaded so that they
10133 work on multiple types.
10135 The definition given in the Intel documentation allows only for the use of
10136 the types @code{int}, @code{long}, @code{long long} or their unsigned
10137 counterparts. GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
10138 size other than the C type @code{_Bool} or the C++ type @code{bool}.
10139 Operations on pointer arguments are performed as if the operands were
10140 of the @code{uintptr_t} type. That is, they are not scaled by the size
10141 of the type to which the pointer points.
10143 These functions are implemented in terms of the @samp{__atomic}
10144 builtins (@pxref{__atomic Builtins}). They should not be used for new
10145 code which should use the @samp{__atomic} builtins instead.
10147 Not all operations are supported by all target processors. If a particular
10148 operation cannot be implemented on the target processor, a warning is
10149 generated and a call to an external function is generated. The external
10150 function carries the same name as the built-in version,
10151 with an additional suffix
10152 @samp{_@var{n}} where @var{n} is the size of the data type.
10154 @c ??? Should we have a mechanism to suppress this warning? This is almost
10155 @c useful for implementing the operation under the control of an external
10158 In most cases, these built-in functions are considered a @dfn{full barrier}.
10160 no memory operand is moved across the operation, either forward or
10161 backward. Further, instructions are issued as necessary to prevent the
10162 processor from speculating loads across the operation and from queuing stores
10163 after the operation.
10165 All of the routines are described in the Intel documentation to take
10166 ``an optional list of variables protected by the memory barrier''. It's
10167 not clear what is meant by that; it could mean that @emph{only} the
10168 listed variables are protected, or it could mean a list of additional
10169 variables to be protected. The list is ignored by GCC which treats it as
10170 empty. GCC interprets an empty list as meaning that all globally
10171 accessible variables should be protected.
10174 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
10175 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
10176 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
10177 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
10178 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
10179 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
10180 @findex __sync_fetch_and_add
10181 @findex __sync_fetch_and_sub
10182 @findex __sync_fetch_and_or
10183 @findex __sync_fetch_and_and
10184 @findex __sync_fetch_and_xor
10185 @findex __sync_fetch_and_nand
10186 These built-in functions perform the operation suggested by the name, and
10187 returns the value that had previously been in memory. That is, operations
10188 on integer operands have the following semantics. Operations on pointer
10189 arguments are performed as if the operands were of the @code{uintptr_t}
10190 type. That is, they are not scaled by the size of the type to which
10191 the pointer points.
10194 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
10195 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
10198 The object pointed to by the first argument must be of integer or pointer
10199 type. It must not be a boolean type.
10201 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
10202 as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
10204 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
10205 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
10206 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
10207 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
10208 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
10209 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
10210 @findex __sync_add_and_fetch
10211 @findex __sync_sub_and_fetch
10212 @findex __sync_or_and_fetch
10213 @findex __sync_and_and_fetch
10214 @findex __sync_xor_and_fetch
10215 @findex __sync_nand_and_fetch
10216 These built-in functions perform the operation suggested by the name, and
10217 return the new value. That is, operations on integer operands have
10218 the following semantics. Operations on pointer operands are performed as
10219 if the operand's type were @code{uintptr_t}.
10222 @{ *ptr @var{op}= value; return *ptr; @}
10223 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
10226 The same constraints on arguments apply as for the corresponding
10227 @code{__sync_op_and_fetch} built-in functions.
10229 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
10230 as @code{*ptr = ~(*ptr & value)} instead of
10231 @code{*ptr = ~*ptr & value}.
10233 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
10234 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
10235 @findex __sync_bool_compare_and_swap
10236 @findex __sync_val_compare_and_swap
10237 These built-in functions perform an atomic compare and swap.
10238 That is, if the current
10239 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
10242 The ``bool'' version returns true if the comparison is successful and
10243 @var{newval} is written. The ``val'' version returns the contents
10244 of @code{*@var{ptr}} before the operation.
10246 @item __sync_synchronize (...)
10247 @findex __sync_synchronize
10248 This built-in function issues a full memory barrier.
10250 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
10251 @findex __sync_lock_test_and_set
10252 This built-in function, as described by Intel, is not a traditional test-and-set
10253 operation, but rather an atomic exchange operation. It writes @var{value}
10254 into @code{*@var{ptr}}, and returns the previous contents of
10257 Many targets have only minimal support for such locks, and do not support
10258 a full exchange operation. In this case, a target may support reduced
10259 functionality here by which the @emph{only} valid value to store is the
10260 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
10261 is implementation defined.
10263 This built-in function is not a full barrier,
10264 but rather an @dfn{acquire barrier}.
10265 This means that references after the operation cannot move to (or be
10266 speculated to) before the operation, but previous memory stores may not
10267 be globally visible yet, and previous memory loads may not yet be
10270 @item void __sync_lock_release (@var{type} *ptr, ...)
10271 @findex __sync_lock_release
10272 This built-in function releases the lock acquired by
10273 @code{__sync_lock_test_and_set}.
10274 Normally this means writing the constant 0 to @code{*@var{ptr}}.
10276 This built-in function is not a full barrier,
10277 but rather a @dfn{release barrier}.
10278 This means that all previous memory stores are globally visible, and all
10279 previous memory loads have been satisfied, but following memory reads
10280 are not prevented from being speculated to before the barrier.
10283 @node __atomic Builtins
10284 @section Built-in Functions for Memory Model Aware Atomic Operations
10286 The following built-in functions approximately match the requirements
10287 for the C++11 memory model. They are all
10288 identified by being prefixed with @samp{__atomic} and most are
10289 overloaded so that they work with multiple types.
10291 These functions are intended to replace the legacy @samp{__sync}
10292 builtins. The main difference is that the memory order that is requested
10293 is a parameter to the functions. New code should always use the
10294 @samp{__atomic} builtins rather than the @samp{__sync} builtins.
10296 Note that the @samp{__atomic} builtins assume that programs will
10297 conform to the C++11 memory model. In particular, they assume
10298 that programs are free of data races. See the C++11 standard for
10299 detailed requirements.
10301 The @samp{__atomic} builtins can be used with any integral scalar or
10302 pointer type that is 1, 2, 4, or 8 bytes in length. 16-byte integral
10303 types are also allowed if @samp{__int128} (@pxref{__int128}) is
10304 supported by the architecture.
10306 The four non-arithmetic functions (load, store, exchange, and
10307 compare_exchange) all have a generic version as well. This generic
10308 version works on any data type. It uses the lock-free built-in function
10309 if the specific data type size makes that possible; otherwise, an
10310 external call is left to be resolved at run time. This external call is
10311 the same format with the addition of a @samp{size_t} parameter inserted
10312 as the first parameter indicating the size of the object being pointed to.
10313 All objects must be the same size.
10315 There are 6 different memory orders that can be specified. These map
10316 to the C++11 memory orders with the same names, see the C++11 standard
10317 or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
10318 on atomic synchronization} for detailed definitions. Individual
10319 targets may also support additional memory orders for use on specific
10320 architectures. Refer to the target documentation for details of
10323 An atomic operation can both constrain code motion and
10324 be mapped to hardware instructions for synchronization between threads
10325 (e.g., a fence). To which extent this happens is controlled by the
10326 memory orders, which are listed here in approximately ascending order of
10327 strength. The description of each memory order is only meant to roughly
10328 illustrate the effects and is not a specification; see the C++11
10329 memory model for precise semantics.
10332 @item __ATOMIC_RELAXED
10333 Implies no inter-thread ordering constraints.
10334 @item __ATOMIC_CONSUME
10335 This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
10336 memory order because of a deficiency in C++11's semantics for
10337 @code{memory_order_consume}.
10338 @item __ATOMIC_ACQUIRE
10339 Creates an inter-thread happens-before constraint from the release (or
10340 stronger) semantic store to this acquire load. Can prevent hoisting
10341 of code to before the operation.
10342 @item __ATOMIC_RELEASE
10343 Creates an inter-thread happens-before constraint to acquire (or stronger)
10344 semantic loads that read from this release store. Can prevent sinking
10345 of code to after the operation.
10346 @item __ATOMIC_ACQ_REL
10347 Combines the effects of both @code{__ATOMIC_ACQUIRE} and
10348 @code{__ATOMIC_RELEASE}.
10349 @item __ATOMIC_SEQ_CST
10350 Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
10353 Note that in the C++11 memory model, @emph{fences} (e.g.,
10354 @samp{__atomic_thread_fence}) take effect in combination with other
10355 atomic operations on specific memory locations (e.g., atomic loads);
10356 operations on specific memory locations do not necessarily affect other
10357 operations in the same way.
10359 Target architectures are encouraged to provide their own patterns for
10360 each of the atomic built-in functions. If no target is provided, the original
10361 non-memory model set of @samp{__sync} atomic built-in functions are
10362 used, along with any required synchronization fences surrounding it in
10363 order to achieve the proper behavior. Execution in this case is subject
10364 to the same restrictions as those built-in functions.
10366 If there is no pattern or mechanism to provide a lock-free instruction
10367 sequence, a call is made to an external routine with the same parameters
10368 to be resolved at run time.
10370 When implementing patterns for these built-in functions, the memory order
10371 parameter can be ignored as long as the pattern implements the most
10372 restrictive @code{__ATOMIC_SEQ_CST} memory order. Any of the other memory
10373 orders execute correctly with this memory order but they may not execute as
10374 efficiently as they could with a more appropriate implementation of the
10375 relaxed requirements.
10377 Note that the C++11 standard allows for the memory order parameter to be
10378 determined at run time rather than at compile time. These built-in
10379 functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
10380 than invoke a runtime library call or inline a switch statement. This is
10381 standard compliant, safe, and the simplest approach for now.
10383 The memory order parameter is a signed int, but only the lower 16 bits are
10384 reserved for the memory order. The remainder of the signed int is reserved
10385 for target use and should be 0. Use of the predefined atomic values
10386 ensures proper usage.
10388 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
10389 This built-in function implements an atomic load operation. It returns the
10390 contents of @code{*@var{ptr}}.
10392 The valid memory order variants are
10393 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
10394 and @code{__ATOMIC_CONSUME}.
10398 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
10399 This is the generic version of an atomic load. It returns the
10400 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
10404 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
10405 This built-in function implements an atomic store operation. It writes
10406 @code{@var{val}} into @code{*@var{ptr}}.
10408 The valid memory order variants are
10409 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
10413 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
10414 This is the generic version of an atomic store. It stores the value
10415 of @code{*@var{val}} into @code{*@var{ptr}}.
10419 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
10420 This built-in function implements an atomic exchange operation. It writes
10421 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
10424 The valid memory order variants are
10425 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
10426 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
10430 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
10431 This is the generic version of an atomic exchange. It stores the
10432 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
10433 of @code{*@var{ptr}} is copied into @code{*@var{ret}}.
10437 @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)
10438 This built-in function implements an atomic compare and exchange operation.
10439 This compares the contents of @code{*@var{ptr}} with the contents of
10440 @code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
10441 operation that writes @var{desired} into @code{*@var{ptr}}. If they are not
10442 equal, the operation is a @emph{read} and the current contents of
10443 @code{*@var{ptr}} are written into @code{*@var{expected}}. @var{weak} is true
10444 for weak compare_exchange, which may fail spuriously, and false for
10445 the strong variation, which never fails spuriously. Many targets
10446 only offer the strong variation and ignore the parameter. When in doubt, use
10447 the strong variation.
10449 If @var{desired} is written into @code{*@var{ptr}} then true is returned
10450 and memory is affected according to the
10451 memory order specified by @var{success_memorder}. There are no
10452 restrictions on what memory order can be used here.
10454 Otherwise, false is returned and memory is affected according
10455 to @var{failure_memorder}. This memory order cannot be
10456 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
10457 stronger order than that specified by @var{success_memorder}.
10461 @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)
10462 This built-in function implements the generic version of
10463 @code{__atomic_compare_exchange}. The function is virtually identical to
10464 @code{__atomic_compare_exchange_n}, except the desired value is also a
10469 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
10470 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
10471 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
10472 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
10473 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
10474 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
10475 These built-in functions perform the operation suggested by the name, and
10476 return the result of the operation. Operations on pointer arguments are
10477 performed as if the operands were of the @code{uintptr_t} type. That is,
10478 they are not scaled by the size of the type to which the pointer points.
10481 @{ *ptr @var{op}= val; return *ptr; @}
10484 The object pointed to by the first argument must be of integer or pointer
10485 type. It must not be a boolean type. All memory orders are valid.
10489 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
10490 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
10491 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
10492 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
10493 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
10494 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
10495 These built-in functions perform the operation suggested by the name, and
10496 return the value that had previously been in @code{*@var{ptr}}. Operations
10497 on pointer arguments are performed as if the operands were of
10498 the @code{uintptr_t} type. That is, they are not scaled by the size of
10499 the type to which the pointer points.
10502 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
10505 The same constraints on arguments apply as for the corresponding
10506 @code{__atomic_op_fetch} built-in functions. All memory orders are valid.
10510 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)
10512 This built-in function performs an atomic test-and-set operation on
10513 the byte at @code{*@var{ptr}}. The byte is set to some implementation
10514 defined nonzero ``set'' value and the return value is @code{true} if and only
10515 if the previous contents were ``set''.
10516 It should be only used for operands of type @code{bool} or @code{char}. For
10517 other types only part of the value may be set.
10519 All memory orders are valid.
10523 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)
10525 This built-in function performs an atomic clear operation on
10526 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} contains 0.
10527 It should be only used for operands of type @code{bool} or @code{char} and
10528 in conjunction with @code{__atomic_test_and_set}.
10529 For other types it may only clear partially. If the type is not @code{bool}
10530 prefer using @code{__atomic_store}.
10532 The valid memory order variants are
10533 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
10534 @code{__ATOMIC_RELEASE}.
10538 @deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)
10540 This built-in function acts as a synchronization fence between threads
10541 based on the specified memory order.
10543 All memory orders are valid.
10547 @deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)
10549 This built-in function acts as a synchronization fence between a thread
10550 and signal handlers based in the same thread.
10552 All memory orders are valid.
10556 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size, void *ptr)
10558 This built-in function returns true if objects of @var{size} bytes always
10559 generate lock-free atomic instructions for the target architecture.
10560 @var{size} must resolve to a compile-time constant and the result also
10561 resolves to a compile-time constant.
10563 @var{ptr} is an optional pointer to the object that may be used to determine
10564 alignment. A value of 0 indicates typical alignment should be used. The
10565 compiler may also ignore this parameter.
10568 if (__atomic_always_lock_free (sizeof (long long), 0))
10573 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)
10575 This built-in function returns true if objects of @var{size} bytes always
10576 generate lock-free atomic instructions for the target architecture. If
10577 the built-in function is not known to be lock-free, a call is made to a
10578 runtime routine named @code{__atomic_is_lock_free}.
10580 @var{ptr} is an optional pointer to the object that may be used to determine
10581 alignment. A value of 0 indicates typical alignment should be used. The
10582 compiler may also ignore this parameter.
10585 @node Integer Overflow Builtins
10586 @section Built-in Functions to Perform Arithmetic with Overflow Checking
10588 The following built-in functions allow performing simple arithmetic operations
10589 together with checking whether the operations overflowed.
10591 @deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10592 @deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
10593 @deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
10594 @deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long long int *res)
10595 @deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
10596 @deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10597 @deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10599 These built-in functions promote the first two operands into infinite precision signed
10600 type and perform addition on those promoted operands. The result is then
10601 cast to the type the third pointer argument points to and stored there.
10602 If the stored result is equal to the infinite precision result, the built-in
10603 functions return false, otherwise they return true. As the addition is
10604 performed in infinite signed precision, these built-in functions have fully defined
10605 behavior for all argument values.
10607 The first built-in function allows arbitrary integral types for operands and
10608 the result type must be pointer to some integral type other than enumerated or
10609 boolean type, the rest of the built-in functions have explicit integer types.
10611 The compiler will attempt to use hardware instructions to implement
10612 these built-in functions where possible, like conditional jump on overflow
10613 after addition, conditional jump on carry etc.
10617 @deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10618 @deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
10619 @deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
10620 @deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long long int *res)
10621 @deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
10622 @deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10623 @deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10625 These built-in functions are similar to the add overflow checking built-in
10626 functions above, except they perform subtraction, subtract the second argument
10627 from the first one, instead of addition.
10631 @deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
10632 @deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
10633 @deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
10634 @deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long long int *res)
10635 @deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
10636 @deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
10637 @deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)
10639 These built-in functions are similar to the add overflow checking built-in
10640 functions above, except they perform multiplication, instead of addition.
10644 The following built-in functions allow checking if simple arithmetic operation
10647 @deftypefn {Built-in Function} bool __builtin_add_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10648 @deftypefnx {Built-in Function} bool __builtin_sub_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10649 @deftypefnx {Built-in Function} bool __builtin_mul_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
10651 These built-in functions are similar to @code{__builtin_add_overflow},
10652 @code{__builtin_sub_overflow}, or @code{__builtin_mul_overflow}, except that
10653 they don't store the result of the arithmetic operation anywhere and the
10654 last argument is not a pointer, but some expression with integral type other
10655 than enumerated or boolean type.
10657 The built-in functions promote the first two operands into infinite precision signed type
10658 and perform addition on those promoted operands. The result is then
10659 cast to the type of the third argument. If the cast result is equal to the infinite
10660 precision result, the built-in functions return false, otherwise they return true.
10661 The value of the third argument is ignored, just the side-effects in the third argument
10662 are evaluated, and no integral argument promotions are performed on the last argument.
10663 If the third argument is a bit-field, the type used for the result cast has the
10664 precision and signedness of the given bit-field, rather than precision and signedness
10665 of the underlying type.
10667 For example, the following macro can be used to portably check, at
10668 compile-time, whether or not adding two constant integers will overflow,
10669 and perform the addition only when it is known to be safe and not to trigger
10670 a @option{-Woverflow} warning.
10673 #define INT_ADD_OVERFLOW_P(a, b) \
10674 __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0)
10677 A = INT_MAX, B = 3,
10678 C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B,
10679 D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0)
10683 The compiler will attempt to use hardware instructions to implement
10684 these built-in functions where possible, like conditional jump on overflow
10685 after addition, conditional jump on carry etc.
10689 @node x86 specific memory model extensions for transactional memory
10690 @section x86-Specific Memory Model Extensions for Transactional Memory
10692 The x86 architecture supports additional memory ordering flags
10693 to mark critical sections for hardware lock elision.
10694 These must be specified in addition to an existing memory order to
10698 @item __ATOMIC_HLE_ACQUIRE
10699 Start lock elision on a lock variable.
10700 Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
10701 @item __ATOMIC_HLE_RELEASE
10702 End lock elision on a lock variable.
10703 Memory order must be @code{__ATOMIC_RELEASE} or stronger.
10706 When a lock acquire fails, it is required for good performance to abort
10707 the transaction quickly. This can be done with a @code{_mm_pause}.
10710 #include <immintrin.h> // For _mm_pause
10714 /* Acquire lock with lock elision */
10715 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
10716 _mm_pause(); /* Abort failed transaction */
10718 /* Free lock with lock elision */
10719 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
10722 @node Object Size Checking
10723 @section Object Size Checking Built-in Functions
10724 @findex __builtin_object_size
10725 @findex __builtin___memcpy_chk
10726 @findex __builtin___mempcpy_chk
10727 @findex __builtin___memmove_chk
10728 @findex __builtin___memset_chk
10729 @findex __builtin___strcpy_chk
10730 @findex __builtin___stpcpy_chk
10731 @findex __builtin___strncpy_chk
10732 @findex __builtin___strcat_chk
10733 @findex __builtin___strncat_chk
10734 @findex __builtin___sprintf_chk
10735 @findex __builtin___snprintf_chk
10736 @findex __builtin___vsprintf_chk
10737 @findex __builtin___vsnprintf_chk
10738 @findex __builtin___printf_chk
10739 @findex __builtin___vprintf_chk
10740 @findex __builtin___fprintf_chk
10741 @findex __builtin___vfprintf_chk
10743 GCC implements a limited buffer overflow protection mechanism that can
10744 prevent some buffer overflow attacks by determining the sizes of objects
10745 into which data is about to be written and preventing the writes when
10746 the size isn't sufficient. The built-in functions described below yield
10747 the best results when used together and when optimization is enabled.
10748 For example, to detect object sizes across function boundaries or to
10749 follow pointer assignments through non-trivial control flow they rely
10750 on various optimization passes enabled with @option{-O2}. However, to
10751 a limited extent, they can be used without optimization as well.
10753 @deftypefn {Built-in Function} {size_t} __builtin_object_size (const void * @var{ptr}, int @var{type})
10754 is a built-in construct that returns a constant number of bytes from
10755 @var{ptr} to the end of the object @var{ptr} pointer points to
10756 (if known at compile time). @code{__builtin_object_size} never evaluates
10757 its arguments for side-effects. If there are any side-effects in them, it
10758 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10759 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
10760 point to and all of them are known at compile time, the returned number
10761 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
10762 0 and minimum if nonzero. If it is not possible to determine which objects
10763 @var{ptr} points to at compile time, @code{__builtin_object_size} should
10764 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
10765 for @var{type} 2 or 3.
10767 @var{type} is an integer constant from 0 to 3. If the least significant
10768 bit is clear, objects are whole variables, if it is set, a closest
10769 surrounding subobject is considered the object a pointer points to.
10770 The second bit determines if maximum or minimum of remaining bytes
10774 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
10775 char *p = &var.buf1[1], *q = &var.b;
10777 /* Here the object p points to is var. */
10778 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
10779 /* The subobject p points to is var.buf1. */
10780 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
10781 /* The object q points to is var. */
10782 assert (__builtin_object_size (q, 0)
10783 == (char *) (&var + 1) - (char *) &var.b);
10784 /* The subobject q points to is var.b. */
10785 assert (__builtin_object_size (q, 1) == sizeof (var.b));
10789 There are built-in functions added for many common string operation
10790 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
10791 built-in is provided. This built-in has an additional last argument,
10792 which is the number of bytes remaining in the object the @var{dest}
10793 argument points to or @code{(size_t) -1} if the size is not known.
10795 The built-in functions are optimized into the normal string functions
10796 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
10797 it is known at compile time that the destination object will not
10798 be overflowed. If the compiler can determine at compile time that the
10799 object will always be overflowed, it issues a warning.
10801 The intended use can be e.g.@:
10805 #define bos0(dest) __builtin_object_size (dest, 0)
10806 #define memcpy(dest, src, n) \
10807 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
10811 /* It is unknown what object p points to, so this is optimized
10812 into plain memcpy - no checking is possible. */
10813 memcpy (p, "abcde", n);
10814 /* Destination is known and length too. It is known at compile
10815 time there will be no overflow. */
10816 memcpy (&buf[5], "abcde", 5);
10817 /* Destination is known, but the length is not known at compile time.
10818 This will result in __memcpy_chk call that can check for overflow
10820 memcpy (&buf[5], "abcde", n);
10821 /* Destination is known and it is known at compile time there will
10822 be overflow. There will be a warning and __memcpy_chk call that
10823 will abort the program at run time. */
10824 memcpy (&buf[6], "abcde", 5);
10827 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
10828 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
10829 @code{strcat} and @code{strncat}.
10831 There are also checking built-in functions for formatted output functions.
10833 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
10834 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10835 const char *fmt, ...);
10836 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
10838 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
10839 const char *fmt, va_list ap);
10842 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
10843 etc.@: functions and can contain implementation specific flags on what
10844 additional security measures the checking function might take, such as
10845 handling @code{%n} differently.
10847 The @var{os} argument is the object size @var{s} points to, like in the
10848 other built-in functions. There is a small difference in the behavior
10849 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
10850 optimized into the non-checking functions only if @var{flag} is 0, otherwise
10851 the checking function is called with @var{os} argument set to
10852 @code{(size_t) -1}.
10854 In addition to this, there are checking built-in functions
10855 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
10856 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
10857 These have just one additional argument, @var{flag}, right before
10858 format string @var{fmt}. If the compiler is able to optimize them to
10859 @code{fputc} etc.@: functions, it does, otherwise the checking function
10860 is called and the @var{flag} argument passed to it.
10862 @node Pointer Bounds Checker builtins
10863 @section Pointer Bounds Checker Built-in Functions
10864 @cindex Pointer Bounds Checker builtins
10865 @findex __builtin___bnd_set_ptr_bounds
10866 @findex __builtin___bnd_narrow_ptr_bounds
10867 @findex __builtin___bnd_copy_ptr_bounds
10868 @findex __builtin___bnd_init_ptr_bounds
10869 @findex __builtin___bnd_null_ptr_bounds
10870 @findex __builtin___bnd_store_ptr_bounds
10871 @findex __builtin___bnd_chk_ptr_lbounds
10872 @findex __builtin___bnd_chk_ptr_ubounds
10873 @findex __builtin___bnd_chk_ptr_bounds
10874 @findex __builtin___bnd_get_ptr_lbound
10875 @findex __builtin___bnd_get_ptr_ubound
10877 GCC provides a set of built-in functions to control Pointer Bounds Checker
10878 instrumentation. Note that all Pointer Bounds Checker builtins can be used
10879 even if you compile with Pointer Bounds Checker off
10880 (@option{-fno-check-pointer-bounds}).
10881 The behavior may differ in such case as documented below.
10883 @deftypefn {Built-in Function} {void *} __builtin___bnd_set_ptr_bounds (const void *@var{q}, size_t @var{size})
10885 This built-in function returns a new pointer with the value of @var{q}, and
10886 associate it with the bounds [@var{q}, @var{q}+@var{size}-1]. With Pointer
10887 Bounds Checker off, the built-in function just returns the first argument.
10890 extern void *__wrap_malloc (size_t n)
10892 void *p = (void *)__real_malloc (n);
10893 if (!p) return __builtin___bnd_null_ptr_bounds (p);
10894 return __builtin___bnd_set_ptr_bounds (p, n);
10900 @deftypefn {Built-in Function} {void *} __builtin___bnd_narrow_ptr_bounds (const void *@var{p}, const void *@var{q}, size_t @var{size})
10902 This built-in function returns a new pointer with the value of @var{p}
10903 and associates it with the narrowed bounds formed by the intersection
10904 of bounds associated with @var{q} and the bounds
10905 [@var{p}, @var{p} + @var{size} - 1].
10906 With Pointer Bounds Checker off, the built-in function just returns the first
10910 void init_objects (object *objs, size_t size)
10913 /* Initialize objects one-by-one passing pointers with bounds of
10914 an object, not the full array of objects. */
10915 for (i = 0; i < size; i++)
10916 init_object (__builtin___bnd_narrow_ptr_bounds (objs + i, objs,
10923 @deftypefn {Built-in Function} {void *} __builtin___bnd_copy_ptr_bounds (const void *@var{q}, const void *@var{r})
10925 This built-in function returns a new pointer with the value of @var{q},
10926 and associates it with the bounds already associated with pointer @var{r}.
10927 With Pointer Bounds Checker off, the built-in function just returns the first
10931 /* Here is a way to get pointer to object's field but
10932 still with the full object's bounds. */
10933 int *field_ptr = __builtin___bnd_copy_ptr_bounds (&objptr->int_field,
10939 @deftypefn {Built-in Function} {void *} __builtin___bnd_init_ptr_bounds (const void *@var{q})
10941 This built-in function returns a new pointer with the value of @var{q}, and
10942 associates it with INIT (allowing full memory access) bounds. With Pointer
10943 Bounds Checker off, the built-in function just returns the first argument.
10947 @deftypefn {Built-in Function} {void *} __builtin___bnd_null_ptr_bounds (const void *@var{q})
10949 This built-in function returns a new pointer with the value of @var{q}, and
10950 associates it with NULL (allowing no memory access) bounds. With Pointer
10951 Bounds Checker off, the built-in function just returns the first argument.
10955 @deftypefn {Built-in Function} void __builtin___bnd_store_ptr_bounds (const void **@var{ptr_addr}, const void *@var{ptr_val})
10957 This built-in function stores the bounds associated with pointer @var{ptr_val}
10958 and location @var{ptr_addr} into Bounds Table. This can be useful to propagate
10959 bounds from legacy code without touching the associated pointer's memory when
10960 pointers are copied as integers. With Pointer Bounds Checker off, the built-in
10961 function call is ignored.
10965 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_lbounds (const void *@var{q})
10967 This built-in function checks if the pointer @var{q} is within the lower
10968 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10969 function call is ignored.
10972 extern void *__wrap_memset (void *dst, int c, size_t len)
10976 __builtin___bnd_chk_ptr_lbounds (dst);
10977 __builtin___bnd_chk_ptr_ubounds ((char *)dst + len - 1);
10978 __real_memset (dst, c, len);
10986 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_ubounds (const void *@var{q})
10988 This built-in function checks if the pointer @var{q} is within the upper
10989 bound of its associated bounds. With Pointer Bounds Checker off, the built-in
10990 function call is ignored.
10994 @deftypefn {Built-in Function} void __builtin___bnd_chk_ptr_bounds (const void *@var{q}, size_t @var{size})
10996 This built-in function checks if [@var{q}, @var{q} + @var{size} - 1] is within
10997 the lower and upper bounds associated with @var{q}. With Pointer Bounds Checker
10998 off, the built-in function call is ignored.
11001 extern void *__wrap_memcpy (void *dst, const void *src, size_t n)
11005 __bnd_chk_ptr_bounds (dst, n);
11006 __bnd_chk_ptr_bounds (src, n);
11007 __real_memcpy (dst, src, n);
11015 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_lbound (const void *@var{q})
11017 This built-in function returns the lower bound associated
11018 with the pointer @var{q}, as a pointer value.
11019 This is useful for debugging using @code{printf}.
11020 With Pointer Bounds Checker off, the built-in function returns 0.
11023 void *lb = __builtin___bnd_get_ptr_lbound (q);
11024 void *ub = __builtin___bnd_get_ptr_ubound (q);
11025 printf ("q = %p lb(q) = %p ub(q) = %p", q, lb, ub);
11030 @deftypefn {Built-in Function} {const void *} __builtin___bnd_get_ptr_ubound (const void *@var{q})
11032 This built-in function returns the upper bound (which is a pointer) associated
11033 with the pointer @var{q}. With Pointer Bounds Checker off,
11034 the built-in function returns -1.
11038 @node Other Builtins
11039 @section Other Built-in Functions Provided by GCC
11040 @cindex built-in functions
11041 @findex __builtin_alloca
11042 @findex __builtin_alloca_with_align
11043 @findex __builtin_alloca_with_align_and_max
11044 @findex __builtin_call_with_static_chain
11045 @findex __builtin_fpclassify
11046 @findex __builtin_isfinite
11047 @findex __builtin_isnormal
11048 @findex __builtin_isgreater
11049 @findex __builtin_isgreaterequal
11050 @findex __builtin_isinf_sign
11051 @findex __builtin_isless
11052 @findex __builtin_islessequal
11053 @findex __builtin_islessgreater
11054 @findex __builtin_isunordered
11055 @findex __builtin_powi
11056 @findex __builtin_powif
11057 @findex __builtin_powil
11218 @findex fprintf_unlocked
11220 @findex fputs_unlocked
11328 @findex nexttowardf
11329 @findex nexttowardl
11337 @findex printf_unlocked
11367 @findex signbitd128
11368 @findex significand
11369 @findex significandf
11370 @findex significandl
11398 @findex strncasecmp
11441 GCC provides a large number of built-in functions other than the ones
11442 mentioned above. Some of these are for internal use in the processing
11443 of exceptions or variable-length argument lists and are not
11444 documented here because they may change from time to time; we do not
11445 recommend general use of these functions.
11447 The remaining functions are provided for optimization purposes.
11449 With the exception of built-ins that have library equivalents such as
11450 the standard C library functions discussed below, or that expand to
11451 library calls, GCC built-in functions are always expanded inline and
11452 thus do not have corresponding entry points and their address cannot
11453 be obtained. Attempting to use them in an expression other than
11454 a function call results in a compile-time error.
11456 @opindex fno-builtin
11457 GCC includes built-in versions of many of the functions in the standard
11458 C library. These functions come in two forms: one whose names start with
11459 the @code{__builtin_} prefix, and the other without. Both forms have the
11460 same type (including prototype), the same address (when their address is
11461 taken), and the same meaning as the C library functions even if you specify
11462 the @option{-fno-builtin} option @pxref{C Dialect Options}). Many of these
11463 functions are only optimized in certain cases; if they are not optimized in
11464 a particular case, a call to the library function is emitted.
11468 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
11469 @option{-std=c99} or @option{-std=c11}), the functions
11470 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
11471 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
11472 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
11473 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
11474 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
11475 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
11476 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
11477 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
11478 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
11479 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
11480 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
11481 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
11482 @code{signbitd64}, @code{signbitd128}, @code{significandf},
11483 @code{significandl}, @code{significand}, @code{sincosf},
11484 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
11485 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
11486 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
11487 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
11489 may be handled as built-in functions.
11490 All these functions have corresponding versions
11491 prefixed with @code{__builtin_}, which may be used even in strict C90
11494 The ISO C99 functions
11495 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
11496 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
11497 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
11498 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
11499 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
11500 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
11501 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
11502 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
11503 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
11504 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
11505 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
11506 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
11507 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
11508 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
11509 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
11510 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
11511 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
11512 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
11513 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
11514 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
11515 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
11516 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
11517 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
11518 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
11519 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
11520 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
11521 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
11522 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
11523 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
11524 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
11525 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
11526 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
11527 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
11528 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
11529 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
11530 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
11531 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
11532 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
11533 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
11534 are handled as built-in functions
11535 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
11537 There are also built-in versions of the ISO C99 functions
11538 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
11539 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
11540 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
11541 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
11542 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
11543 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
11544 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
11545 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
11546 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
11547 that are recognized in any mode since ISO C90 reserves these names for
11548 the purpose to which ISO C99 puts them. All these functions have
11549 corresponding versions prefixed with @code{__builtin_}.
11551 There are also built-in functions @code{__builtin_fabsf@var{n}},
11552 @code{__builtin_fabsf@var{n}x}, @code{__builtin_copysignf@var{n}} and
11553 @code{__builtin_copysignf@var{n}x}, corresponding to the TS 18661-3
11554 functions @code{fabsf@var{n}}, @code{fabsf@var{n}x},
11555 @code{copysignf@var{n}} and @code{copysignf@var{n}x}, for supported
11556 types @code{_Float@var{n}} and @code{_Float@var{n}x}.
11558 There are also GNU extension functions @code{clog10}, @code{clog10f} and
11559 @code{clog10l} which names are reserved by ISO C99 for future use.
11560 All these functions have versions prefixed with @code{__builtin_}.
11562 The ISO C94 functions
11563 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
11564 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
11565 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
11567 are handled as built-in functions
11568 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
11570 The ISO C90 functions
11571 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
11572 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
11573 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
11574 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
11575 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
11576 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
11577 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
11578 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
11579 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
11580 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
11581 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
11582 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
11583 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
11584 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
11585 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
11586 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
11587 are all recognized as built-in functions unless
11588 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
11589 is specified for an individual function). All of these functions have
11590 corresponding versions prefixed with @code{__builtin_}.
11592 GCC provides built-in versions of the ISO C99 floating-point comparison
11593 macros that avoid raising exceptions for unordered operands. They have
11594 the same names as the standard macros ( @code{isgreater},
11595 @code{isgreaterequal}, @code{isless}, @code{islessequal},
11596 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
11597 prefixed. We intend for a library implementor to be able to simply
11598 @code{#define} each standard macro to its built-in equivalent.
11599 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
11600 @code{isinf_sign}, @code{isnormal} and @code{signbit} built-ins used with
11601 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
11602 built-in functions appear both with and without the @code{__builtin_} prefix.
11604 @deftypefn {Built-in Function} void *__builtin_alloca (size_t size)
11605 The @code{__builtin_alloca} function must be called at block scope.
11606 The function allocates an object @var{size} bytes large on the stack
11607 of the calling function. The object is aligned on the default stack
11608 alignment boundary for the target determined by the
11609 @code{__BIGGEST_ALIGNMENT__} macro. The @code{__builtin_alloca}
11610 function returns a pointer to the first byte of the allocated object.
11611 The lifetime of the allocated object ends just before the calling
11612 function returns to its caller. This is so even when
11613 @code{__builtin_alloca} is called within a nested block.
11615 For example, the following function allocates eight objects of @code{n}
11616 bytes each on the stack, storing a pointer to each in consecutive elements
11617 of the array @code{a}. It then passes the array to function @code{g}
11618 which can safely use the storage pointed to by each of the array elements.
11621 void f (unsigned n)
11624 for (int i = 0; i != 8; ++i)
11625 a [i] = __builtin_alloca (n);
11627 g (a, n); // @r{safe}
11631 Since the @code{__builtin_alloca} function doesn't validate its argument
11632 it is the responsibility of its caller to make sure the argument doesn't
11633 cause it to exceed the stack size limit.
11634 The @code{__builtin_alloca} function is provided to make it possible to
11635 allocate on the stack arrays of bytes with an upper bound that may be
11636 computed at run time. Since C99 Variable Length Arrays offer
11637 similar functionality under a portable, more convenient, and safer
11638 interface they are recommended instead, in both C99 and C++ programs
11639 where GCC provides them as an extension.
11640 @xref{Variable Length}, for details.
11644 @deftypefn {Built-in Function} void *__builtin_alloca_with_align (size_t size, size_t alignment)
11645 The @code{__builtin_alloca_with_align} function must be called at block
11646 scope. The function allocates an object @var{size} bytes large on
11647 the stack of the calling function. The allocated object is aligned on
11648 the boundary specified by the argument @var{alignment} whose unit is given
11649 in bits (not bytes). The @var{size} argument must be positive and not
11650 exceed the stack size limit. The @var{alignment} argument must be a constant
11651 integer expression that evaluates to a power of 2 greater than or equal to
11652 @code{CHAR_BIT} and less than some unspecified maximum. Invocations
11653 with other values are rejected with an error indicating the valid bounds.
11654 The function returns a pointer to the first byte of the allocated object.
11655 The lifetime of the allocated object ends at the end of the block in which
11656 the function was called. The allocated storage is released no later than
11657 just before the calling function returns to its caller, but may be released
11658 at the end of the block in which the function was called.
11660 For example, in the following function the call to @code{g} is unsafe
11661 because when @code{overalign} is non-zero, the space allocated by
11662 @code{__builtin_alloca_with_align} may have been released at the end
11663 of the @code{if} statement in which it was called.
11666 void f (unsigned n, bool overalign)
11670 p = __builtin_alloca_with_align (n, 64 /* bits */);
11672 p = __builtin_alloc (n);
11674 g (p, n); // @r{unsafe}
11678 Since the @code{__builtin_alloca_with_align} function doesn't validate its
11679 @var{size} argument it is the responsibility of its caller to make sure
11680 the argument doesn't cause it to exceed the stack size limit.
11681 The @code{__builtin_alloca_with_align} function is provided to make
11682 it possible to allocate on the stack overaligned arrays of bytes with
11683 an upper bound that may be computed at run time. Since C99
11684 Variable Length Arrays offer the same functionality under
11685 a portable, more convenient, and safer interface they are recommended
11686 instead, in both C99 and C++ programs where GCC provides them as
11687 an extension. @xref{Variable Length}, for details.
11691 @deftypefn {Built-in Function} void *__builtin_alloca_with_align_and_max (size_t size, size_t alignment, size_t max_size)
11692 Similar to @code{__builtin_alloca_with_align} but takes an extra argument
11693 specifying an upper bound for @var{size} in case its value cannot be computed
11694 at compile time, for use by @option{-fstack-usage}, @option{-Wstack-usage}
11695 and @option{-Walloca-larger-than}. @var{max_size} must be a constant integer
11696 expression, it has no effect on code generation and no attempt is made to
11697 check its compatibility with @var{size}.
11701 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
11703 You can use the built-in function @code{__builtin_types_compatible_p} to
11704 determine whether two types are the same.
11706 This built-in function returns 1 if the unqualified versions of the
11707 types @var{type1} and @var{type2} (which are types, not expressions) are
11708 compatible, 0 otherwise. The result of this built-in function can be
11709 used in integer constant expressions.
11711 This built-in function ignores top level qualifiers (e.g., @code{const},
11712 @code{volatile}). For example, @code{int} is equivalent to @code{const
11715 The type @code{int[]} and @code{int[5]} are compatible. On the other
11716 hand, @code{int} and @code{char *} are not compatible, even if the size
11717 of their types, on the particular architecture are the same. Also, the
11718 amount of pointer indirection is taken into account when determining
11719 similarity. Consequently, @code{short *} is not similar to
11720 @code{short **}. Furthermore, two types that are typedefed are
11721 considered compatible if their underlying types are compatible.
11723 An @code{enum} type is not considered to be compatible with another
11724 @code{enum} type even if both are compatible with the same integer
11725 type; this is what the C standard specifies.
11726 For example, @code{enum @{foo, bar@}} is not similar to
11727 @code{enum @{hot, dog@}}.
11729 You typically use this function in code whose execution varies
11730 depending on the arguments' types. For example:
11735 typeof (x) tmp = (x); \
11736 if (__builtin_types_compatible_p (typeof (x), long double)) \
11737 tmp = foo_long_double (tmp); \
11738 else if (__builtin_types_compatible_p (typeof (x), double)) \
11739 tmp = foo_double (tmp); \
11740 else if (__builtin_types_compatible_p (typeof (x), float)) \
11741 tmp = foo_float (tmp); \
11748 @emph{Note:} This construct is only available for C@.
11752 @deftypefn {Built-in Function} @var{type} __builtin_call_with_static_chain (@var{call_exp}, @var{pointer_exp})
11754 The @var{call_exp} expression must be a function call, and the
11755 @var{pointer_exp} expression must be a pointer. The @var{pointer_exp}
11756 is passed to the function call in the target's static chain location.
11757 The result of builtin is the result of the function call.
11759 @emph{Note:} This builtin is only available for C@.
11760 This builtin can be used to call Go closures from C.
11764 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
11766 You can use the built-in function @code{__builtin_choose_expr} to
11767 evaluate code depending on the value of a constant expression. This
11768 built-in function returns @var{exp1} if @var{const_exp}, which is an
11769 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
11771 This built-in function is analogous to the @samp{? :} operator in C,
11772 except that the expression returned has its type unaltered by promotion
11773 rules. Also, the built-in function does not evaluate the expression
11774 that is not chosen. For example, if @var{const_exp} evaluates to true,
11775 @var{exp2} is not evaluated even if it has side-effects.
11777 This built-in function can return an lvalue if the chosen argument is an
11780 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
11781 type. Similarly, if @var{exp2} is returned, its return type is the same
11788 __builtin_choose_expr ( \
11789 __builtin_types_compatible_p (typeof (x), double), \
11791 __builtin_choose_expr ( \
11792 __builtin_types_compatible_p (typeof (x), float), \
11794 /* @r{The void expression results in a compile-time error} \
11795 @r{when assigning the result to something.} */ \
11799 @emph{Note:} This construct is only available for C@. Furthermore, the
11800 unused expression (@var{exp1} or @var{exp2} depending on the value of
11801 @var{const_exp}) may still generate syntax errors. This may change in
11806 @deftypefn {Built-in Function} @var{type} __builtin_tgmath (@var{functions}, @var{arguments})
11808 The built-in function @code{__builtin_tgmath}, available only for C
11809 and Objective-C, calls a function determined according to the rules of
11810 @code{<tgmath.h>} macros. It is intended to be used in
11811 implementations of that header, so that expansions of macros from that
11812 header only expand each of their arguments once, to avoid problems
11813 when calls to such macros are nested inside the arguments of other
11814 calls to such macros; in addition, it results in better diagnostics
11815 for invalid calls to @code{<tgmath.h>} macros than implementations
11816 using other GNU C language features. For example, the @code{pow}
11817 type-generic macro might be defined as:
11820 #define pow(a, b) __builtin_tgmath (powf, pow, powl, \
11821 cpowf, cpow, cpowl, a, b)
11824 The arguments to @code{__builtin_tgmath} are at least two pointers to
11825 functions, followed by the arguments to the type-generic macro (which
11826 will be passed as arguments to the selected function). All the
11827 pointers to functions must be pointers to prototyped functions, none
11828 of which may have variable arguments, and all of which must have the
11829 same number of parameters; the number of parameters of the first
11830 function determines how many arguments to @code{__builtin_tgmath} are
11831 interpreted as function pointers, and how many as the arguments to the
11834 The types of the specified functions must all be different, but
11835 related to each other in the same way as a set of functions that may
11836 be selected between by a macro in @code{<tgmath.h>}. This means that
11837 the functions are parameterized by a floating-point type @var{t},
11838 different for each such function. The function return types may all
11839 be the same type, or they may be @var{t} for each function, or they
11840 may be the real type corresponding to @var{t} for each function (if
11841 some of the types @var{t} are complex). Likewise, for each parameter
11842 position, the type of the parameter in that position may always be the
11843 same type, or may be @var{t} for each function (this case must apply
11844 for at least one parameter position), or may be the real type
11845 corresponding to @var{t} for each function.
11847 The standard rules for @code{<tgmath.h>} macros are used to find a
11848 common type @var{u} from the types of the arguments for parameters
11849 whose types vary between the functions; complex integer types (a GNU
11850 extension) are treated like @code{_Complex double} for this purpose.
11851 If the function return types vary, or are all the same integer type,
11852 the function called is the one for which @var{t} is @var{u}, and it is
11853 an error if there is no such function. If the function return types
11854 are all the same floating-point type, the type-generic macro is taken
11855 to be one of those from TS 18661 that rounds the result to a narrower
11856 type; if there is a function for which @var{t} is @var{u}, it is
11857 called, and otherwise the first function, if any, for which @var{t}
11858 has at least the range and precision of @var{u} is called, and it is
11859 an error if there is no such function.
11863 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
11865 The built-in function @code{__builtin_complex} is provided for use in
11866 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
11867 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
11868 real binary floating-point type, and the result has the corresponding
11869 complex type with real and imaginary parts @var{real} and @var{imag}.
11870 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
11871 infinities, NaNs and negative zeros are involved.
11875 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
11876 You can use the built-in function @code{__builtin_constant_p} to
11877 determine if a value is known to be constant at compile time and hence
11878 that GCC can perform constant-folding on expressions involving that
11879 value. The argument of the function is the value to test. The function
11880 returns the integer 1 if the argument is known to be a compile-time
11881 constant and 0 if it is not known to be a compile-time constant. A
11882 return of 0 does not indicate that the value is @emph{not} a constant,
11883 but merely that GCC cannot prove it is a constant with the specified
11884 value of the @option{-O} option.
11886 You typically use this function in an embedded application where
11887 memory is a critical resource. If you have some complex calculation,
11888 you may want it to be folded if it involves constants, but need to call
11889 a function if it does not. For example:
11892 #define Scale_Value(X) \
11893 (__builtin_constant_p (X) \
11894 ? ((X) * SCALE + OFFSET) : Scale (X))
11897 You may use this built-in function in either a macro or an inline
11898 function. However, if you use it in an inlined function and pass an
11899 argument of the function as the argument to the built-in, GCC
11900 never returns 1 when you call the inline function with a string constant
11901 or compound literal (@pxref{Compound Literals}) and does not return 1
11902 when you pass a constant numeric value to the inline function unless you
11903 specify the @option{-O} option.
11905 You may also use @code{__builtin_constant_p} in initializers for static
11906 data. For instance, you can write
11909 static const int table[] = @{
11910 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
11916 This is an acceptable initializer even if @var{EXPRESSION} is not a
11917 constant expression, including the case where
11918 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
11919 folded to a constant but @var{EXPRESSION} contains operands that are
11920 not otherwise permitted in a static initializer (for example,
11921 @code{0 && foo ()}). GCC must be more conservative about evaluating the
11922 built-in in this case, because it has no opportunity to perform
11926 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
11927 @opindex fprofile-arcs
11928 You may use @code{__builtin_expect} to provide the compiler with
11929 branch prediction information. In general, you should prefer to
11930 use actual profile feedback for this (@option{-fprofile-arcs}), as
11931 programmers are notoriously bad at predicting how their programs
11932 actually perform. However, there are applications in which this
11933 data is hard to collect.
11935 The return value is the value of @var{exp}, which should be an integral
11936 expression. The semantics of the built-in are that it is expected that
11937 @var{exp} == @var{c}. For example:
11940 if (__builtin_expect (x, 0))
11945 indicates that we do not expect to call @code{foo}, since
11946 we expect @code{x} to be zero. Since you are limited to integral
11947 expressions for @var{exp}, you should use constructions such as
11950 if (__builtin_expect (ptr != NULL, 1))
11955 when testing pointer or floating-point values.
11958 @deftypefn {Built-in Function} void __builtin_trap (void)
11959 This function causes the program to exit abnormally. GCC implements
11960 this function by using a target-dependent mechanism (such as
11961 intentionally executing an illegal instruction) or by calling
11962 @code{abort}. The mechanism used may vary from release to release so
11963 you should not rely on any particular implementation.
11966 @deftypefn {Built-in Function} void __builtin_unreachable (void)
11967 If control flow reaches the point of the @code{__builtin_unreachable},
11968 the program is undefined. It is useful in situations where the
11969 compiler cannot deduce the unreachability of the code.
11971 One such case is immediately following an @code{asm} statement that
11972 either never terminates, or one that transfers control elsewhere
11973 and never returns. In this example, without the
11974 @code{__builtin_unreachable}, GCC issues a warning that control
11975 reaches the end of a non-void function. It also generates code
11976 to return after the @code{asm}.
11979 int f (int c, int v)
11987 asm("jmp error_handler");
11988 __builtin_unreachable ();
11994 Because the @code{asm} statement unconditionally transfers control out
11995 of the function, control never reaches the end of the function
11996 body. The @code{__builtin_unreachable} is in fact unreachable and
11997 communicates this fact to the compiler.
11999 Another use for @code{__builtin_unreachable} is following a call a
12000 function that never returns but that is not declared
12001 @code{__attribute__((noreturn))}, as in this example:
12004 void function_that_never_returns (void);
12014 function_that_never_returns ();
12015 __builtin_unreachable ();
12022 @deftypefn {Built-in Function} {void *} __builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
12023 This function returns its first argument, and allows the compiler
12024 to assume that the returned pointer is at least @var{align} bytes
12025 aligned. This built-in can have either two or three arguments,
12026 if it has three, the third argument should have integer type, and
12027 if it is nonzero means misalignment offset. For example:
12030 void *x = __builtin_assume_aligned (arg, 16);
12034 means that the compiler can assume @code{x}, set to @code{arg}, is at least
12035 16-byte aligned, while:
12038 void *x = __builtin_assume_aligned (arg, 32, 8);
12042 means that the compiler can assume for @code{x}, set to @code{arg}, that
12043 @code{(char *) x - 8} is 32-byte aligned.
12046 @deftypefn {Built-in Function} int __builtin_LINE ()
12047 This function is the equivalent of the preprocessor @code{__LINE__}
12048 macro and returns a constant integer expression that evaluates to
12049 the line number of the invocation of the built-in. When used as a C++
12050 default argument for a function @var{F}, it returns the line number
12051 of the call to @var{F}.
12054 @deftypefn {Built-in Function} {const char *} __builtin_FUNCTION ()
12055 This function is the equivalent of the @code{__FUNCTION__} symbol
12056 and returns an address constant pointing to the name of the function
12057 from which the built-in was invoked, or the empty string if
12058 the invocation is not at function scope. When used as a C++ default
12059 argument for a function @var{F}, it returns the name of @var{F}'s
12060 caller or the empty string if the call was not made at function
12064 @deftypefn {Built-in Function} {const char *} __builtin_FILE ()
12065 This function is the equivalent of the preprocessor @code{__FILE__}
12066 macro and returns an address constant pointing to the file name
12067 containing the invocation of the built-in, or the empty string if
12068 the invocation is not at function scope. When used as a C++ default
12069 argument for a function @var{F}, it returns the file name of the call
12070 to @var{F} or the empty string if the call was not made at function
12073 For example, in the following, each call to function @code{foo} will
12074 print a line similar to @code{"file.c:123: foo: message"} with the name
12075 of the file and the line number of the @code{printf} call, the name of
12076 the function @code{foo}, followed by the word @code{message}.
12080 function (const char *func = __builtin_FUNCTION ())
12087 printf ("%s:%i: %s: message\n", file (), line (), function ());
12093 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
12094 This function is used to flush the processor's instruction cache for
12095 the region of memory between @var{begin} inclusive and @var{end}
12096 exclusive. Some targets require that the instruction cache be
12097 flushed, after modifying memory containing code, in order to obtain
12098 deterministic behavior.
12100 If the target does not require instruction cache flushes,
12101 @code{__builtin___clear_cache} has no effect. Otherwise either
12102 instructions are emitted in-line to clear the instruction cache or a
12103 call to the @code{__clear_cache} function in libgcc is made.
12106 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
12107 This function is used to minimize cache-miss latency by moving data into
12108 a cache before it is accessed.
12109 You can insert calls to @code{__builtin_prefetch} into code for which
12110 you know addresses of data in memory that is likely to be accessed soon.
12111 If the target supports them, data prefetch instructions are generated.
12112 If the prefetch is done early enough before the access then the data will
12113 be in the cache by the time it is accessed.
12115 The value of @var{addr} is the address of the memory to prefetch.
12116 There are two optional arguments, @var{rw} and @var{locality}.
12117 The value of @var{rw} is a compile-time constant one or zero; one
12118 means that the prefetch is preparing for a write to the memory address
12119 and zero, the default, means that the prefetch is preparing for a read.
12120 The value @var{locality} must be a compile-time constant integer between
12121 zero and three. A value of zero means that the data has no temporal
12122 locality, so it need not be left in the cache after the access. A value
12123 of three means that the data has a high degree of temporal locality and
12124 should be left in all levels of cache possible. Values of one and two
12125 mean, respectively, a low or moderate degree of temporal locality. The
12129 for (i = 0; i < n; i++)
12131 a[i] = a[i] + b[i];
12132 __builtin_prefetch (&a[i+j], 1, 1);
12133 __builtin_prefetch (&b[i+j], 0, 1);
12138 Data prefetch does not generate faults if @var{addr} is invalid, but
12139 the address expression itself must be valid. For example, a prefetch
12140 of @code{p->next} does not fault if @code{p->next} is not a valid
12141 address, but evaluation faults if @code{p} is not a valid address.
12143 If the target does not support data prefetch, the address expression
12144 is evaluated if it includes side effects but no other code is generated
12145 and GCC does not issue a warning.
12148 @deftypefn {Built-in Function} double __builtin_huge_val (void)
12149 Returns a positive infinity, if supported by the floating-point format,
12150 else @code{DBL_MAX}. This function is suitable for implementing the
12151 ISO C macro @code{HUGE_VAL}.
12154 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
12155 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
12158 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
12159 Similar to @code{__builtin_huge_val}, except the return
12160 type is @code{long double}.
12163 @deftypefn {Built-in Function} _Float@var{n} __builtin_huge_valf@var{n} (void)
12164 Similar to @code{__builtin_huge_val}, except the return type is
12165 @code{_Float@var{n}}.
12168 @deftypefn {Built-in Function} _Float@var{n}x __builtin_huge_valf@var{n}x (void)
12169 Similar to @code{__builtin_huge_val}, except the return type is
12170 @code{_Float@var{n}x}.
12173 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
12174 This built-in implements the C99 fpclassify functionality. The first
12175 five int arguments should be the target library's notion of the
12176 possible FP classes and are used for return values. They must be
12177 constant values and they must appear in this order: @code{FP_NAN},
12178 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
12179 @code{FP_ZERO}. The ellipsis is for exactly one floating-point value
12180 to classify. GCC treats the last argument as type-generic, which
12181 means it does not do default promotion from float to double.
12184 @deftypefn {Built-in Function} double __builtin_inf (void)
12185 Similar to @code{__builtin_huge_val}, except a warning is generated
12186 if the target floating-point format does not support infinities.
12189 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
12190 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
12193 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
12194 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
12197 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
12198 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
12201 @deftypefn {Built-in Function} float __builtin_inff (void)
12202 Similar to @code{__builtin_inf}, except the return type is @code{float}.
12203 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
12206 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
12207 Similar to @code{__builtin_inf}, except the return
12208 type is @code{long double}.
12211 @deftypefn {Built-in Function} _Float@var{n} __builtin_inff@var{n} (void)
12212 Similar to @code{__builtin_inf}, except the return
12213 type is @code{_Float@var{n}}.
12216 @deftypefn {Built-in Function} _Float@var{n} __builtin_inff@var{n}x (void)
12217 Similar to @code{__builtin_inf}, except the return
12218 type is @code{_Float@var{n}x}.
12221 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
12222 Similar to @code{isinf}, except the return value is -1 for
12223 an argument of @code{-Inf} and 1 for an argument of @code{+Inf}.
12224 Note while the parameter list is an
12225 ellipsis, this function only accepts exactly one floating-point
12226 argument. GCC treats this parameter as type-generic, which means it
12227 does not do default promotion from float to double.
12230 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
12231 This is an implementation of the ISO C99 function @code{nan}.
12233 Since ISO C99 defines this function in terms of @code{strtod}, which we
12234 do not implement, a description of the parsing is in order. The string
12235 is parsed as by @code{strtol}; that is, the base is recognized by
12236 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
12237 in the significand such that the least significant bit of the number
12238 is at the least significant bit of the significand. The number is
12239 truncated to fit the significand field provided. The significand is
12240 forced to be a quiet NaN@.
12242 This function, if given a string literal all of which would have been
12243 consumed by @code{strtol}, is evaluated early enough that it is considered a
12244 compile-time constant.
12247 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
12248 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
12251 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
12252 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
12255 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
12256 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
12259 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
12260 Similar to @code{__builtin_nan}, except the return type is @code{float}.
12263 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
12264 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
12267 @deftypefn {Built-in Function} _Float@var{n} __builtin_nanf@var{n} (const char *str)
12268 Similar to @code{__builtin_nan}, except the return type is
12269 @code{_Float@var{n}}.
12272 @deftypefn {Built-in Function} _Float@var{n}x __builtin_nanf@var{n}x (const char *str)
12273 Similar to @code{__builtin_nan}, except the return type is
12274 @code{_Float@var{n}x}.
12277 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
12278 Similar to @code{__builtin_nan}, except the significand is forced
12279 to be a signaling NaN@. The @code{nans} function is proposed by
12280 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
12283 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
12284 Similar to @code{__builtin_nans}, except the return type is @code{float}.
12287 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
12288 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
12291 @deftypefn {Built-in Function} _Float@var{n} __builtin_nansf@var{n} (const char *str)
12292 Similar to @code{__builtin_nans}, except the return type is
12293 @code{_Float@var{n}}.
12296 @deftypefn {Built-in Function} _Float@var{n}x __builtin_nansf@var{n}x (const char *str)
12297 Similar to @code{__builtin_nans}, except the return type is
12298 @code{_Float@var{n}x}.
12301 @deftypefn {Built-in Function} int __builtin_ffs (int x)
12302 Returns one plus the index of the least significant 1-bit of @var{x}, or
12303 if @var{x} is zero, returns zero.
12306 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
12307 Returns the number of leading 0-bits in @var{x}, starting at the most
12308 significant bit position. If @var{x} is 0, the result is undefined.
12311 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
12312 Returns the number of trailing 0-bits in @var{x}, starting at the least
12313 significant bit position. If @var{x} is 0, the result is undefined.
12316 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
12317 Returns the number of leading redundant sign bits in @var{x}, i.e.@: the
12318 number of bits following the most significant bit that are identical
12319 to it. There are no special cases for 0 or other values.
12322 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
12323 Returns the number of 1-bits in @var{x}.
12326 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
12327 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
12331 @deftypefn {Built-in Function} int __builtin_ffsl (long)
12332 Similar to @code{__builtin_ffs}, except the argument type is
12336 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
12337 Similar to @code{__builtin_clz}, except the argument type is
12338 @code{unsigned long}.
12341 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
12342 Similar to @code{__builtin_ctz}, except the argument type is
12343 @code{unsigned long}.
12346 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
12347 Similar to @code{__builtin_clrsb}, except the argument type is
12351 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
12352 Similar to @code{__builtin_popcount}, except the argument type is
12353 @code{unsigned long}.
12356 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
12357 Similar to @code{__builtin_parity}, except the argument type is
12358 @code{unsigned long}.
12361 @deftypefn {Built-in Function} int __builtin_ffsll (long long)
12362 Similar to @code{__builtin_ffs}, except the argument type is
12366 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
12367 Similar to @code{__builtin_clz}, except the argument type is
12368 @code{unsigned long long}.
12371 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
12372 Similar to @code{__builtin_ctz}, except the argument type is
12373 @code{unsigned long long}.
12376 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
12377 Similar to @code{__builtin_clrsb}, except the argument type is
12381 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
12382 Similar to @code{__builtin_popcount}, except the argument type is
12383 @code{unsigned long long}.
12386 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
12387 Similar to @code{__builtin_parity}, except the argument type is
12388 @code{unsigned long long}.
12391 @deftypefn {Built-in Function} double __builtin_powi (double, int)
12392 Returns the first argument raised to the power of the second. Unlike the
12393 @code{pow} function no guarantees about precision and rounding are made.
12396 @deftypefn {Built-in Function} float __builtin_powif (float, int)
12397 Similar to @code{__builtin_powi}, except the argument and return types
12401 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
12402 Similar to @code{__builtin_powi}, except the argument and return types
12403 are @code{long double}.
12406 @deftypefn {Built-in Function} uint16_t __builtin_bswap16 (uint16_t x)
12407 Returns @var{x} with the order of the bytes reversed; for example,
12408 @code{0xaabb} becomes @code{0xbbaa}. Byte here always means
12412 @deftypefn {Built-in Function} uint32_t __builtin_bswap32 (uint32_t x)
12413 Similar to @code{__builtin_bswap16}, except the argument and return types
12417 @deftypefn {Built-in Function} uint64_t __builtin_bswap64 (uint64_t x)
12418 Similar to @code{__builtin_bswap32}, except the argument and return types
12422 @node Target Builtins
12423 @section Built-in Functions Specific to Particular Target Machines
12425 On some target machines, GCC supports many built-in functions specific
12426 to those machines. Generally these generate calls to specific machine
12427 instructions, but allow the compiler to schedule those calls.
12430 * AArch64 Built-in Functions::
12431 * Alpha Built-in Functions::
12432 * Altera Nios II Built-in Functions::
12433 * ARC Built-in Functions::
12434 * ARC SIMD Built-in Functions::
12435 * ARM iWMMXt Built-in Functions::
12436 * ARM C Language Extensions (ACLE)::
12437 * ARM Floating Point Status and Control Intrinsics::
12438 * ARM ARMv8-M Security Extensions::
12439 * AVR Built-in Functions::
12440 * Blackfin Built-in Functions::
12441 * FR-V Built-in Functions::
12442 * MIPS DSP Built-in Functions::
12443 * MIPS Paired-Single Support::
12444 * MIPS Loongson Built-in Functions::
12445 * MIPS SIMD Architecture (MSA) Support::
12446 * Other MIPS Built-in Functions::
12447 * MSP430 Built-in Functions::
12448 * NDS32 Built-in Functions::
12449 * picoChip Built-in Functions::
12450 * PowerPC Built-in Functions::
12451 * PowerPC AltiVec/VSX Built-in Functions::
12452 * PowerPC Hardware Transactional Memory Built-in Functions::
12453 * PowerPC Atomic Memory Operation Functions::
12454 * RX Built-in Functions::
12455 * S/390 System z Built-in Functions::
12456 * SH Built-in Functions::
12457 * SPARC VIS Built-in Functions::
12458 * SPU Built-in Functions::
12459 * TI C6X Built-in Functions::
12460 * TILE-Gx Built-in Functions::
12461 * TILEPro Built-in Functions::
12462 * x86 Built-in Functions::
12463 * x86 transactional memory intrinsics::
12466 @node AArch64 Built-in Functions
12467 @subsection AArch64 Built-in Functions
12469 These built-in functions are available for the AArch64 family of
12472 unsigned int __builtin_aarch64_get_fpcr ()
12473 void __builtin_aarch64_set_fpcr (unsigned int)
12474 unsigned int __builtin_aarch64_get_fpsr ()
12475 void __builtin_aarch64_set_fpsr (unsigned int)
12478 @node Alpha Built-in Functions
12479 @subsection Alpha Built-in Functions
12481 These built-in functions are available for the Alpha family of
12482 processors, depending on the command-line switches used.
12484 The following built-in functions are always available. They
12485 all generate the machine instruction that is part of the name.
12488 long __builtin_alpha_implver (void)
12489 long __builtin_alpha_rpcc (void)
12490 long __builtin_alpha_amask (long)
12491 long __builtin_alpha_cmpbge (long, long)
12492 long __builtin_alpha_extbl (long, long)
12493 long __builtin_alpha_extwl (long, long)
12494 long __builtin_alpha_extll (long, long)
12495 long __builtin_alpha_extql (long, long)
12496 long __builtin_alpha_extwh (long, long)
12497 long __builtin_alpha_extlh (long, long)
12498 long __builtin_alpha_extqh (long, long)
12499 long __builtin_alpha_insbl (long, long)
12500 long __builtin_alpha_inswl (long, long)
12501 long __builtin_alpha_insll (long, long)
12502 long __builtin_alpha_insql (long, long)
12503 long __builtin_alpha_inswh (long, long)
12504 long __builtin_alpha_inslh (long, long)
12505 long __builtin_alpha_insqh (long, long)
12506 long __builtin_alpha_mskbl (long, long)
12507 long __builtin_alpha_mskwl (long, long)
12508 long __builtin_alpha_mskll (long, long)
12509 long __builtin_alpha_mskql (long, long)
12510 long __builtin_alpha_mskwh (long, long)
12511 long __builtin_alpha_msklh (long, long)
12512 long __builtin_alpha_mskqh (long, long)
12513 long __builtin_alpha_umulh (long, long)
12514 long __builtin_alpha_zap (long, long)
12515 long __builtin_alpha_zapnot (long, long)
12518 The following built-in functions are always with @option{-mmax}
12519 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
12520 later. They all generate the machine instruction that is part
12524 long __builtin_alpha_pklb (long)
12525 long __builtin_alpha_pkwb (long)
12526 long __builtin_alpha_unpkbl (long)
12527 long __builtin_alpha_unpkbw (long)
12528 long __builtin_alpha_minub8 (long, long)
12529 long __builtin_alpha_minsb8 (long, long)
12530 long __builtin_alpha_minuw4 (long, long)
12531 long __builtin_alpha_minsw4 (long, long)
12532 long __builtin_alpha_maxub8 (long, long)
12533 long __builtin_alpha_maxsb8 (long, long)
12534 long __builtin_alpha_maxuw4 (long, long)
12535 long __builtin_alpha_maxsw4 (long, long)
12536 long __builtin_alpha_perr (long, long)
12539 The following built-in functions are always with @option{-mcix}
12540 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
12541 later. They all generate the machine instruction that is part
12545 long __builtin_alpha_cttz (long)
12546 long __builtin_alpha_ctlz (long)
12547 long __builtin_alpha_ctpop (long)
12550 The following built-in functions are available on systems that use the OSF/1
12551 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
12552 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
12553 @code{rdval} and @code{wrval}.
12556 void *__builtin_thread_pointer (void)
12557 void __builtin_set_thread_pointer (void *)
12560 @node Altera Nios II Built-in Functions
12561 @subsection Altera Nios II Built-in Functions
12563 These built-in functions are available for the Altera Nios II
12564 family of processors.
12566 The following built-in functions are always available. They
12567 all generate the machine instruction that is part of the name.
12570 int __builtin_ldbio (volatile const void *)
12571 int __builtin_ldbuio (volatile const void *)
12572 int __builtin_ldhio (volatile const void *)
12573 int __builtin_ldhuio (volatile const void *)
12574 int __builtin_ldwio (volatile const void *)
12575 void __builtin_stbio (volatile void *, int)
12576 void __builtin_sthio (volatile void *, int)
12577 void __builtin_stwio (volatile void *, int)
12578 void __builtin_sync (void)
12579 int __builtin_rdctl (int)
12580 int __builtin_rdprs (int, int)
12581 void __builtin_wrctl (int, int)
12582 void __builtin_flushd (volatile void *)
12583 void __builtin_flushda (volatile void *)
12584 int __builtin_wrpie (int);
12585 void __builtin_eni (int);
12586 int __builtin_ldex (volatile const void *)
12587 int __builtin_stex (volatile void *, int)
12588 int __builtin_ldsex (volatile const void *)
12589 int __builtin_stsex (volatile void *, int)
12592 The following built-in functions are always available. They
12593 all generate a Nios II Custom Instruction. The name of the
12594 function represents the types that the function takes and
12595 returns. The letter before the @code{n} is the return type
12596 or void if absent. The @code{n} represents the first parameter
12597 to all the custom instructions, the custom instruction number.
12598 The two letters after the @code{n} represent the up to two
12599 parameters to the function.
12601 The letters represent the following data types:
12604 @code{void} for return type and no parameter for parameter types.
12607 @code{int} for return type and parameter type
12610 @code{float} for return type and parameter type
12613 @code{void *} for return type and parameter type
12617 And the function names are:
12619 void __builtin_custom_n (void)
12620 void __builtin_custom_ni (int)
12621 void __builtin_custom_nf (float)
12622 void __builtin_custom_np (void *)
12623 void __builtin_custom_nii (int, int)
12624 void __builtin_custom_nif (int, float)
12625 void __builtin_custom_nip (int, void *)
12626 void __builtin_custom_nfi (float, int)
12627 void __builtin_custom_nff (float, float)
12628 void __builtin_custom_nfp (float, void *)
12629 void __builtin_custom_npi (void *, int)
12630 void __builtin_custom_npf (void *, float)
12631 void __builtin_custom_npp (void *, void *)
12632 int __builtin_custom_in (void)
12633 int __builtin_custom_ini (int)
12634 int __builtin_custom_inf (float)
12635 int __builtin_custom_inp (void *)
12636 int __builtin_custom_inii (int, int)
12637 int __builtin_custom_inif (int, float)
12638 int __builtin_custom_inip (int, void *)
12639 int __builtin_custom_infi (float, int)
12640 int __builtin_custom_inff (float, float)
12641 int __builtin_custom_infp (float, void *)
12642 int __builtin_custom_inpi (void *, int)
12643 int __builtin_custom_inpf (void *, float)
12644 int __builtin_custom_inpp (void *, void *)
12645 float __builtin_custom_fn (void)
12646 float __builtin_custom_fni (int)
12647 float __builtin_custom_fnf (float)
12648 float __builtin_custom_fnp (void *)
12649 float __builtin_custom_fnii (int, int)
12650 float __builtin_custom_fnif (int, float)
12651 float __builtin_custom_fnip (int, void *)
12652 float __builtin_custom_fnfi (float, int)
12653 float __builtin_custom_fnff (float, float)
12654 float __builtin_custom_fnfp (float, void *)
12655 float __builtin_custom_fnpi (void *, int)
12656 float __builtin_custom_fnpf (void *, float)
12657 float __builtin_custom_fnpp (void *, void *)
12658 void * __builtin_custom_pn (void)
12659 void * __builtin_custom_pni (int)
12660 void * __builtin_custom_pnf (float)
12661 void * __builtin_custom_pnp (void *)
12662 void * __builtin_custom_pnii (int, int)
12663 void * __builtin_custom_pnif (int, float)
12664 void * __builtin_custom_pnip (int, void *)
12665 void * __builtin_custom_pnfi (float, int)
12666 void * __builtin_custom_pnff (float, float)
12667 void * __builtin_custom_pnfp (float, void *)
12668 void * __builtin_custom_pnpi (void *, int)
12669 void * __builtin_custom_pnpf (void *, float)
12670 void * __builtin_custom_pnpp (void *, void *)
12673 @node ARC Built-in Functions
12674 @subsection ARC Built-in Functions
12676 The following built-in functions are provided for ARC targets. The
12677 built-ins generate the corresponding assembly instructions. In the
12678 examples given below, the generated code often requires an operand or
12679 result to be in a register. Where necessary further code will be
12680 generated to ensure this is true, but for brevity this is not
12681 described in each case.
12683 @emph{Note:} Using a built-in to generate an instruction not supported
12684 by a target may cause problems. At present the compiler is not
12685 guaranteed to detect such misuse, and as a result an internal compiler
12686 error may be generated.
12688 @deftypefn {Built-in Function} int __builtin_arc_aligned (void *@var{val}, int @var{alignval})
12689 Return 1 if @var{val} is known to have the byte alignment given
12690 by @var{alignval}, otherwise return 0.
12691 Note that this is different from
12693 __alignof__(*(char *)@var{val}) >= alignval
12695 because __alignof__ sees only the type of the dereference, whereas
12696 __builtin_arc_align uses alignment information from the pointer
12697 as well as from the pointed-to type.
12698 The information available will depend on optimization level.
12701 @deftypefn {Built-in Function} void __builtin_arc_brk (void)
12708 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_core_read (unsigned int @var{regno})
12709 The operand is the number of a register to be read. Generates:
12711 mov @var{dest}, r@var{regno}
12713 where the value in @var{dest} will be the result returned from the
12717 @deftypefn {Built-in Function} void __builtin_arc_core_write (unsigned int @var{regno}, unsigned int @var{val})
12718 The first operand is the number of a register to be written, the
12719 second operand is a compile time constant to write into that
12720 register. Generates:
12722 mov r@var{regno}, @var{val}
12726 @deftypefn {Built-in Function} int __builtin_arc_divaw (int @var{a}, int @var{b})
12727 Only available if either @option{-mcpu=ARC700} or @option{-meA} is set.
12730 divaw @var{dest}, @var{a}, @var{b}
12732 where the value in @var{dest} will be the result returned from the
12736 @deftypefn {Built-in Function} void __builtin_arc_flag (unsigned int @var{a})
12743 @deftypefn {Built-in Function} {unsigned int} __builtin_arc_lr (unsigned int @var{auxr})
12744 The operand, @var{auxv}, is the address of an auxiliary register and
12745 must be a compile time constant. Generates:
12747 lr @var{dest}, [@var{auxr}]
12749 Where the value in @var{dest} will be the result returned from the
12753 @deftypefn {Built-in Function} void __builtin_arc_mul64 (int @var{a}, int @var{b})
12754 Only available with @option{-mmul64}. Generates:
12756 mul64 @var{a}, @var{b}
12760 @deftypefn {Built-in Function} void __builtin_arc_mulu64 (unsigned int @var{a}, unsigned int @var{b})
12761 Only available with @option{-mmul64}. Generates:
12763 mulu64 @var{a}, @var{b}
12767 @deftypefn {Built-in Function} void __builtin_arc_nop (void)
12774 @deftypefn {Built-in Function} int __builtin_arc_norm (int @var{src})
12775 Only valid if the @samp{norm} instruction is available through the
12776 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
12779 norm @var{dest}, @var{src}
12781 Where the value in @var{dest} will be the result returned from the
12785 @deftypefn {Built-in Function} {short int} __builtin_arc_normw (short int @var{src})
12786 Only valid if the @samp{normw} instruction is available through the
12787 @option{-mnorm} option or by default with @option{-mcpu=ARC700}.
12790 normw @var{dest}, @var{src}
12792 Where the value in @var{dest} will be the result returned from the
12796 @deftypefn {Built-in Function} void __builtin_arc_rtie (void)
12803 @deftypefn {Built-in Function} void __builtin_arc_sleep (int @var{a}
12810 @deftypefn {Built-in Function} void __builtin_arc_sr (unsigned int @var{auxr}, unsigned int @var{val})
12811 The first argument, @var{auxv}, is the address of an auxiliary
12812 register, the second argument, @var{val}, is a compile time constant
12813 to be written to the register. Generates:
12815 sr @var{auxr}, [@var{val}]
12819 @deftypefn {Built-in Function} int __builtin_arc_swap (int @var{src})
12820 Only valid with @option{-mswap}. Generates:
12822 swap @var{dest}, @var{src}
12824 Where the value in @var{dest} will be the result returned from the
12828 @deftypefn {Built-in Function} void __builtin_arc_swi (void)
12835 @deftypefn {Built-in Function} void __builtin_arc_sync (void)
12836 Only available with @option{-mcpu=ARC700}. Generates:
12842 @deftypefn {Built-in Function} void __builtin_arc_trap_s (unsigned int @var{c})
12843 Only available with @option{-mcpu=ARC700}. Generates:
12849 @deftypefn {Built-in Function} void __builtin_arc_unimp_s (void)
12850 Only available with @option{-mcpu=ARC700}. Generates:
12856 The instructions generated by the following builtins are not
12857 considered as candidates for scheduling. They are not moved around by
12858 the compiler during scheduling, and thus can be expected to appear
12859 where they are put in the C code:
12861 __builtin_arc_brk()
12862 __builtin_arc_core_read()
12863 __builtin_arc_core_write()
12864 __builtin_arc_flag()
12866 __builtin_arc_sleep()
12868 __builtin_arc_swi()
12871 @node ARC SIMD Built-in Functions
12872 @subsection ARC SIMD Built-in Functions
12874 SIMD builtins provided by the compiler can be used to generate the
12875 vector instructions. This section describes the available builtins
12876 and their usage in programs. With the @option{-msimd} option, the
12877 compiler provides 128-bit vector types, which can be specified using
12878 the @code{vector_size} attribute. The header file @file{arc-simd.h}
12879 can be included to use the following predefined types:
12881 typedef int __v4si __attribute__((vector_size(16)));
12882 typedef short __v8hi __attribute__((vector_size(16)));
12885 These types can be used to define 128-bit variables. The built-in
12886 functions listed in the following section can be used on these
12887 variables to generate the vector operations.
12889 For all builtins, @code{__builtin_arc_@var{someinsn}}, the header file
12890 @file{arc-simd.h} also provides equivalent macros called
12891 @code{_@var{someinsn}} that can be used for programming ease and
12892 improved readability. The following macros for DMA control are also
12895 #define _setup_dma_in_channel_reg _vdiwr
12896 #define _setup_dma_out_channel_reg _vdowr
12899 The following is a complete list of all the SIMD built-ins provided
12900 for ARC, grouped by calling signature.
12902 The following take two @code{__v8hi} arguments and return a
12903 @code{__v8hi} result:
12905 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi)
12906 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi)
12907 __v8hi __builtin_arc_vand (__v8hi, __v8hi)
12908 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi)
12909 __v8hi __builtin_arc_vavb (__v8hi, __v8hi)
12910 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi)
12911 __v8hi __builtin_arc_vbic (__v8hi, __v8hi)
12912 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi)
12913 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi)
12914 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi)
12915 __v8hi __builtin_arc_veqw (__v8hi, __v8hi)
12916 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi)
12917 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi)
12918 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi)
12919 __v8hi __builtin_arc_vlew (__v8hi, __v8hi)
12920 __v8hi __builtin_arc_vltw (__v8hi, __v8hi)
12921 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi)
12922 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi)
12923 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi)
12924 __v8hi __builtin_arc_vminw (__v8hi, __v8hi)
12925 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi)
12926 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi)
12927 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi)
12928 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi)
12929 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi)
12930 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi)
12931 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi)
12932 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi)
12933 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi)
12934 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi)
12935 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi)
12936 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi)
12937 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi)
12938 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi)
12939 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi)
12940 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi)
12941 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi)
12942 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi)
12943 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi)
12944 __v8hi __builtin_arc_vnew (__v8hi, __v8hi)
12945 __v8hi __builtin_arc_vor (__v8hi, __v8hi)
12946 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi)
12947 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi)
12948 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi)
12949 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi)
12950 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi)
12951 __v8hi __builtin_arc_vxor (__v8hi, __v8hi)
12952 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi)
12955 The following take one @code{__v8hi} and one @code{int} argument and return a
12956 @code{__v8hi} result:
12959 __v8hi __builtin_arc_vbaddw (__v8hi, int)
12960 __v8hi __builtin_arc_vbmaxw (__v8hi, int)
12961 __v8hi __builtin_arc_vbminw (__v8hi, int)
12962 __v8hi __builtin_arc_vbmulaw (__v8hi, int)
12963 __v8hi __builtin_arc_vbmulfw (__v8hi, int)
12964 __v8hi __builtin_arc_vbmulw (__v8hi, int)
12965 __v8hi __builtin_arc_vbrsubw (__v8hi, int)
12966 __v8hi __builtin_arc_vbsubw (__v8hi, int)
12969 The following take one @code{__v8hi} argument and one @code{int} argument which
12970 must be a 3-bit compile time constant indicating a register number
12971 I0-I7. They return a @code{__v8hi} result.
12973 __v8hi __builtin_arc_vasrw (__v8hi, const int)
12974 __v8hi __builtin_arc_vsr8 (__v8hi, const int)
12975 __v8hi __builtin_arc_vsr8aw (__v8hi, const int)
12978 The following take one @code{__v8hi} argument and one @code{int}
12979 argument which must be a 6-bit compile time constant. They return a
12980 @code{__v8hi} result.
12982 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int)
12983 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int)
12984 __v8hi __builtin_arc_vasrrwi (__v8hi, const int)
12985 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int)
12986 __v8hi __builtin_arc_vasrwi (__v8hi, const int)
12987 __v8hi __builtin_arc_vsr8awi (__v8hi, const int)
12988 __v8hi __builtin_arc_vsr8i (__v8hi, const int)
12991 The following take one @code{__v8hi} argument and one @code{int} argument which
12992 must be a 8-bit compile time constant. They return a @code{__v8hi}
12995 __v8hi __builtin_arc_vd6tapf (__v8hi, const int)
12996 __v8hi __builtin_arc_vmvaw (__v8hi, const int)
12997 __v8hi __builtin_arc_vmvw (__v8hi, const int)
12998 __v8hi __builtin_arc_vmvzw (__v8hi, const int)
13001 The following take two @code{int} arguments, the second of which which
13002 must be a 8-bit compile time constant. They return a @code{__v8hi}
13005 __v8hi __builtin_arc_vmovaw (int, const int)
13006 __v8hi __builtin_arc_vmovw (int, const int)
13007 __v8hi __builtin_arc_vmovzw (int, const int)
13010 The following take a single @code{__v8hi} argument and return a
13011 @code{__v8hi} result:
13013 __v8hi __builtin_arc_vabsaw (__v8hi)
13014 __v8hi __builtin_arc_vabsw (__v8hi)
13015 __v8hi __builtin_arc_vaddsuw (__v8hi)
13016 __v8hi __builtin_arc_vexch1 (__v8hi)
13017 __v8hi __builtin_arc_vexch2 (__v8hi)
13018 __v8hi __builtin_arc_vexch4 (__v8hi)
13019 __v8hi __builtin_arc_vsignw (__v8hi)
13020 __v8hi __builtin_arc_vupbaw (__v8hi)
13021 __v8hi __builtin_arc_vupbw (__v8hi)
13022 __v8hi __builtin_arc_vupsbaw (__v8hi)
13023 __v8hi __builtin_arc_vupsbw (__v8hi)
13026 The following take two @code{int} arguments and return no result:
13028 void __builtin_arc_vdirun (int, int)
13029 void __builtin_arc_vdorun (int, int)
13032 The following take two @code{int} arguments and return no result. The
13033 first argument must a 3-bit compile time constant indicating one of
13034 the DR0-DR7 DMA setup channels:
13036 void __builtin_arc_vdiwr (const int, int)
13037 void __builtin_arc_vdowr (const int, int)
13040 The following take an @code{int} argument and return no result:
13042 void __builtin_arc_vendrec (int)
13043 void __builtin_arc_vrec (int)
13044 void __builtin_arc_vrecrun (int)
13045 void __builtin_arc_vrun (int)
13048 The following take a @code{__v8hi} argument and two @code{int}
13049 arguments and return a @code{__v8hi} result. The second argument must
13050 be a 3-bit compile time constants, indicating one the registers I0-I7,
13051 and the third argument must be an 8-bit compile time constant.
13053 @emph{Note:} Although the equivalent hardware instructions do not take
13054 an SIMD register as an operand, these builtins overwrite the relevant
13055 bits of the @code{__v8hi} register provided as the first argument with
13056 the value loaded from the @code{[Ib, u8]} location in the SDM.
13059 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int)
13060 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int)
13061 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int)
13062 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int)
13065 The following take two @code{int} arguments and return a @code{__v8hi}
13066 result. The first argument must be a 3-bit compile time constants,
13067 indicating one the registers I0-I7, and the second argument must be an
13068 8-bit compile time constant.
13071 __v8hi __builtin_arc_vld128 (const int, const int)
13072 __v8hi __builtin_arc_vld64w (const int, const int)
13075 The following take a @code{__v8hi} argument and two @code{int}
13076 arguments and return no result. The second argument must be a 3-bit
13077 compile time constants, indicating one the registers I0-I7, and the
13078 third argument must be an 8-bit compile time constant.
13081 void __builtin_arc_vst128 (__v8hi, const int, const int)
13082 void __builtin_arc_vst64 (__v8hi, const int, const int)
13085 The following take a @code{__v8hi} argument and three @code{int}
13086 arguments and return no result. The second argument must be a 3-bit
13087 compile-time constant, identifying the 16-bit sub-register to be
13088 stored, the third argument must be a 3-bit compile time constants,
13089 indicating one the registers I0-I7, and the fourth argument must be an
13090 8-bit compile time constant.
13093 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int)
13094 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int)
13097 @node ARM iWMMXt Built-in Functions
13098 @subsection ARM iWMMXt Built-in Functions
13100 These built-in functions are available for the ARM family of
13101 processors when the @option{-mcpu=iwmmxt} switch is used:
13104 typedef int v2si __attribute__ ((vector_size (8)));
13105 typedef short v4hi __attribute__ ((vector_size (8)));
13106 typedef char v8qi __attribute__ ((vector_size (8)));
13108 int __builtin_arm_getwcgr0 (void)
13109 void __builtin_arm_setwcgr0 (int)
13110 int __builtin_arm_getwcgr1 (void)
13111 void __builtin_arm_setwcgr1 (int)
13112 int __builtin_arm_getwcgr2 (void)
13113 void __builtin_arm_setwcgr2 (int)
13114 int __builtin_arm_getwcgr3 (void)
13115 void __builtin_arm_setwcgr3 (int)
13116 int __builtin_arm_textrmsb (v8qi, int)
13117 int __builtin_arm_textrmsh (v4hi, int)
13118 int __builtin_arm_textrmsw (v2si, int)
13119 int __builtin_arm_textrmub (v8qi, int)
13120 int __builtin_arm_textrmuh (v4hi, int)
13121 int __builtin_arm_textrmuw (v2si, int)
13122 v8qi __builtin_arm_tinsrb (v8qi, int, int)
13123 v4hi __builtin_arm_tinsrh (v4hi, int, int)
13124 v2si __builtin_arm_tinsrw (v2si, int, int)
13125 long long __builtin_arm_tmia (long long, int, int)
13126 long long __builtin_arm_tmiabb (long long, int, int)
13127 long long __builtin_arm_tmiabt (long long, int, int)
13128 long long __builtin_arm_tmiaph (long long, int, int)
13129 long long __builtin_arm_tmiatb (long long, int, int)
13130 long long __builtin_arm_tmiatt (long long, int, int)
13131 int __builtin_arm_tmovmskb (v8qi)
13132 int __builtin_arm_tmovmskh (v4hi)
13133 int __builtin_arm_tmovmskw (v2si)
13134 long long __builtin_arm_waccb (v8qi)
13135 long long __builtin_arm_wacch (v4hi)
13136 long long __builtin_arm_waccw (v2si)
13137 v8qi __builtin_arm_waddb (v8qi, v8qi)
13138 v8qi __builtin_arm_waddbss (v8qi, v8qi)
13139 v8qi __builtin_arm_waddbus (v8qi, v8qi)
13140 v4hi __builtin_arm_waddh (v4hi, v4hi)
13141 v4hi __builtin_arm_waddhss (v4hi, v4hi)
13142 v4hi __builtin_arm_waddhus (v4hi, v4hi)
13143 v2si __builtin_arm_waddw (v2si, v2si)
13144 v2si __builtin_arm_waddwss (v2si, v2si)
13145 v2si __builtin_arm_waddwus (v2si, v2si)
13146 v8qi __builtin_arm_walign (v8qi, v8qi, int)
13147 long long __builtin_arm_wand(long long, long long)
13148 long long __builtin_arm_wandn (long long, long long)
13149 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
13150 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
13151 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
13152 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
13153 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
13154 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
13155 v2si __builtin_arm_wcmpeqw (v2si, v2si)
13156 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
13157 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
13158 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
13159 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
13160 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
13161 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
13162 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
13163 long long __builtin_arm_wmacsz (v4hi, v4hi)
13164 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
13165 long long __builtin_arm_wmacuz (v4hi, v4hi)
13166 v4hi __builtin_arm_wmadds (v4hi, v4hi)
13167 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
13168 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
13169 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
13170 v2si __builtin_arm_wmaxsw (v2si, v2si)
13171 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
13172 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
13173 v2si __builtin_arm_wmaxuw (v2si, v2si)
13174 v8qi __builtin_arm_wminsb (v8qi, v8qi)
13175 v4hi __builtin_arm_wminsh (v4hi, v4hi)
13176 v2si __builtin_arm_wminsw (v2si, v2si)
13177 v8qi __builtin_arm_wminub (v8qi, v8qi)
13178 v4hi __builtin_arm_wminuh (v4hi, v4hi)
13179 v2si __builtin_arm_wminuw (v2si, v2si)
13180 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
13181 v4hi __builtin_arm_wmulul (v4hi, v4hi)
13182 v4hi __builtin_arm_wmulum (v4hi, v4hi)
13183 long long __builtin_arm_wor (long long, long long)
13184 v2si __builtin_arm_wpackdss (long long, long long)
13185 v2si __builtin_arm_wpackdus (long long, long long)
13186 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
13187 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
13188 v4hi __builtin_arm_wpackwss (v2si, v2si)
13189 v4hi __builtin_arm_wpackwus (v2si, v2si)
13190 long long __builtin_arm_wrord (long long, long long)
13191 long long __builtin_arm_wrordi (long long, int)
13192 v4hi __builtin_arm_wrorh (v4hi, long long)
13193 v4hi __builtin_arm_wrorhi (v4hi, int)
13194 v2si __builtin_arm_wrorw (v2si, long long)
13195 v2si __builtin_arm_wrorwi (v2si, int)
13196 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi)
13197 v2si __builtin_arm_wsadbz (v8qi, v8qi)
13198 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi)
13199 v2si __builtin_arm_wsadhz (v4hi, v4hi)
13200 v4hi __builtin_arm_wshufh (v4hi, int)
13201 long long __builtin_arm_wslld (long long, long long)
13202 long long __builtin_arm_wslldi (long long, int)
13203 v4hi __builtin_arm_wsllh (v4hi, long long)
13204 v4hi __builtin_arm_wsllhi (v4hi, int)
13205 v2si __builtin_arm_wsllw (v2si, long long)
13206 v2si __builtin_arm_wsllwi (v2si, int)
13207 long long __builtin_arm_wsrad (long long, long long)
13208 long long __builtin_arm_wsradi (long long, int)
13209 v4hi __builtin_arm_wsrah (v4hi, long long)
13210 v4hi __builtin_arm_wsrahi (v4hi, int)
13211 v2si __builtin_arm_wsraw (v2si, long long)
13212 v2si __builtin_arm_wsrawi (v2si, int)
13213 long long __builtin_arm_wsrld (long long, long long)
13214 long long __builtin_arm_wsrldi (long long, int)
13215 v4hi __builtin_arm_wsrlh (v4hi, long long)
13216 v4hi __builtin_arm_wsrlhi (v4hi, int)
13217 v2si __builtin_arm_wsrlw (v2si, long long)
13218 v2si __builtin_arm_wsrlwi (v2si, int)
13219 v8qi __builtin_arm_wsubb (v8qi, v8qi)
13220 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
13221 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
13222 v4hi __builtin_arm_wsubh (v4hi, v4hi)
13223 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
13224 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
13225 v2si __builtin_arm_wsubw (v2si, v2si)
13226 v2si __builtin_arm_wsubwss (v2si, v2si)
13227 v2si __builtin_arm_wsubwus (v2si, v2si)
13228 v4hi __builtin_arm_wunpckehsb (v8qi)
13229 v2si __builtin_arm_wunpckehsh (v4hi)
13230 long long __builtin_arm_wunpckehsw (v2si)
13231 v4hi __builtin_arm_wunpckehub (v8qi)
13232 v2si __builtin_arm_wunpckehuh (v4hi)
13233 long long __builtin_arm_wunpckehuw (v2si)
13234 v4hi __builtin_arm_wunpckelsb (v8qi)
13235 v2si __builtin_arm_wunpckelsh (v4hi)
13236 long long __builtin_arm_wunpckelsw (v2si)
13237 v4hi __builtin_arm_wunpckelub (v8qi)
13238 v2si __builtin_arm_wunpckeluh (v4hi)
13239 long long __builtin_arm_wunpckeluw (v2si)
13240 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
13241 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
13242 v2si __builtin_arm_wunpckihw (v2si, v2si)
13243 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
13244 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
13245 v2si __builtin_arm_wunpckilw (v2si, v2si)
13246 long long __builtin_arm_wxor (long long, long long)
13247 long long __builtin_arm_wzero ()
13251 @node ARM C Language Extensions (ACLE)
13252 @subsection ARM C Language Extensions (ACLE)
13254 GCC implements extensions for C as described in the ARM C Language
13255 Extensions (ACLE) specification, which can be found at
13256 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0053c/IHI0053C_acle_2_0.pdf}.
13258 As a part of ACLE, GCC implements extensions for Advanced SIMD as described in
13259 the ARM C Language Extensions Specification. The complete list of Advanced SIMD
13260 intrinsics can be found at
13261 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ihi0073a/IHI0073A_arm_neon_intrinsics_ref.pdf}.
13262 The built-in intrinsics for the Advanced SIMD extension are available when
13265 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. Both
13266 back ends support CRC32 intrinsics and the ARM back end supports the
13267 Coprocessor intrinsics, all from @file{arm_acle.h}. The ARM back end's 16-bit
13268 floating-point Advanced SIMD intrinsics currently comply to ACLE v1.1.
13269 AArch64's back end does not have support for 16-bit floating point Advanced SIMD
13272 See @ref{ARM Options} and @ref{AArch64 Options} for more information on the
13273 availability of extensions.
13275 @node ARM Floating Point Status and Control Intrinsics
13276 @subsection ARM Floating Point Status and Control Intrinsics
13278 These built-in functions are available for the ARM family of
13279 processors with floating-point unit.
13282 unsigned int __builtin_arm_get_fpscr ()
13283 void __builtin_arm_set_fpscr (unsigned int)
13286 @node ARM ARMv8-M Security Extensions
13287 @subsection ARM ARMv8-M Security Extensions
13289 GCC implements the ARMv8-M Security Extensions as described in the ARMv8-M
13290 Security Extensions: Requirements on Development Tools Engineering
13291 Specification, which can be found at
13292 @uref{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
13294 As part of the Security Extensions GCC implements two new function attributes:
13295 @code{cmse_nonsecure_entry} and @code{cmse_nonsecure_call}.
13297 As part of the Security Extensions GCC implements the intrinsics below. FPTR
13298 is used here to mean any function pointer type.
13301 cmse_address_info_t cmse_TT (void *)
13302 cmse_address_info_t cmse_TT_fptr (FPTR)
13303 cmse_address_info_t cmse_TTT (void *)
13304 cmse_address_info_t cmse_TTT_fptr (FPTR)
13305 cmse_address_info_t cmse_TTA (void *)
13306 cmse_address_info_t cmse_TTA_fptr (FPTR)
13307 cmse_address_info_t cmse_TTAT (void *)
13308 cmse_address_info_t cmse_TTAT_fptr (FPTR)
13309 void * cmse_check_address_range (void *, size_t, int)
13310 typeof(p) cmse_nsfptr_create (FPTR p)
13311 intptr_t cmse_is_nsfptr (FPTR)
13312 int cmse_nonsecure_caller (void)
13315 @node AVR Built-in Functions
13316 @subsection AVR Built-in Functions
13318 For each built-in function for AVR, there is an equally named,
13319 uppercase built-in macro defined. That way users can easily query if
13320 or if not a specific built-in is implemented or not. For example, if
13321 @code{__builtin_avr_nop} is available the macro
13322 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
13326 @item void __builtin_avr_nop (void)
13327 @itemx void __builtin_avr_sei (void)
13328 @itemx void __builtin_avr_cli (void)
13329 @itemx void __builtin_avr_sleep (void)
13330 @itemx void __builtin_avr_wdr (void)
13331 @itemx unsigned char __builtin_avr_swap (unsigned char)
13332 @itemx unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
13333 @itemx int __builtin_avr_fmuls (char, char)
13334 @itemx int __builtin_avr_fmulsu (char, unsigned char)
13335 These built-in functions map to the respective machine
13336 instruction, i.e.@: @code{nop}, @code{sei}, @code{cli}, @code{sleep},
13337 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
13338 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
13339 as library call if no hardware multiplier is available.
13341 @item void __builtin_avr_delay_cycles (unsigned long ticks)
13342 Delay execution for @var{ticks} cycles. Note that this
13343 built-in does not take into account the effect of interrupts that
13344 might increase delay time. @var{ticks} must be a compile-time
13345 integer constant; delays with a variable number of cycles are not supported.
13347 @item char __builtin_avr_flash_segment (const __memx void*)
13348 This built-in takes a byte address to the 24-bit
13349 @ref{AVR Named Address Spaces,address space} @code{__memx} and returns
13350 the number of the flash segment (the 64 KiB chunk) where the address
13351 points to. Counting starts at @code{0}.
13352 If the address does not point to flash memory, return @code{-1}.
13354 @item uint8_t __builtin_avr_insert_bits (uint32_t map, uint8_t bits, uint8_t val)
13355 Insert bits from @var{bits} into @var{val} and return the resulting
13356 value. The nibbles of @var{map} determine how the insertion is
13357 performed: Let @var{X} be the @var{n}-th nibble of @var{map}
13359 @item If @var{X} is @code{0xf},
13360 then the @var{n}-th bit of @var{val} is returned unaltered.
13362 @item If X is in the range 0@dots{}7,
13363 then the @var{n}-th result bit is set to the @var{X}-th bit of @var{bits}
13365 @item If X is in the range 8@dots{}@code{0xe},
13366 then the @var{n}-th result bit is undefined.
13370 One typical use case for this built-in is adjusting input and
13371 output values to non-contiguous port layouts. Some examples:
13374 // same as val, bits is unused
13375 __builtin_avr_insert_bits (0xffffffff, bits, val)
13379 // same as bits, val is unused
13380 __builtin_avr_insert_bits (0x76543210, bits, val)
13384 // same as rotating bits by 4
13385 __builtin_avr_insert_bits (0x32107654, bits, 0)
13389 // high nibble of result is the high nibble of val
13390 // low nibble of result is the low nibble of bits
13391 __builtin_avr_insert_bits (0xffff3210, bits, val)
13395 // reverse the bit order of bits
13396 __builtin_avr_insert_bits (0x01234567, bits, 0)
13399 @item void __builtin_avr_nops (unsigned count)
13400 Insert @var{count} @code{NOP} instructions.
13401 The number of instructions must be a compile-time integer constant.
13406 There are many more AVR-specific built-in functions that are used to
13407 implement the ISO/IEC TR 18037 ``Embedded C'' fixed-point functions of
13408 section 7.18a.6. You don't need to use these built-ins directly.
13409 Instead, use the declarations as supplied by the @code{stdfix.h} header
13413 #include <stdfix.h>
13415 // Re-interpret the bit representation of unsigned 16-bit
13416 // integer @var{uval} as Q-format 0.16 value.
13417 unsigned fract get_bits (uint_ur_t uval)
13419 return urbits (uval);
13423 @node Blackfin Built-in Functions
13424 @subsection Blackfin Built-in Functions
13426 Currently, there are two Blackfin-specific built-in functions. These are
13427 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
13428 using inline assembly; by using these built-in functions the compiler can
13429 automatically add workarounds for hardware errata involving these
13430 instructions. These functions are named as follows:
13433 void __builtin_bfin_csync (void)
13434 void __builtin_bfin_ssync (void)
13437 @node FR-V Built-in Functions
13438 @subsection FR-V Built-in Functions
13440 GCC provides many FR-V-specific built-in functions. In general,
13441 these functions are intended to be compatible with those described
13442 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
13443 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
13444 @code{__MBTOHE}, the GCC forms of which pass 128-bit values by
13445 pointer rather than by value.
13447 Most of the functions are named after specific FR-V instructions.
13448 Such functions are said to be ``directly mapped'' and are summarized
13449 here in tabular form.
13453 * Directly-mapped Integer Functions::
13454 * Directly-mapped Media Functions::
13455 * Raw read/write Functions::
13456 * Other Built-in Functions::
13459 @node Argument Types
13460 @subsubsection Argument Types
13462 The arguments to the built-in functions can be divided into three groups:
13463 register numbers, compile-time constants and run-time values. In order
13464 to make this classification clear at a glance, the arguments and return
13465 values are given the following pseudo types:
13467 @multitable @columnfractions .20 .30 .15 .35
13468 @item Pseudo type @tab Real C type @tab Constant? @tab Description
13469 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
13470 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
13471 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
13472 @item @code{uw2} @tab @code{unsigned long long} @tab No
13473 @tab an unsigned doubleword
13474 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
13475 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
13476 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
13477 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
13480 These pseudo types are not defined by GCC, they are simply a notational
13481 convenience used in this manual.
13483 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
13484 and @code{sw2} are evaluated at run time. They correspond to
13485 register operands in the underlying FR-V instructions.
13487 @code{const} arguments represent immediate operands in the underlying
13488 FR-V instructions. They must be compile-time constants.
13490 @code{acc} arguments are evaluated at compile time and specify the number
13491 of an accumulator register. For example, an @code{acc} argument of 2
13492 selects the ACC2 register.
13494 @code{iacc} arguments are similar to @code{acc} arguments but specify the
13495 number of an IACC register. See @pxref{Other Built-in Functions}
13498 @node Directly-mapped Integer Functions
13499 @subsubsection Directly-Mapped Integer Functions
13501 The functions listed below map directly to FR-V I-type instructions.
13503 @multitable @columnfractions .45 .32 .23
13504 @item Function prototype @tab Example usage @tab Assembly output
13505 @item @code{sw1 __ADDSS (sw1, sw1)}
13506 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
13507 @tab @code{ADDSS @var{a},@var{b},@var{c}}
13508 @item @code{sw1 __SCAN (sw1, sw1)}
13509 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
13510 @tab @code{SCAN @var{a},@var{b},@var{c}}
13511 @item @code{sw1 __SCUTSS (sw1)}
13512 @tab @code{@var{b} = __SCUTSS (@var{a})}
13513 @tab @code{SCUTSS @var{a},@var{b}}
13514 @item @code{sw1 __SLASS (sw1, sw1)}
13515 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
13516 @tab @code{SLASS @var{a},@var{b},@var{c}}
13517 @item @code{void __SMASS (sw1, sw1)}
13518 @tab @code{__SMASS (@var{a}, @var{b})}
13519 @tab @code{SMASS @var{a},@var{b}}
13520 @item @code{void __SMSSS (sw1, sw1)}
13521 @tab @code{__SMSSS (@var{a}, @var{b})}
13522 @tab @code{SMSSS @var{a},@var{b}}
13523 @item @code{void __SMU (sw1, sw1)}
13524 @tab @code{__SMU (@var{a}, @var{b})}
13525 @tab @code{SMU @var{a},@var{b}}
13526 @item @code{sw2 __SMUL (sw1, sw1)}
13527 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
13528 @tab @code{SMUL @var{a},@var{b},@var{c}}
13529 @item @code{sw1 __SUBSS (sw1, sw1)}
13530 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
13531 @tab @code{SUBSS @var{a},@var{b},@var{c}}
13532 @item @code{uw2 __UMUL (uw1, uw1)}
13533 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
13534 @tab @code{UMUL @var{a},@var{b},@var{c}}
13537 @node Directly-mapped Media Functions
13538 @subsubsection Directly-Mapped Media Functions
13540 The functions listed below map directly to FR-V M-type instructions.
13542 @multitable @columnfractions .45 .32 .23
13543 @item Function prototype @tab Example usage @tab Assembly output
13544 @item @code{uw1 __MABSHS (sw1)}
13545 @tab @code{@var{b} = __MABSHS (@var{a})}
13546 @tab @code{MABSHS @var{a},@var{b}}
13547 @item @code{void __MADDACCS (acc, acc)}
13548 @tab @code{__MADDACCS (@var{b}, @var{a})}
13549 @tab @code{MADDACCS @var{a},@var{b}}
13550 @item @code{sw1 __MADDHSS (sw1, sw1)}
13551 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
13552 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
13553 @item @code{uw1 __MADDHUS (uw1, uw1)}
13554 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
13555 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
13556 @item @code{uw1 __MAND (uw1, uw1)}
13557 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
13558 @tab @code{MAND @var{a},@var{b},@var{c}}
13559 @item @code{void __MASACCS (acc, acc)}
13560 @tab @code{__MASACCS (@var{b}, @var{a})}
13561 @tab @code{MASACCS @var{a},@var{b}}
13562 @item @code{uw1 __MAVEH (uw1, uw1)}
13563 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
13564 @tab @code{MAVEH @var{a},@var{b},@var{c}}
13565 @item @code{uw2 __MBTOH (uw1)}
13566 @tab @code{@var{b} = __MBTOH (@var{a})}
13567 @tab @code{MBTOH @var{a},@var{b}}
13568 @item @code{void __MBTOHE (uw1 *, uw1)}
13569 @tab @code{__MBTOHE (&@var{b}, @var{a})}
13570 @tab @code{MBTOHE @var{a},@var{b}}
13571 @item @code{void __MCLRACC (acc)}
13572 @tab @code{__MCLRACC (@var{a})}
13573 @tab @code{MCLRACC @var{a}}
13574 @item @code{void __MCLRACCA (void)}
13575 @tab @code{__MCLRACCA ()}
13576 @tab @code{MCLRACCA}
13577 @item @code{uw1 __Mcop1 (uw1, uw1)}
13578 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
13579 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
13580 @item @code{uw1 __Mcop2 (uw1, uw1)}
13581 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
13582 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
13583 @item @code{uw1 __MCPLHI (uw2, const)}
13584 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
13585 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
13586 @item @code{uw1 __MCPLI (uw2, const)}
13587 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
13588 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
13589 @item @code{void __MCPXIS (acc, sw1, sw1)}
13590 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
13591 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
13592 @item @code{void __MCPXIU (acc, uw1, uw1)}
13593 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
13594 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
13595 @item @code{void __MCPXRS (acc, sw1, sw1)}
13596 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
13597 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
13598 @item @code{void __MCPXRU (acc, uw1, uw1)}
13599 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
13600 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
13601 @item @code{uw1 __MCUT (acc, uw1)}
13602 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
13603 @tab @code{MCUT @var{a},@var{b},@var{c}}
13604 @item @code{uw1 __MCUTSS (acc, sw1)}
13605 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
13606 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
13607 @item @code{void __MDADDACCS (acc, acc)}
13608 @tab @code{__MDADDACCS (@var{b}, @var{a})}
13609 @tab @code{MDADDACCS @var{a},@var{b}}
13610 @item @code{void __MDASACCS (acc, acc)}
13611 @tab @code{__MDASACCS (@var{b}, @var{a})}
13612 @tab @code{MDASACCS @var{a},@var{b}}
13613 @item @code{uw2 __MDCUTSSI (acc, const)}
13614 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
13615 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
13616 @item @code{uw2 __MDPACKH (uw2, uw2)}
13617 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
13618 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
13619 @item @code{uw2 __MDROTLI (uw2, const)}
13620 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
13621 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
13622 @item @code{void __MDSUBACCS (acc, acc)}
13623 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
13624 @tab @code{MDSUBACCS @var{a},@var{b}}
13625 @item @code{void __MDUNPACKH (uw1 *, uw2)}
13626 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
13627 @tab @code{MDUNPACKH @var{a},@var{b}}
13628 @item @code{uw2 __MEXPDHD (uw1, const)}
13629 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
13630 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
13631 @item @code{uw1 __MEXPDHW (uw1, const)}
13632 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
13633 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
13634 @item @code{uw1 __MHDSETH (uw1, const)}
13635 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
13636 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
13637 @item @code{sw1 __MHDSETS (const)}
13638 @tab @code{@var{b} = __MHDSETS (@var{a})}
13639 @tab @code{MHDSETS #@var{a},@var{b}}
13640 @item @code{uw1 __MHSETHIH (uw1, const)}
13641 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
13642 @tab @code{MHSETHIH #@var{a},@var{b}}
13643 @item @code{sw1 __MHSETHIS (sw1, const)}
13644 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
13645 @tab @code{MHSETHIS #@var{a},@var{b}}
13646 @item @code{uw1 __MHSETLOH (uw1, const)}
13647 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
13648 @tab @code{MHSETLOH #@var{a},@var{b}}
13649 @item @code{sw1 __MHSETLOS (sw1, const)}
13650 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
13651 @tab @code{MHSETLOS #@var{a},@var{b}}
13652 @item @code{uw1 __MHTOB (uw2)}
13653 @tab @code{@var{b} = __MHTOB (@var{a})}
13654 @tab @code{MHTOB @var{a},@var{b}}
13655 @item @code{void __MMACHS (acc, sw1, sw1)}
13656 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
13657 @tab @code{MMACHS @var{a},@var{b},@var{c}}
13658 @item @code{void __MMACHU (acc, uw1, uw1)}
13659 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
13660 @tab @code{MMACHU @var{a},@var{b},@var{c}}
13661 @item @code{void __MMRDHS (acc, sw1, sw1)}
13662 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
13663 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
13664 @item @code{void __MMRDHU (acc, uw1, uw1)}
13665 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
13666 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
13667 @item @code{void __MMULHS (acc, sw1, sw1)}
13668 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
13669 @tab @code{MMULHS @var{a},@var{b},@var{c}}
13670 @item @code{void __MMULHU (acc, uw1, uw1)}
13671 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
13672 @tab @code{MMULHU @var{a},@var{b},@var{c}}
13673 @item @code{void __MMULXHS (acc, sw1, sw1)}
13674 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
13675 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
13676 @item @code{void __MMULXHU (acc, uw1, uw1)}
13677 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
13678 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
13679 @item @code{uw1 __MNOT (uw1)}
13680 @tab @code{@var{b} = __MNOT (@var{a})}
13681 @tab @code{MNOT @var{a},@var{b}}
13682 @item @code{uw1 __MOR (uw1, uw1)}
13683 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
13684 @tab @code{MOR @var{a},@var{b},@var{c}}
13685 @item @code{uw1 __MPACKH (uh, uh)}
13686 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
13687 @tab @code{MPACKH @var{a},@var{b},@var{c}}
13688 @item @code{sw2 __MQADDHSS (sw2, sw2)}
13689 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
13690 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
13691 @item @code{uw2 __MQADDHUS (uw2, uw2)}
13692 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
13693 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
13694 @item @code{void __MQCPXIS (acc, sw2, sw2)}
13695 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
13696 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
13697 @item @code{void __MQCPXIU (acc, uw2, uw2)}
13698 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
13699 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
13700 @item @code{void __MQCPXRS (acc, sw2, sw2)}
13701 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
13702 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
13703 @item @code{void __MQCPXRU (acc, uw2, uw2)}
13704 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
13705 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
13706 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
13707 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
13708 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
13709 @item @code{sw2 __MQLMTHS (sw2, sw2)}
13710 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
13711 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
13712 @item @code{void __MQMACHS (acc, sw2, sw2)}
13713 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
13714 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
13715 @item @code{void __MQMACHU (acc, uw2, uw2)}
13716 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
13717 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
13718 @item @code{void __MQMACXHS (acc, sw2, sw2)}
13719 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
13720 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
13721 @item @code{void __MQMULHS (acc, sw2, sw2)}
13722 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
13723 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
13724 @item @code{void __MQMULHU (acc, uw2, uw2)}
13725 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
13726 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
13727 @item @code{void __MQMULXHS (acc, sw2, sw2)}
13728 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
13729 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
13730 @item @code{void __MQMULXHU (acc, uw2, uw2)}
13731 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
13732 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
13733 @item @code{sw2 __MQSATHS (sw2, sw2)}
13734 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
13735 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
13736 @item @code{uw2 __MQSLLHI (uw2, int)}
13737 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
13738 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
13739 @item @code{sw2 __MQSRAHI (sw2, int)}
13740 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
13741 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
13742 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
13743 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
13744 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
13745 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
13746 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
13747 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
13748 @item @code{void __MQXMACHS (acc, sw2, sw2)}
13749 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
13750 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
13751 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
13752 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
13753 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
13754 @item @code{uw1 __MRDACC (acc)}
13755 @tab @code{@var{b} = __MRDACC (@var{a})}
13756 @tab @code{MRDACC @var{a},@var{b}}
13757 @item @code{uw1 __MRDACCG (acc)}
13758 @tab @code{@var{b} = __MRDACCG (@var{a})}
13759 @tab @code{MRDACCG @var{a},@var{b}}
13760 @item @code{uw1 __MROTLI (uw1, const)}
13761 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
13762 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
13763 @item @code{uw1 __MROTRI (uw1, const)}
13764 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
13765 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
13766 @item @code{sw1 __MSATHS (sw1, sw1)}
13767 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
13768 @tab @code{MSATHS @var{a},@var{b},@var{c}}
13769 @item @code{uw1 __MSATHU (uw1, uw1)}
13770 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
13771 @tab @code{MSATHU @var{a},@var{b},@var{c}}
13772 @item @code{uw1 __MSLLHI (uw1, const)}
13773 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
13774 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
13775 @item @code{sw1 __MSRAHI (sw1, const)}
13776 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
13777 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
13778 @item @code{uw1 __MSRLHI (uw1, const)}
13779 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
13780 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
13781 @item @code{void __MSUBACCS (acc, acc)}
13782 @tab @code{__MSUBACCS (@var{b}, @var{a})}
13783 @tab @code{MSUBACCS @var{a},@var{b}}
13784 @item @code{sw1 __MSUBHSS (sw1, sw1)}
13785 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
13786 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
13787 @item @code{uw1 __MSUBHUS (uw1, uw1)}
13788 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
13789 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
13790 @item @code{void __MTRAP (void)}
13791 @tab @code{__MTRAP ()}
13793 @item @code{uw2 __MUNPACKH (uw1)}
13794 @tab @code{@var{b} = __MUNPACKH (@var{a})}
13795 @tab @code{MUNPACKH @var{a},@var{b}}
13796 @item @code{uw1 __MWCUT (uw2, uw1)}
13797 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
13798 @tab @code{MWCUT @var{a},@var{b},@var{c}}
13799 @item @code{void __MWTACC (acc, uw1)}
13800 @tab @code{__MWTACC (@var{b}, @var{a})}
13801 @tab @code{MWTACC @var{a},@var{b}}
13802 @item @code{void __MWTACCG (acc, uw1)}
13803 @tab @code{__MWTACCG (@var{b}, @var{a})}
13804 @tab @code{MWTACCG @var{a},@var{b}}
13805 @item @code{uw1 __MXOR (uw1, uw1)}
13806 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
13807 @tab @code{MXOR @var{a},@var{b},@var{c}}
13810 @node Raw read/write Functions
13811 @subsubsection Raw Read/Write Functions
13813 This sections describes built-in functions related to read and write
13814 instructions to access memory. These functions generate
13815 @code{membar} instructions to flush the I/O load and stores where
13816 appropriate, as described in Fujitsu's manual described above.
13820 @item unsigned char __builtin_read8 (void *@var{data})
13821 @item unsigned short __builtin_read16 (void *@var{data})
13822 @item unsigned long __builtin_read32 (void *@var{data})
13823 @item unsigned long long __builtin_read64 (void *@var{data})
13825 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
13826 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
13827 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
13828 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
13831 @node Other Built-in Functions
13832 @subsubsection Other Built-in Functions
13834 This section describes built-in functions that are not named after
13835 a specific FR-V instruction.
13838 @item sw2 __IACCreadll (iacc @var{reg})
13839 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
13840 for future expansion and must be 0.
13842 @item sw1 __IACCreadl (iacc @var{reg})
13843 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
13844 Other values of @var{reg} are rejected as invalid.
13846 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
13847 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
13848 is reserved for future expansion and must be 0.
13850 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
13851 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
13852 is 1. Other values of @var{reg} are rejected as invalid.
13854 @item void __data_prefetch0 (const void *@var{x})
13855 Use the @code{dcpl} instruction to load the contents of address @var{x}
13856 into the data cache.
13858 @item void __data_prefetch (const void *@var{x})
13859 Use the @code{nldub} instruction to load the contents of address @var{x}
13860 into the data cache. The instruction is issued in slot I1@.
13863 @node MIPS DSP Built-in Functions
13864 @subsection MIPS DSP Built-in Functions
13866 The MIPS DSP Application-Specific Extension (ASE) includes new
13867 instructions that are designed to improve the performance of DSP and
13868 media applications. It provides instructions that operate on packed
13869 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
13871 GCC supports MIPS DSP operations using both the generic
13872 vector extensions (@pxref{Vector Extensions}) and a collection of
13873 MIPS-specific built-in functions. Both kinds of support are
13874 enabled by the @option{-mdsp} command-line option.
13876 Revision 2 of the ASE was introduced in the second half of 2006.
13877 This revision adds extra instructions to the original ASE, but is
13878 otherwise backwards-compatible with it. You can select revision 2
13879 using the command-line option @option{-mdspr2}; this option implies
13882 The SCOUNT and POS bits of the DSP control register are global. The
13883 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
13884 POS bits. During optimization, the compiler does not delete these
13885 instructions and it does not delete calls to functions containing
13886 these instructions.
13888 At present, GCC only provides support for operations on 32-bit
13889 vectors. The vector type associated with 8-bit integer data is
13890 usually called @code{v4i8}, the vector type associated with Q7
13891 is usually called @code{v4q7}, the vector type associated with 16-bit
13892 integer data is usually called @code{v2i16}, and the vector type
13893 associated with Q15 is usually called @code{v2q15}. They can be
13894 defined in C as follows:
13897 typedef signed char v4i8 __attribute__ ((vector_size(4)));
13898 typedef signed char v4q7 __attribute__ ((vector_size(4)));
13899 typedef short v2i16 __attribute__ ((vector_size(4)));
13900 typedef short v2q15 __attribute__ ((vector_size(4)));
13903 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
13904 initialized in the same way as aggregates. For example:
13907 v4i8 a = @{1, 2, 3, 4@};
13909 b = (v4i8) @{5, 6, 7, 8@};
13911 v2q15 c = @{0x0fcb, 0x3a75@};
13913 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
13916 @emph{Note:} The CPU's endianness determines the order in which values
13917 are packed. On little-endian targets, the first value is the least
13918 significant and the last value is the most significant. The opposite
13919 order applies to big-endian targets. For example, the code above
13920 sets the lowest byte of @code{a} to @code{1} on little-endian targets
13921 and @code{4} on big-endian targets.
13923 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
13924 representation. As shown in this example, the integer representation
13925 of a Q7 value can be obtained by multiplying the fractional value by
13926 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
13927 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
13930 The table below lists the @code{v4i8} and @code{v2q15} operations for which
13931 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
13932 and @code{c} and @code{d} are @code{v2q15} values.
13934 @multitable @columnfractions .50 .50
13935 @item C code @tab MIPS instruction
13936 @item @code{a + b} @tab @code{addu.qb}
13937 @item @code{c + d} @tab @code{addq.ph}
13938 @item @code{a - b} @tab @code{subu.qb}
13939 @item @code{c - d} @tab @code{subq.ph}
13942 The table below lists the @code{v2i16} operation for which
13943 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
13944 @code{v2i16} values.
13946 @multitable @columnfractions .50 .50
13947 @item C code @tab MIPS instruction
13948 @item @code{e * f} @tab @code{mul.ph}
13951 It is easier to describe the DSP built-in functions if we first define
13952 the following types:
13957 typedef unsigned int ui32;
13958 typedef long long a64;
13961 @code{q31} and @code{i32} are actually the same as @code{int}, but we
13962 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
13963 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
13964 @code{long long}, but we use @code{a64} to indicate values that are
13965 placed in one of the four DSP accumulators (@code{$ac0},
13966 @code{$ac1}, @code{$ac2} or @code{$ac3}).
13968 Also, some built-in functions prefer or require immediate numbers as
13969 parameters, because the corresponding DSP instructions accept both immediate
13970 numbers and register operands, or accept immediate numbers only. The
13971 immediate parameters are listed as follows.
13979 imm0_255: 0 to 255.
13980 imm_n32_31: -32 to 31.
13981 imm_n512_511: -512 to 511.
13984 The following built-in functions map directly to a particular MIPS DSP
13985 instruction. Please refer to the architecture specification
13986 for details on what each instruction does.
13989 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
13990 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
13991 q31 __builtin_mips_addq_s_w (q31, q31)
13992 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
13993 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
13994 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
13995 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
13996 q31 __builtin_mips_subq_s_w (q31, q31)
13997 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
13998 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
13999 i32 __builtin_mips_addsc (i32, i32)
14000 i32 __builtin_mips_addwc (i32, i32)
14001 i32 __builtin_mips_modsub (i32, i32)
14002 i32 __builtin_mips_raddu_w_qb (v4i8)
14003 v2q15 __builtin_mips_absq_s_ph (v2q15)
14004 q31 __builtin_mips_absq_s_w (q31)
14005 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
14006 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
14007 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
14008 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
14009 q31 __builtin_mips_preceq_w_phl (v2q15)
14010 q31 __builtin_mips_preceq_w_phr (v2q15)
14011 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
14012 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
14013 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
14014 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
14015 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
14016 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
14017 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
14018 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
14019 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
14020 v4i8 __builtin_mips_shll_qb (v4i8, i32)
14021 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
14022 v2q15 __builtin_mips_shll_ph (v2q15, i32)
14023 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
14024 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
14025 q31 __builtin_mips_shll_s_w (q31, imm0_31)
14026 q31 __builtin_mips_shll_s_w (q31, i32)
14027 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
14028 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
14029 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
14030 v2q15 __builtin_mips_shra_ph (v2q15, i32)
14031 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
14032 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
14033 q31 __builtin_mips_shra_r_w (q31, imm0_31)
14034 q31 __builtin_mips_shra_r_w (q31, i32)
14035 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
14036 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
14037 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
14038 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
14039 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
14040 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
14041 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
14042 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
14043 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
14044 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
14045 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
14046 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
14047 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
14048 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
14049 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
14050 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
14051 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
14052 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
14053 i32 __builtin_mips_bitrev (i32)
14054 i32 __builtin_mips_insv (i32, i32)
14055 v4i8 __builtin_mips_repl_qb (imm0_255)
14056 v4i8 __builtin_mips_repl_qb (i32)
14057 v2q15 __builtin_mips_repl_ph (imm_n512_511)
14058 v2q15 __builtin_mips_repl_ph (i32)
14059 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
14060 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
14061 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
14062 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
14063 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
14064 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
14065 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
14066 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
14067 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
14068 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
14069 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
14070 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
14071 i32 __builtin_mips_extr_w (a64, imm0_31)
14072 i32 __builtin_mips_extr_w (a64, i32)
14073 i32 __builtin_mips_extr_r_w (a64, imm0_31)
14074 i32 __builtin_mips_extr_s_h (a64, i32)
14075 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
14076 i32 __builtin_mips_extr_rs_w (a64, i32)
14077 i32 __builtin_mips_extr_s_h (a64, imm0_31)
14078 i32 __builtin_mips_extr_r_w (a64, i32)
14079 i32 __builtin_mips_extp (a64, imm0_31)
14080 i32 __builtin_mips_extp (a64, i32)
14081 i32 __builtin_mips_extpdp (a64, imm0_31)
14082 i32 __builtin_mips_extpdp (a64, i32)
14083 a64 __builtin_mips_shilo (a64, imm_n32_31)
14084 a64 __builtin_mips_shilo (a64, i32)
14085 a64 __builtin_mips_mthlip (a64, i32)
14086 void __builtin_mips_wrdsp (i32, imm0_63)
14087 i32 __builtin_mips_rddsp (imm0_63)
14088 i32 __builtin_mips_lbux (void *, i32)
14089 i32 __builtin_mips_lhx (void *, i32)
14090 i32 __builtin_mips_lwx (void *, i32)
14091 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
14092 i32 __builtin_mips_bposge32 (void)
14093 a64 __builtin_mips_madd (a64, i32, i32);
14094 a64 __builtin_mips_maddu (a64, ui32, ui32);
14095 a64 __builtin_mips_msub (a64, i32, i32);
14096 a64 __builtin_mips_msubu (a64, ui32, ui32);
14097 a64 __builtin_mips_mult (i32, i32);
14098 a64 __builtin_mips_multu (ui32, ui32);
14101 The following built-in functions map directly to a particular MIPS DSP REV 2
14102 instruction. Please refer to the architecture specification
14103 for details on what each instruction does.
14106 v4q7 __builtin_mips_absq_s_qb (v4q7);
14107 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
14108 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
14109 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
14110 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
14111 i32 __builtin_mips_append (i32, i32, imm0_31);
14112 i32 __builtin_mips_balign (i32, i32, imm0_3);
14113 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
14114 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
14115 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
14116 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
14117 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
14118 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
14119 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
14120 q31 __builtin_mips_mulq_rs_w (q31, q31);
14121 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
14122 q31 __builtin_mips_mulq_s_w (q31, q31);
14123 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
14124 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
14125 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
14126 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
14127 i32 __builtin_mips_prepend (i32, i32, imm0_31);
14128 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
14129 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
14130 v4i8 __builtin_mips_shra_qb (v4i8, i32);
14131 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
14132 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
14133 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
14134 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
14135 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
14136 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
14137 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
14138 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
14139 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
14140 q31 __builtin_mips_addqh_w (q31, q31);
14141 q31 __builtin_mips_addqh_r_w (q31, q31);
14142 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
14143 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
14144 q31 __builtin_mips_subqh_w (q31, q31);
14145 q31 __builtin_mips_subqh_r_w (q31, q31);
14146 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
14147 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
14148 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
14149 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
14150 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
14151 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
14155 @node MIPS Paired-Single Support
14156 @subsection MIPS Paired-Single Support
14158 The MIPS64 architecture includes a number of instructions that
14159 operate on pairs of single-precision floating-point values.
14160 Each pair is packed into a 64-bit floating-point register,
14161 with one element being designated the ``upper half'' and
14162 the other being designated the ``lower half''.
14164 GCC supports paired-single operations using both the generic
14165 vector extensions (@pxref{Vector Extensions}) and a collection of
14166 MIPS-specific built-in functions. Both kinds of support are
14167 enabled by the @option{-mpaired-single} command-line option.
14169 The vector type associated with paired-single values is usually
14170 called @code{v2sf}. It can be defined in C as follows:
14173 typedef float v2sf __attribute__ ((vector_size (8)));
14176 @code{v2sf} values are initialized in the same way as aggregates.
14180 v2sf a = @{1.5, 9.1@};
14183 b = (v2sf) @{e, f@};
14186 @emph{Note:} The CPU's endianness determines which value is stored in
14187 the upper half of a register and which value is stored in the lower half.
14188 On little-endian targets, the first value is the lower one and the second
14189 value is the upper one. The opposite order applies to big-endian targets.
14190 For example, the code above sets the lower half of @code{a} to
14191 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
14193 @node MIPS Loongson Built-in Functions
14194 @subsection MIPS Loongson Built-in Functions
14196 GCC provides intrinsics to access the SIMD instructions provided by the
14197 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
14198 available after inclusion of the @code{loongson.h} header file,
14199 operate on the following 64-bit vector types:
14202 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
14203 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
14204 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
14205 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
14206 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
14207 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
14210 The intrinsics provided are listed below; each is named after the
14211 machine instruction to which it corresponds, with suffixes added as
14212 appropriate to distinguish intrinsics that expand to the same machine
14213 instruction yet have different argument types. Refer to the architecture
14214 documentation for a description of the functionality of each
14218 int16x4_t packsswh (int32x2_t s, int32x2_t t);
14219 int8x8_t packsshb (int16x4_t s, int16x4_t t);
14220 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
14221 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
14222 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
14223 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
14224 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
14225 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
14226 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
14227 uint64_t paddd_u (uint64_t s, uint64_t t);
14228 int64_t paddd_s (int64_t s, int64_t t);
14229 int16x4_t paddsh (int16x4_t s, int16x4_t t);
14230 int8x8_t paddsb (int8x8_t s, int8x8_t t);
14231 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
14232 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
14233 uint64_t pandn_ud (uint64_t s, uint64_t t);
14234 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
14235 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
14236 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
14237 int64_t pandn_sd (int64_t s, int64_t t);
14238 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
14239 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
14240 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
14241 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
14242 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
14243 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
14244 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
14245 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
14246 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
14247 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
14248 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
14249 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
14250 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
14251 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
14252 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
14253 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
14254 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
14255 uint16x4_t pextrh_u (uint16x4_t s, int field);
14256 int16x4_t pextrh_s (int16x4_t s, int field);
14257 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
14258 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
14259 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
14260 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
14261 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
14262 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
14263 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
14264 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
14265 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
14266 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
14267 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
14268 int16x4_t pminsh (int16x4_t s, int16x4_t t);
14269 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
14270 uint8x8_t pmovmskb_u (uint8x8_t s);
14271 int8x8_t pmovmskb_s (int8x8_t s);
14272 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
14273 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
14274 int16x4_t pmullh (int16x4_t s, int16x4_t t);
14275 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
14276 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
14277 uint16x4_t biadd (uint8x8_t s);
14278 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
14279 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
14280 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
14281 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
14282 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
14283 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
14284 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
14285 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
14286 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
14287 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
14288 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
14289 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
14290 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
14291 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
14292 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
14293 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
14294 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
14295 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
14296 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
14297 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
14298 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
14299 uint64_t psubd_u (uint64_t s, uint64_t t);
14300 int64_t psubd_s (int64_t s, int64_t t);
14301 int16x4_t psubsh (int16x4_t s, int16x4_t t);
14302 int8x8_t psubsb (int8x8_t s, int8x8_t t);
14303 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
14304 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
14305 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
14306 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
14307 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
14308 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
14309 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
14310 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
14311 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
14312 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
14313 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
14314 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
14315 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
14316 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
14320 * Paired-Single Arithmetic::
14321 * Paired-Single Built-in Functions::
14322 * MIPS-3D Built-in Functions::
14325 @node Paired-Single Arithmetic
14326 @subsubsection Paired-Single Arithmetic
14328 The table below lists the @code{v2sf} operations for which hardware
14329 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
14330 values and @code{x} is an integral value.
14332 @multitable @columnfractions .50 .50
14333 @item C code @tab MIPS instruction
14334 @item @code{a + b} @tab @code{add.ps}
14335 @item @code{a - b} @tab @code{sub.ps}
14336 @item @code{-a} @tab @code{neg.ps}
14337 @item @code{a * b} @tab @code{mul.ps}
14338 @item @code{a * b + c} @tab @code{madd.ps}
14339 @item @code{a * b - c} @tab @code{msub.ps}
14340 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
14341 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
14342 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
14345 Note that the multiply-accumulate instructions can be disabled
14346 using the command-line option @code{-mno-fused-madd}.
14348 @node Paired-Single Built-in Functions
14349 @subsubsection Paired-Single Built-in Functions
14351 The following paired-single functions map directly to a particular
14352 MIPS instruction. Please refer to the architecture specification
14353 for details on what each instruction does.
14356 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
14357 Pair lower lower (@code{pll.ps}).
14359 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
14360 Pair upper lower (@code{pul.ps}).
14362 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
14363 Pair lower upper (@code{plu.ps}).
14365 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
14366 Pair upper upper (@code{puu.ps}).
14368 @item v2sf __builtin_mips_cvt_ps_s (float, float)
14369 Convert pair to paired single (@code{cvt.ps.s}).
14371 @item float __builtin_mips_cvt_s_pl (v2sf)
14372 Convert pair lower to single (@code{cvt.s.pl}).
14374 @item float __builtin_mips_cvt_s_pu (v2sf)
14375 Convert pair upper to single (@code{cvt.s.pu}).
14377 @item v2sf __builtin_mips_abs_ps (v2sf)
14378 Absolute value (@code{abs.ps}).
14380 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
14381 Align variable (@code{alnv.ps}).
14383 @emph{Note:} The value of the third parameter must be 0 or 4
14384 modulo 8, otherwise the result is unpredictable. Please read the
14385 instruction description for details.
14388 The following multi-instruction functions are also available.
14389 In each case, @var{cond} can be any of the 16 floating-point conditions:
14390 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
14391 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
14392 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
14395 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14396 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14397 Conditional move based on floating-point comparison (@code{c.@var{cond}.ps},
14398 @code{movt.ps}/@code{movf.ps}).
14400 The @code{movt} functions return the value @var{x} computed by:
14403 c.@var{cond}.ps @var{cc},@var{a},@var{b}
14404 mov.ps @var{x},@var{c}
14405 movt.ps @var{x},@var{d},@var{cc}
14408 The @code{movf} functions are similar but use @code{movf.ps} instead
14411 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14412 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14413 Comparison of two paired-single values (@code{c.@var{cond}.ps},
14414 @code{bc1t}/@code{bc1f}).
14416 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
14417 and return either the upper or lower half of the result. For example:
14421 if (__builtin_mips_upper_c_eq_ps (a, b))
14422 upper_halves_are_equal ();
14424 upper_halves_are_unequal ();
14426 if (__builtin_mips_lower_c_eq_ps (a, b))
14427 lower_halves_are_equal ();
14429 lower_halves_are_unequal ();
14433 @node MIPS-3D Built-in Functions
14434 @subsubsection MIPS-3D Built-in Functions
14436 The MIPS-3D Application-Specific Extension (ASE) includes additional
14437 paired-single instructions that are designed to improve the performance
14438 of 3D graphics operations. Support for these instructions is controlled
14439 by the @option{-mips3d} command-line option.
14441 The functions listed below map directly to a particular MIPS-3D
14442 instruction. Please refer to the architecture specification for
14443 more details on what each instruction does.
14446 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
14447 Reduction add (@code{addr.ps}).
14449 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
14450 Reduction multiply (@code{mulr.ps}).
14452 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
14453 Convert paired single to paired word (@code{cvt.pw.ps}).
14455 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
14456 Convert paired word to paired single (@code{cvt.ps.pw}).
14458 @item float __builtin_mips_recip1_s (float)
14459 @itemx double __builtin_mips_recip1_d (double)
14460 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
14461 Reduced-precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
14463 @item float __builtin_mips_recip2_s (float, float)
14464 @itemx double __builtin_mips_recip2_d (double, double)
14465 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
14466 Reduced-precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
14468 @item float __builtin_mips_rsqrt1_s (float)
14469 @itemx double __builtin_mips_rsqrt1_d (double)
14470 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
14471 Reduced-precision reciprocal square root (sequence step 1)
14472 (@code{rsqrt1.@var{fmt}}).
14474 @item float __builtin_mips_rsqrt2_s (float, float)
14475 @itemx double __builtin_mips_rsqrt2_d (double, double)
14476 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
14477 Reduced-precision reciprocal square root (sequence step 2)
14478 (@code{rsqrt2.@var{fmt}}).
14481 The following multi-instruction functions are also available.
14482 In each case, @var{cond} can be any of the 16 floating-point conditions:
14483 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
14484 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
14485 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
14488 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
14489 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
14490 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
14491 @code{bc1t}/@code{bc1f}).
14493 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
14494 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
14499 if (__builtin_mips_cabs_eq_s (a, b))
14505 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14506 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14507 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
14508 @code{bc1t}/@code{bc1f}).
14510 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
14511 and return either the upper or lower half of the result. For example:
14515 if (__builtin_mips_upper_cabs_eq_ps (a, b))
14516 upper_halves_are_equal ();
14518 upper_halves_are_unequal ();
14520 if (__builtin_mips_lower_cabs_eq_ps (a, b))
14521 lower_halves_are_equal ();
14523 lower_halves_are_unequal ();
14526 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14527 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14528 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
14529 @code{movt.ps}/@code{movf.ps}).
14531 The @code{movt} functions return the value @var{x} computed by:
14534 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
14535 mov.ps @var{x},@var{c}
14536 movt.ps @var{x},@var{d},@var{cc}
14539 The @code{movf} functions are similar but use @code{movf.ps} instead
14542 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14543 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14544 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14545 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
14546 Comparison of two paired-single values
14547 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
14548 @code{bc1any2t}/@code{bc1any2f}).
14550 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
14551 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
14552 result is true and the @code{all} forms return true if both results are true.
14557 if (__builtin_mips_any_c_eq_ps (a, b))
14562 if (__builtin_mips_all_c_eq_ps (a, b))
14568 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14569 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14570 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14571 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
14572 Comparison of four paired-single values
14573 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
14574 @code{bc1any4t}/@code{bc1any4f}).
14576 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
14577 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
14578 The @code{any} forms return true if any of the four results are true
14579 and the @code{all} forms return true if all four results are true.
14584 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
14589 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
14596 @node MIPS SIMD Architecture (MSA) Support
14597 @subsection MIPS SIMD Architecture (MSA) Support
14600 * MIPS SIMD Architecture Built-in Functions::
14603 GCC provides intrinsics to access the SIMD instructions provided by the
14604 MSA MIPS SIMD Architecture. The interface is made available by including
14605 @code{<msa.h>} and using @option{-mmsa -mhard-float -mfp64 -mnan=2008}.
14606 For each @code{__builtin_msa_*}, there is a shortened name of the intrinsic,
14609 MSA implements 128-bit wide vector registers, operating on 8-, 16-, 32- and
14610 64-bit integer, 16- and 32-bit fixed-point, or 32- and 64-bit floating point
14611 data elements. The following vectors typedefs are included in @code{msa.h}:
14613 @item @code{v16i8}, a vector of sixteen signed 8-bit integers;
14614 @item @code{v16u8}, a vector of sixteen unsigned 8-bit integers;
14615 @item @code{v8i16}, a vector of eight signed 16-bit integers;
14616 @item @code{v8u16}, a vector of eight unsigned 16-bit integers;
14617 @item @code{v4i32}, a vector of four signed 32-bit integers;
14618 @item @code{v4u32}, a vector of four unsigned 32-bit integers;
14619 @item @code{v2i64}, a vector of two signed 64-bit integers;
14620 @item @code{v2u64}, a vector of two unsigned 64-bit integers;
14621 @item @code{v4f32}, a vector of four 32-bit floats;
14622 @item @code{v2f64}, a vector of two 64-bit doubles.
14625 Instructions and corresponding built-ins may have additional restrictions and/or
14626 input/output values manipulated:
14628 @item @code{imm0_1}, an integer literal in range 0 to 1;
14629 @item @code{imm0_3}, an integer literal in range 0 to 3;
14630 @item @code{imm0_7}, an integer literal in range 0 to 7;
14631 @item @code{imm0_15}, an integer literal in range 0 to 15;
14632 @item @code{imm0_31}, an integer literal in range 0 to 31;
14633 @item @code{imm0_63}, an integer literal in range 0 to 63;
14634 @item @code{imm0_255}, an integer literal in range 0 to 255;
14635 @item @code{imm_n16_15}, an integer literal in range -16 to 15;
14636 @item @code{imm_n512_511}, an integer literal in range -512 to 511;
14637 @item @code{imm_n1024_1022}, an integer literal in range -512 to 511 left
14638 shifted by 1 bit, i.e., -1024, -1022, @dots{}, 1020, 1022;
14639 @item @code{imm_n2048_2044}, an integer literal in range -512 to 511 left
14640 shifted by 2 bits, i.e., -2048, -2044, @dots{}, 2040, 2044;
14641 @item @code{imm_n4096_4088}, an integer literal in range -512 to 511 left
14642 shifted by 3 bits, i.e., -4096, -4088, @dots{}, 4080, 4088;
14643 @item @code{imm1_4}, an integer literal in range 1 to 4;
14644 @item @code{i32, i64, u32, u64, f32, f64}, defined as follows:
14650 #if __LONG_MAX__ == __LONG_LONG_MAX__
14653 typedef long long i64;
14656 typedef unsigned int u32;
14657 #if __LONG_MAX__ == __LONG_LONG_MAX__
14658 typedef unsigned long u64;
14660 typedef unsigned long long u64;
14663 typedef double f64;
14668 @node MIPS SIMD Architecture Built-in Functions
14669 @subsubsection MIPS SIMD Architecture Built-in Functions
14671 The intrinsics provided are listed below; each is named after the
14672 machine instruction.
14675 v16i8 __builtin_msa_add_a_b (v16i8, v16i8);
14676 v8i16 __builtin_msa_add_a_h (v8i16, v8i16);
14677 v4i32 __builtin_msa_add_a_w (v4i32, v4i32);
14678 v2i64 __builtin_msa_add_a_d (v2i64, v2i64);
14680 v16i8 __builtin_msa_adds_a_b (v16i8, v16i8);
14681 v8i16 __builtin_msa_adds_a_h (v8i16, v8i16);
14682 v4i32 __builtin_msa_adds_a_w (v4i32, v4i32);
14683 v2i64 __builtin_msa_adds_a_d (v2i64, v2i64);
14685 v16i8 __builtin_msa_adds_s_b (v16i8, v16i8);
14686 v8i16 __builtin_msa_adds_s_h (v8i16, v8i16);
14687 v4i32 __builtin_msa_adds_s_w (v4i32, v4i32);
14688 v2i64 __builtin_msa_adds_s_d (v2i64, v2i64);
14690 v16u8 __builtin_msa_adds_u_b (v16u8, v16u8);
14691 v8u16 __builtin_msa_adds_u_h (v8u16, v8u16);
14692 v4u32 __builtin_msa_adds_u_w (v4u32, v4u32);
14693 v2u64 __builtin_msa_adds_u_d (v2u64, v2u64);
14695 v16i8 __builtin_msa_addv_b (v16i8, v16i8);
14696 v8i16 __builtin_msa_addv_h (v8i16, v8i16);
14697 v4i32 __builtin_msa_addv_w (v4i32, v4i32);
14698 v2i64 __builtin_msa_addv_d (v2i64, v2i64);
14700 v16i8 __builtin_msa_addvi_b (v16i8, imm0_31);
14701 v8i16 __builtin_msa_addvi_h (v8i16, imm0_31);
14702 v4i32 __builtin_msa_addvi_w (v4i32, imm0_31);
14703 v2i64 __builtin_msa_addvi_d (v2i64, imm0_31);
14705 v16u8 __builtin_msa_and_v (v16u8, v16u8);
14707 v16u8 __builtin_msa_andi_b (v16u8, imm0_255);
14709 v16i8 __builtin_msa_asub_s_b (v16i8, v16i8);
14710 v8i16 __builtin_msa_asub_s_h (v8i16, v8i16);
14711 v4i32 __builtin_msa_asub_s_w (v4i32, v4i32);
14712 v2i64 __builtin_msa_asub_s_d (v2i64, v2i64);
14714 v16u8 __builtin_msa_asub_u_b (v16u8, v16u8);
14715 v8u16 __builtin_msa_asub_u_h (v8u16, v8u16);
14716 v4u32 __builtin_msa_asub_u_w (v4u32, v4u32);
14717 v2u64 __builtin_msa_asub_u_d (v2u64, v2u64);
14719 v16i8 __builtin_msa_ave_s_b (v16i8, v16i8);
14720 v8i16 __builtin_msa_ave_s_h (v8i16, v8i16);
14721 v4i32 __builtin_msa_ave_s_w (v4i32, v4i32);
14722 v2i64 __builtin_msa_ave_s_d (v2i64, v2i64);
14724 v16u8 __builtin_msa_ave_u_b (v16u8, v16u8);
14725 v8u16 __builtin_msa_ave_u_h (v8u16, v8u16);
14726 v4u32 __builtin_msa_ave_u_w (v4u32, v4u32);
14727 v2u64 __builtin_msa_ave_u_d (v2u64, v2u64);
14729 v16i8 __builtin_msa_aver_s_b (v16i8, v16i8);
14730 v8i16 __builtin_msa_aver_s_h (v8i16, v8i16);
14731 v4i32 __builtin_msa_aver_s_w (v4i32, v4i32);
14732 v2i64 __builtin_msa_aver_s_d (v2i64, v2i64);
14734 v16u8 __builtin_msa_aver_u_b (v16u8, v16u8);
14735 v8u16 __builtin_msa_aver_u_h (v8u16, v8u16);
14736 v4u32 __builtin_msa_aver_u_w (v4u32, v4u32);
14737 v2u64 __builtin_msa_aver_u_d (v2u64, v2u64);
14739 v16u8 __builtin_msa_bclr_b (v16u8, v16u8);
14740 v8u16 __builtin_msa_bclr_h (v8u16, v8u16);
14741 v4u32 __builtin_msa_bclr_w (v4u32, v4u32);
14742 v2u64 __builtin_msa_bclr_d (v2u64, v2u64);
14744 v16u8 __builtin_msa_bclri_b (v16u8, imm0_7);
14745 v8u16 __builtin_msa_bclri_h (v8u16, imm0_15);
14746 v4u32 __builtin_msa_bclri_w (v4u32, imm0_31);
14747 v2u64 __builtin_msa_bclri_d (v2u64, imm0_63);
14749 v16u8 __builtin_msa_binsl_b (v16u8, v16u8, v16u8);
14750 v8u16 __builtin_msa_binsl_h (v8u16, v8u16, v8u16);
14751 v4u32 __builtin_msa_binsl_w (v4u32, v4u32, v4u32);
14752 v2u64 __builtin_msa_binsl_d (v2u64, v2u64, v2u64);
14754 v16u8 __builtin_msa_binsli_b (v16u8, v16u8, imm0_7);
14755 v8u16 __builtin_msa_binsli_h (v8u16, v8u16, imm0_15);
14756 v4u32 __builtin_msa_binsli_w (v4u32, v4u32, imm0_31);
14757 v2u64 __builtin_msa_binsli_d (v2u64, v2u64, imm0_63);
14759 v16u8 __builtin_msa_binsr_b (v16u8, v16u8, v16u8);
14760 v8u16 __builtin_msa_binsr_h (v8u16, v8u16, v8u16);
14761 v4u32 __builtin_msa_binsr_w (v4u32, v4u32, v4u32);
14762 v2u64 __builtin_msa_binsr_d (v2u64, v2u64, v2u64);
14764 v16u8 __builtin_msa_binsri_b (v16u8, v16u8, imm0_7);
14765 v8u16 __builtin_msa_binsri_h (v8u16, v8u16, imm0_15);
14766 v4u32 __builtin_msa_binsri_w (v4u32, v4u32, imm0_31);
14767 v2u64 __builtin_msa_binsri_d (v2u64, v2u64, imm0_63);
14769 v16u8 __builtin_msa_bmnz_v (v16u8, v16u8, v16u8);
14771 v16u8 __builtin_msa_bmnzi_b (v16u8, v16u8, imm0_255);
14773 v16u8 __builtin_msa_bmz_v (v16u8, v16u8, v16u8);
14775 v16u8 __builtin_msa_bmzi_b (v16u8, v16u8, imm0_255);
14777 v16u8 __builtin_msa_bneg_b (v16u8, v16u8);
14778 v8u16 __builtin_msa_bneg_h (v8u16, v8u16);
14779 v4u32 __builtin_msa_bneg_w (v4u32, v4u32);
14780 v2u64 __builtin_msa_bneg_d (v2u64, v2u64);
14782 v16u8 __builtin_msa_bnegi_b (v16u8, imm0_7);
14783 v8u16 __builtin_msa_bnegi_h (v8u16, imm0_15);
14784 v4u32 __builtin_msa_bnegi_w (v4u32, imm0_31);
14785 v2u64 __builtin_msa_bnegi_d (v2u64, imm0_63);
14787 i32 __builtin_msa_bnz_b (v16u8);
14788 i32 __builtin_msa_bnz_h (v8u16);
14789 i32 __builtin_msa_bnz_w (v4u32);
14790 i32 __builtin_msa_bnz_d (v2u64);
14792 i32 __builtin_msa_bnz_v (v16u8);
14794 v16u8 __builtin_msa_bsel_v (v16u8, v16u8, v16u8);
14796 v16u8 __builtin_msa_bseli_b (v16u8, v16u8, imm0_255);
14798 v16u8 __builtin_msa_bset_b (v16u8, v16u8);
14799 v8u16 __builtin_msa_bset_h (v8u16, v8u16);
14800 v4u32 __builtin_msa_bset_w (v4u32, v4u32);
14801 v2u64 __builtin_msa_bset_d (v2u64, v2u64);
14803 v16u8 __builtin_msa_bseti_b (v16u8, imm0_7);
14804 v8u16 __builtin_msa_bseti_h (v8u16, imm0_15);
14805 v4u32 __builtin_msa_bseti_w (v4u32, imm0_31);
14806 v2u64 __builtin_msa_bseti_d (v2u64, imm0_63);
14808 i32 __builtin_msa_bz_b (v16u8);
14809 i32 __builtin_msa_bz_h (v8u16);
14810 i32 __builtin_msa_bz_w (v4u32);
14811 i32 __builtin_msa_bz_d (v2u64);
14813 i32 __builtin_msa_bz_v (v16u8);
14815 v16i8 __builtin_msa_ceq_b (v16i8, v16i8);
14816 v8i16 __builtin_msa_ceq_h (v8i16, v8i16);
14817 v4i32 __builtin_msa_ceq_w (v4i32, v4i32);
14818 v2i64 __builtin_msa_ceq_d (v2i64, v2i64);
14820 v16i8 __builtin_msa_ceqi_b (v16i8, imm_n16_15);
14821 v8i16 __builtin_msa_ceqi_h (v8i16, imm_n16_15);
14822 v4i32 __builtin_msa_ceqi_w (v4i32, imm_n16_15);
14823 v2i64 __builtin_msa_ceqi_d (v2i64, imm_n16_15);
14825 i32 __builtin_msa_cfcmsa (imm0_31);
14827 v16i8 __builtin_msa_cle_s_b (v16i8, v16i8);
14828 v8i16 __builtin_msa_cle_s_h (v8i16, v8i16);
14829 v4i32 __builtin_msa_cle_s_w (v4i32, v4i32);
14830 v2i64 __builtin_msa_cle_s_d (v2i64, v2i64);
14832 v16i8 __builtin_msa_cle_u_b (v16u8, v16u8);
14833 v8i16 __builtin_msa_cle_u_h (v8u16, v8u16);
14834 v4i32 __builtin_msa_cle_u_w (v4u32, v4u32);
14835 v2i64 __builtin_msa_cle_u_d (v2u64, v2u64);
14837 v16i8 __builtin_msa_clei_s_b (v16i8, imm_n16_15);
14838 v8i16 __builtin_msa_clei_s_h (v8i16, imm_n16_15);
14839 v4i32 __builtin_msa_clei_s_w (v4i32, imm_n16_15);
14840 v2i64 __builtin_msa_clei_s_d (v2i64, imm_n16_15);
14842 v16i8 __builtin_msa_clei_u_b (v16u8, imm0_31);
14843 v8i16 __builtin_msa_clei_u_h (v8u16, imm0_31);
14844 v4i32 __builtin_msa_clei_u_w (v4u32, imm0_31);
14845 v2i64 __builtin_msa_clei_u_d (v2u64, imm0_31);
14847 v16i8 __builtin_msa_clt_s_b (v16i8, v16i8);
14848 v8i16 __builtin_msa_clt_s_h (v8i16, v8i16);
14849 v4i32 __builtin_msa_clt_s_w (v4i32, v4i32);
14850 v2i64 __builtin_msa_clt_s_d (v2i64, v2i64);
14852 v16i8 __builtin_msa_clt_u_b (v16u8, v16u8);
14853 v8i16 __builtin_msa_clt_u_h (v8u16, v8u16);
14854 v4i32 __builtin_msa_clt_u_w (v4u32, v4u32);
14855 v2i64 __builtin_msa_clt_u_d (v2u64, v2u64);
14857 v16i8 __builtin_msa_clti_s_b (v16i8, imm_n16_15);
14858 v8i16 __builtin_msa_clti_s_h (v8i16, imm_n16_15);
14859 v4i32 __builtin_msa_clti_s_w (v4i32, imm_n16_15);
14860 v2i64 __builtin_msa_clti_s_d (v2i64, imm_n16_15);
14862 v16i8 __builtin_msa_clti_u_b (v16u8, imm0_31);
14863 v8i16 __builtin_msa_clti_u_h (v8u16, imm0_31);
14864 v4i32 __builtin_msa_clti_u_w (v4u32, imm0_31);
14865 v2i64 __builtin_msa_clti_u_d (v2u64, imm0_31);
14867 i32 __builtin_msa_copy_s_b (v16i8, imm0_15);
14868 i32 __builtin_msa_copy_s_h (v8i16, imm0_7);
14869 i32 __builtin_msa_copy_s_w (v4i32, imm0_3);
14870 i64 __builtin_msa_copy_s_d (v2i64, imm0_1);
14872 u32 __builtin_msa_copy_u_b (v16i8, imm0_15);
14873 u32 __builtin_msa_copy_u_h (v8i16, imm0_7);
14874 u32 __builtin_msa_copy_u_w (v4i32, imm0_3);
14875 u64 __builtin_msa_copy_u_d (v2i64, imm0_1);
14877 void __builtin_msa_ctcmsa (imm0_31, i32);
14879 v16i8 __builtin_msa_div_s_b (v16i8, v16i8);
14880 v8i16 __builtin_msa_div_s_h (v8i16, v8i16);
14881 v4i32 __builtin_msa_div_s_w (v4i32, v4i32);
14882 v2i64 __builtin_msa_div_s_d (v2i64, v2i64);
14884 v16u8 __builtin_msa_div_u_b (v16u8, v16u8);
14885 v8u16 __builtin_msa_div_u_h (v8u16, v8u16);
14886 v4u32 __builtin_msa_div_u_w (v4u32, v4u32);
14887 v2u64 __builtin_msa_div_u_d (v2u64, v2u64);
14889 v8i16 __builtin_msa_dotp_s_h (v16i8, v16i8);
14890 v4i32 __builtin_msa_dotp_s_w (v8i16, v8i16);
14891 v2i64 __builtin_msa_dotp_s_d (v4i32, v4i32);
14893 v8u16 __builtin_msa_dotp_u_h (v16u8, v16u8);
14894 v4u32 __builtin_msa_dotp_u_w (v8u16, v8u16);
14895 v2u64 __builtin_msa_dotp_u_d (v4u32, v4u32);
14897 v8i16 __builtin_msa_dpadd_s_h (v8i16, v16i8, v16i8);
14898 v4i32 __builtin_msa_dpadd_s_w (v4i32, v8i16, v8i16);
14899 v2i64 __builtin_msa_dpadd_s_d (v2i64, v4i32, v4i32);
14901 v8u16 __builtin_msa_dpadd_u_h (v8u16, v16u8, v16u8);
14902 v4u32 __builtin_msa_dpadd_u_w (v4u32, v8u16, v8u16);
14903 v2u64 __builtin_msa_dpadd_u_d (v2u64, v4u32, v4u32);
14905 v8i16 __builtin_msa_dpsub_s_h (v8i16, v16i8, v16i8);
14906 v4i32 __builtin_msa_dpsub_s_w (v4i32, v8i16, v8i16);
14907 v2i64 __builtin_msa_dpsub_s_d (v2i64, v4i32, v4i32);
14909 v8i16 __builtin_msa_dpsub_u_h (v8i16, v16u8, v16u8);
14910 v4i32 __builtin_msa_dpsub_u_w (v4i32, v8u16, v8u16);
14911 v2i64 __builtin_msa_dpsub_u_d (v2i64, v4u32, v4u32);
14913 v4f32 __builtin_msa_fadd_w (v4f32, v4f32);
14914 v2f64 __builtin_msa_fadd_d (v2f64, v2f64);
14916 v4i32 __builtin_msa_fcaf_w (v4f32, v4f32);
14917 v2i64 __builtin_msa_fcaf_d (v2f64, v2f64);
14919 v4i32 __builtin_msa_fceq_w (v4f32, v4f32);
14920 v2i64 __builtin_msa_fceq_d (v2f64, v2f64);
14922 v4i32 __builtin_msa_fclass_w (v4f32);
14923 v2i64 __builtin_msa_fclass_d (v2f64);
14925 v4i32 __builtin_msa_fcle_w (v4f32, v4f32);
14926 v2i64 __builtin_msa_fcle_d (v2f64, v2f64);
14928 v4i32 __builtin_msa_fclt_w (v4f32, v4f32);
14929 v2i64 __builtin_msa_fclt_d (v2f64, v2f64);
14931 v4i32 __builtin_msa_fcne_w (v4f32, v4f32);
14932 v2i64 __builtin_msa_fcne_d (v2f64, v2f64);
14934 v4i32 __builtin_msa_fcor_w (v4f32, v4f32);
14935 v2i64 __builtin_msa_fcor_d (v2f64, v2f64);
14937 v4i32 __builtin_msa_fcueq_w (v4f32, v4f32);
14938 v2i64 __builtin_msa_fcueq_d (v2f64, v2f64);
14940 v4i32 __builtin_msa_fcule_w (v4f32, v4f32);
14941 v2i64 __builtin_msa_fcule_d (v2f64, v2f64);
14943 v4i32 __builtin_msa_fcult_w (v4f32, v4f32);
14944 v2i64 __builtin_msa_fcult_d (v2f64, v2f64);
14946 v4i32 __builtin_msa_fcun_w (v4f32, v4f32);
14947 v2i64 __builtin_msa_fcun_d (v2f64, v2f64);
14949 v4i32 __builtin_msa_fcune_w (v4f32, v4f32);
14950 v2i64 __builtin_msa_fcune_d (v2f64, v2f64);
14952 v4f32 __builtin_msa_fdiv_w (v4f32, v4f32);
14953 v2f64 __builtin_msa_fdiv_d (v2f64, v2f64);
14955 v8i16 __builtin_msa_fexdo_h (v4f32, v4f32);
14956 v4f32 __builtin_msa_fexdo_w (v2f64, v2f64);
14958 v4f32 __builtin_msa_fexp2_w (v4f32, v4i32);
14959 v2f64 __builtin_msa_fexp2_d (v2f64, v2i64);
14961 v4f32 __builtin_msa_fexupl_w (v8i16);
14962 v2f64 __builtin_msa_fexupl_d (v4f32);
14964 v4f32 __builtin_msa_fexupr_w (v8i16);
14965 v2f64 __builtin_msa_fexupr_d (v4f32);
14967 v4f32 __builtin_msa_ffint_s_w (v4i32);
14968 v2f64 __builtin_msa_ffint_s_d (v2i64);
14970 v4f32 __builtin_msa_ffint_u_w (v4u32);
14971 v2f64 __builtin_msa_ffint_u_d (v2u64);
14973 v4f32 __builtin_msa_ffql_w (v8i16);
14974 v2f64 __builtin_msa_ffql_d (v4i32);
14976 v4f32 __builtin_msa_ffqr_w (v8i16);
14977 v2f64 __builtin_msa_ffqr_d (v4i32);
14979 v16i8 __builtin_msa_fill_b (i32);
14980 v8i16 __builtin_msa_fill_h (i32);
14981 v4i32 __builtin_msa_fill_w (i32);
14982 v2i64 __builtin_msa_fill_d (i64);
14984 v4f32 __builtin_msa_flog2_w (v4f32);
14985 v2f64 __builtin_msa_flog2_d (v2f64);
14987 v4f32 __builtin_msa_fmadd_w (v4f32, v4f32, v4f32);
14988 v2f64 __builtin_msa_fmadd_d (v2f64, v2f64, v2f64);
14990 v4f32 __builtin_msa_fmax_w (v4f32, v4f32);
14991 v2f64 __builtin_msa_fmax_d (v2f64, v2f64);
14993 v4f32 __builtin_msa_fmax_a_w (v4f32, v4f32);
14994 v2f64 __builtin_msa_fmax_a_d (v2f64, v2f64);
14996 v4f32 __builtin_msa_fmin_w (v4f32, v4f32);
14997 v2f64 __builtin_msa_fmin_d (v2f64, v2f64);
14999 v4f32 __builtin_msa_fmin_a_w (v4f32, v4f32);
15000 v2f64 __builtin_msa_fmin_a_d (v2f64, v2f64);
15002 v4f32 __builtin_msa_fmsub_w (v4f32, v4f32, v4f32);
15003 v2f64 __builtin_msa_fmsub_d (v2f64, v2f64, v2f64);
15005 v4f32 __builtin_msa_fmul_w (v4f32, v4f32);
15006 v2f64 __builtin_msa_fmul_d (v2f64, v2f64);
15008 v4f32 __builtin_msa_frint_w (v4f32);
15009 v2f64 __builtin_msa_frint_d (v2f64);
15011 v4f32 __builtin_msa_frcp_w (v4f32);
15012 v2f64 __builtin_msa_frcp_d (v2f64);
15014 v4f32 __builtin_msa_frsqrt_w (v4f32);
15015 v2f64 __builtin_msa_frsqrt_d (v2f64);
15017 v4i32 __builtin_msa_fsaf_w (v4f32, v4f32);
15018 v2i64 __builtin_msa_fsaf_d (v2f64, v2f64);
15020 v4i32 __builtin_msa_fseq_w (v4f32, v4f32);
15021 v2i64 __builtin_msa_fseq_d (v2f64, v2f64);
15023 v4i32 __builtin_msa_fsle_w (v4f32, v4f32);
15024 v2i64 __builtin_msa_fsle_d (v2f64, v2f64);
15026 v4i32 __builtin_msa_fslt_w (v4f32, v4f32);
15027 v2i64 __builtin_msa_fslt_d (v2f64, v2f64);
15029 v4i32 __builtin_msa_fsne_w (v4f32, v4f32);
15030 v2i64 __builtin_msa_fsne_d (v2f64, v2f64);
15032 v4i32 __builtin_msa_fsor_w (v4f32, v4f32);
15033 v2i64 __builtin_msa_fsor_d (v2f64, v2f64);
15035 v4f32 __builtin_msa_fsqrt_w (v4f32);
15036 v2f64 __builtin_msa_fsqrt_d (v2f64);
15038 v4f32 __builtin_msa_fsub_w (v4f32, v4f32);
15039 v2f64 __builtin_msa_fsub_d (v2f64, v2f64);
15041 v4i32 __builtin_msa_fsueq_w (v4f32, v4f32);
15042 v2i64 __builtin_msa_fsueq_d (v2f64, v2f64);
15044 v4i32 __builtin_msa_fsule_w (v4f32, v4f32);
15045 v2i64 __builtin_msa_fsule_d (v2f64, v2f64);
15047 v4i32 __builtin_msa_fsult_w (v4f32, v4f32);
15048 v2i64 __builtin_msa_fsult_d (v2f64, v2f64);
15050 v4i32 __builtin_msa_fsun_w (v4f32, v4f32);
15051 v2i64 __builtin_msa_fsun_d (v2f64, v2f64);
15053 v4i32 __builtin_msa_fsune_w (v4f32, v4f32);
15054 v2i64 __builtin_msa_fsune_d (v2f64, v2f64);
15056 v4i32 __builtin_msa_ftint_s_w (v4f32);
15057 v2i64 __builtin_msa_ftint_s_d (v2f64);
15059 v4u32 __builtin_msa_ftint_u_w (v4f32);
15060 v2u64 __builtin_msa_ftint_u_d (v2f64);
15062 v8i16 __builtin_msa_ftq_h (v4f32, v4f32);
15063 v4i32 __builtin_msa_ftq_w (v2f64, v2f64);
15065 v4i32 __builtin_msa_ftrunc_s_w (v4f32);
15066 v2i64 __builtin_msa_ftrunc_s_d (v2f64);
15068 v4u32 __builtin_msa_ftrunc_u_w (v4f32);
15069 v2u64 __builtin_msa_ftrunc_u_d (v2f64);
15071 v8i16 __builtin_msa_hadd_s_h (v16i8, v16i8);
15072 v4i32 __builtin_msa_hadd_s_w (v8i16, v8i16);
15073 v2i64 __builtin_msa_hadd_s_d (v4i32, v4i32);
15075 v8u16 __builtin_msa_hadd_u_h (v16u8, v16u8);
15076 v4u32 __builtin_msa_hadd_u_w (v8u16, v8u16);
15077 v2u64 __builtin_msa_hadd_u_d (v4u32, v4u32);
15079 v8i16 __builtin_msa_hsub_s_h (v16i8, v16i8);
15080 v4i32 __builtin_msa_hsub_s_w (v8i16, v8i16);
15081 v2i64 __builtin_msa_hsub_s_d (v4i32, v4i32);
15083 v8i16 __builtin_msa_hsub_u_h (v16u8, v16u8);
15084 v4i32 __builtin_msa_hsub_u_w (v8u16, v8u16);
15085 v2i64 __builtin_msa_hsub_u_d (v4u32, v4u32);
15087 v16i8 __builtin_msa_ilvev_b (v16i8, v16i8);
15088 v8i16 __builtin_msa_ilvev_h (v8i16, v8i16);
15089 v4i32 __builtin_msa_ilvev_w (v4i32, v4i32);
15090 v2i64 __builtin_msa_ilvev_d (v2i64, v2i64);
15092 v16i8 __builtin_msa_ilvl_b (v16i8, v16i8);
15093 v8i16 __builtin_msa_ilvl_h (v8i16, v8i16);
15094 v4i32 __builtin_msa_ilvl_w (v4i32, v4i32);
15095 v2i64 __builtin_msa_ilvl_d (v2i64, v2i64);
15097 v16i8 __builtin_msa_ilvod_b (v16i8, v16i8);
15098 v8i16 __builtin_msa_ilvod_h (v8i16, v8i16);
15099 v4i32 __builtin_msa_ilvod_w (v4i32, v4i32);
15100 v2i64 __builtin_msa_ilvod_d (v2i64, v2i64);
15102 v16i8 __builtin_msa_ilvr_b (v16i8, v16i8);
15103 v8i16 __builtin_msa_ilvr_h (v8i16, v8i16);
15104 v4i32 __builtin_msa_ilvr_w (v4i32, v4i32);
15105 v2i64 __builtin_msa_ilvr_d (v2i64, v2i64);
15107 v16i8 __builtin_msa_insert_b (v16i8, imm0_15, i32);
15108 v8i16 __builtin_msa_insert_h (v8i16, imm0_7, i32);
15109 v4i32 __builtin_msa_insert_w (v4i32, imm0_3, i32);
15110 v2i64 __builtin_msa_insert_d (v2i64, imm0_1, i64);
15112 v16i8 __builtin_msa_insve_b (v16i8, imm0_15, v16i8);
15113 v8i16 __builtin_msa_insve_h (v8i16, imm0_7, v8i16);
15114 v4i32 __builtin_msa_insve_w (v4i32, imm0_3, v4i32);
15115 v2i64 __builtin_msa_insve_d (v2i64, imm0_1, v2i64);
15117 v16i8 __builtin_msa_ld_b (void *, imm_n512_511);
15118 v8i16 __builtin_msa_ld_h (void *, imm_n1024_1022);
15119 v4i32 __builtin_msa_ld_w (void *, imm_n2048_2044);
15120 v2i64 __builtin_msa_ld_d (void *, imm_n4096_4088);
15122 v16i8 __builtin_msa_ldi_b (imm_n512_511);
15123 v8i16 __builtin_msa_ldi_h (imm_n512_511);
15124 v4i32 __builtin_msa_ldi_w (imm_n512_511);
15125 v2i64 __builtin_msa_ldi_d (imm_n512_511);
15127 v8i16 __builtin_msa_madd_q_h (v8i16, v8i16, v8i16);
15128 v4i32 __builtin_msa_madd_q_w (v4i32, v4i32, v4i32);
15130 v8i16 __builtin_msa_maddr_q_h (v8i16, v8i16, v8i16);
15131 v4i32 __builtin_msa_maddr_q_w (v4i32, v4i32, v4i32);
15133 v16i8 __builtin_msa_maddv_b (v16i8, v16i8, v16i8);
15134 v8i16 __builtin_msa_maddv_h (v8i16, v8i16, v8i16);
15135 v4i32 __builtin_msa_maddv_w (v4i32, v4i32, v4i32);
15136 v2i64 __builtin_msa_maddv_d (v2i64, v2i64, v2i64);
15138 v16i8 __builtin_msa_max_a_b (v16i8, v16i8);
15139 v8i16 __builtin_msa_max_a_h (v8i16, v8i16);
15140 v4i32 __builtin_msa_max_a_w (v4i32, v4i32);
15141 v2i64 __builtin_msa_max_a_d (v2i64, v2i64);
15143 v16i8 __builtin_msa_max_s_b (v16i8, v16i8);
15144 v8i16 __builtin_msa_max_s_h (v8i16, v8i16);
15145 v4i32 __builtin_msa_max_s_w (v4i32, v4i32);
15146 v2i64 __builtin_msa_max_s_d (v2i64, v2i64);
15148 v16u8 __builtin_msa_max_u_b (v16u8, v16u8);
15149 v8u16 __builtin_msa_max_u_h (v8u16, v8u16);
15150 v4u32 __builtin_msa_max_u_w (v4u32, v4u32);
15151 v2u64 __builtin_msa_max_u_d (v2u64, v2u64);
15153 v16i8 __builtin_msa_maxi_s_b (v16i8, imm_n16_15);
15154 v8i16 __builtin_msa_maxi_s_h (v8i16, imm_n16_15);
15155 v4i32 __builtin_msa_maxi_s_w (v4i32, imm_n16_15);
15156 v2i64 __builtin_msa_maxi_s_d (v2i64, imm_n16_15);
15158 v16u8 __builtin_msa_maxi_u_b (v16u8, imm0_31);
15159 v8u16 __builtin_msa_maxi_u_h (v8u16, imm0_31);
15160 v4u32 __builtin_msa_maxi_u_w (v4u32, imm0_31);
15161 v2u64 __builtin_msa_maxi_u_d (v2u64, imm0_31);
15163 v16i8 __builtin_msa_min_a_b (v16i8, v16i8);
15164 v8i16 __builtin_msa_min_a_h (v8i16, v8i16);
15165 v4i32 __builtin_msa_min_a_w (v4i32, v4i32);
15166 v2i64 __builtin_msa_min_a_d (v2i64, v2i64);
15168 v16i8 __builtin_msa_min_s_b (v16i8, v16i8);
15169 v8i16 __builtin_msa_min_s_h (v8i16, v8i16);
15170 v4i32 __builtin_msa_min_s_w (v4i32, v4i32);
15171 v2i64 __builtin_msa_min_s_d (v2i64, v2i64);
15173 v16u8 __builtin_msa_min_u_b (v16u8, v16u8);
15174 v8u16 __builtin_msa_min_u_h (v8u16, v8u16);
15175 v4u32 __builtin_msa_min_u_w (v4u32, v4u32);
15176 v2u64 __builtin_msa_min_u_d (v2u64, v2u64);
15178 v16i8 __builtin_msa_mini_s_b (v16i8, imm_n16_15);
15179 v8i16 __builtin_msa_mini_s_h (v8i16, imm_n16_15);
15180 v4i32 __builtin_msa_mini_s_w (v4i32, imm_n16_15);
15181 v2i64 __builtin_msa_mini_s_d (v2i64, imm_n16_15);
15183 v16u8 __builtin_msa_mini_u_b (v16u8, imm0_31);
15184 v8u16 __builtin_msa_mini_u_h (v8u16, imm0_31);
15185 v4u32 __builtin_msa_mini_u_w (v4u32, imm0_31);
15186 v2u64 __builtin_msa_mini_u_d (v2u64, imm0_31);
15188 v16i8 __builtin_msa_mod_s_b (v16i8, v16i8);
15189 v8i16 __builtin_msa_mod_s_h (v8i16, v8i16);
15190 v4i32 __builtin_msa_mod_s_w (v4i32, v4i32);
15191 v2i64 __builtin_msa_mod_s_d (v2i64, v2i64);
15193 v16u8 __builtin_msa_mod_u_b (v16u8, v16u8);
15194 v8u16 __builtin_msa_mod_u_h (v8u16, v8u16);
15195 v4u32 __builtin_msa_mod_u_w (v4u32, v4u32);
15196 v2u64 __builtin_msa_mod_u_d (v2u64, v2u64);
15198 v16i8 __builtin_msa_move_v (v16i8);
15200 v8i16 __builtin_msa_msub_q_h (v8i16, v8i16, v8i16);
15201 v4i32 __builtin_msa_msub_q_w (v4i32, v4i32, v4i32);
15203 v8i16 __builtin_msa_msubr_q_h (v8i16, v8i16, v8i16);
15204 v4i32 __builtin_msa_msubr_q_w (v4i32, v4i32, v4i32);
15206 v16i8 __builtin_msa_msubv_b (v16i8, v16i8, v16i8);
15207 v8i16 __builtin_msa_msubv_h (v8i16, v8i16, v8i16);
15208 v4i32 __builtin_msa_msubv_w (v4i32, v4i32, v4i32);
15209 v2i64 __builtin_msa_msubv_d (v2i64, v2i64, v2i64);
15211 v8i16 __builtin_msa_mul_q_h (v8i16, v8i16);
15212 v4i32 __builtin_msa_mul_q_w (v4i32, v4i32);
15214 v8i16 __builtin_msa_mulr_q_h (v8i16, v8i16);
15215 v4i32 __builtin_msa_mulr_q_w (v4i32, v4i32);
15217 v16i8 __builtin_msa_mulv_b (v16i8, v16i8);
15218 v8i16 __builtin_msa_mulv_h (v8i16, v8i16);
15219 v4i32 __builtin_msa_mulv_w (v4i32, v4i32);
15220 v2i64 __builtin_msa_mulv_d (v2i64, v2i64);
15222 v16i8 __builtin_msa_nloc_b (v16i8);
15223 v8i16 __builtin_msa_nloc_h (v8i16);
15224 v4i32 __builtin_msa_nloc_w (v4i32);
15225 v2i64 __builtin_msa_nloc_d (v2i64);
15227 v16i8 __builtin_msa_nlzc_b (v16i8);
15228 v8i16 __builtin_msa_nlzc_h (v8i16);
15229 v4i32 __builtin_msa_nlzc_w (v4i32);
15230 v2i64 __builtin_msa_nlzc_d (v2i64);
15232 v16u8 __builtin_msa_nor_v (v16u8, v16u8);
15234 v16u8 __builtin_msa_nori_b (v16u8, imm0_255);
15236 v16u8 __builtin_msa_or_v (v16u8, v16u8);
15238 v16u8 __builtin_msa_ori_b (v16u8, imm0_255);
15240 v16i8 __builtin_msa_pckev_b (v16i8, v16i8);
15241 v8i16 __builtin_msa_pckev_h (v8i16, v8i16);
15242 v4i32 __builtin_msa_pckev_w (v4i32, v4i32);
15243 v2i64 __builtin_msa_pckev_d (v2i64, v2i64);
15245 v16i8 __builtin_msa_pckod_b (v16i8, v16i8);
15246 v8i16 __builtin_msa_pckod_h (v8i16, v8i16);
15247 v4i32 __builtin_msa_pckod_w (v4i32, v4i32);
15248 v2i64 __builtin_msa_pckod_d (v2i64, v2i64);
15250 v16i8 __builtin_msa_pcnt_b (v16i8);
15251 v8i16 __builtin_msa_pcnt_h (v8i16);
15252 v4i32 __builtin_msa_pcnt_w (v4i32);
15253 v2i64 __builtin_msa_pcnt_d (v2i64);
15255 v16i8 __builtin_msa_sat_s_b (v16i8, imm0_7);
15256 v8i16 __builtin_msa_sat_s_h (v8i16, imm0_15);
15257 v4i32 __builtin_msa_sat_s_w (v4i32, imm0_31);
15258 v2i64 __builtin_msa_sat_s_d (v2i64, imm0_63);
15260 v16u8 __builtin_msa_sat_u_b (v16u8, imm0_7);
15261 v8u16 __builtin_msa_sat_u_h (v8u16, imm0_15);
15262 v4u32 __builtin_msa_sat_u_w (v4u32, imm0_31);
15263 v2u64 __builtin_msa_sat_u_d (v2u64, imm0_63);
15265 v16i8 __builtin_msa_shf_b (v16i8, imm0_255);
15266 v8i16 __builtin_msa_shf_h (v8i16, imm0_255);
15267 v4i32 __builtin_msa_shf_w (v4i32, imm0_255);
15269 v16i8 __builtin_msa_sld_b (v16i8, v16i8, i32);
15270 v8i16 __builtin_msa_sld_h (v8i16, v8i16, i32);
15271 v4i32 __builtin_msa_sld_w (v4i32, v4i32, i32);
15272 v2i64 __builtin_msa_sld_d (v2i64, v2i64, i32);
15274 v16i8 __builtin_msa_sldi_b (v16i8, v16i8, imm0_15);
15275 v8i16 __builtin_msa_sldi_h (v8i16, v8i16, imm0_7);
15276 v4i32 __builtin_msa_sldi_w (v4i32, v4i32, imm0_3);
15277 v2i64 __builtin_msa_sldi_d (v2i64, v2i64, imm0_1);
15279 v16i8 __builtin_msa_sll_b (v16i8, v16i8);
15280 v8i16 __builtin_msa_sll_h (v8i16, v8i16);
15281 v4i32 __builtin_msa_sll_w (v4i32, v4i32);
15282 v2i64 __builtin_msa_sll_d (v2i64, v2i64);
15284 v16i8 __builtin_msa_slli_b (v16i8, imm0_7);
15285 v8i16 __builtin_msa_slli_h (v8i16, imm0_15);
15286 v4i32 __builtin_msa_slli_w (v4i32, imm0_31);
15287 v2i64 __builtin_msa_slli_d (v2i64, imm0_63);
15289 v16i8 __builtin_msa_splat_b (v16i8, i32);
15290 v8i16 __builtin_msa_splat_h (v8i16, i32);
15291 v4i32 __builtin_msa_splat_w (v4i32, i32);
15292 v2i64 __builtin_msa_splat_d (v2i64, i32);
15294 v16i8 __builtin_msa_splati_b (v16i8, imm0_15);
15295 v8i16 __builtin_msa_splati_h (v8i16, imm0_7);
15296 v4i32 __builtin_msa_splati_w (v4i32, imm0_3);
15297 v2i64 __builtin_msa_splati_d (v2i64, imm0_1);
15299 v16i8 __builtin_msa_sra_b (v16i8, v16i8);
15300 v8i16 __builtin_msa_sra_h (v8i16, v8i16);
15301 v4i32 __builtin_msa_sra_w (v4i32, v4i32);
15302 v2i64 __builtin_msa_sra_d (v2i64, v2i64);
15304 v16i8 __builtin_msa_srai_b (v16i8, imm0_7);
15305 v8i16 __builtin_msa_srai_h (v8i16, imm0_15);
15306 v4i32 __builtin_msa_srai_w (v4i32, imm0_31);
15307 v2i64 __builtin_msa_srai_d (v2i64, imm0_63);
15309 v16i8 __builtin_msa_srar_b (v16i8, v16i8);
15310 v8i16 __builtin_msa_srar_h (v8i16, v8i16);
15311 v4i32 __builtin_msa_srar_w (v4i32, v4i32);
15312 v2i64 __builtin_msa_srar_d (v2i64, v2i64);
15314 v16i8 __builtin_msa_srari_b (v16i8, imm0_7);
15315 v8i16 __builtin_msa_srari_h (v8i16, imm0_15);
15316 v4i32 __builtin_msa_srari_w (v4i32, imm0_31);
15317 v2i64 __builtin_msa_srari_d (v2i64, imm0_63);
15319 v16i8 __builtin_msa_srl_b (v16i8, v16i8);
15320 v8i16 __builtin_msa_srl_h (v8i16, v8i16);
15321 v4i32 __builtin_msa_srl_w (v4i32, v4i32);
15322 v2i64 __builtin_msa_srl_d (v2i64, v2i64);
15324 v16i8 __builtin_msa_srli_b (v16i8, imm0_7);
15325 v8i16 __builtin_msa_srli_h (v8i16, imm0_15);
15326 v4i32 __builtin_msa_srli_w (v4i32, imm0_31);
15327 v2i64 __builtin_msa_srli_d (v2i64, imm0_63);
15329 v16i8 __builtin_msa_srlr_b (v16i8, v16i8);
15330 v8i16 __builtin_msa_srlr_h (v8i16, v8i16);
15331 v4i32 __builtin_msa_srlr_w (v4i32, v4i32);
15332 v2i64 __builtin_msa_srlr_d (v2i64, v2i64);
15334 v16i8 __builtin_msa_srlri_b (v16i8, imm0_7);
15335 v8i16 __builtin_msa_srlri_h (v8i16, imm0_15);
15336 v4i32 __builtin_msa_srlri_w (v4i32, imm0_31);
15337 v2i64 __builtin_msa_srlri_d (v2i64, imm0_63);
15339 void __builtin_msa_st_b (v16i8, void *, imm_n512_511);
15340 void __builtin_msa_st_h (v8i16, void *, imm_n1024_1022);
15341 void __builtin_msa_st_w (v4i32, void *, imm_n2048_2044);
15342 void __builtin_msa_st_d (v2i64, void *, imm_n4096_4088);
15344 v16i8 __builtin_msa_subs_s_b (v16i8, v16i8);
15345 v8i16 __builtin_msa_subs_s_h (v8i16, v8i16);
15346 v4i32 __builtin_msa_subs_s_w (v4i32, v4i32);
15347 v2i64 __builtin_msa_subs_s_d (v2i64, v2i64);
15349 v16u8 __builtin_msa_subs_u_b (v16u8, v16u8);
15350 v8u16 __builtin_msa_subs_u_h (v8u16, v8u16);
15351 v4u32 __builtin_msa_subs_u_w (v4u32, v4u32);
15352 v2u64 __builtin_msa_subs_u_d (v2u64, v2u64);
15354 v16u8 __builtin_msa_subsus_u_b (v16u8, v16i8);
15355 v8u16 __builtin_msa_subsus_u_h (v8u16, v8i16);
15356 v4u32 __builtin_msa_subsus_u_w (v4u32, v4i32);
15357 v2u64 __builtin_msa_subsus_u_d (v2u64, v2i64);
15359 v16i8 __builtin_msa_subsuu_s_b (v16u8, v16u8);
15360 v8i16 __builtin_msa_subsuu_s_h (v8u16, v8u16);
15361 v4i32 __builtin_msa_subsuu_s_w (v4u32, v4u32);
15362 v2i64 __builtin_msa_subsuu_s_d (v2u64, v2u64);
15364 v16i8 __builtin_msa_subv_b (v16i8, v16i8);
15365 v8i16 __builtin_msa_subv_h (v8i16, v8i16);
15366 v4i32 __builtin_msa_subv_w (v4i32, v4i32);
15367 v2i64 __builtin_msa_subv_d (v2i64, v2i64);
15369 v16i8 __builtin_msa_subvi_b (v16i8, imm0_31);
15370 v8i16 __builtin_msa_subvi_h (v8i16, imm0_31);
15371 v4i32 __builtin_msa_subvi_w (v4i32, imm0_31);
15372 v2i64 __builtin_msa_subvi_d (v2i64, imm0_31);
15374 v16i8 __builtin_msa_vshf_b (v16i8, v16i8, v16i8);
15375 v8i16 __builtin_msa_vshf_h (v8i16, v8i16, v8i16);
15376 v4i32 __builtin_msa_vshf_w (v4i32, v4i32, v4i32);
15377 v2i64 __builtin_msa_vshf_d (v2i64, v2i64, v2i64);
15379 v16u8 __builtin_msa_xor_v (v16u8, v16u8);
15381 v16u8 __builtin_msa_xori_b (v16u8, imm0_255);
15384 @node Other MIPS Built-in Functions
15385 @subsection Other MIPS Built-in Functions
15387 GCC provides other MIPS-specific built-in functions:
15390 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
15391 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
15392 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
15393 when this function is available.
15395 @item unsigned int __builtin_mips_get_fcsr (void)
15396 @itemx void __builtin_mips_set_fcsr (unsigned int @var{value})
15397 Get and set the contents of the floating-point control and status register
15398 (FPU control register 31). These functions are only available in hard-float
15399 code but can be called in both MIPS16 and non-MIPS16 contexts.
15401 @code{__builtin_mips_set_fcsr} can be used to change any bit of the
15402 register except the condition codes, which GCC assumes are preserved.
15405 @node MSP430 Built-in Functions
15406 @subsection MSP430 Built-in Functions
15408 GCC provides a couple of special builtin functions to aid in the
15409 writing of interrupt handlers in C.
15412 @item __bic_SR_register_on_exit (int @var{mask})
15413 This clears the indicated bits in the saved copy of the status register
15414 currently residing on the stack. This only works inside interrupt
15415 handlers and the changes to the status register will only take affect
15416 once the handler returns.
15418 @item __bis_SR_register_on_exit (int @var{mask})
15419 This sets the indicated bits in the saved copy of the status register
15420 currently residing on the stack. This only works inside interrupt
15421 handlers and the changes to the status register will only take affect
15422 once the handler returns.
15424 @item __delay_cycles (long long @var{cycles})
15425 This inserts an instruction sequence that takes exactly @var{cycles}
15426 cycles (between 0 and about 17E9) to complete. The inserted sequence
15427 may use jumps, loops, or no-ops, and does not interfere with any other
15428 instructions. Note that @var{cycles} must be a compile-time constant
15429 integer - that is, you must pass a number, not a variable that may be
15430 optimized to a constant later. The number of cycles delayed by this
15434 @node NDS32 Built-in Functions
15435 @subsection NDS32 Built-in Functions
15437 These built-in functions are available for the NDS32 target:
15439 @deftypefn {Built-in Function} void __builtin_nds32_isync (int *@var{addr})
15440 Insert an ISYNC instruction into the instruction stream where
15441 @var{addr} is an instruction address for serialization.
15444 @deftypefn {Built-in Function} void __builtin_nds32_isb (void)
15445 Insert an ISB instruction into the instruction stream.
15448 @deftypefn {Built-in Function} int __builtin_nds32_mfsr (int @var{sr})
15449 Return the content of a system register which is mapped by @var{sr}.
15452 @deftypefn {Built-in Function} int __builtin_nds32_mfusr (int @var{usr})
15453 Return the content of a user space register which is mapped by @var{usr}.
15456 @deftypefn {Built-in Function} void __builtin_nds32_mtsr (int @var{value}, int @var{sr})
15457 Move the @var{value} to a system register which is mapped by @var{sr}.
15460 @deftypefn {Built-in Function} void __builtin_nds32_mtusr (int @var{value}, int @var{usr})
15461 Move the @var{value} to a user space register which is mapped by @var{usr}.
15464 @deftypefn {Built-in Function} void __builtin_nds32_setgie_en (void)
15465 Enable global interrupt.
15468 @deftypefn {Built-in Function} void __builtin_nds32_setgie_dis (void)
15469 Disable global interrupt.
15472 @node picoChip Built-in Functions
15473 @subsection picoChip Built-in Functions
15475 GCC provides an interface to selected machine instructions from the
15476 picoChip instruction set.
15479 @item int __builtin_sbc (int @var{value})
15480 Sign bit count. Return the number of consecutive bits in @var{value}
15481 that have the same value as the sign bit. The result is the number of
15482 leading sign bits minus one, giving the number of redundant sign bits in
15485 @item int __builtin_byteswap (int @var{value})
15486 Byte swap. Return the result of swapping the upper and lower bytes of
15489 @item int __builtin_brev (int @var{value})
15490 Bit reversal. Return the result of reversing the bits in
15491 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
15494 @item int __builtin_adds (int @var{x}, int @var{y})
15495 Saturating addition. Return the result of adding @var{x} and @var{y},
15496 storing the value 32767 if the result overflows.
15498 @item int __builtin_subs (int @var{x}, int @var{y})
15499 Saturating subtraction. Return the result of subtracting @var{y} from
15500 @var{x}, storing the value @minus{}32768 if the result overflows.
15502 @item void __builtin_halt (void)
15503 Halt. The processor stops execution. This built-in is useful for
15504 implementing assertions.
15508 @node PowerPC Built-in Functions
15509 @subsection PowerPC Built-in Functions
15511 The following built-in functions are always available and can be used to
15512 check the PowerPC target platform type:
15514 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
15515 This function is a @code{nop} on the PowerPC platform and is included solely
15516 to maintain API compatibility with the x86 builtins.
15519 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
15520 This function returns a value of @code{1} if the run-time CPU is of type
15521 @var{cpuname} and returns @code{0} otherwise
15523 The @code{__builtin_cpu_is} function requires GLIBC 2.23 or newer
15524 which exports the hardware capability bits. GCC defines the macro
15525 @code{__BUILTIN_CPU_SUPPORTS__} if the @code{__builtin_cpu_supports}
15526 built-in function is fully supported.
15528 If GCC was configured to use a GLIBC before 2.23, the built-in
15529 function @code{__builtin_cpu_is} always returns a 0 and the compiler
15532 The following CPU names can be detected:
15536 IBM POWER9 Server CPU.
15538 IBM POWER8 Server CPU.
15540 IBM POWER7 Server CPU.
15542 IBM POWER6 Server CPU (RAW mode).
15544 IBM POWER6 Server CPU (Architected mode).
15546 IBM POWER5+ Server CPU.
15548 IBM POWER5 Server CPU.
15550 IBM 970 Server CPU (ie, Apple G5).
15552 IBM POWER4 Server CPU.
15554 IBM A2 64-bit Embedded CPU
15556 IBM PowerPC 476FP 32-bit Embedded CPU.
15558 IBM PowerPC 464 32-bit Embedded CPU.
15560 PowerPC 440 32-bit Embedded CPU.
15562 PowerPC 405 32-bit Embedded CPU.
15564 IBM PowerPC Cell Broadband Engine Architecture CPU.
15567 Here is an example:
15569 #ifdef __BUILTIN_CPU_SUPPORTS__
15570 if (__builtin_cpu_is ("power8"))
15572 do_power8 (); // POWER8 specific implementation.
15577 do_generic (); // Generic implementation.
15582 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
15583 This function returns a value of @code{1} if the run-time CPU supports the HWCAP
15584 feature @var{feature} and returns @code{0} otherwise.
15586 The @code{__builtin_cpu_supports} function requires GLIBC 2.23 or
15587 newer which exports the hardware capability bits. GCC defines the
15588 macro @code{__BUILTIN_CPU_SUPPORTS__} if the
15589 @code{__builtin_cpu_supports} built-in function is fully supported.
15591 If GCC was configured to use a GLIBC before 2.23, the built-in
15592 function @code{__builtin_cpu_suports} always returns a 0 and the
15593 compiler issues a warning.
15595 The following features can be
15600 4xx CPU has a Multiply Accumulator.
15602 CPU has a SIMD/Vector Unit.
15604 CPU supports ISA 2.05 (eg, POWER6)
15606 CPU supports ISA 2.06 (eg, POWER7)
15608 CPU supports ISA 2.07 (eg, POWER8)
15610 CPU supports ISA 3.0 (eg, POWER9)
15612 CPU supports the set of compatible performance monitoring events.
15614 CPU supports the Embedded ISA category.
15616 CPU has a CELL broadband engine.
15618 CPU has a decimal floating point unit.
15620 CPU supports the data stream control register.
15622 CPU supports event base branching.
15624 CPU has a SPE double precision floating point unit.
15626 CPU has a SPE single precision floating point unit.
15628 CPU has a floating point unit.
15630 CPU has hardware transaction memory instructions.
15632 Kernel aborts hardware transactions when a syscall is made.
15634 CPU supports icache snooping capabilities.
15636 CPU supports 128-bit IEEE binary floating point instructions.
15638 CPU supports the integer select instruction.
15640 CPU has a memory management unit.
15642 CPU does not have a timebase (eg, 601 and 403gx).
15644 CPU supports the PA Semi 6T CORE ISA.
15646 CPU supports ISA 2.00 (eg, POWER4)
15648 CPU supports ISA 2.02 (eg, POWER5)
15650 CPU supports ISA 2.03 (eg, POWER5+)
15652 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
15654 CPU supports 32-bit mode execution.
15656 CPU supports the old POWER ISA (eg, 601)
15658 CPU supports 64-bit mode execution.
15660 CPU supports a little-endian mode that uses address swizzling.
15662 CPU support simultaneous multi-threading.
15664 CPU has a signal processing extension unit.
15666 CPU supports the target address register.
15668 CPU supports true little-endian mode.
15670 CPU has unified I/D cache.
15672 CPU supports the vector cryptography instructions.
15674 CPU supports the vector-scalar extension.
15677 Here is an example:
15679 #ifdef __BUILTIN_CPU_SUPPORTS__
15680 if (__builtin_cpu_supports ("fpu"))
15682 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
15687 dst = __fadd (src1, src2); // Software FP addition function.
15692 These built-in functions are available for the PowerPC family of
15695 float __builtin_recipdivf (float, float);
15696 float __builtin_rsqrtf (float);
15697 double __builtin_recipdiv (double, double);
15698 double __builtin_rsqrt (double);
15699 uint64_t __builtin_ppc_get_timebase ();
15700 unsigned long __builtin_ppc_mftb ();
15701 double __builtin_unpack_longdouble (long double, int);
15702 long double __builtin_pack_longdouble (double, double);
15705 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
15706 @code{__builtin_rsqrtf} functions generate multiple instructions to
15707 implement the reciprocal sqrt functionality using reciprocal sqrt
15708 estimate instructions.
15710 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
15711 functions generate multiple instructions to implement division using
15712 the reciprocal estimate instructions.
15714 The @code{__builtin_ppc_get_timebase} and @code{__builtin_ppc_mftb}
15715 functions generate instructions to read the Time Base Register. The
15716 @code{__builtin_ppc_get_timebase} function may generate multiple
15717 instructions and always returns the 64 bits of the Time Base Register.
15718 The @code{__builtin_ppc_mftb} function always generates one instruction and
15719 returns the Time Base Register value as an unsigned long, throwing away
15720 the most significant word on 32-bit environments.
15722 Additional built-in functions are available for the 64-bit PowerPC
15723 family of processors, for efficient use of 128-bit floating point
15724 (@code{__float128}) values.
15726 Previous versions of GCC supported some 'q' builtins for IEEE 128-bit
15727 floating point. These functions are now mapped into the equivalent
15728 'f128' builtin functions.
15731 __builtin_fabsq is mapped into __builtin_fabsf128
15732 __builtin_copysignq is mapped into __builtin_copysignf128
15733 __builtin_infq is mapped into __builtin_inff128
15734 __builtin_huge_valq is mapped into __builtin_huge_valf128
15735 __builtin_nanq is mapped into __builtin_nanf128
15736 __builtin_nansq is mapped into __builtin_nansf128
15739 The following built-in functions are available on Linux 64-bit systems
15740 that use the ISA 3.0 instruction set.
15743 @item __float128 __builtin_sqrtf128 (__float128)
15744 Perform a 128-bit IEEE floating point square root operation.
15745 @findex __builtin_sqrtf128
15747 @item __float128 __builtin_fmaf128 (__float128, __float128, __float128)
15748 Perform a 128-bit IEEE floating point fused multiply and add operation.
15749 @findex __builtin_fmaf128
15751 @item __float128 __builtin_addf128_round_to_odd (__float128, __float128)
15752 Perform a 128-bit IEEE floating point add using round to odd as the
15754 @findex __builtin_addf128_round_to_odd
15756 @item __float128 __builtin_subf128_round_to_odd (__float128, __float128)
15757 Perform a 128-bit IEEE floating point subtract using round to odd as
15759 @findex __builtin_subf128_round_to_odd
15761 @item __float128 __builtin_mulf128_round_to_odd (__float128, __float128)
15762 Perform a 128-bit IEEE floating point multiply using round to odd as
15764 @findex __builtin_mulf128_round_to_odd
15766 @item __float128 __builtin_divf128_round_to_odd (__float128, __float128)
15767 Perform a 128-bit IEEE floating point divide using round to odd as
15769 @findex __builtin_divf128_round_to_odd
15771 @item __float128 __builtin_sqrtf128_round_to_odd (__float128)
15772 Perform a 128-bit IEEE floating point square root using round to odd
15773 as the rounding mode.
15774 @findex __builtin_sqrtf128_round_to_odd
15776 @item __float128 __builtin_fmaf128 (__float128, __float128, __float128)
15777 Perform a 128-bit IEEE floating point fused multiply and add operation
15778 using round to odd as the rounding mode.
15779 @findex __builtin_fmaf128_round_to_odd
15781 @item double __builtin_truncf128_round_to_odd (__float128)
15782 Convert a 128-bit IEEE floating point value to @code{double} using
15783 round to odd as the rounding mode.
15784 @findex __builtin_truncf128_round_to_odd
15787 The following built-in functions are available for the PowerPC family
15788 of processors, starting with ISA 2.05 or later (@option{-mcpu=power6}
15789 or @option{-mcmpb}):
15791 unsigned long long __builtin_cmpb (unsigned long long int, unsigned long long int);
15792 unsigned int __builtin_cmpb (unsigned int, unsigned int);
15795 The @code{__builtin_cmpb} function
15796 performs a byte-wise compare on the contents of its two arguments,
15797 returning the result of the byte-wise comparison as the returned
15798 value. For each byte comparison, the corresponding byte of the return
15799 value holds 0xff if the input bytes are equal and 0 if the input bytes
15800 are not equal. If either of the arguments to this built-in function
15801 is wider than 32 bits, the function call expands into the form that
15802 expects @code{unsigned long long int} arguments
15803 which is only available on 64-bit targets.
15805 The following built-in functions are available for the PowerPC family
15806 of processors, starting with ISA 2.06 or later (@option{-mcpu=power7}
15807 or @option{-mpopcntd}):
15809 long __builtin_bpermd (long, long);
15810 int __builtin_divwe (int, int);
15811 int __builtin_divweo (int, int);
15812 unsigned int __builtin_divweu (unsigned int, unsigned int);
15813 unsigned int __builtin_divweuo (unsigned int, unsigned int);
15814 long __builtin_divde (long, long);
15815 long __builtin_divdeo (long, long);
15816 unsigned long __builtin_divdeu (unsigned long, unsigned long);
15817 unsigned long __builtin_divdeuo (unsigned long, unsigned long);
15818 unsigned int cdtbcd (unsigned int);
15819 unsigned int cbcdtd (unsigned int);
15820 unsigned int addg6s (unsigned int, unsigned int);
15821 void __builtin_rs6000_speculation_barrier (void);
15824 The @code{__builtin_divde}, @code{__builtin_divdeo},
15825 @code{__builtin_divdeu}, @code{__builtin_divdeou} functions require a
15826 64-bit environment support ISA 2.06 or later.
15828 The following built-in functions are available for the PowerPC family
15829 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
15831 long long __builtin_darn (void);
15832 long long __builtin_darn_raw (void);
15833 int __builtin_darn_32 (void);
15835 unsigned int scalar_extract_exp (double source);
15836 unsigned long long int scalar_extract_exp (__ieee128 source);
15838 unsigned long long int scalar_extract_sig (double source);
15839 unsigned __int128 scalar_extract_sig (__ieee128 source);
15842 scalar_insert_exp (unsigned long long int significand, unsigned long long int exponent);
15844 scalar_insert_exp (double significand, unsigned long long int exponent);
15847 scalar_insert_exp (unsigned __int128 significand, unsigned long long int exponent);
15849 scalar_insert_exp (ieee_128 significand, unsigned long long int exponent);
15851 int scalar_cmp_exp_gt (double arg1, double arg2);
15852 int scalar_cmp_exp_lt (double arg1, double arg2);
15853 int scalar_cmp_exp_eq (double arg1, double arg2);
15854 int scalar_cmp_exp_unordered (double arg1, double arg2);
15856 bool scalar_test_data_class (float source, const int condition);
15857 bool scalar_test_data_class (double source, const int condition);
15858 bool scalar_test_data_class (__ieee128 source, const int condition);
15860 bool scalar_test_neg (float source);
15861 bool scalar_test_neg (double source);
15862 bool scalar_test_neg (__ieee128 source);
15864 int __builtin_byte_in_set (unsigned char u, unsigned long long set);
15865 int __builtin_byte_in_range (unsigned char u, unsigned int range);
15866 int __builtin_byte_in_either_range (unsigned char u, unsigned int ranges);
15868 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal64 value);
15869 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal128 value);
15870 int __builtin_dfp_dtstsfi_lt_dd (unsigned int comparison, _Decimal64 value);
15871 int __builtin_dfp_dtstsfi_lt_td (unsigned int comparison, _Decimal128 value);
15873 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal64 value);
15874 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal128 value);
15875 int __builtin_dfp_dtstsfi_gt_dd (unsigned int comparison, _Decimal64 value);
15876 int __builtin_dfp_dtstsfi_gt_td (unsigned int comparison, _Decimal128 value);
15878 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal64 value);
15879 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal128 value);
15880 int __builtin_dfp_dtstsfi_eq_dd (unsigned int comparison, _Decimal64 value);
15881 int __builtin_dfp_dtstsfi_eq_td (unsigned int comparison, _Decimal128 value);
15883 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal64 value);
15884 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal128 value);
15885 int __builtin_dfp_dtstsfi_ov_dd (unsigned int comparison, _Decimal64 value);
15886 int __builtin_dfp_dtstsfi_ov_td (unsigned int comparison, _Decimal128 value);
15889 The @code{__builtin_darn} and @code{__builtin_darn_raw}
15890 functions require a
15891 64-bit environment supporting ISA 3.0 or later.
15892 The @code{__builtin_darn} function provides a 64-bit conditioned
15893 random number. The @code{__builtin_darn_raw} function provides a
15894 64-bit raw random number. The @code{__builtin_darn_32} function
15895 provides a 32-bit random number.
15897 The @code{scalar_extract_exp} and @code{scalar_extract_sig}
15898 functions require a 64-bit environment supporting ISA 3.0 or later.
15899 The @code{scalar_extract_exp} and @code{scalar_extract_sig} built-in
15900 functions return the significand and the biased exponent value
15901 respectively of their @code{source} arguments.
15902 When supplied with a 64-bit @code{source} argument, the
15903 result returned by @code{scalar_extract_sig} has
15904 the @code{0x0010000000000000} bit set if the
15905 function's @code{source} argument is in normalized form.
15906 Otherwise, this bit is set to 0.
15907 When supplied with a 128-bit @code{source} argument, the
15908 @code{0x00010000000000000000000000000000} bit of the result is
15910 Note that the sign of the significand is not represented in the result
15911 returned from the @code{scalar_extract_sig} function. Use the
15912 @code{scalar_test_neg} function to test the sign of its @code{double}
15915 The @code{scalar_insert_exp}
15916 functions require a 64-bit environment supporting ISA 3.0 or later.
15917 When supplied with a 64-bit first argument, the
15918 @code{scalar_insert_exp} built-in function returns a double-precision
15919 floating point value that is constructed by assembling the values of its
15920 @code{significand} and @code{exponent} arguments. The sign of the
15921 result is copied from the most significant bit of the
15922 @code{significand} argument. The significand and exponent components
15923 of the result are composed of the least significant 11 bits of the
15924 @code{exponent} argument and the least significant 52 bits of the
15925 @code{significand} argument respectively.
15927 When supplied with a 128-bit first argument, the
15928 @code{scalar_insert_exp} built-in function returns a quad-precision
15929 ieee floating point value. The sign bit of the result is copied from
15930 the most significant bit of the @code{significand} argument.
15931 The significand and exponent components of the result are composed of
15932 the least significant 15 bits of the @code{exponent} argument and the
15933 least significant 112 bits of the @code{significand} argument respectively.
15935 The @code{scalar_cmp_exp_gt}, @code{scalar_cmp_exp_lt},
15936 @code{scalar_cmp_exp_eq}, and @code{scalar_cmp_exp_unordered} built-in
15937 functions return a non-zero value if @code{arg1} is greater than, less
15938 than, equal to, or not comparable to @code{arg2} respectively. The
15939 arguments are not comparable if one or the other equals NaN (not a
15942 The @code{scalar_test_data_class} built-in function returns 1
15943 if any of the condition tests enabled by the value of the
15944 @code{condition} variable are true, and 0 otherwise. The
15945 @code{condition} argument must be a compile-time constant integer with
15946 value not exceeding 127. The
15947 @code{condition} argument is encoded as a bitmask with each bit
15948 enabling the testing of a different condition, as characterized by the
15952 0x20 Test for +Infinity
15953 0x10 Test for -Infinity
15954 0x08 Test for +Zero
15955 0x04 Test for -Zero
15956 0x02 Test for +Denormal
15957 0x01 Test for -Denormal
15960 The @code{scalar_test_neg} built-in function returns 1 if its
15961 @code{source} argument holds a negative value, 0 otherwise.
15963 The @code{__builtin_byte_in_set} function requires a
15964 64-bit environment supporting ISA 3.0 or later. This function returns
15965 a non-zero value if and only if its @code{u} argument exactly equals one of
15966 the eight bytes contained within its 64-bit @code{set} argument.
15968 The @code{__builtin_byte_in_range} and
15969 @code{__builtin_byte_in_either_range} require an environment
15970 supporting ISA 3.0 or later. For these two functions, the
15971 @code{range} argument is encoded as 4 bytes, organized as
15972 @code{hi_1:lo_1:hi_2:lo_2}.
15973 The @code{__builtin_byte_in_range} function returns a
15974 non-zero value if and only if its @code{u} argument is within the
15975 range bounded between @code{lo_2} and @code{hi_2} inclusive.
15976 The @code{__builtin_byte_in_either_range} function returns non-zero if
15977 and only if its @code{u} argument is within either the range bounded
15978 between @code{lo_1} and @code{hi_1} inclusive or the range bounded
15979 between @code{lo_2} and @code{hi_2} inclusive.
15981 The @code{__builtin_dfp_dtstsfi_lt} function returns a non-zero value
15982 if and only if the number of signficant digits of its @code{value} argument
15983 is less than its @code{comparison} argument. The
15984 @code{__builtin_dfp_dtstsfi_lt_dd} and
15985 @code{__builtin_dfp_dtstsfi_lt_td} functions behave similarly, but
15986 require that the type of the @code{value} argument be
15987 @code{__Decimal64} and @code{__Decimal128} respectively.
15989 The @code{__builtin_dfp_dtstsfi_gt} function returns a non-zero value
15990 if and only if the number of signficant digits of its @code{value} argument
15991 is greater than its @code{comparison} argument. The
15992 @code{__builtin_dfp_dtstsfi_gt_dd} and
15993 @code{__builtin_dfp_dtstsfi_gt_td} functions behave similarly, but
15994 require that the type of the @code{value} argument be
15995 @code{__Decimal64} and @code{__Decimal128} respectively.
15997 The @code{__builtin_dfp_dtstsfi_eq} function returns a non-zero value
15998 if and only if the number of signficant digits of its @code{value} argument
15999 equals its @code{comparison} argument. The
16000 @code{__builtin_dfp_dtstsfi_eq_dd} and
16001 @code{__builtin_dfp_dtstsfi_eq_td} functions behave similarly, but
16002 require that the type of the @code{value} argument be
16003 @code{__Decimal64} and @code{__Decimal128} respectively.
16005 The @code{__builtin_dfp_dtstsfi_ov} function returns a non-zero value
16006 if and only if its @code{value} argument has an undefined number of
16007 significant digits, such as when @code{value} is an encoding of @code{NaN}.
16008 The @code{__builtin_dfp_dtstsfi_ov_dd} and
16009 @code{__builtin_dfp_dtstsfi_ov_td} functions behave similarly, but
16010 require that the type of the @code{value} argument be
16011 @code{__Decimal64} and @code{__Decimal128} respectively.
16013 The following built-in functions are also available for the PowerPC family
16014 of processors, starting with ISA 3.0 or later
16015 (@option{-mcpu=power9}). These string functions are described
16016 separately in order to group the descriptions closer to the function
16019 int vec_all_nez (vector signed char, vector signed char);
16020 int vec_all_nez (vector unsigned char, vector unsigned char);
16021 int vec_all_nez (vector signed short, vector signed short);
16022 int vec_all_nez (vector unsigned short, vector unsigned short);
16023 int vec_all_nez (vector signed int, vector signed int);
16024 int vec_all_nez (vector unsigned int, vector unsigned int);
16026 int vec_any_eqz (vector signed char, vector signed char);
16027 int vec_any_eqz (vector unsigned char, vector unsigned char);
16028 int vec_any_eqz (vector signed short, vector signed short);
16029 int vec_any_eqz (vector unsigned short, vector unsigned short);
16030 int vec_any_eqz (vector signed int, vector signed int);
16031 int vec_any_eqz (vector unsigned int, vector unsigned int);
16033 vector bool char vec_cmpnez (vector signed char arg1, vector signed char arg2);
16034 vector bool char vec_cmpnez (vector unsigned char arg1, vector unsigned char arg2);
16035 vector bool short vec_cmpnez (vector signed short arg1, vector signed short arg2);
16036 vector bool short vec_cmpnez (vector unsigned short arg1, vector unsigned short arg2);
16037 vector bool int vec_cmpnez (vector signed int arg1, vector signed int arg2);
16038 vector bool int vec_cmpnez (vector unsigned int, vector unsigned int);
16040 vector signed char vec_cnttz (vector signed char);
16041 vector unsigned char vec_cnttz (vector unsigned char);
16042 vector signed short vec_cnttz (vector signed short);
16043 vector unsigned short vec_cnttz (vector unsigned short);
16044 vector signed int vec_cnttz (vector signed int);
16045 vector unsigned int vec_cnttz (vector unsigned int);
16046 vector signed long long vec_cnttz (vector signed long long);
16047 vector unsigned long long vec_cnttz (vector unsigned long long);
16049 signed int vec_cntlz_lsbb (vector signed char);
16050 signed int vec_cntlz_lsbb (vector unsigned char);
16052 signed int vec_cnttz_lsbb (vector signed char);
16053 signed int vec_cnttz_lsbb (vector unsigned char);
16055 unsigned int vec_first_match_index (vector signed char, vector signed char);
16056 unsigned int vec_first_match_index (vector unsigned char,
16057 vector unsigned char);
16058 unsigned int vec_first_match_index (vector signed int, vector signed int);
16059 unsigned int vec_first_match_index (vector unsigned int, vector unsigned int);
16060 unsigned int vec_first_match_index (vector signed short, vector signed short);
16061 unsigned int vec_first_match_index (vector unsigned short,
16062 vector unsigned short);
16063 unsigned int vec_first_match_or_eos_index (vector signed char,
16064 vector signed char);
16065 unsigned int vec_first_match_or_eos_index (vector unsigned char,
16066 vector unsigned char);
16067 unsigned int vec_first_match_or_eos_index (vector signed int,
16068 vector signed int);
16069 unsigned int vec_first_match_or_eos_index (vector unsigned int,
16070 vector unsigned int);
16071 unsigned int vec_first_match_or_eos_index (vector signed short,
16072 vector signed short);
16073 unsigned int vec_first_match_or_eos_index (vector unsigned short,
16074 vector unsigned short);
16075 unsigned int vec_first_mismatch_index (vector signed char,
16076 vector signed char);
16077 unsigned int vec_first_mismatch_index (vector unsigned char,
16078 vector unsigned char);
16079 unsigned int vec_first_mismatch_index (vector signed int,
16080 vector signed int);
16081 unsigned int vec_first_mismatch_index (vector unsigned int,
16082 vector unsigned int);
16083 unsigned int vec_first_mismatch_index (vector signed short,
16084 vector signed short);
16085 unsigned int vec_first_mismatch_index (vector unsigned short,
16086 vector unsigned short);
16087 unsigned int vec_first_mismatch_or_eos_index (vector signed char,
16088 vector signed char);
16089 unsigned int vec_first_mismatch_or_eos_index (vector unsigned char,
16090 vector unsigned char);
16091 unsigned int vec_first_mismatch_or_eos_index (vector signed int,
16092 vector signed int);
16093 unsigned int vec_first_mismatch_or_eos_index (vector unsigned int,
16094 vector unsigned int);
16095 unsigned int vec_first_mismatch_or_eos_index (vector signed short,
16096 vector signed short);
16097 unsigned int vec_first_mismatch_or_eos_index (vector unsigned short,
16098 vector unsigned short);
16100 vector unsigned short vec_pack_to_short_fp32 (vector float, vector float);
16102 vector signed char vec_xl_be (signed long long, signed char *);
16103 vector unsigned char vec_xl_be (signed long long, unsigned char *);
16104 vector signed int vec_xl_be (signed long long, signed int *);
16105 vector unsigned int vec_xl_be (signed long long, unsigned int *);
16106 vector signed __int128 vec_xl_be (signed long long, signed __int128 *);
16107 vector unsigned __int128 vec_xl_be (signed long long, unsigned __int128 *);
16108 vector signed long long vec_xl_be (signed long long, signed long long *);
16109 vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
16110 vector signed short vec_xl_be (signed long long, signed short *);
16111 vector unsigned short vec_xl_be (signed long long, unsigned short *);
16112 vector double vec_xl_be (signed long long, double *);
16113 vector float vec_xl_be (signed long long, float *);
16115 vector signed char vec_xl_len (signed char *addr, size_t len);
16116 vector unsigned char vec_xl_len (unsigned char *addr, size_t len);
16117 vector signed int vec_xl_len (signed int *addr, size_t len);
16118 vector unsigned int vec_xl_len (unsigned int *addr, size_t len);
16119 vector signed __int128 vec_xl_len (signed __int128 *addr, size_t len);
16120 vector unsigned __int128 vec_xl_len (unsigned __int128 *addr, size_t len);
16121 vector signed long long vec_xl_len (signed long long *addr, size_t len);
16122 vector unsigned long long vec_xl_len (unsigned long long *addr, size_t len);
16123 vector signed short vec_xl_len (signed short *addr, size_t len);
16124 vector unsigned short vec_xl_len (unsigned short *addr, size_t len);
16125 vector double vec_xl_len (double *addr, size_t len);
16126 vector float vec_xl_len (float *addr, size_t len);
16128 vector unsigned char vec_xl_len_r (unsigned char *addr, size_t len);
16130 void vec_xst_len (vector signed char data, signed char *addr, size_t len);
16131 void vec_xst_len (vector unsigned char data, unsigned char *addr, size_t len);
16132 void vec_xst_len (vector signed int data, signed int *addr, size_t len);
16133 void vec_xst_len (vector unsigned int data, unsigned int *addr, size_t len);
16134 void vec_xst_len (vector unsigned __int128 data, unsigned __int128 *addr, size_t len);
16135 void vec_xst_len (vector signed long long data, signed long long *addr, size_t len);
16136 void vec_xst_len (vector unsigned long long data, unsigned long long *addr, size_t len);
16137 void vec_xst_len (vector signed short data, signed short *addr, size_t len);
16138 void vec_xst_len (vector unsigned short data, unsigned short *addr, size_t len);
16139 void vec_xst_len (vector signed __int128 data, signed __int128 *addr, size_t len);
16140 void vec_xst_len (vector double data, double *addr, size_t len);
16141 void vec_xst_len (vector float data, float *addr, size_t len);
16143 void vec_xst_len_r (vector unsigned char data, unsigned char *addr, size_t len);
16145 signed char vec_xlx (unsigned int index, vector signed char data);
16146 unsigned char vec_xlx (unsigned int index, vector unsigned char data);
16147 signed short vec_xlx (unsigned int index, vector signed short data);
16148 unsigned short vec_xlx (unsigned int index, vector unsigned short data);
16149 signed int vec_xlx (unsigned int index, vector signed int data);
16150 unsigned int vec_xlx (unsigned int index, vector unsigned int data);
16151 float vec_xlx (unsigned int index, vector float data);
16153 signed char vec_xrx (unsigned int index, vector signed char data);
16154 unsigned char vec_xrx (unsigned int index, vector unsigned char data);
16155 signed short vec_xrx (unsigned int index, vector signed short data);
16156 unsigned short vec_xrx (unsigned int index, vector unsigned short data);
16157 signed int vec_xrx (unsigned int index, vector signed int data);
16158 unsigned int vec_xrx (unsigned int index, vector unsigned int data);
16159 float vec_xrx (unsigned int index, vector float data);
16162 The @code{vec_all_nez}, @code{vec_any_eqz}, and @code{vec_cmpnez}
16163 perform pairwise comparisons between the elements at the same
16164 positions within their two vector arguments.
16165 The @code{vec_all_nez} function returns a
16166 non-zero value if and only if all pairwise comparisons are not
16167 equal and no element of either vector argument contains a zero.
16168 The @code{vec_any_eqz} function returns a
16169 non-zero value if and only if at least one pairwise comparison is equal
16170 or if at least one element of either vector argument contains a zero.
16171 The @code{vec_cmpnez} function returns a vector of the same type as
16172 its two arguments, within which each element consists of all ones to
16173 denote that either the corresponding elements of the incoming arguments are
16174 not equal or that at least one of the corresponding elements contains
16175 zero. Otherwise, the element of the returned vector contains all zeros.
16177 The @code{vec_cntlz_lsbb} function returns the count of the number of
16178 consecutive leading byte elements (starting from position 0 within the
16179 supplied vector argument) for which the least-significant bit
16180 equals zero. The @code{vec_cnttz_lsbb} function returns the count of
16181 the number of consecutive trailing byte elements (starting from
16182 position 15 and counting backwards within the supplied vector
16183 argument) for which the least-significant bit equals zero.
16185 The @code{vec_xl_len} and @code{vec_xst_len} functions require a
16186 64-bit environment supporting ISA 3.0 or later. The @code{vec_xl_len}
16187 function loads a variable length vector from memory. The
16188 @code{vec_xst_len} function stores a variable length vector to memory.
16189 With both the @code{vec_xl_len} and @code{vec_xst_len} functions, the
16190 @code{addr} argument represents the memory address to or from which
16191 data will be transferred, and the
16192 @code{len} argument represents the number of bytes to be
16193 transferred, as computed by the C expression @code{min((len & 0xff), 16)}.
16194 If this expression's value is not a multiple of the vector element's
16195 size, the behavior of this function is undefined.
16196 In the case that the underlying computer is configured to run in
16197 big-endian mode, the data transfer moves bytes 0 to @code{(len - 1)} of
16198 the corresponding vector. In little-endian mode, the data transfer
16199 moves bytes @code{(16 - len)} to @code{15} of the corresponding
16200 vector. For the load function, any bytes of the result vector that
16201 are not loaded from memory are set to zero.
16202 The value of the @code{addr} argument need not be aligned on a
16203 multiple of the vector's element size.
16205 The @code{vec_xlx} and @code{vec_xrx} functions extract the single
16206 element selected by the @code{index} argument from the vector
16207 represented by the @code{data} argument. The @code{index} argument
16208 always specifies a byte offset, regardless of the size of the vector
16209 element. With @code{vec_xlx}, @code{index} is the offset of the first
16210 byte of the element to be extracted. With @code{vec_xrx}, @code{index}
16211 represents the last byte of the element to be extracted, measured
16212 from the right end of the vector. In other words, the last byte of
16213 the element to be extracted is found at position @code{(15 - index)}.
16214 There is no requirement that @code{index} be a multiple of the vector
16215 element size. However, if the size of the vector element added to
16216 @code{index} is greater than 15, the content of the returned value is
16219 The following built-in functions are available for the PowerPC family
16220 of processors when hardware decimal floating point
16221 (@option{-mhard-dfp}) is available:
16223 long long __builtin_dxex (_Decimal64);
16224 long long __builtin_dxexq (_Decimal128);
16225 _Decimal64 __builtin_ddedpd (int, _Decimal64);
16226 _Decimal128 __builtin_ddedpdq (int, _Decimal128);
16227 _Decimal64 __builtin_denbcd (int, _Decimal64);
16228 _Decimal128 __builtin_denbcdq (int, _Decimal128);
16229 _Decimal64 __builtin_diex (long long, _Decimal64);
16230 _Decimal128 _builtin_diexq (long long, _Decimal128);
16231 _Decimal64 __builtin_dscli (_Decimal64, int);
16232 _Decimal128 __builtin_dscliq (_Decimal128, int);
16233 _Decimal64 __builtin_dscri (_Decimal64, int);
16234 _Decimal128 __builtin_dscriq (_Decimal128, int);
16235 unsigned long long __builtin_unpack_dec128 (_Decimal128, int);
16236 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long);
16239 The following built-in functions are available for the PowerPC family
16240 of processors when the Vector Scalar (vsx) instruction set is
16243 unsigned long long __builtin_unpack_vector_int128 (vector __int128_t, int);
16244 vector __int128_t __builtin_pack_vector_int128 (unsigned long long,
16245 unsigned long long);
16248 @node PowerPC AltiVec/VSX Built-in Functions
16249 @subsection PowerPC AltiVec Built-in Functions
16251 GCC provides an interface for the PowerPC family of processors to access
16252 the AltiVec operations described in Motorola's AltiVec Programming
16253 Interface Manual. The interface is made available by including
16254 @code{<altivec.h>} and using @option{-maltivec} and
16255 @option{-mabi=altivec}. The interface supports the following vector
16259 vector unsigned char
16263 vector unsigned short
16264 vector signed short
16268 vector unsigned int
16274 If @option{-mvsx} is used the following additional vector types are
16278 vector unsigned long
16283 The long types are only implemented for 64-bit code generation, and
16284 the long type is only used in the floating point/integer conversion
16287 GCC's implementation of the high-level language interface available from
16288 C and C++ code differs from Motorola's documentation in several ways.
16293 A vector constant is a list of constant expressions within curly braces.
16296 A vector initializer requires no cast if the vector constant is of the
16297 same type as the variable it is initializing.
16300 If @code{signed} or @code{unsigned} is omitted, the signedness of the
16301 vector type is the default signedness of the base type. The default
16302 varies depending on the operating system, so a portable program should
16303 always specify the signedness.
16306 Compiling with @option{-maltivec} adds keywords @code{__vector},
16307 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
16308 @code{bool}. When compiling ISO C, the context-sensitive substitution
16309 of the keywords @code{vector}, @code{pixel} and @code{bool} is
16310 disabled. To use them, you must include @code{<altivec.h>} instead.
16313 GCC allows using a @code{typedef} name as the type specifier for a
16317 For C, overloaded functions are implemented with macros so the following
16321 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
16325 Since @code{vec_add} is a macro, the vector constant in the example
16326 is treated as four separate arguments. Wrap the entire argument in
16327 parentheses for this to work.
16330 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
16331 Internally, GCC uses built-in functions to achieve the functionality in
16332 the aforementioned header file, but they are not supported and are
16333 subject to change without notice.
16335 GCC complies with the OpenPOWER 64-Bit ELF V2 ABI Specification,
16336 which may be found at
16337 @uref{http://openpowerfoundation.org/wp-content/uploads/resources/leabi-prd/content/index.html}.
16338 Appendix A of this document lists the vector API interfaces that must be
16339 provided by compliant compilers. Programmers should preferentially use
16340 the interfaces described therein. However, historically GCC has provided
16341 additional interfaces for access to vector instructions. These are
16342 briefly described below.
16344 The following interfaces are supported for the generic and specific
16345 AltiVec operations and the AltiVec predicates. In cases where there
16346 is a direct mapping between generic and specific operations, only the
16347 generic names are shown here, although the specific operations can also
16350 Arguments that are documented as @code{const int} require literal
16351 integral values within the range required for that operation.
16354 vector signed char vec_abs (vector signed char);
16355 vector signed short vec_abs (vector signed short);
16356 vector signed int vec_abs (vector signed int);
16357 vector float vec_abs (vector float);
16359 vector signed char vec_abss (vector signed char);
16360 vector signed short vec_abss (vector signed short);
16361 vector signed int vec_abss (vector signed int);
16363 vector signed char vec_add (vector bool char, vector signed char);
16364 vector signed char vec_add (vector signed char, vector bool char);
16365 vector signed char vec_add (vector signed char, vector signed char);
16366 vector unsigned char vec_add (vector bool char, vector unsigned char);
16367 vector unsigned char vec_add (vector unsigned char, vector bool char);
16368 vector unsigned char vec_add (vector unsigned char,
16369 vector unsigned char);
16370 vector signed short vec_add (vector bool short, vector signed short);
16371 vector signed short vec_add (vector signed short, vector bool short);
16372 vector signed short vec_add (vector signed short, vector signed short);
16373 vector unsigned short vec_add (vector bool short,
16374 vector unsigned short);
16375 vector unsigned short vec_add (vector unsigned short,
16376 vector bool short);
16377 vector unsigned short vec_add (vector unsigned short,
16378 vector unsigned short);
16379 vector signed int vec_add (vector bool int, vector signed int);
16380 vector signed int vec_add (vector signed int, vector bool int);
16381 vector signed int vec_add (vector signed int, vector signed int);
16382 vector unsigned int vec_add (vector bool int, vector unsigned int);
16383 vector unsigned int vec_add (vector unsigned int, vector bool int);
16384 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
16385 vector float vec_add (vector float, vector float);
16387 vector float vec_vaddfp (vector float, vector float);
16389 vector signed int vec_vadduwm (vector bool int, vector signed int);
16390 vector signed int vec_vadduwm (vector signed int, vector bool int);
16391 vector signed int vec_vadduwm (vector signed int, vector signed int);
16392 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
16393 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
16394 vector unsigned int vec_vadduwm (vector unsigned int,
16395 vector unsigned int);
16397 vector signed short vec_vadduhm (vector bool short,
16398 vector signed short);
16399 vector signed short vec_vadduhm (vector signed short,
16400 vector bool short);
16401 vector signed short vec_vadduhm (vector signed short,
16402 vector signed short);
16403 vector unsigned short vec_vadduhm (vector bool short,
16404 vector unsigned short);
16405 vector unsigned short vec_vadduhm (vector unsigned short,
16406 vector bool short);
16407 vector unsigned short vec_vadduhm (vector unsigned short,
16408 vector unsigned short);
16410 vector signed char vec_vaddubm (vector bool char, vector signed char);
16411 vector signed char vec_vaddubm (vector signed char, vector bool char);
16412 vector signed char vec_vaddubm (vector signed char, vector signed char);
16413 vector unsigned char vec_vaddubm (vector bool char,
16414 vector unsigned char);
16415 vector unsigned char vec_vaddubm (vector unsigned char,
16417 vector unsigned char vec_vaddubm (vector unsigned char,
16418 vector unsigned char);
16420 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
16422 vector unsigned char vec_adds (vector bool char, vector unsigned char);
16423 vector unsigned char vec_adds (vector unsigned char, vector bool char);
16424 vector unsigned char vec_adds (vector unsigned char,
16425 vector unsigned char);
16426 vector signed char vec_adds (vector bool char, vector signed char);
16427 vector signed char vec_adds (vector signed char, vector bool char);
16428 vector signed char vec_adds (vector signed char, vector signed char);
16429 vector unsigned short vec_adds (vector bool short,
16430 vector unsigned short);
16431 vector unsigned short vec_adds (vector unsigned short,
16432 vector bool short);
16433 vector unsigned short vec_adds (vector unsigned short,
16434 vector unsigned short);
16435 vector signed short vec_adds (vector bool short, vector signed short);
16436 vector signed short vec_adds (vector signed short, vector bool short);
16437 vector signed short vec_adds (vector signed short, vector signed short);
16438 vector unsigned int vec_adds (vector bool int, vector unsigned int);
16439 vector unsigned int vec_adds (vector unsigned int, vector bool int);
16440 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
16441 vector signed int vec_adds (vector bool int, vector signed int);
16442 vector signed int vec_adds (vector signed int, vector bool int);
16443 vector signed int vec_adds (vector signed int, vector signed int);
16445 vector signed int vec_vaddsws (vector bool int, vector signed int);
16446 vector signed int vec_vaddsws (vector signed int, vector bool int);
16447 vector signed int vec_vaddsws (vector signed int, vector signed int);
16449 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
16450 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
16451 vector unsigned int vec_vadduws (vector unsigned int,
16452 vector unsigned int);
16454 vector signed short vec_vaddshs (vector bool short,
16455 vector signed short);
16456 vector signed short vec_vaddshs (vector signed short,
16457 vector bool short);
16458 vector signed short vec_vaddshs (vector signed short,
16459 vector signed short);
16461 vector unsigned short vec_vadduhs (vector bool short,
16462 vector unsigned short);
16463 vector unsigned short vec_vadduhs (vector unsigned short,
16464 vector bool short);
16465 vector unsigned short vec_vadduhs (vector unsigned short,
16466 vector unsigned short);
16468 vector signed char vec_vaddsbs (vector bool char, vector signed char);
16469 vector signed char vec_vaddsbs (vector signed char, vector bool char);
16470 vector signed char vec_vaddsbs (vector signed char, vector signed char);
16472 vector unsigned char vec_vaddubs (vector bool char,
16473 vector unsigned char);
16474 vector unsigned char vec_vaddubs (vector unsigned char,
16476 vector unsigned char vec_vaddubs (vector unsigned char,
16477 vector unsigned char);
16479 vector float vec_and (vector float, vector float);
16480 vector float vec_and (vector float, vector bool int);
16481 vector float vec_and (vector bool int, vector float);
16482 vector bool long long vec_and (vector bool long long int,
16483 vector bool long long);
16484 vector bool int vec_and (vector bool int, vector bool int);
16485 vector signed int vec_and (vector bool int, vector signed int);
16486 vector signed int vec_and (vector signed int, vector bool int);
16487 vector signed int vec_and (vector signed int, vector signed int);
16488 vector unsigned int vec_and (vector bool int, vector unsigned int);
16489 vector unsigned int vec_and (vector unsigned int, vector bool int);
16490 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
16491 vector bool short vec_and (vector bool short, vector bool short);
16492 vector signed short vec_and (vector bool short, vector signed short);
16493 vector signed short vec_and (vector signed short, vector bool short);
16494 vector signed short vec_and (vector signed short, vector signed short);
16495 vector unsigned short vec_and (vector bool short,
16496 vector unsigned short);
16497 vector unsigned short vec_and (vector unsigned short,
16498 vector bool short);
16499 vector unsigned short vec_and (vector unsigned short,
16500 vector unsigned short);
16501 vector signed char vec_and (vector bool char, vector signed char);
16502 vector bool char vec_and (vector bool char, vector bool char);
16503 vector signed char vec_and (vector signed char, vector bool char);
16504 vector signed char vec_and (vector signed char, vector signed char);
16505 vector unsigned char vec_and (vector bool char, vector unsigned char);
16506 vector unsigned char vec_and (vector unsigned char, vector bool char);
16507 vector unsigned char vec_and (vector unsigned char,
16508 vector unsigned char);
16510 vector float vec_andc (vector float, vector float);
16511 vector float vec_andc (vector float, vector bool int);
16512 vector float vec_andc (vector bool int, vector float);
16513 vector bool int vec_andc (vector bool int, vector bool int);
16514 vector signed int vec_andc (vector bool int, vector signed int);
16515 vector signed int vec_andc (vector signed int, vector bool int);
16516 vector signed int vec_andc (vector signed int, vector signed int);
16517 vector unsigned int vec_andc (vector bool int, vector unsigned int);
16518 vector unsigned int vec_andc (vector unsigned int, vector bool int);
16519 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
16520 vector bool short vec_andc (vector bool short, vector bool short);
16521 vector signed short vec_andc (vector bool short, vector signed short);
16522 vector signed short vec_andc (vector signed short, vector bool short);
16523 vector signed short vec_andc (vector signed short, vector signed short);
16524 vector unsigned short vec_andc (vector bool short,
16525 vector unsigned short);
16526 vector unsigned short vec_andc (vector unsigned short,
16527 vector bool short);
16528 vector unsigned short vec_andc (vector unsigned short,
16529 vector unsigned short);
16530 vector signed char vec_andc (vector bool char, vector signed char);
16531 vector bool char vec_andc (vector bool char, vector bool char);
16532 vector signed char vec_andc (vector signed char, vector bool char);
16533 vector signed char vec_andc (vector signed char, vector signed char);
16534 vector unsigned char vec_andc (vector bool char, vector unsigned char);
16535 vector unsigned char vec_andc (vector unsigned char, vector bool char);
16536 vector unsigned char vec_andc (vector unsigned char,
16537 vector unsigned char);
16539 vector unsigned char vec_avg (vector unsigned char,
16540 vector unsigned char);
16541 vector signed char vec_avg (vector signed char, vector signed char);
16542 vector unsigned short vec_avg (vector unsigned short,
16543 vector unsigned short);
16544 vector signed short vec_avg (vector signed short, vector signed short);
16545 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
16546 vector signed int vec_avg (vector signed int, vector signed int);
16548 vector signed int vec_vavgsw (vector signed int, vector signed int);
16550 vector unsigned int vec_vavguw (vector unsigned int,
16551 vector unsigned int);
16553 vector signed short vec_vavgsh (vector signed short,
16554 vector signed short);
16556 vector unsigned short vec_vavguh (vector unsigned short,
16557 vector unsigned short);
16559 vector signed char vec_vavgsb (vector signed char, vector signed char);
16561 vector unsigned char vec_vavgub (vector unsigned char,
16562 vector unsigned char);
16564 vector float vec_copysign (vector float);
16566 vector float vec_ceil (vector float);
16568 vector signed int vec_cmpb (vector float, vector float);
16570 vector bool char vec_cmpeq (vector bool char, vector bool char);
16571 vector bool short vec_cmpeq (vector bool short, vector bool short);
16572 vector bool int vec_cmpeq (vector bool int, vector bool int);
16573 vector bool char vec_cmpeq (vector signed char, vector signed char);
16574 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
16575 vector bool short vec_cmpeq (vector signed short, vector signed short);
16576 vector bool short vec_cmpeq (vector unsigned short,
16577 vector unsigned short);
16578 vector bool int vec_cmpeq (vector signed int, vector signed int);
16579 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
16580 vector bool int vec_cmpeq (vector float, vector float);
16582 vector bool int vec_vcmpeqfp (vector float, vector float);
16584 vector bool int vec_vcmpequw (vector signed int, vector signed int);
16585 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
16587 vector bool short vec_vcmpequh (vector signed short,
16588 vector signed short);
16589 vector bool short vec_vcmpequh (vector unsigned short,
16590 vector unsigned short);
16592 vector bool char vec_vcmpequb (vector signed char, vector signed char);
16593 vector bool char vec_vcmpequb (vector unsigned char,
16594 vector unsigned char);
16596 vector bool int vec_cmpge (vector float, vector float);
16598 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
16599 vector bool char vec_cmpgt (vector signed char, vector signed char);
16600 vector bool short vec_cmpgt (vector unsigned short,
16601 vector unsigned short);
16602 vector bool short vec_cmpgt (vector signed short, vector signed short);
16603 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
16604 vector bool int vec_cmpgt (vector signed int, vector signed int);
16605 vector bool int vec_cmpgt (vector float, vector float);
16607 vector bool int vec_vcmpgtfp (vector float, vector float);
16609 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
16611 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
16613 vector bool short vec_vcmpgtsh (vector signed short,
16614 vector signed short);
16616 vector bool short vec_vcmpgtuh (vector unsigned short,
16617 vector unsigned short);
16619 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
16621 vector bool char vec_vcmpgtub (vector unsigned char,
16622 vector unsigned char);
16624 vector bool int vec_cmple (vector float, vector float);
16626 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
16627 vector bool char vec_cmplt (vector signed char, vector signed char);
16628 vector bool short vec_cmplt (vector unsigned short,
16629 vector unsigned short);
16630 vector bool short vec_cmplt (vector signed short, vector signed short);
16631 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
16632 vector bool int vec_cmplt (vector signed int, vector signed int);
16633 vector bool int vec_cmplt (vector float, vector float);
16635 vector float vec_cpsgn (vector float, vector float);
16637 vector float vec_ctf (vector unsigned int, const int);
16638 vector float vec_ctf (vector signed int, const int);
16639 vector double vec_ctf (vector unsigned long, const int);
16640 vector double vec_ctf (vector signed long, const int);
16642 vector float vec_vcfsx (vector signed int, const int);
16644 vector float vec_vcfux (vector unsigned int, const int);
16646 vector signed int vec_cts (vector float, const int);
16647 vector signed long vec_cts (vector double, const int);
16649 vector unsigned int vec_ctu (vector float, const int);
16650 vector unsigned long vec_ctu (vector double, const int);
16652 vector double vec_doublee (vector float);
16653 vector double vec_doublee (vector signed int);
16654 vector double vec_doublee (vector unsigned int);
16656 vector double vec_doubleo (vector float);
16657 vector double vec_doubleo (vector signed int);
16658 vector double vec_doubleo (vector unsigned int);
16660 vector double vec_doubleh (vector float);
16661 vector double vec_doubleh (vector signed int);
16662 vector double vec_doubleh (vector unsigned int);
16664 vector double vec_doublel (vector float);
16665 vector double vec_doublel (vector signed int);
16666 vector double vec_doublel (vector unsigned int);
16668 void vec_dss (const int);
16670 void vec_dssall (void);
16672 void vec_dst (const vector unsigned char *, int, const int);
16673 void vec_dst (const vector signed char *, int, const int);
16674 void vec_dst (const vector bool char *, int, const int);
16675 void vec_dst (const vector unsigned short *, int, const int);
16676 void vec_dst (const vector signed short *, int, const int);
16677 void vec_dst (const vector bool short *, int, const int);
16678 void vec_dst (const vector pixel *, int, const int);
16679 void vec_dst (const vector unsigned int *, int, const int);
16680 void vec_dst (const vector signed int *, int, const int);
16681 void vec_dst (const vector bool int *, int, const int);
16682 void vec_dst (const vector float *, int, const int);
16683 void vec_dst (const unsigned char *, int, const int);
16684 void vec_dst (const signed char *, int, const int);
16685 void vec_dst (const unsigned short *, int, const int);
16686 void vec_dst (const short *, int, const int);
16687 void vec_dst (const unsigned int *, int, const int);
16688 void vec_dst (const int *, int, const int);
16689 void vec_dst (const unsigned long *, int, const int);
16690 void vec_dst (const long *, int, const int);
16691 void vec_dst (const float *, int, const int);
16693 void vec_dstst (const vector unsigned char *, int, const int);
16694 void vec_dstst (const vector signed char *, int, const int);
16695 void vec_dstst (const vector bool char *, int, const int);
16696 void vec_dstst (const vector unsigned short *, int, const int);
16697 void vec_dstst (const vector signed short *, int, const int);
16698 void vec_dstst (const vector bool short *, int, const int);
16699 void vec_dstst (const vector pixel *, int, const int);
16700 void vec_dstst (const vector unsigned int *, int, const int);
16701 void vec_dstst (const vector signed int *, int, const int);
16702 void vec_dstst (const vector bool int *, int, const int);
16703 void vec_dstst (const vector float *, int, const int);
16704 void vec_dstst (const unsigned char *, int, const int);
16705 void vec_dstst (const signed char *, int, const int);
16706 void vec_dstst (const unsigned short *, int, const int);
16707 void vec_dstst (const short *, int, const int);
16708 void vec_dstst (const unsigned int *, int, const int);
16709 void vec_dstst (const int *, int, const int);
16710 void vec_dstst (const unsigned long *, int, const int);
16711 void vec_dstst (const long *, int, const int);
16712 void vec_dstst (const float *, int, const int);
16714 void vec_dststt (const vector unsigned char *, int, const int);
16715 void vec_dststt (const vector signed char *, int, const int);
16716 void vec_dststt (const vector bool char *, int, const int);
16717 void vec_dststt (const vector unsigned short *, int, const int);
16718 void vec_dststt (const vector signed short *, int, const int);
16719 void vec_dststt (const vector bool short *, int, const int);
16720 void vec_dststt (const vector pixel *, int, const int);
16721 void vec_dststt (const vector unsigned int *, int, const int);
16722 void vec_dststt (const vector signed int *, int, const int);
16723 void vec_dststt (const vector bool int *, int, const int);
16724 void vec_dststt (const vector float *, int, const int);
16725 void vec_dststt (const unsigned char *, int, const int);
16726 void vec_dststt (const signed char *, int, const int);
16727 void vec_dststt (const unsigned short *, int, const int);
16728 void vec_dststt (const short *, int, const int);
16729 void vec_dststt (const unsigned int *, int, const int);
16730 void vec_dststt (const int *, int, const int);
16731 void vec_dststt (const unsigned long *, int, const int);
16732 void vec_dststt (const long *, int, const int);
16733 void vec_dststt (const float *, int, const int);
16735 void vec_dstt (const vector unsigned char *, int, const int);
16736 void vec_dstt (const vector signed char *, int, const int);
16737 void vec_dstt (const vector bool char *, int, const int);
16738 void vec_dstt (const vector unsigned short *, int, const int);
16739 void vec_dstt (const vector signed short *, int, const int);
16740 void vec_dstt (const vector bool short *, int, const int);
16741 void vec_dstt (const vector pixel *, int, const int);
16742 void vec_dstt (const vector unsigned int *, int, const int);
16743 void vec_dstt (const vector signed int *, int, const int);
16744 void vec_dstt (const vector bool int *, int, const int);
16745 void vec_dstt (const vector float *, int, const int);
16746 void vec_dstt (const unsigned char *, int, const int);
16747 void vec_dstt (const signed char *, int, const int);
16748 void vec_dstt (const unsigned short *, int, const int);
16749 void vec_dstt (const short *, int, const int);
16750 void vec_dstt (const unsigned int *, int, const int);
16751 void vec_dstt (const int *, int, const int);
16752 void vec_dstt (const unsigned long *, int, const int);
16753 void vec_dstt (const long *, int, const int);
16754 void vec_dstt (const float *, int, const int);
16756 vector float vec_expte (vector float);
16758 vector float vec_floor (vector float);
16760 vector float vec_float (vector signed int);
16761 vector float vec_float (vector unsigned int);
16763 vector float vec_float2 (vector signed long long, vector signed long long);
16764 vector float vec_float2 (vector unsigned long long, vector signed long long);
16766 vector float vec_floate (vector double);
16767 vector float vec_floate (vector signed long long);
16768 vector float vec_floate (vector unsigned long long);
16770 vector float vec_floato (vector double);
16771 vector float vec_floato (vector signed long long);
16772 vector float vec_floato (vector unsigned long long);
16774 vector float vec_ld (int, const vector float *);
16775 vector float vec_ld (int, const float *);
16776 vector bool int vec_ld (int, const vector bool int *);
16777 vector signed int vec_ld (int, const vector signed int *);
16778 vector signed int vec_ld (int, const int *);
16779 vector signed int vec_ld (int, const long *);
16780 vector unsigned int vec_ld (int, const vector unsigned int *);
16781 vector unsigned int vec_ld (int, const unsigned int *);
16782 vector unsigned int vec_ld (int, const unsigned long *);
16783 vector bool short vec_ld (int, const vector bool short *);
16784 vector pixel vec_ld (int, const vector pixel *);
16785 vector signed short vec_ld (int, const vector signed short *);
16786 vector signed short vec_ld (int, const short *);
16787 vector unsigned short vec_ld (int, const vector unsigned short *);
16788 vector unsigned short vec_ld (int, const unsigned short *);
16789 vector bool char vec_ld (int, const vector bool char *);
16790 vector signed char vec_ld (int, const vector signed char *);
16791 vector signed char vec_ld (int, const signed char *);
16792 vector unsigned char vec_ld (int, const vector unsigned char *);
16793 vector unsigned char vec_ld (int, const unsigned char *);
16795 vector signed char vec_lde (int, const signed char *);
16796 vector unsigned char vec_lde (int, const unsigned char *);
16797 vector signed short vec_lde (int, const short *);
16798 vector unsigned short vec_lde (int, const unsigned short *);
16799 vector float vec_lde (int, const float *);
16800 vector signed int vec_lde (int, const int *);
16801 vector unsigned int vec_lde (int, const unsigned int *);
16802 vector signed int vec_lde (int, const long *);
16803 vector unsigned int vec_lde (int, const unsigned long *);
16805 vector float vec_lvewx (int, float *);
16806 vector signed int vec_lvewx (int, int *);
16807 vector unsigned int vec_lvewx (int, unsigned int *);
16808 vector signed int vec_lvewx (int, long *);
16809 vector unsigned int vec_lvewx (int, unsigned long *);
16811 vector signed short vec_lvehx (int, short *);
16812 vector unsigned short vec_lvehx (int, unsigned short *);
16814 vector signed char vec_lvebx (int, char *);
16815 vector unsigned char vec_lvebx (int, unsigned char *);
16817 vector float vec_ldl (int, const vector float *);
16818 vector float vec_ldl (int, const float *);
16819 vector bool int vec_ldl (int, const vector bool int *);
16820 vector signed int vec_ldl (int, const vector signed int *);
16821 vector signed int vec_ldl (int, const int *);
16822 vector signed int vec_ldl (int, const long *);
16823 vector unsigned int vec_ldl (int, const vector unsigned int *);
16824 vector unsigned int vec_ldl (int, const unsigned int *);
16825 vector unsigned int vec_ldl (int, const unsigned long *);
16826 vector bool short vec_ldl (int, const vector bool short *);
16827 vector pixel vec_ldl (int, const vector pixel *);
16828 vector signed short vec_ldl (int, const vector signed short *);
16829 vector signed short vec_ldl (int, const short *);
16830 vector unsigned short vec_ldl (int, const vector unsigned short *);
16831 vector unsigned short vec_ldl (int, const unsigned short *);
16832 vector bool char vec_ldl (int, const vector bool char *);
16833 vector signed char vec_ldl (int, const vector signed char *);
16834 vector signed char vec_ldl (int, const signed char *);
16835 vector unsigned char vec_ldl (int, const vector unsigned char *);
16836 vector unsigned char vec_ldl (int, const unsigned char *);
16838 vector float vec_loge (vector float);
16840 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
16841 vector unsigned char vec_lvsl (int, const volatile signed char *);
16842 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
16843 vector unsigned char vec_lvsl (int, const volatile short *);
16844 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
16845 vector unsigned char vec_lvsl (int, const volatile int *);
16846 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
16847 vector unsigned char vec_lvsl (int, const volatile long *);
16848 vector unsigned char vec_lvsl (int, const volatile float *);
16850 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
16851 vector unsigned char vec_lvsr (int, const volatile signed char *);
16852 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
16853 vector unsigned char vec_lvsr (int, const volatile short *);
16854 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
16855 vector unsigned char vec_lvsr (int, const volatile int *);
16856 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
16857 vector unsigned char vec_lvsr (int, const volatile long *);
16858 vector unsigned char vec_lvsr (int, const volatile float *);
16860 vector float vec_madd (vector float, vector float, vector float);
16862 vector signed short vec_madds (vector signed short,
16863 vector signed short,
16864 vector signed short);
16866 vector unsigned char vec_max (vector bool char, vector unsigned char);
16867 vector unsigned char vec_max (vector unsigned char, vector bool char);
16868 vector unsigned char vec_max (vector unsigned char,
16869 vector unsigned char);
16870 vector signed char vec_max (vector bool char, vector signed char);
16871 vector signed char vec_max (vector signed char, vector bool char);
16872 vector signed char vec_max (vector signed char, vector signed char);
16873 vector unsigned short vec_max (vector bool short,
16874 vector unsigned short);
16875 vector unsigned short vec_max (vector unsigned short,
16876 vector bool short);
16877 vector unsigned short vec_max (vector unsigned short,
16878 vector unsigned short);
16879 vector signed short vec_max (vector bool short, vector signed short);
16880 vector signed short vec_max (vector signed short, vector bool short);
16881 vector signed short vec_max (vector signed short, vector signed short);
16882 vector unsigned int vec_max (vector bool int, vector unsigned int);
16883 vector unsigned int vec_max (vector unsigned int, vector bool int);
16884 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
16885 vector signed int vec_max (vector bool int, vector signed int);
16886 vector signed int vec_max (vector signed int, vector bool int);
16887 vector signed int vec_max (vector signed int, vector signed int);
16888 vector float vec_max (vector float, vector float);
16890 vector float vec_vmaxfp (vector float, vector float);
16892 vector signed int vec_vmaxsw (vector bool int, vector signed int);
16893 vector signed int vec_vmaxsw (vector signed int, vector bool int);
16894 vector signed int vec_vmaxsw (vector signed int, vector signed int);
16896 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
16897 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
16898 vector unsigned int vec_vmaxuw (vector unsigned int,
16899 vector unsigned int);
16901 vector signed short vec_vmaxsh (vector bool short, vector signed short);
16902 vector signed short vec_vmaxsh (vector signed short, vector bool short);
16903 vector signed short vec_vmaxsh (vector signed short,
16904 vector signed short);
16906 vector unsigned short vec_vmaxuh (vector bool short,
16907 vector unsigned short);
16908 vector unsigned short vec_vmaxuh (vector unsigned short,
16909 vector bool short);
16910 vector unsigned short vec_vmaxuh (vector unsigned short,
16911 vector unsigned short);
16913 vector signed char vec_vmaxsb (vector bool char, vector signed char);
16914 vector signed char vec_vmaxsb (vector signed char, vector bool char);
16915 vector signed char vec_vmaxsb (vector signed char, vector signed char);
16917 vector unsigned char vec_vmaxub (vector bool char,
16918 vector unsigned char);
16919 vector unsigned char vec_vmaxub (vector unsigned char,
16921 vector unsigned char vec_vmaxub (vector unsigned char,
16922 vector unsigned char);
16924 vector bool char vec_mergeh (vector bool char, vector bool char);
16925 vector signed char vec_mergeh (vector signed char, vector signed char);
16926 vector unsigned char vec_mergeh (vector unsigned char,
16927 vector unsigned char);
16928 vector bool short vec_mergeh (vector bool short, vector bool short);
16929 vector pixel vec_mergeh (vector pixel, vector pixel);
16930 vector signed short vec_mergeh (vector signed short,
16931 vector signed short);
16932 vector unsigned short vec_mergeh (vector unsigned short,
16933 vector unsigned short);
16934 vector float vec_mergeh (vector float, vector float);
16935 vector bool int vec_mergeh (vector bool int, vector bool int);
16936 vector signed int vec_mergeh (vector signed int, vector signed int);
16937 vector unsigned int vec_mergeh (vector unsigned int,
16938 vector unsigned int);
16940 vector float vec_vmrghw (vector float, vector float);
16941 vector bool int vec_vmrghw (vector bool int, vector bool int);
16942 vector signed int vec_vmrghw (vector signed int, vector signed int);
16943 vector unsigned int vec_vmrghw (vector unsigned int,
16944 vector unsigned int);
16946 vector bool short vec_vmrghh (vector bool short, vector bool short);
16947 vector signed short vec_vmrghh (vector signed short,
16948 vector signed short);
16949 vector unsigned short vec_vmrghh (vector unsigned short,
16950 vector unsigned short);
16951 vector pixel vec_vmrghh (vector pixel, vector pixel);
16953 vector bool char vec_vmrghb (vector bool char, vector bool char);
16954 vector signed char vec_vmrghb (vector signed char, vector signed char);
16955 vector unsigned char vec_vmrghb (vector unsigned char,
16956 vector unsigned char);
16958 vector bool char vec_mergel (vector bool char, vector bool char);
16959 vector signed char vec_mergel (vector signed char, vector signed char);
16960 vector unsigned char vec_mergel (vector unsigned char,
16961 vector unsigned char);
16962 vector bool short vec_mergel (vector bool short, vector bool short);
16963 vector pixel vec_mergel (vector pixel, vector pixel);
16964 vector signed short vec_mergel (vector signed short,
16965 vector signed short);
16966 vector unsigned short vec_mergel (vector unsigned short,
16967 vector unsigned short);
16968 vector float vec_mergel (vector float, vector float);
16969 vector bool int vec_mergel (vector bool int, vector bool int);
16970 vector signed int vec_mergel (vector signed int, vector signed int);
16971 vector unsigned int vec_mergel (vector unsigned int,
16972 vector unsigned int);
16974 vector float vec_vmrglw (vector float, vector float);
16975 vector signed int vec_vmrglw (vector signed int, vector signed int);
16976 vector unsigned int vec_vmrglw (vector unsigned int,
16977 vector unsigned int);
16978 vector bool int vec_vmrglw (vector bool int, vector bool int);
16980 vector bool short vec_vmrglh (vector bool short, vector bool short);
16981 vector signed short vec_vmrglh (vector signed short,
16982 vector signed short);
16983 vector unsigned short vec_vmrglh (vector unsigned short,
16984 vector unsigned short);
16985 vector pixel vec_vmrglh (vector pixel, vector pixel);
16987 vector bool char vec_vmrglb (vector bool char, vector bool char);
16988 vector signed char vec_vmrglb (vector signed char, vector signed char);
16989 vector unsigned char vec_vmrglb (vector unsigned char,
16990 vector unsigned char);
16992 vector unsigned short vec_mfvscr (void);
16994 vector unsigned char vec_min (vector bool char, vector unsigned char);
16995 vector unsigned char vec_min (vector unsigned char, vector bool char);
16996 vector unsigned char vec_min (vector unsigned char,
16997 vector unsigned char);
16998 vector signed char vec_min (vector bool char, vector signed char);
16999 vector signed char vec_min (vector signed char, vector bool char);
17000 vector signed char vec_min (vector signed char, vector signed char);
17001 vector unsigned short vec_min (vector bool short,
17002 vector unsigned short);
17003 vector unsigned short vec_min (vector unsigned short,
17004 vector bool short);
17005 vector unsigned short vec_min (vector unsigned short,
17006 vector unsigned short);
17007 vector signed short vec_min (vector bool short, vector signed short);
17008 vector signed short vec_min (vector signed short, vector bool short);
17009 vector signed short vec_min (vector signed short, vector signed short);
17010 vector unsigned int vec_min (vector bool int, vector unsigned int);
17011 vector unsigned int vec_min (vector unsigned int, vector bool int);
17012 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
17013 vector signed int vec_min (vector bool int, vector signed int);
17014 vector signed int vec_min (vector signed int, vector bool int);
17015 vector signed int vec_min (vector signed int, vector signed int);
17016 vector float vec_min (vector float, vector float);
17018 vector float vec_vminfp (vector float, vector float);
17020 vector signed int vec_vminsw (vector bool int, vector signed int);
17021 vector signed int vec_vminsw (vector signed int, vector bool int);
17022 vector signed int vec_vminsw (vector signed int, vector signed int);
17024 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
17025 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
17026 vector unsigned int vec_vminuw (vector unsigned int,
17027 vector unsigned int);
17029 vector signed short vec_vminsh (vector bool short, vector signed short);
17030 vector signed short vec_vminsh (vector signed short, vector bool short);
17031 vector signed short vec_vminsh (vector signed short,
17032 vector signed short);
17034 vector unsigned short vec_vminuh (vector bool short,
17035 vector unsigned short);
17036 vector unsigned short vec_vminuh (vector unsigned short,
17037 vector bool short);
17038 vector unsigned short vec_vminuh (vector unsigned short,
17039 vector unsigned short);
17041 vector signed char vec_vminsb (vector bool char, vector signed char);
17042 vector signed char vec_vminsb (vector signed char, vector bool char);
17043 vector signed char vec_vminsb (vector signed char, vector signed char);
17045 vector unsigned char vec_vminub (vector bool char,
17046 vector unsigned char);
17047 vector unsigned char vec_vminub (vector unsigned char,
17049 vector unsigned char vec_vminub (vector unsigned char,
17050 vector unsigned char);
17052 vector signed short vec_mladd (vector signed short,
17053 vector signed short,
17054 vector signed short);
17055 vector signed short vec_mladd (vector signed short,
17056 vector unsigned short,
17057 vector unsigned short);
17058 vector signed short vec_mladd (vector unsigned short,
17059 vector signed short,
17060 vector signed short);
17061 vector unsigned short vec_mladd (vector unsigned short,
17062 vector unsigned short,
17063 vector unsigned short);
17065 vector signed short vec_mradds (vector signed short,
17066 vector signed short,
17067 vector signed short);
17069 vector unsigned int vec_msum (vector unsigned char,
17070 vector unsigned char,
17071 vector unsigned int);
17072 vector signed int vec_msum (vector signed char,
17073 vector unsigned char,
17074 vector signed int);
17075 vector unsigned int vec_msum (vector unsigned short,
17076 vector unsigned short,
17077 vector unsigned int);
17078 vector signed int vec_msum (vector signed short,
17079 vector signed short,
17080 vector signed int);
17082 vector signed int vec_vmsumshm (vector signed short,
17083 vector signed short,
17084 vector signed int);
17086 vector unsigned int vec_vmsumuhm (vector unsigned short,
17087 vector unsigned short,
17088 vector unsigned int);
17090 vector signed int vec_vmsummbm (vector signed char,
17091 vector unsigned char,
17092 vector signed int);
17094 vector unsigned int vec_vmsumubm (vector unsigned char,
17095 vector unsigned char,
17096 vector unsigned int);
17098 vector unsigned int vec_msums (vector unsigned short,
17099 vector unsigned short,
17100 vector unsigned int);
17101 vector signed int vec_msums (vector signed short,
17102 vector signed short,
17103 vector signed int);
17105 vector signed int vec_vmsumshs (vector signed short,
17106 vector signed short,
17107 vector signed int);
17109 vector unsigned int vec_vmsumuhs (vector unsigned short,
17110 vector unsigned short,
17111 vector unsigned int);
17113 void vec_mtvscr (vector signed int);
17114 void vec_mtvscr (vector unsigned int);
17115 void vec_mtvscr (vector bool int);
17116 void vec_mtvscr (vector signed short);
17117 void vec_mtvscr (vector unsigned short);
17118 void vec_mtvscr (vector bool short);
17119 void vec_mtvscr (vector pixel);
17120 void vec_mtvscr (vector signed char);
17121 void vec_mtvscr (vector unsigned char);
17122 void vec_mtvscr (vector bool char);
17124 vector unsigned short vec_mule (vector unsigned char,
17125 vector unsigned char);
17126 vector signed short vec_mule (vector signed char,
17127 vector signed char);
17128 vector unsigned int vec_mule (vector unsigned short,
17129 vector unsigned short);
17130 vector signed int vec_mule (vector signed short, vector signed short);
17131 vector unsigned long long vec_mule (vector unsigned int,
17132 vector unsigned int);
17133 vector signed long long vec_mule (vector signed int,
17134 vector signed int);
17136 vector signed int vec_vmulesh (vector signed short,
17137 vector signed short);
17139 vector unsigned int vec_vmuleuh (vector unsigned short,
17140 vector unsigned short);
17142 vector signed short vec_vmulesb (vector signed char,
17143 vector signed char);
17145 vector unsigned short vec_vmuleub (vector unsigned char,
17146 vector unsigned char);
17148 vector unsigned short vec_mulo (vector unsigned char,
17149 vector unsigned char);
17150 vector signed short vec_mulo (vector signed char, vector signed char);
17151 vector unsigned int vec_mulo (vector unsigned short,
17152 vector unsigned short);
17153 vector signed int vec_mulo (vector signed short, vector signed short);
17154 vector unsigned long long vec_mulo (vector unsigned int,
17155 vector unsigned int);
17156 vector signed long long vec_mulo (vector signed int,
17157 vector signed int);
17159 vector signed int vec_vmulosh (vector signed short,
17160 vector signed short);
17162 vector unsigned int vec_vmulouh (vector unsigned short,
17163 vector unsigned short);
17165 vector signed short vec_vmulosb (vector signed char,
17166 vector signed char);
17168 vector unsigned short vec_vmuloub (vector unsigned char,
17169 vector unsigned char);
17171 vector float vec_nmsub (vector float, vector float, vector float);
17173 vector signed char vec_nabs (vector signed char);
17174 vector signed short vec_nabs (vector signed short);
17175 vector signed int vec_nabs (vector signed int);
17176 vector float vec_nabs (vector float);
17177 vector double vec_nabs (vector double);
17179 vector signed char vec_neg (vector signed char);
17180 vector signed short vec_neg (vector signed short);
17181 vector signed int vec_neg (vector signed int);
17182 vector signed long long vec_neg (vector signed long long);
17183 vector float char vec_neg (vector float);
17184 vector double vec_neg (vector double);
17186 vector float vec_nor (vector float, vector float);
17187 vector signed int vec_nor (vector signed int, vector signed int);
17188 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
17189 vector bool int vec_nor (vector bool int, vector bool int);
17190 vector signed short vec_nor (vector signed short, vector signed short);
17191 vector unsigned short vec_nor (vector unsigned short,
17192 vector unsigned short);
17193 vector bool short vec_nor (vector bool short, vector bool short);
17194 vector signed char vec_nor (vector signed char, vector signed char);
17195 vector unsigned char vec_nor (vector unsigned char,
17196 vector unsigned char);
17197 vector bool char vec_nor (vector bool char, vector bool char);
17199 vector float vec_or (vector float, vector float);
17200 vector float vec_or (vector float, vector bool int);
17201 vector float vec_or (vector bool int, vector float);
17202 vector bool int vec_or (vector bool int, vector bool int);
17203 vector signed int vec_or (vector bool int, vector signed int);
17204 vector signed int vec_or (vector signed int, vector bool int);
17205 vector signed int vec_or (vector signed int, vector signed int);
17206 vector unsigned int vec_or (vector bool int, vector unsigned int);
17207 vector unsigned int vec_or (vector unsigned int, vector bool int);
17208 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
17209 vector bool short vec_or (vector bool short, vector bool short);
17210 vector signed short vec_or (vector bool short, vector signed short);
17211 vector signed short vec_or (vector signed short, vector bool short);
17212 vector signed short vec_or (vector signed short, vector signed short);
17213 vector unsigned short vec_or (vector bool short, vector unsigned short);
17214 vector unsigned short vec_or (vector unsigned short, vector bool short);
17215 vector unsigned short vec_or (vector unsigned short,
17216 vector unsigned short);
17217 vector signed char vec_or (vector bool char, vector signed char);
17218 vector bool char vec_or (vector bool char, vector bool char);
17219 vector signed char vec_or (vector signed char, vector bool char);
17220 vector signed char vec_or (vector signed char, vector signed char);
17221 vector unsigned char vec_or (vector bool char, vector unsigned char);
17222 vector unsigned char vec_or (vector unsigned char, vector bool char);
17223 vector unsigned char vec_or (vector unsigned char,
17224 vector unsigned char);
17226 vector signed char vec_pack (vector signed short, vector signed short);
17227 vector unsigned char vec_pack (vector unsigned short,
17228 vector unsigned short);
17229 vector bool char vec_pack (vector bool short, vector bool short);
17230 vector signed short vec_pack (vector signed int, vector signed int);
17231 vector unsigned short vec_pack (vector unsigned int,
17232 vector unsigned int);
17233 vector bool short vec_pack (vector bool int, vector bool int);
17235 vector bool short vec_vpkuwum (vector bool int, vector bool int);
17236 vector signed short vec_vpkuwum (vector signed int, vector signed int);
17237 vector unsigned short vec_vpkuwum (vector unsigned int,
17238 vector unsigned int);
17240 vector bool char vec_vpkuhum (vector bool short, vector bool short);
17241 vector signed char vec_vpkuhum (vector signed short,
17242 vector signed short);
17243 vector unsigned char vec_vpkuhum (vector unsigned short,
17244 vector unsigned short);
17246 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
17248 vector unsigned char vec_packs (vector unsigned short,
17249 vector unsigned short);
17250 vector signed char vec_packs (vector signed short, vector signed short);
17251 vector unsigned short vec_packs (vector unsigned int,
17252 vector unsigned int);
17253 vector signed short vec_packs (vector signed int, vector signed int);
17255 vector signed short vec_vpkswss (vector signed int, vector signed int);
17257 vector unsigned short vec_vpkuwus (vector unsigned int,
17258 vector unsigned int);
17260 vector signed char vec_vpkshss (vector signed short,
17261 vector signed short);
17263 vector unsigned char vec_vpkuhus (vector unsigned short,
17264 vector unsigned short);
17266 vector unsigned char vec_packsu (vector unsigned short,
17267 vector unsigned short);
17268 vector unsigned char vec_packsu (vector signed short,
17269 vector signed short);
17270 vector unsigned short vec_packsu (vector unsigned int,
17271 vector unsigned int);
17272 vector unsigned short vec_packsu (vector signed int, vector signed int);
17274 vector unsigned short vec_vpkswus (vector signed int,
17275 vector signed int);
17277 vector unsigned char vec_vpkshus (vector signed short,
17278 vector signed short);
17280 vector float vec_perm (vector float,
17282 vector unsigned char);
17283 vector signed int vec_perm (vector signed int,
17285 vector unsigned char);
17286 vector unsigned int vec_perm (vector unsigned int,
17287 vector unsigned int,
17288 vector unsigned char);
17289 vector bool int vec_perm (vector bool int,
17291 vector unsigned char);
17292 vector signed short vec_perm (vector signed short,
17293 vector signed short,
17294 vector unsigned char);
17295 vector unsigned short vec_perm (vector unsigned short,
17296 vector unsigned short,
17297 vector unsigned char);
17298 vector bool short vec_perm (vector bool short,
17300 vector unsigned char);
17301 vector pixel vec_perm (vector pixel,
17303 vector unsigned char);
17304 vector signed char vec_perm (vector signed char,
17305 vector signed char,
17306 vector unsigned char);
17307 vector unsigned char vec_perm (vector unsigned char,
17308 vector unsigned char,
17309 vector unsigned char);
17310 vector bool char vec_perm (vector bool char,
17312 vector unsigned char);
17314 vector float vec_re (vector float);
17316 vector bool char vec_reve (vector bool char);
17317 vector signed char vec_reve (vector signed char);
17318 vector unsigned char vec_reve (vector unsigned char);
17319 vector bool int vec_reve (vector bool int);
17320 vector signed int vec_reve (vector signed int);
17321 vector unsigned int vec_reve (vector unsigned int);
17322 vector bool long long vec_reve (vector bool long long);
17323 vector signed long long vec_reve (vector signed long long);
17324 vector unsigned long long vec_reve (vector unsigned long long);
17325 vector bool short vec_reve (vector bool short);
17326 vector signed short vec_reve (vector signed short);
17327 vector unsigned short vec_reve (vector unsigned short);
17329 vector signed char vec_rl (vector signed char,
17330 vector unsigned char);
17331 vector unsigned char vec_rl (vector unsigned char,
17332 vector unsigned char);
17333 vector signed short vec_rl (vector signed short, vector unsigned short);
17334 vector unsigned short vec_rl (vector unsigned short,
17335 vector unsigned short);
17336 vector signed int vec_rl (vector signed int, vector unsigned int);
17337 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
17339 vector signed int vec_vrlw (vector signed int, vector unsigned int);
17340 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
17342 vector signed short vec_vrlh (vector signed short,
17343 vector unsigned short);
17344 vector unsigned short vec_vrlh (vector unsigned short,
17345 vector unsigned short);
17347 vector signed char vec_vrlb (vector signed char, vector unsigned char);
17348 vector unsigned char vec_vrlb (vector unsigned char,
17349 vector unsigned char);
17351 vector float vec_round (vector float);
17353 vector float vec_recip (vector float, vector float);
17355 vector float vec_rsqrt (vector float);
17357 vector float vec_rsqrte (vector float);
17359 vector float vec_sel (vector float, vector float, vector bool int);
17360 vector float vec_sel (vector float, vector float, vector unsigned int);
17361 vector signed int vec_sel (vector signed int,
17364 vector signed int vec_sel (vector signed int,
17366 vector unsigned int);
17367 vector unsigned int vec_sel (vector unsigned int,
17368 vector unsigned int,
17370 vector unsigned int vec_sel (vector unsigned int,
17371 vector unsigned int,
17372 vector unsigned int);
17373 vector bool int vec_sel (vector bool int,
17376 vector bool int vec_sel (vector bool int,
17378 vector unsigned int);
17379 vector signed short vec_sel (vector signed short,
17380 vector signed short,
17381 vector bool short);
17382 vector signed short vec_sel (vector signed short,
17383 vector signed short,
17384 vector unsigned short);
17385 vector unsigned short vec_sel (vector unsigned short,
17386 vector unsigned short,
17387 vector bool short);
17388 vector unsigned short vec_sel (vector unsigned short,
17389 vector unsigned short,
17390 vector unsigned short);
17391 vector bool short vec_sel (vector bool short,
17393 vector bool short);
17394 vector bool short vec_sel (vector bool short,
17396 vector unsigned short);
17397 vector signed char vec_sel (vector signed char,
17398 vector signed char,
17400 vector signed char vec_sel (vector signed char,
17401 vector signed char,
17402 vector unsigned char);
17403 vector unsigned char vec_sel (vector unsigned char,
17404 vector unsigned char,
17406 vector unsigned char vec_sel (vector unsigned char,
17407 vector unsigned char,
17408 vector unsigned char);
17409 vector bool char vec_sel (vector bool char,
17412 vector bool char vec_sel (vector bool char,
17414 vector unsigned char);
17416 vector signed long long vec_signed (vector double);
17417 vector signed int vec_signed (vector float);
17419 vector signed int vec_signede (vector double);
17420 vector signed int vec_signedo (vector double);
17421 vector signed int vec_signed2 (vector double, vector double);
17423 vector signed char vec_sl (vector signed char,
17424 vector unsigned char);
17425 vector unsigned char vec_sl (vector unsigned char,
17426 vector unsigned char);
17427 vector signed short vec_sl (vector signed short, vector unsigned short);
17428 vector unsigned short vec_sl (vector unsigned short,
17429 vector unsigned short);
17430 vector signed int vec_sl (vector signed int, vector unsigned int);
17431 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
17433 vector signed int vec_vslw (vector signed int, vector unsigned int);
17434 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
17436 vector signed short vec_vslh (vector signed short,
17437 vector unsigned short);
17438 vector unsigned short vec_vslh (vector unsigned short,
17439 vector unsigned short);
17441 vector signed char vec_vslb (vector signed char, vector unsigned char);
17442 vector unsigned char vec_vslb (vector unsigned char,
17443 vector unsigned char);
17445 vector float vec_sld (vector float, vector float, const int);
17446 vector double vec_sld (vector double, vector double, const int);
17448 vector signed int vec_sld (vector signed int,
17451 vector unsigned int vec_sld (vector unsigned int,
17452 vector unsigned int,
17454 vector bool int vec_sld (vector bool int,
17457 vector signed short vec_sld (vector signed short,
17458 vector signed short,
17460 vector unsigned short vec_sld (vector unsigned short,
17461 vector unsigned short,
17463 vector bool short vec_sld (vector bool short,
17466 vector pixel vec_sld (vector pixel,
17469 vector signed char vec_sld (vector signed char,
17470 vector signed char,
17472 vector unsigned char vec_sld (vector unsigned char,
17473 vector unsigned char,
17475 vector bool char vec_sld (vector bool char,
17478 vector bool long long int vec_sld (vector bool long long int,
17479 vector bool long long int, const int);
17480 vector long long int vec_sld (vector long long int,
17481 vector long long int, const int);
17482 vector unsigned long long int vec_sld (vector unsigned long long int,
17483 vector unsigned long long int,
17486 vector signed char vec_sldw (vector signed char,
17487 vector signed char,
17489 vector unsigned char vec_sldw (vector unsigned char,
17490 vector unsigned char,
17492 vector signed short vec_sldw (vector signed short,
17493 vector signed short,
17495 vector unsigned short vec_sldw (vector unsigned short,
17496 vector unsigned short,
17498 vector signed int vec_sldw (vector signed int,
17501 vector unsigned int vec_sldw (vector unsigned int,
17502 vector unsigned int,
17504 vector signed long long vec_sldw (vector signed long long,
17505 vector signed long long,
17507 vector unsigned long long vec_sldw (vector unsigned long long,
17508 vector unsigned long long,
17511 vector signed int vec_sll (vector signed int,
17512 vector unsigned int);
17513 vector signed int vec_sll (vector signed int,
17514 vector unsigned short);
17515 vector signed int vec_sll (vector signed int,
17516 vector unsigned char);
17517 vector unsigned int vec_sll (vector unsigned int,
17518 vector unsigned int);
17519 vector unsigned int vec_sll (vector unsigned int,
17520 vector unsigned short);
17521 vector unsigned int vec_sll (vector unsigned int,
17522 vector unsigned char);
17523 vector bool int vec_sll (vector bool int,
17524 vector unsigned int);
17525 vector bool int vec_sll (vector bool int,
17526 vector unsigned short);
17527 vector bool int vec_sll (vector bool int,
17528 vector unsigned char);
17529 vector signed short vec_sll (vector signed short,
17530 vector unsigned int);
17531 vector signed short vec_sll (vector signed short,
17532 vector unsigned short);
17533 vector signed short vec_sll (vector signed short,
17534 vector unsigned char);
17535 vector unsigned short vec_sll (vector unsigned short,
17536 vector unsigned int);
17537 vector unsigned short vec_sll (vector unsigned short,
17538 vector unsigned short);
17539 vector unsigned short vec_sll (vector unsigned short,
17540 vector unsigned char);
17541 vector long long int vec_sll (vector long long int,
17542 vector unsigned char);
17543 vector unsigned long long int vec_sll (vector unsigned long long int,
17544 vector unsigned char);
17545 vector bool short vec_sll (vector bool short, vector unsigned int);
17546 vector bool short vec_sll (vector bool short, vector unsigned short);
17547 vector bool short vec_sll (vector bool short, vector unsigned char);
17548 vector pixel vec_sll (vector pixel, vector unsigned int);
17549 vector pixel vec_sll (vector pixel, vector unsigned short);
17550 vector pixel vec_sll (vector pixel, vector unsigned char);
17551 vector signed char vec_sll (vector signed char, vector unsigned int);
17552 vector signed char vec_sll (vector signed char, vector unsigned short);
17553 vector signed char vec_sll (vector signed char, vector unsigned char);
17554 vector unsigned char vec_sll (vector unsigned char,
17555 vector unsigned int);
17556 vector unsigned char vec_sll (vector unsigned char,
17557 vector unsigned short);
17558 vector unsigned char vec_sll (vector unsigned char,
17559 vector unsigned char);
17560 vector bool char vec_sll (vector bool char, vector unsigned int);
17561 vector bool char vec_sll (vector bool char, vector unsigned short);
17562 vector bool char vec_sll (vector bool char, vector unsigned char);
17564 vector float vec_slo (vector float, vector signed char);
17565 vector float vec_slo (vector float, vector unsigned char);
17566 vector signed int vec_slo (vector signed int, vector signed char);
17567 vector signed int vec_slo (vector signed int, vector unsigned char);
17568 vector unsigned int vec_slo (vector unsigned int, vector signed char);
17569 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
17570 vector signed short vec_slo (vector signed short, vector signed char);
17571 vector signed short vec_slo (vector signed short, vector unsigned char);
17572 vector unsigned short vec_slo (vector unsigned short,
17573 vector signed char);
17574 vector unsigned short vec_slo (vector unsigned short,
17575 vector unsigned char);
17576 vector pixel vec_slo (vector pixel, vector signed char);
17577 vector pixel vec_slo (vector pixel, vector unsigned char);
17578 vector signed char vec_slo (vector signed char, vector signed char);
17579 vector signed char vec_slo (vector signed char, vector unsigned char);
17580 vector unsigned char vec_slo (vector unsigned char, vector signed char);
17581 vector unsigned char vec_slo (vector unsigned char,
17582 vector unsigned char);
17583 vector signed long long vec_slo (vector signed long long, vector signed char);
17584 vector signed long long vec_slo (vector signed long long, vector unsigned char);
17585 vector unsigned long long vec_slo (vector unsigned long long, vector signed char);
17586 vector unsigned long long vec_slo (vector unsigned long long, vector unsigned char);
17588 vector signed char vec_splat (vector signed char, const int);
17589 vector unsigned char vec_splat (vector unsigned char, const int);
17590 vector bool char vec_splat (vector bool char, const int);
17591 vector signed short vec_splat (vector signed short, const int);
17592 vector unsigned short vec_splat (vector unsigned short, const int);
17593 vector bool short vec_splat (vector bool short, const int);
17594 vector pixel vec_splat (vector pixel, const int);
17595 vector float vec_splat (vector float, const int);
17596 vector signed int vec_splat (vector signed int, const int);
17597 vector unsigned int vec_splat (vector unsigned int, const int);
17598 vector bool int vec_splat (vector bool int, const int);
17599 vector signed long vec_splat (vector signed long, const int);
17600 vector unsigned long vec_splat (vector unsigned long, const int);
17602 vector signed char vec_splats (signed char);
17603 vector unsigned char vec_splats (unsigned char);
17604 vector signed short vec_splats (signed short);
17605 vector unsigned short vec_splats (unsigned short);
17606 vector signed int vec_splats (signed int);
17607 vector unsigned int vec_splats (unsigned int);
17608 vector float vec_splats (float);
17610 vector float vec_vspltw (vector float, const int);
17611 vector signed int vec_vspltw (vector signed int, const int);
17612 vector unsigned int vec_vspltw (vector unsigned int, const int);
17613 vector bool int vec_vspltw (vector bool int, const int);
17615 vector bool short vec_vsplth (vector bool short, const int);
17616 vector signed short vec_vsplth (vector signed short, const int);
17617 vector unsigned short vec_vsplth (vector unsigned short, const int);
17618 vector pixel vec_vsplth (vector pixel, const int);
17620 vector signed char vec_vspltb (vector signed char, const int);
17621 vector unsigned char vec_vspltb (vector unsigned char, const int);
17622 vector bool char vec_vspltb (vector bool char, const int);
17624 vector signed char vec_splat_s8 (const int);
17626 vector signed short vec_splat_s16 (const int);
17628 vector signed int vec_splat_s32 (const int);
17630 vector unsigned char vec_splat_u8 (const int);
17632 vector unsigned short vec_splat_u16 (const int);
17634 vector unsigned int vec_splat_u32 (const int);
17636 vector signed char vec_sr (vector signed char, vector unsigned char);
17637 vector unsigned char vec_sr (vector unsigned char,
17638 vector unsigned char);
17639 vector signed short vec_sr (vector signed short,
17640 vector unsigned short);
17641 vector unsigned short vec_sr (vector unsigned short,
17642 vector unsigned short);
17643 vector signed int vec_sr (vector signed int, vector unsigned int);
17644 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
17646 vector signed int vec_vsrw (vector signed int, vector unsigned int);
17647 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
17649 vector signed short vec_vsrh (vector signed short,
17650 vector unsigned short);
17651 vector unsigned short vec_vsrh (vector unsigned short,
17652 vector unsigned short);
17654 vector signed char vec_vsrb (vector signed char, vector unsigned char);
17655 vector unsigned char vec_vsrb (vector unsigned char,
17656 vector unsigned char);
17658 vector signed char vec_sra (vector signed char, vector unsigned char);
17659 vector unsigned char vec_sra (vector unsigned char,
17660 vector unsigned char);
17661 vector signed short vec_sra (vector signed short,
17662 vector unsigned short);
17663 vector unsigned short vec_sra (vector unsigned short,
17664 vector unsigned short);
17665 vector signed int vec_sra (vector signed int, vector unsigned int);
17666 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
17668 vector signed int vec_vsraw (vector signed int, vector unsigned int);
17669 vector unsigned int vec_vsraw (vector unsigned int,
17670 vector unsigned int);
17672 vector signed short vec_vsrah (vector signed short,
17673 vector unsigned short);
17674 vector unsigned short vec_vsrah (vector unsigned short,
17675 vector unsigned short);
17677 vector signed char vec_vsrab (vector signed char, vector unsigned char);
17678 vector unsigned char vec_vsrab (vector unsigned char,
17679 vector unsigned char);
17681 vector signed int vec_srl (vector signed int, vector unsigned int);
17682 vector signed int vec_srl (vector signed int, vector unsigned short);
17683 vector signed int vec_srl (vector signed int, vector unsigned char);
17684 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
17685 vector unsigned int vec_srl (vector unsigned int,
17686 vector unsigned short);
17687 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
17688 vector bool int vec_srl (vector bool int, vector unsigned int);
17689 vector bool int vec_srl (vector bool int, vector unsigned short);
17690 vector bool int vec_srl (vector bool int, vector unsigned char);
17691 vector signed short vec_srl (vector signed short, vector unsigned int);
17692 vector signed short vec_srl (vector signed short,
17693 vector unsigned short);
17694 vector signed short vec_srl (vector signed short, vector unsigned char);
17695 vector unsigned short vec_srl (vector unsigned short,
17696 vector unsigned int);
17697 vector unsigned short vec_srl (vector unsigned short,
17698 vector unsigned short);
17699 vector unsigned short vec_srl (vector unsigned short,
17700 vector unsigned char);
17701 vector long long int vec_srl (vector long long int,
17702 vector unsigned char);
17703 vector unsigned long long int vec_srl (vector unsigned long long int,
17704 vector unsigned char);
17705 vector bool short vec_srl (vector bool short, vector unsigned int);
17706 vector bool short vec_srl (vector bool short, vector unsigned short);
17707 vector bool short vec_srl (vector bool short, vector unsigned char);
17708 vector pixel vec_srl (vector pixel, vector unsigned int);
17709 vector pixel vec_srl (vector pixel, vector unsigned short);
17710 vector pixel vec_srl (vector pixel, vector unsigned char);
17711 vector signed char vec_srl (vector signed char, vector unsigned int);
17712 vector signed char vec_srl (vector signed char, vector unsigned short);
17713 vector signed char vec_srl (vector signed char, vector unsigned char);
17714 vector unsigned char vec_srl (vector unsigned char,
17715 vector unsigned int);
17716 vector unsigned char vec_srl (vector unsigned char,
17717 vector unsigned short);
17718 vector unsigned char vec_srl (vector unsigned char,
17719 vector unsigned char);
17720 vector bool char vec_srl (vector bool char, vector unsigned int);
17721 vector bool char vec_srl (vector bool char, vector unsigned short);
17722 vector bool char vec_srl (vector bool char, vector unsigned char);
17724 vector float vec_sro (vector float, vector signed char);
17725 vector float vec_sro (vector float, vector unsigned char);
17726 vector signed int vec_sro (vector signed int, vector signed char);
17727 vector signed int vec_sro (vector signed int, vector unsigned char);
17728 vector unsigned int vec_sro (vector unsigned int, vector signed char);
17729 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
17730 vector signed short vec_sro (vector signed short, vector signed char);
17731 vector signed short vec_sro (vector signed short, vector unsigned char);
17732 vector unsigned short vec_sro (vector unsigned short,
17733 vector signed char);
17734 vector unsigned short vec_sro (vector unsigned short,
17735 vector unsigned char);
17736 vector long long int vec_sro (vector long long int,
17738 vector long long int vec_sro (vector long long int,
17739 vector unsigned char);
17740 vector unsigned long long int vec_sro (vector unsigned long long int,
17742 vector unsigned long long int vec_sro (vector unsigned long long int,
17743 vector unsigned char);
17744 vector pixel vec_sro (vector pixel, vector signed char);
17745 vector pixel vec_sro (vector pixel, vector unsigned char);
17746 vector signed char vec_sro (vector signed char, vector signed char);
17747 vector signed char vec_sro (vector signed char, vector unsigned char);
17748 vector unsigned char vec_sro (vector unsigned char, vector signed char);
17749 vector unsigned char vec_sro (vector unsigned char,
17750 vector unsigned char);
17752 void vec_st (vector float, int, vector float *);
17753 void vec_st (vector float, int, float *);
17754 void vec_st (vector signed int, int, vector signed int *);
17755 void vec_st (vector signed int, int, int *);
17756 void vec_st (vector unsigned int, int, vector unsigned int *);
17757 void vec_st (vector unsigned int, int, unsigned int *);
17758 void vec_st (vector bool int, int, vector bool int *);
17759 void vec_st (vector bool int, int, unsigned int *);
17760 void vec_st (vector bool int, int, int *);
17761 void vec_st (vector signed short, int, vector signed short *);
17762 void vec_st (vector signed short, int, short *);
17763 void vec_st (vector unsigned short, int, vector unsigned short *);
17764 void vec_st (vector unsigned short, int, unsigned short *);
17765 void vec_st (vector bool short, int, vector bool short *);
17766 void vec_st (vector bool short, int, unsigned short *);
17767 void vec_st (vector pixel, int, vector pixel *);
17768 void vec_st (vector pixel, int, unsigned short *);
17769 void vec_st (vector pixel, int, short *);
17770 void vec_st (vector bool short, int, short *);
17771 void vec_st (vector signed char, int, vector signed char *);
17772 void vec_st (vector signed char, int, signed char *);
17773 void vec_st (vector unsigned char, int, vector unsigned char *);
17774 void vec_st (vector unsigned char, int, unsigned char *);
17775 void vec_st (vector bool char, int, vector bool char *);
17776 void vec_st (vector bool char, int, unsigned char *);
17777 void vec_st (vector bool char, int, signed char *);
17779 void vec_ste (vector signed char, int, signed char *);
17780 void vec_ste (vector unsigned char, int, unsigned char *);
17781 void vec_ste (vector bool char, int, signed char *);
17782 void vec_ste (vector bool char, int, unsigned char *);
17783 void vec_ste (vector signed short, int, short *);
17784 void vec_ste (vector unsigned short, int, unsigned short *);
17785 void vec_ste (vector bool short, int, short *);
17786 void vec_ste (vector bool short, int, unsigned short *);
17787 void vec_ste (vector pixel, int, short *);
17788 void vec_ste (vector pixel, int, unsigned short *);
17789 void vec_ste (vector float, int, float *);
17790 void vec_ste (vector signed int, int, int *);
17791 void vec_ste (vector unsigned int, int, unsigned int *);
17792 void vec_ste (vector bool int, int, int *);
17793 void vec_ste (vector bool int, int, unsigned int *);
17795 void vec_stvewx (vector float, int, float *);
17796 void vec_stvewx (vector signed int, int, int *);
17797 void vec_stvewx (vector unsigned int, int, unsigned int *);
17798 void vec_stvewx (vector bool int, int, int *);
17799 void vec_stvewx (vector bool int, int, unsigned int *);
17801 void vec_stvehx (vector signed short, int, short *);
17802 void vec_stvehx (vector unsigned short, int, unsigned short *);
17803 void vec_stvehx (vector bool short, int, short *);
17804 void vec_stvehx (vector bool short, int, unsigned short *);
17805 void vec_stvehx (vector pixel, int, short *);
17806 void vec_stvehx (vector pixel, int, unsigned short *);
17808 void vec_stvebx (vector signed char, int, signed char *);
17809 void vec_stvebx (vector unsigned char, int, unsigned char *);
17810 void vec_stvebx (vector bool char, int, signed char *);
17811 void vec_stvebx (vector bool char, int, unsigned char *);
17813 void vec_stl (vector float, int, vector float *);
17814 void vec_stl (vector float, int, float *);
17815 void vec_stl (vector signed int, int, vector signed int *);
17816 void vec_stl (vector signed int, int, int *);
17817 void vec_stl (vector unsigned int, int, vector unsigned int *);
17818 void vec_stl (vector unsigned int, int, unsigned int *);
17819 void vec_stl (vector bool int, int, vector bool int *);
17820 void vec_stl (vector bool int, int, unsigned int *);
17821 void vec_stl (vector bool int, int, int *);
17822 void vec_stl (vector signed short, int, vector signed short *);
17823 void vec_stl (vector signed short, int, short *);
17824 void vec_stl (vector unsigned short, int, vector unsigned short *);
17825 void vec_stl (vector unsigned short, int, unsigned short *);
17826 void vec_stl (vector bool short, int, vector bool short *);
17827 void vec_stl (vector bool short, int, unsigned short *);
17828 void vec_stl (vector bool short, int, short *);
17829 void vec_stl (vector pixel, int, vector pixel *);
17830 void vec_stl (vector pixel, int, unsigned short *);
17831 void vec_stl (vector pixel, int, short *);
17832 void vec_stl (vector signed char, int, vector signed char *);
17833 void vec_stl (vector signed char, int, signed char *);
17834 void vec_stl (vector unsigned char, int, vector unsigned char *);
17835 void vec_stl (vector unsigned char, int, unsigned char *);
17836 void vec_stl (vector bool char, int, vector bool char *);
17837 void vec_stl (vector bool char, int, unsigned char *);
17838 void vec_stl (vector bool char, int, signed char *);
17840 vector signed char vec_sub (vector bool char, vector signed char);
17841 vector signed char vec_sub (vector signed char, vector bool char);
17842 vector signed char vec_sub (vector signed char, vector signed char);
17843 vector unsigned char vec_sub (vector bool char, vector unsigned char);
17844 vector unsigned char vec_sub (vector unsigned char, vector bool char);
17845 vector unsigned char vec_sub (vector unsigned char,
17846 vector unsigned char);
17847 vector signed short vec_sub (vector bool short, vector signed short);
17848 vector signed short vec_sub (vector signed short, vector bool short);
17849 vector signed short vec_sub (vector signed short, vector signed short);
17850 vector unsigned short vec_sub (vector bool short,
17851 vector unsigned short);
17852 vector unsigned short vec_sub (vector unsigned short,
17853 vector bool short);
17854 vector unsigned short vec_sub (vector unsigned short,
17855 vector unsigned short);
17856 vector signed int vec_sub (vector bool int, vector signed int);
17857 vector signed int vec_sub (vector signed int, vector bool int);
17858 vector signed int vec_sub (vector signed int, vector signed int);
17859 vector unsigned int vec_sub (vector bool int, vector unsigned int);
17860 vector unsigned int vec_sub (vector unsigned int, vector bool int);
17861 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
17862 vector float vec_sub (vector float, vector float);
17864 vector float vec_vsubfp (vector float, vector float);
17866 vector signed int vec_vsubuwm (vector bool int, vector signed int);
17867 vector signed int vec_vsubuwm (vector signed int, vector bool int);
17868 vector signed int vec_vsubuwm (vector signed int, vector signed int);
17869 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
17870 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
17871 vector unsigned int vec_vsubuwm (vector unsigned int,
17872 vector unsigned int);
17874 vector signed short vec_vsubuhm (vector bool short,
17875 vector signed short);
17876 vector signed short vec_vsubuhm (vector signed short,
17877 vector bool short);
17878 vector signed short vec_vsubuhm (vector signed short,
17879 vector signed short);
17880 vector unsigned short vec_vsubuhm (vector bool short,
17881 vector unsigned short);
17882 vector unsigned short vec_vsubuhm (vector unsigned short,
17883 vector bool short);
17884 vector unsigned short vec_vsubuhm (vector unsigned short,
17885 vector unsigned short);
17887 vector signed char vec_vsububm (vector bool char, vector signed char);
17888 vector signed char vec_vsububm (vector signed char, vector bool char);
17889 vector signed char vec_vsububm (vector signed char, vector signed char);
17890 vector unsigned char vec_vsububm (vector bool char,
17891 vector unsigned char);
17892 vector unsigned char vec_vsububm (vector unsigned char,
17894 vector unsigned char vec_vsububm (vector unsigned char,
17895 vector unsigned char);
17897 vector signed int vec_subc (vector signed int, vector signed int);
17898 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
17899 vector signed __int128 vec_subc (vector signed __int128,
17900 vector signed __int128);
17901 vector unsigned __int128 vec_subc (vector unsigned __int128,
17902 vector unsigned __int128);
17904 vector signed int vec_sube (vector signed int, vector signed int,
17905 vector signed int);
17906 vector unsigned int vec_sube (vector unsigned int, vector unsigned int,
17907 vector unsigned int);
17908 vector signed __int128 vec_sube (vector signed __int128,
17909 vector signed __int128,
17910 vector signed __int128);
17911 vector unsigned __int128 vec_sube (vector unsigned __int128,
17912 vector unsigned __int128,
17913 vector unsigned __int128);
17915 vector signed int vec_subec (vector signed int, vector signed int,
17916 vector signed int);
17917 vector unsigned int vec_subec (vector unsigned int, vector unsigned int,
17918 vector unsigned int);
17919 vector signed __int128 vec_subec (vector signed __int128,
17920 vector signed __int128,
17921 vector signed __int128);
17922 vector unsigned __int128 vec_subec (vector unsigned __int128,
17923 vector unsigned __int128,
17924 vector unsigned __int128);
17926 vector unsigned char vec_subs (vector bool char, vector unsigned char);
17927 vector unsigned char vec_subs (vector unsigned char, vector bool char);
17928 vector unsigned char vec_subs (vector unsigned char,
17929 vector unsigned char);
17930 vector signed char vec_subs (vector bool char, vector signed char);
17931 vector signed char vec_subs (vector signed char, vector bool char);
17932 vector signed char vec_subs (vector signed char, vector signed char);
17933 vector unsigned short vec_subs (vector bool short,
17934 vector unsigned short);
17935 vector unsigned short vec_subs (vector unsigned short,
17936 vector bool short);
17937 vector unsigned short vec_subs (vector unsigned short,
17938 vector unsigned short);
17939 vector signed short vec_subs (vector bool short, vector signed short);
17940 vector signed short vec_subs (vector signed short, vector bool short);
17941 vector signed short vec_subs (vector signed short, vector signed short);
17942 vector unsigned int vec_subs (vector bool int, vector unsigned int);
17943 vector unsigned int vec_subs (vector unsigned int, vector bool int);
17944 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
17945 vector signed int vec_subs (vector bool int, vector signed int);
17946 vector signed int vec_subs (vector signed int, vector bool int);
17947 vector signed int vec_subs (vector signed int, vector signed int);
17949 vector signed int vec_vsubsws (vector bool int, vector signed int);
17950 vector signed int vec_vsubsws (vector signed int, vector bool int);
17951 vector signed int vec_vsubsws (vector signed int, vector signed int);
17953 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
17954 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
17955 vector unsigned int vec_vsubuws (vector unsigned int,
17956 vector unsigned int);
17958 vector signed short vec_vsubshs (vector bool short,
17959 vector signed short);
17960 vector signed short vec_vsubshs (vector signed short,
17961 vector bool short);
17962 vector signed short vec_vsubshs (vector signed short,
17963 vector signed short);
17965 vector unsigned short vec_vsubuhs (vector bool short,
17966 vector unsigned short);
17967 vector unsigned short vec_vsubuhs (vector unsigned short,
17968 vector bool short);
17969 vector unsigned short vec_vsubuhs (vector unsigned short,
17970 vector unsigned short);
17972 vector signed char vec_vsubsbs (vector bool char, vector signed char);
17973 vector signed char vec_vsubsbs (vector signed char, vector bool char);
17974 vector signed char vec_vsubsbs (vector signed char, vector signed char);
17976 vector unsigned char vec_vsububs (vector bool char,
17977 vector unsigned char);
17978 vector unsigned char vec_vsububs (vector unsigned char,
17980 vector unsigned char vec_vsububs (vector unsigned char,
17981 vector unsigned char);
17983 vector unsigned int vec_sum4s (vector unsigned char,
17984 vector unsigned int);
17985 vector signed int vec_sum4s (vector signed char, vector signed int);
17986 vector signed int vec_sum4s (vector signed short, vector signed int);
17988 vector signed int vec_vsum4shs (vector signed short, vector signed int);
17990 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
17992 vector unsigned int vec_vsum4ubs (vector unsigned char,
17993 vector unsigned int);
17995 vector signed int vec_sum2s (vector signed int, vector signed int);
17997 vector signed int vec_sums (vector signed int, vector signed int);
17999 vector float vec_trunc (vector float);
18001 vector signed long long vec_unsigned (vector double);
18002 vector signed int vec_unsigned (vector float);
18004 vector signed int vec_unsignede (vector double);
18005 vector signed int vec_unsignedo (vector double);
18006 vector signed int vec_unsigned2 (vector double, vector double);
18008 vector signed short vec_unpackh (vector signed char);
18009 vector bool short vec_unpackh (vector bool char);
18010 vector signed int vec_unpackh (vector signed short);
18011 vector bool int vec_unpackh (vector bool short);
18012 vector unsigned int vec_unpackh (vector pixel);
18013 vector double vec_unpackh (vector float);
18015 vector bool int vec_vupkhsh (vector bool short);
18016 vector signed int vec_vupkhsh (vector signed short);
18018 vector unsigned int vec_vupkhpx (vector pixel);
18020 vector bool short vec_vupkhsb (vector bool char);
18021 vector signed short vec_vupkhsb (vector signed char);
18023 vector signed short vec_unpackl (vector signed char);
18024 vector bool short vec_unpackl (vector bool char);
18025 vector unsigned int vec_unpackl (vector pixel);
18026 vector signed int vec_unpackl (vector signed short);
18027 vector bool int vec_unpackl (vector bool short);
18028 vector double vec_unpackl (vector float);
18030 vector unsigned int vec_vupklpx (vector pixel);
18032 vector bool int vec_vupklsh (vector bool short);
18033 vector signed int vec_vupklsh (vector signed short);
18035 vector bool short vec_vupklsb (vector bool char);
18036 vector signed short vec_vupklsb (vector signed char);
18038 vector float vec_xor (vector float, vector float);
18039 vector float vec_xor (vector float, vector bool int);
18040 vector float vec_xor (vector bool int, vector float);
18041 vector bool int vec_xor (vector bool int, vector bool int);
18042 vector signed int vec_xor (vector bool int, vector signed int);
18043 vector signed int vec_xor (vector signed int, vector bool int);
18044 vector signed int vec_xor (vector signed int, vector signed int);
18045 vector unsigned int vec_xor (vector bool int, vector unsigned int);
18046 vector unsigned int vec_xor (vector unsigned int, vector bool int);
18047 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
18048 vector bool short vec_xor (vector bool short, vector bool short);
18049 vector signed short vec_xor (vector bool short, vector signed short);
18050 vector signed short vec_xor (vector signed short, vector bool short);
18051 vector signed short vec_xor (vector signed short, vector signed short);
18052 vector unsigned short vec_xor (vector bool short,
18053 vector unsigned short);
18054 vector unsigned short vec_xor (vector unsigned short,
18055 vector bool short);
18056 vector unsigned short vec_xor (vector unsigned short,
18057 vector unsigned short);
18058 vector signed char vec_xor (vector bool char, vector signed char);
18059 vector bool char vec_xor (vector bool char, vector bool char);
18060 vector signed char vec_xor (vector signed char, vector bool char);
18061 vector signed char vec_xor (vector signed char, vector signed char);
18062 vector unsigned char vec_xor (vector bool char, vector unsigned char);
18063 vector unsigned char vec_xor (vector unsigned char, vector bool char);
18064 vector unsigned char vec_xor (vector unsigned char,
18065 vector unsigned char);
18067 int vec_all_eq (vector signed char, vector bool char);
18068 int vec_all_eq (vector signed char, vector signed char);
18069 int vec_all_eq (vector unsigned char, vector bool char);
18070 int vec_all_eq (vector unsigned char, vector unsigned char);
18071 int vec_all_eq (vector bool char, vector bool char);
18072 int vec_all_eq (vector bool char, vector unsigned char);
18073 int vec_all_eq (vector bool char, vector signed char);
18074 int vec_all_eq (vector signed short, vector bool short);
18075 int vec_all_eq (vector signed short, vector signed short);
18076 int vec_all_eq (vector unsigned short, vector bool short);
18077 int vec_all_eq (vector unsigned short, vector unsigned short);
18078 int vec_all_eq (vector bool short, vector bool short);
18079 int vec_all_eq (vector bool short, vector unsigned short);
18080 int vec_all_eq (vector bool short, vector signed short);
18081 int vec_all_eq (vector pixel, vector pixel);
18082 int vec_all_eq (vector signed int, vector bool int);
18083 int vec_all_eq (vector signed int, vector signed int);
18084 int vec_all_eq (vector unsigned int, vector bool int);
18085 int vec_all_eq (vector unsigned int, vector unsigned int);
18086 int vec_all_eq (vector bool int, vector bool int);
18087 int vec_all_eq (vector bool int, vector unsigned int);
18088 int vec_all_eq (vector bool int, vector signed int);
18089 int vec_all_eq (vector float, vector float);
18091 int vec_all_ge (vector bool char, vector unsigned char);
18092 int vec_all_ge (vector unsigned char, vector bool char);
18093 int vec_all_ge (vector unsigned char, vector unsigned char);
18094 int vec_all_ge (vector bool char, vector signed char);
18095 int vec_all_ge (vector signed char, vector bool char);
18096 int vec_all_ge (vector signed char, vector signed char);
18097 int vec_all_ge (vector bool short, vector unsigned short);
18098 int vec_all_ge (vector unsigned short, vector bool short);
18099 int vec_all_ge (vector unsigned short, vector unsigned short);
18100 int vec_all_ge (vector signed short, vector signed short);
18101 int vec_all_ge (vector bool short, vector signed short);
18102 int vec_all_ge (vector signed short, vector bool short);
18103 int vec_all_ge (vector bool int, vector unsigned int);
18104 int vec_all_ge (vector unsigned int, vector bool int);
18105 int vec_all_ge (vector unsigned int, vector unsigned int);
18106 int vec_all_ge (vector bool int, vector signed int);
18107 int vec_all_ge (vector signed int, vector bool int);
18108 int vec_all_ge (vector signed int, vector signed int);
18109 int vec_all_ge (vector float, vector float);
18111 int vec_all_gt (vector bool char, vector unsigned char);
18112 int vec_all_gt (vector unsigned char, vector bool char);
18113 int vec_all_gt (vector unsigned char, vector unsigned char);
18114 int vec_all_gt (vector bool char, vector signed char);
18115 int vec_all_gt (vector signed char, vector bool char);
18116 int vec_all_gt (vector signed char, vector signed char);
18117 int vec_all_gt (vector bool short, vector unsigned short);
18118 int vec_all_gt (vector unsigned short, vector bool short);
18119 int vec_all_gt (vector unsigned short, vector unsigned short);
18120 int vec_all_gt (vector bool short, vector signed short);
18121 int vec_all_gt (vector signed short, vector bool short);
18122 int vec_all_gt (vector signed short, vector signed short);
18123 int vec_all_gt (vector bool int, vector unsigned int);
18124 int vec_all_gt (vector unsigned int, vector bool int);
18125 int vec_all_gt (vector unsigned int, vector unsigned int);
18126 int vec_all_gt (vector bool int, vector signed int);
18127 int vec_all_gt (vector signed int, vector bool int);
18128 int vec_all_gt (vector signed int, vector signed int);
18129 int vec_all_gt (vector float, vector float);
18131 int vec_all_in (vector float, vector float);
18133 int vec_all_le (vector bool char, vector unsigned char);
18134 int vec_all_le (vector unsigned char, vector bool char);
18135 int vec_all_le (vector unsigned char, vector unsigned char);
18136 int vec_all_le (vector bool char, vector signed char);
18137 int vec_all_le (vector signed char, vector bool char);
18138 int vec_all_le (vector signed char, vector signed char);
18139 int vec_all_le (vector bool short, vector unsigned short);
18140 int vec_all_le (vector unsigned short, vector bool short);
18141 int vec_all_le (vector unsigned short, vector unsigned short);
18142 int vec_all_le (vector bool short, vector signed short);
18143 int vec_all_le (vector signed short, vector bool short);
18144 int vec_all_le (vector signed short, vector signed short);
18145 int vec_all_le (vector bool int, vector unsigned int);
18146 int vec_all_le (vector unsigned int, vector bool int);
18147 int vec_all_le (vector unsigned int, vector unsigned int);
18148 int vec_all_le (vector bool int, vector signed int);
18149 int vec_all_le (vector signed int, vector bool int);
18150 int vec_all_le (vector signed int, vector signed int);
18151 int vec_all_le (vector float, vector float);
18153 int vec_all_lt (vector bool char, vector unsigned char);
18154 int vec_all_lt (vector unsigned char, vector bool char);
18155 int vec_all_lt (vector unsigned char, vector unsigned char);
18156 int vec_all_lt (vector bool char, vector signed char);
18157 int vec_all_lt (vector signed char, vector bool char);
18158 int vec_all_lt (vector signed char, vector signed char);
18159 int vec_all_lt (vector bool short, vector unsigned short);
18160 int vec_all_lt (vector unsigned short, vector bool short);
18161 int vec_all_lt (vector unsigned short, vector unsigned short);
18162 int vec_all_lt (vector bool short, vector signed short);
18163 int vec_all_lt (vector signed short, vector bool short);
18164 int vec_all_lt (vector signed short, vector signed short);
18165 int vec_all_lt (vector bool int, vector unsigned int);
18166 int vec_all_lt (vector unsigned int, vector bool int);
18167 int vec_all_lt (vector unsigned int, vector unsigned int);
18168 int vec_all_lt (vector bool int, vector signed int);
18169 int vec_all_lt (vector signed int, vector bool int);
18170 int vec_all_lt (vector signed int, vector signed int);
18171 int vec_all_lt (vector float, vector float);
18173 int vec_all_nan (vector float);
18175 int vec_all_ne (vector signed char, vector bool char);
18176 int vec_all_ne (vector signed char, vector signed char);
18177 int vec_all_ne (vector unsigned char, vector bool char);
18178 int vec_all_ne (vector unsigned char, vector unsigned char);
18179 int vec_all_ne (vector bool char, vector bool char);
18180 int vec_all_ne (vector bool char, vector unsigned char);
18181 int vec_all_ne (vector bool char, vector signed char);
18182 int vec_all_ne (vector signed short, vector bool short);
18183 int vec_all_ne (vector signed short, vector signed short);
18184 int vec_all_ne (vector unsigned short, vector bool short);
18185 int vec_all_ne (vector unsigned short, vector unsigned short);
18186 int vec_all_ne (vector bool short, vector bool short);
18187 int vec_all_ne (vector bool short, vector unsigned short);
18188 int vec_all_ne (vector bool short, vector signed short);
18189 int vec_all_ne (vector pixel, vector pixel);
18190 int vec_all_ne (vector signed int, vector bool int);
18191 int vec_all_ne (vector signed int, vector signed int);
18192 int vec_all_ne (vector unsigned int, vector bool int);
18193 int vec_all_ne (vector unsigned int, vector unsigned int);
18194 int vec_all_ne (vector bool int, vector bool int);
18195 int vec_all_ne (vector bool int, vector unsigned int);
18196 int vec_all_ne (vector bool int, vector signed int);
18197 int vec_all_ne (vector float, vector float);
18199 int vec_all_nge (vector float, vector float);
18201 int vec_all_ngt (vector float, vector float);
18203 int vec_all_nle (vector float, vector float);
18205 int vec_all_nlt (vector float, vector float);
18207 int vec_all_numeric (vector float);
18209 int vec_any_eq (vector signed char, vector bool char);
18210 int vec_any_eq (vector signed char, vector signed char);
18211 int vec_any_eq (vector unsigned char, vector bool char);
18212 int vec_any_eq (vector unsigned char, vector unsigned char);
18213 int vec_any_eq (vector bool char, vector bool char);
18214 int vec_any_eq (vector bool char, vector unsigned char);
18215 int vec_any_eq (vector bool char, vector signed char);
18216 int vec_any_eq (vector signed short, vector bool short);
18217 int vec_any_eq (vector signed short, vector signed short);
18218 int vec_any_eq (vector unsigned short, vector bool short);
18219 int vec_any_eq (vector unsigned short, vector unsigned short);
18220 int vec_any_eq (vector bool short, vector bool short);
18221 int vec_any_eq (vector bool short, vector unsigned short);
18222 int vec_any_eq (vector bool short, vector signed short);
18223 int vec_any_eq (vector pixel, vector pixel);
18224 int vec_any_eq (vector signed int, vector bool int);
18225 int vec_any_eq (vector signed int, vector signed int);
18226 int vec_any_eq (vector unsigned int, vector bool int);
18227 int vec_any_eq (vector unsigned int, vector unsigned int);
18228 int vec_any_eq (vector bool int, vector bool int);
18229 int vec_any_eq (vector bool int, vector unsigned int);
18230 int vec_any_eq (vector bool int, vector signed int);
18231 int vec_any_eq (vector float, vector float);
18233 int vec_any_ge (vector signed char, vector bool char);
18234 int vec_any_ge (vector unsigned char, vector bool char);
18235 int vec_any_ge (vector unsigned char, vector unsigned char);
18236 int vec_any_ge (vector signed char, vector signed char);
18237 int vec_any_ge (vector bool char, vector unsigned char);
18238 int vec_any_ge (vector bool char, vector signed char);
18239 int vec_any_ge (vector unsigned short, vector bool short);
18240 int vec_any_ge (vector unsigned short, vector unsigned short);
18241 int vec_any_ge (vector signed short, vector signed short);
18242 int vec_any_ge (vector signed short, vector bool short);
18243 int vec_any_ge (vector bool short, vector unsigned short);
18244 int vec_any_ge (vector bool short, vector signed short);
18245 int vec_any_ge (vector signed int, vector bool int);
18246 int vec_any_ge (vector unsigned int, vector bool int);
18247 int vec_any_ge (vector unsigned int, vector unsigned int);
18248 int vec_any_ge (vector signed int, vector signed int);
18249 int vec_any_ge (vector bool int, vector unsigned int);
18250 int vec_any_ge (vector bool int, vector signed int);
18251 int vec_any_ge (vector float, vector float);
18253 int vec_any_gt (vector bool char, vector unsigned char);
18254 int vec_any_gt (vector unsigned char, vector bool char);
18255 int vec_any_gt (vector unsigned char, vector unsigned char);
18256 int vec_any_gt (vector bool char, vector signed char);
18257 int vec_any_gt (vector signed char, vector bool char);
18258 int vec_any_gt (vector signed char, vector signed char);
18259 int vec_any_gt (vector bool short, vector unsigned short);
18260 int vec_any_gt (vector unsigned short, vector bool short);
18261 int vec_any_gt (vector unsigned short, vector unsigned short);
18262 int vec_any_gt (vector bool short, vector signed short);
18263 int vec_any_gt (vector signed short, vector bool short);
18264 int vec_any_gt (vector signed short, vector signed short);
18265 int vec_any_gt (vector bool int, vector unsigned int);
18266 int vec_any_gt (vector unsigned int, vector bool int);
18267 int vec_any_gt (vector unsigned int, vector unsigned int);
18268 int vec_any_gt (vector bool int, vector signed int);
18269 int vec_any_gt (vector signed int, vector bool int);
18270 int vec_any_gt (vector signed int, vector signed int);
18271 int vec_any_gt (vector float, vector float);
18273 int vec_any_le (vector bool char, vector unsigned char);
18274 int vec_any_le (vector unsigned char, vector bool char);
18275 int vec_any_le (vector unsigned char, vector unsigned char);
18276 int vec_any_le (vector bool char, vector signed char);
18277 int vec_any_le (vector signed char, vector bool char);
18278 int vec_any_le (vector signed char, vector signed char);
18279 int vec_any_le (vector bool short, vector unsigned short);
18280 int vec_any_le (vector unsigned short, vector bool short);
18281 int vec_any_le (vector unsigned short, vector unsigned short);
18282 int vec_any_le (vector bool short, vector signed short);
18283 int vec_any_le (vector signed short, vector bool short);
18284 int vec_any_le (vector signed short, vector signed short);
18285 int vec_any_le (vector bool int, vector unsigned int);
18286 int vec_any_le (vector unsigned int, vector bool int);
18287 int vec_any_le (vector unsigned int, vector unsigned int);
18288 int vec_any_le (vector bool int, vector signed int);
18289 int vec_any_le (vector signed int, vector bool int);
18290 int vec_any_le (vector signed int, vector signed int);
18291 int vec_any_le (vector float, vector float);
18293 int vec_any_lt (vector bool char, vector unsigned char);
18294 int vec_any_lt (vector unsigned char, vector bool char);
18295 int vec_any_lt (vector unsigned char, vector unsigned char);
18296 int vec_any_lt (vector bool char, vector signed char);
18297 int vec_any_lt (vector signed char, vector bool char);
18298 int vec_any_lt (vector signed char, vector signed char);
18299 int vec_any_lt (vector bool short, vector unsigned short);
18300 int vec_any_lt (vector unsigned short, vector bool short);
18301 int vec_any_lt (vector unsigned short, vector unsigned short);
18302 int vec_any_lt (vector bool short, vector signed short);
18303 int vec_any_lt (vector signed short, vector bool short);
18304 int vec_any_lt (vector signed short, vector signed short);
18305 int vec_any_lt (vector bool int, vector unsigned int);
18306 int vec_any_lt (vector unsigned int, vector bool int);
18307 int vec_any_lt (vector unsigned int, vector unsigned int);
18308 int vec_any_lt (vector bool int, vector signed int);
18309 int vec_any_lt (vector signed int, vector bool int);
18310 int vec_any_lt (vector signed int, vector signed int);
18311 int vec_any_lt (vector float, vector float);
18313 int vec_any_nan (vector float);
18315 int vec_any_ne (vector signed char, vector bool char);
18316 int vec_any_ne (vector signed char, vector signed char);
18317 int vec_any_ne (vector unsigned char, vector bool char);
18318 int vec_any_ne (vector unsigned char, vector unsigned char);
18319 int vec_any_ne (vector bool char, vector bool char);
18320 int vec_any_ne (vector bool char, vector unsigned char);
18321 int vec_any_ne (vector bool char, vector signed char);
18322 int vec_any_ne (vector signed short, vector bool short);
18323 int vec_any_ne (vector signed short, vector signed short);
18324 int vec_any_ne (vector unsigned short, vector bool short);
18325 int vec_any_ne (vector unsigned short, vector unsigned short);
18326 int vec_any_ne (vector bool short, vector bool short);
18327 int vec_any_ne (vector bool short, vector unsigned short);
18328 int vec_any_ne (vector bool short, vector signed short);
18329 int vec_any_ne (vector pixel, vector pixel);
18330 int vec_any_ne (vector signed int, vector bool int);
18331 int vec_any_ne (vector signed int, vector signed int);
18332 int vec_any_ne (vector unsigned int, vector bool int);
18333 int vec_any_ne (vector unsigned int, vector unsigned int);
18334 int vec_any_ne (vector bool int, vector bool int);
18335 int vec_any_ne (vector bool int, vector unsigned int);
18336 int vec_any_ne (vector bool int, vector signed int);
18337 int vec_any_ne (vector float, vector float);
18339 int vec_any_nge (vector float, vector float);
18341 int vec_any_ngt (vector float, vector float);
18343 int vec_any_nle (vector float, vector float);
18345 int vec_any_nlt (vector float, vector float);
18347 int vec_any_numeric (vector float);
18349 int vec_any_out (vector float, vector float);
18352 If the vector/scalar (VSX) instruction set is available, the following
18353 additional functions are available:
18356 vector double vec_abs (vector double);
18357 vector double vec_add (vector double, vector double);
18358 vector double vec_and (vector double, vector double);
18359 vector double vec_and (vector double, vector bool long);
18360 vector double vec_and (vector bool long, vector double);
18361 vector long vec_and (vector long, vector long);
18362 vector long vec_and (vector long, vector bool long);
18363 vector long vec_and (vector bool long, vector long);
18364 vector unsigned long vec_and (vector unsigned long, vector unsigned long);
18365 vector unsigned long vec_and (vector unsigned long, vector bool long);
18366 vector unsigned long vec_and (vector bool long, vector unsigned long);
18367 vector double vec_andc (vector double, vector double);
18368 vector double vec_andc (vector double, vector bool long);
18369 vector double vec_andc (vector bool long, vector double);
18370 vector long vec_andc (vector long, vector long);
18371 vector long vec_andc (vector long, vector bool long);
18372 vector long vec_andc (vector bool long, vector long);
18373 vector unsigned long vec_andc (vector unsigned long, vector unsigned long);
18374 vector unsigned long vec_andc (vector unsigned long, vector bool long);
18375 vector unsigned long vec_andc (vector bool long, vector unsigned long);
18376 vector double vec_ceil (vector double);
18377 vector bool long vec_cmpeq (vector double, vector double);
18378 vector bool long vec_cmpge (vector double, vector double);
18379 vector bool long vec_cmpgt (vector double, vector double);
18380 vector bool long vec_cmple (vector double, vector double);
18381 vector bool long vec_cmplt (vector double, vector double);
18382 vector double vec_cpsgn (vector double, vector double);
18383 vector float vec_div (vector float, vector float);
18384 vector double vec_div (vector double, vector double);
18385 vector long vec_div (vector long, vector long);
18386 vector unsigned long vec_div (vector unsigned long, vector unsigned long);
18387 vector double vec_floor (vector double);
18388 vector double vec_ld (int, const vector double *);
18389 vector double vec_ld (int, const double *);
18390 vector double vec_ldl (int, const vector double *);
18391 vector double vec_ldl (int, const double *);
18392 vector unsigned char vec_lvsl (int, const volatile double *);
18393 vector unsigned char vec_lvsr (int, const volatile double *);
18394 vector double vec_madd (vector double, vector double, vector double);
18395 vector double vec_max (vector double, vector double);
18396 vector signed long vec_mergeh (vector signed long, vector signed long);
18397 vector signed long vec_mergeh (vector signed long, vector bool long);
18398 vector signed long vec_mergeh (vector bool long, vector signed long);
18399 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long);
18400 vector unsigned long vec_mergeh (vector unsigned long, vector bool long);
18401 vector unsigned long vec_mergeh (vector bool long, vector unsigned long);
18402 vector signed long vec_mergel (vector signed long, vector signed long);
18403 vector signed long vec_mergel (vector signed long, vector bool long);
18404 vector signed long vec_mergel (vector bool long, vector signed long);
18405 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long);
18406 vector unsigned long vec_mergel (vector unsigned long, vector bool long);
18407 vector unsigned long vec_mergel (vector bool long, vector unsigned long);
18408 vector double vec_min (vector double, vector double);
18409 vector float vec_msub (vector float, vector float, vector float);
18410 vector double vec_msub (vector double, vector double, vector double);
18411 vector float vec_mul (vector float, vector float);
18412 vector double vec_mul (vector double, vector double);
18413 vector long vec_mul (vector long, vector long);
18414 vector unsigned long vec_mul (vector unsigned long, vector unsigned long);
18415 vector float vec_nearbyint (vector float);
18416 vector double vec_nearbyint (vector double);
18417 vector float vec_nmadd (vector float, vector float, vector float);
18418 vector double vec_nmadd (vector double, vector double, vector double);
18419 vector double vec_nmsub (vector double, vector double, vector double);
18420 vector double vec_nor (vector double, vector double);
18421 vector long vec_nor (vector long, vector long);
18422 vector long vec_nor (vector long, vector bool long);
18423 vector long vec_nor (vector bool long, vector long);
18424 vector unsigned long vec_nor (vector unsigned long, vector unsigned long);
18425 vector unsigned long vec_nor (vector unsigned long, vector bool long);
18426 vector unsigned long vec_nor (vector bool long, vector unsigned long);
18427 vector double vec_or (vector double, vector double);
18428 vector double vec_or (vector double, vector bool long);
18429 vector double vec_or (vector bool long, vector double);
18430 vector long vec_or (vector long, vector long);
18431 vector long vec_or (vector long, vector bool long);
18432 vector long vec_or (vector bool long, vector long);
18433 vector unsigned long vec_or (vector unsigned long, vector unsigned long);
18434 vector unsigned long vec_or (vector unsigned long, vector bool long);
18435 vector unsigned long vec_or (vector bool long, vector unsigned long);
18436 vector double vec_perm (vector double, vector double, vector unsigned char);
18437 vector long vec_perm (vector long, vector long, vector unsigned char);
18438 vector unsigned long vec_perm (vector unsigned long, vector unsigned long,
18439 vector unsigned char);
18440 vector double vec_rint (vector double);
18441 vector double vec_recip (vector double, vector double);
18442 vector double vec_rsqrt (vector double);
18443 vector double vec_rsqrte (vector double);
18444 vector double vec_sel (vector double, vector double, vector bool long);
18445 vector double vec_sel (vector double, vector double, vector unsigned long);
18446 vector long vec_sel (vector long, vector long, vector long);
18447 vector long vec_sel (vector long, vector long, vector unsigned long);
18448 vector long vec_sel (vector long, vector long, vector bool long);
18449 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
18451 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
18452 vector unsigned long);
18453 vector unsigned long vec_sel (vector unsigned long, vector unsigned long,
18455 vector double vec_splats (double);
18456 vector signed long vec_splats (signed long);
18457 vector unsigned long vec_splats (unsigned long);
18458 vector float vec_sqrt (vector float);
18459 vector double vec_sqrt (vector double);
18460 void vec_st (vector double, int, vector double *);
18461 void vec_st (vector double, int, double *);
18462 vector double vec_sub (vector double, vector double);
18463 vector double vec_trunc (vector double);
18464 vector double vec_xl (int, vector double *);
18465 vector double vec_xl (int, double *);
18466 vector long long vec_xl (int, vector long long *);
18467 vector long long vec_xl (int, long long *);
18468 vector unsigned long long vec_xl (int, vector unsigned long long *);
18469 vector unsigned long long vec_xl (int, unsigned long long *);
18470 vector float vec_xl (int, vector float *);
18471 vector float vec_xl (int, float *);
18472 vector int vec_xl (int, vector int *);
18473 vector int vec_xl (int, int *);
18474 vector unsigned int vec_xl (int, vector unsigned int *);
18475 vector unsigned int vec_xl (int, unsigned int *);
18476 vector double vec_xor (vector double, vector double);
18477 vector double vec_xor (vector double, vector bool long);
18478 vector double vec_xor (vector bool long, vector double);
18479 vector long vec_xor (vector long, vector long);
18480 vector long vec_xor (vector long, vector bool long);
18481 vector long vec_xor (vector bool long, vector long);
18482 vector unsigned long vec_xor (vector unsigned long, vector unsigned long);
18483 vector unsigned long vec_xor (vector unsigned long, vector bool long);
18484 vector unsigned long vec_xor (vector bool long, vector unsigned long);
18485 void vec_xst (vector double, int, vector double *);
18486 void vec_xst (vector double, int, double *);
18487 void vec_xst (vector long long, int, vector long long *);
18488 void vec_xst (vector long long, int, long long *);
18489 void vec_xst (vector unsigned long long, int, vector unsigned long long *);
18490 void vec_xst (vector unsigned long long, int, unsigned long long *);
18491 void vec_xst (vector float, int, vector float *);
18492 void vec_xst (vector float, int, float *);
18493 void vec_xst (vector int, int, vector int *);
18494 void vec_xst (vector int, int, int *);
18495 void vec_xst (vector unsigned int, int, vector unsigned int *);
18496 void vec_xst (vector unsigned int, int, unsigned int *);
18497 int vec_all_eq (vector double, vector double);
18498 int vec_all_ge (vector double, vector double);
18499 int vec_all_gt (vector double, vector double);
18500 int vec_all_le (vector double, vector double);
18501 int vec_all_lt (vector double, vector double);
18502 int vec_all_nan (vector double);
18503 int vec_all_ne (vector double, vector double);
18504 int vec_all_nge (vector double, vector double);
18505 int vec_all_ngt (vector double, vector double);
18506 int vec_all_nle (vector double, vector double);
18507 int vec_all_nlt (vector double, vector double);
18508 int vec_all_numeric (vector double);
18509 int vec_any_eq (vector double, vector double);
18510 int vec_any_ge (vector double, vector double);
18511 int vec_any_gt (vector double, vector double);
18512 int vec_any_le (vector double, vector double);
18513 int vec_any_lt (vector double, vector double);
18514 int vec_any_nan (vector double);
18515 int vec_any_ne (vector double, vector double);
18516 int vec_any_nge (vector double, vector double);
18517 int vec_any_ngt (vector double, vector double);
18518 int vec_any_nle (vector double, vector double);
18519 int vec_any_nlt (vector double, vector double);
18520 int vec_any_numeric (vector double);
18522 vector double vec_vsx_ld (int, const vector double *);
18523 vector double vec_vsx_ld (int, const double *);
18524 vector float vec_vsx_ld (int, const vector float *);
18525 vector float vec_vsx_ld (int, const float *);
18526 vector bool int vec_vsx_ld (int, const vector bool int *);
18527 vector signed int vec_vsx_ld (int, const vector signed int *);
18528 vector signed int vec_vsx_ld (int, const int *);
18529 vector signed int vec_vsx_ld (int, const long *);
18530 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
18531 vector unsigned int vec_vsx_ld (int, const unsigned int *);
18532 vector unsigned int vec_vsx_ld (int, const unsigned long *);
18533 vector bool short vec_vsx_ld (int, const vector bool short *);
18534 vector pixel vec_vsx_ld (int, const vector pixel *);
18535 vector signed short vec_vsx_ld (int, const vector signed short *);
18536 vector signed short vec_vsx_ld (int, const short *);
18537 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
18538 vector unsigned short vec_vsx_ld (int, const unsigned short *);
18539 vector bool char vec_vsx_ld (int, const vector bool char *);
18540 vector signed char vec_vsx_ld (int, const vector signed char *);
18541 vector signed char vec_vsx_ld (int, const signed char *);
18542 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
18543 vector unsigned char vec_vsx_ld (int, const unsigned char *);
18545 void vec_vsx_st (vector double, int, vector double *);
18546 void vec_vsx_st (vector double, int, double *);
18547 void vec_vsx_st (vector float, int, vector float *);
18548 void vec_vsx_st (vector float, int, float *);
18549 void vec_vsx_st (vector signed int, int, vector signed int *);
18550 void vec_vsx_st (vector signed int, int, int *);
18551 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
18552 void vec_vsx_st (vector unsigned int, int, unsigned int *);
18553 void vec_vsx_st (vector bool int, int, vector bool int *);
18554 void vec_vsx_st (vector bool int, int, unsigned int *);
18555 void vec_vsx_st (vector bool int, int, int *);
18556 void vec_vsx_st (vector signed short, int, vector signed short *);
18557 void vec_vsx_st (vector signed short, int, short *);
18558 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
18559 void vec_vsx_st (vector unsigned short, int, unsigned short *);
18560 void vec_vsx_st (vector bool short, int, vector bool short *);
18561 void vec_vsx_st (vector bool short, int, unsigned short *);
18562 void vec_vsx_st (vector pixel, int, vector pixel *);
18563 void vec_vsx_st (vector pixel, int, unsigned short *);
18564 void vec_vsx_st (vector pixel, int, short *);
18565 void vec_vsx_st (vector bool short, int, short *);
18566 void vec_vsx_st (vector signed char, int, vector signed char *);
18567 void vec_vsx_st (vector signed char, int, signed char *);
18568 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
18569 void vec_vsx_st (vector unsigned char, int, unsigned char *);
18570 void vec_vsx_st (vector bool char, int, vector bool char *);
18571 void vec_vsx_st (vector bool char, int, unsigned char *);
18572 void vec_vsx_st (vector bool char, int, signed char *);
18574 vector double vec_xxpermdi (vector double, vector double, const int);
18575 vector float vec_xxpermdi (vector float, vector float, const int);
18576 vector long long vec_xxpermdi (vector long long, vector long long, const int);
18577 vector unsigned long long vec_xxpermdi (vector unsigned long long,
18578 vector unsigned long long, const int);
18579 vector int vec_xxpermdi (vector int, vector int, const int);
18580 vector unsigned int vec_xxpermdi (vector unsigned int,
18581 vector unsigned int, const int);
18582 vector short vec_xxpermdi (vector short, vector short, const int);
18583 vector unsigned short vec_xxpermdi (vector unsigned short,
18584 vector unsigned short, const int);
18585 vector signed char vec_xxpermdi (vector signed char, vector signed char,
18587 vector unsigned char vec_xxpermdi (vector unsigned char,
18588 vector unsigned char, const int);
18590 vector double vec_xxsldi (vector double, vector double, int);
18591 vector float vec_xxsldi (vector float, vector float, int);
18592 vector long long vec_xxsldi (vector long long, vector long long, int);
18593 vector unsigned long long vec_xxsldi (vector unsigned long long,
18594 vector unsigned long long, int);
18595 vector int vec_xxsldi (vector int, vector int, int);
18596 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int);
18597 vector short vec_xxsldi (vector short, vector short, int);
18598 vector unsigned short vec_xxsldi (vector unsigned short,
18599 vector unsigned short, int);
18600 vector signed char vec_xxsldi (vector signed char, vector signed char, int);
18601 vector unsigned char vec_xxsldi (vector unsigned char,
18602 vector unsigned char, int);
18605 Note that the @samp{vec_ld} and @samp{vec_st} built-in functions always
18606 generate the AltiVec @samp{LVX} and @samp{STVX} instructions even
18607 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
18608 @samp{vec_vsx_st} built-in functions always generate the VSX @samp{LXVD2X},
18609 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
18611 If the ISA 2.07 additions to the vector/scalar (power8-vector)
18612 instruction set are available, the following additional functions are
18613 available for both 32-bit and 64-bit targets. For 64-bit targets, you
18614 can use @var{vector long} instead of @var{vector long long},
18615 @var{vector bool long} instead of @var{vector bool long long}, and
18616 @var{vector unsigned long} instead of @var{vector unsigned long long}.
18619 vector long long vec_abs (vector long long);
18621 vector long long vec_add (vector long long, vector long long);
18622 vector unsigned long long vec_add (vector unsigned long long,
18623 vector unsigned long long);
18625 int vec_all_eq (vector long long, vector long long);
18626 int vec_all_eq (vector unsigned long long, vector unsigned long long);
18627 int vec_all_ge (vector long long, vector long long);
18628 int vec_all_ge (vector unsigned long long, vector unsigned long long);
18629 int vec_all_gt (vector long long, vector long long);
18630 int vec_all_gt (vector unsigned long long, vector unsigned long long);
18631 int vec_all_le (vector long long, vector long long);
18632 int vec_all_le (vector unsigned long long, vector unsigned long long);
18633 int vec_all_lt (vector long long, vector long long);
18634 int vec_all_lt (vector unsigned long long, vector unsigned long long);
18635 int vec_all_ne (vector long long, vector long long);
18636 int vec_all_ne (vector unsigned long long, vector unsigned long long);
18638 int vec_any_eq (vector long long, vector long long);
18639 int vec_any_eq (vector unsigned long long, vector unsigned long long);
18640 int vec_any_ge (vector long long, vector long long);
18641 int vec_any_ge (vector unsigned long long, vector unsigned long long);
18642 int vec_any_gt (vector long long, vector long long);
18643 int vec_any_gt (vector unsigned long long, vector unsigned long long);
18644 int vec_any_le (vector long long, vector long long);
18645 int vec_any_le (vector unsigned long long, vector unsigned long long);
18646 int vec_any_lt (vector long long, vector long long);
18647 int vec_any_lt (vector unsigned long long, vector unsigned long long);
18648 int vec_any_ne (vector long long, vector long long);
18649 int vec_any_ne (vector unsigned long long, vector unsigned long long);
18651 vector bool long long vec_cmpeq (vector bool long long, vector bool long long);
18653 vector long long vec_eqv (vector long long, vector long long);
18654 vector long long vec_eqv (vector bool long long, vector long long);
18655 vector long long vec_eqv (vector long long, vector bool long long);
18656 vector unsigned long long vec_eqv (vector unsigned long long,
18657 vector unsigned long long);
18658 vector unsigned long long vec_eqv (vector bool long long,
18659 vector unsigned long long);
18660 vector unsigned long long vec_eqv (vector unsigned long long,
18661 vector bool long long);
18662 vector int vec_eqv (vector int, vector int);
18663 vector int vec_eqv (vector bool int, vector int);
18664 vector int vec_eqv (vector int, vector bool int);
18665 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int);
18666 vector unsigned int vec_eqv (vector bool unsigned int,
18667 vector unsigned int);
18668 vector unsigned int vec_eqv (vector unsigned int,
18669 vector bool unsigned int);
18670 vector short vec_eqv (vector short, vector short);
18671 vector short vec_eqv (vector bool short, vector short);
18672 vector short vec_eqv (vector short, vector bool short);
18673 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short);
18674 vector unsigned short vec_eqv (vector bool unsigned short,
18675 vector unsigned short);
18676 vector unsigned short vec_eqv (vector unsigned short,
18677 vector bool unsigned short);
18678 vector signed char vec_eqv (vector signed char, vector signed char);
18679 vector signed char vec_eqv (vector bool signed char, vector signed char);
18680 vector signed char vec_eqv (vector signed char, vector bool signed char);
18681 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char);
18682 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char);
18683 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char);
18685 vector long long vec_max (vector long long, vector long long);
18686 vector unsigned long long vec_max (vector unsigned long long,
18687 vector unsigned long long);
18689 vector signed int vec_mergee (vector signed int, vector signed int);
18690 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int);
18691 vector bool int vec_mergee (vector bool int, vector bool int);
18693 vector signed int vec_mergeo (vector signed int, vector signed int);
18694 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int);
18695 vector bool int vec_mergeo (vector bool int, vector bool int);
18697 vector long long vec_min (vector long long, vector long long);
18698 vector unsigned long long vec_min (vector unsigned long long,
18699 vector unsigned long long);
18701 vector signed long long vec_nabs (vector signed long long);
18703 vector long long vec_nand (vector long long, vector long long);
18704 vector long long vec_nand (vector bool long long, vector long long);
18705 vector long long vec_nand (vector long long, vector bool long long);
18706 vector unsigned long long vec_nand (vector unsigned long long,
18707 vector unsigned long long);
18708 vector unsigned long long vec_nand (vector bool long long,
18709 vector unsigned long long);
18710 vector unsigned long long vec_nand (vector unsigned long long,
18711 vector bool long long);
18712 vector int vec_nand (vector int, vector int);
18713 vector int vec_nand (vector bool int, vector int);
18714 vector int vec_nand (vector int, vector bool int);
18715 vector unsigned int vec_nand (vector unsigned int, vector unsigned int);
18716 vector unsigned int vec_nand (vector bool unsigned int,
18717 vector unsigned int);
18718 vector unsigned int vec_nand (vector unsigned int,
18719 vector bool unsigned int);
18720 vector short vec_nand (vector short, vector short);
18721 vector short vec_nand (vector bool short, vector short);
18722 vector short vec_nand (vector short, vector bool short);
18723 vector unsigned short vec_nand (vector unsigned short, vector unsigned short);
18724 vector unsigned short vec_nand (vector bool unsigned short,
18725 vector unsigned short);
18726 vector unsigned short vec_nand (vector unsigned short,
18727 vector bool unsigned short);
18728 vector signed char vec_nand (vector signed char, vector signed char);
18729 vector signed char vec_nand (vector bool signed char, vector signed char);
18730 vector signed char vec_nand (vector signed char, vector bool signed char);
18731 vector unsigned char vec_nand (vector unsigned char, vector unsigned char);
18732 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char);
18733 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char);
18735 vector long long vec_orc (vector long long, vector long long);
18736 vector long long vec_orc (vector bool long long, vector long long);
18737 vector long long vec_orc (vector long long, vector bool long long);
18738 vector unsigned long long vec_orc (vector unsigned long long,
18739 vector unsigned long long);
18740 vector unsigned long long vec_orc (vector bool long long,
18741 vector unsigned long long);
18742 vector unsigned long long vec_orc (vector unsigned long long,
18743 vector bool long long);
18744 vector int vec_orc (vector int, vector int);
18745 vector int vec_orc (vector bool int, vector int);
18746 vector int vec_orc (vector int, vector bool int);
18747 vector unsigned int vec_orc (vector unsigned int, vector unsigned int);
18748 vector unsigned int vec_orc (vector bool unsigned int,
18749 vector unsigned int);
18750 vector unsigned int vec_orc (vector unsigned int,
18751 vector bool unsigned int);
18752 vector short vec_orc (vector short, vector short);
18753 vector short vec_orc (vector bool short, vector short);
18754 vector short vec_orc (vector short, vector bool short);
18755 vector unsigned short vec_orc (vector unsigned short, vector unsigned short);
18756 vector unsigned short vec_orc (vector bool unsigned short,
18757 vector unsigned short);
18758 vector unsigned short vec_orc (vector unsigned short,
18759 vector bool unsigned short);
18760 vector signed char vec_orc (vector signed char, vector signed char);
18761 vector signed char vec_orc (vector bool signed char, vector signed char);
18762 vector signed char vec_orc (vector signed char, vector bool signed char);
18763 vector unsigned char vec_orc (vector unsigned char, vector unsigned char);
18764 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char);
18765 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char);
18767 vector int vec_pack (vector long long, vector long long);
18768 vector unsigned int vec_pack (vector unsigned long long,
18769 vector unsigned long long);
18770 vector bool int vec_pack (vector bool long long, vector bool long long);
18771 vector float vec_pack (vector double, vector double);
18773 vector int vec_packs (vector long long, vector long long);
18774 vector unsigned int vec_packs (vector unsigned long long,
18775 vector unsigned long long);
18777 test_vsi_packsu_vssi_vssi (vector signed short x,
18779 vector unsigned char vec_packsu (vector signed short, vector signed short )
18780 vector unsigned char vec_packsu (vector unsigned short, vector unsigned short )
18781 vector unsigned short int vec_packsu (vector signed int, vector signed int);
18782 vector unsigned short int vec_packsu (vector unsigned int,
18783 vector unsigned int);
18784 vector unsigned int vec_packsu (vector long long, vector long long);
18785 vector unsigned int vec_packsu (vector unsigned long long,
18786 vector unsigned long long);
18787 vector unsigned int vec_packsu (vector signed long long,
18788 vector signed long long);
18790 vector unsigned char vec_popcnt (vector signed char);
18791 vector unsigned char vec_popcnt (vector unsigned char);
18792 vector unsigned short vec_popcnt (vector signed short);
18793 vector unsigned short vec_popcnt (vector unsigned short);
18794 vector unsigned int vec_popcnt (vector signed int);
18795 vector unsigned int vec_popcnt (vector unsigned int);
18796 vector unsigned long long vec_popcnt (vector signed long long);
18797 vector unsigned long long vec_popcnt (vector unsigned long long);
18799 vector long long vec_rl (vector long long,
18800 vector unsigned long long);
18801 vector long long vec_rl (vector unsigned long long,
18802 vector unsigned long long);
18804 vector long long vec_sl (vector long long, vector unsigned long long);
18805 vector long long vec_sl (vector unsigned long long,
18806 vector unsigned long long);
18808 vector long long vec_sr (vector long long, vector unsigned long long);
18809 vector unsigned long long char vec_sr (vector unsigned long long,
18810 vector unsigned long long);
18812 vector long long vec_sra (vector long long, vector unsigned long long);
18813 vector unsigned long long vec_sra (vector unsigned long long,
18814 vector unsigned long long);
18816 vector long long vec_sub (vector long long, vector long long);
18817 vector unsigned long long vec_sub (vector unsigned long long,
18818 vector unsigned long long);
18820 vector long long vec_unpackh (vector int);
18821 vector unsigned long long vec_unpackh (vector unsigned int);
18823 vector long long vec_unpackl (vector int);
18824 vector unsigned long long vec_unpackl (vector unsigned int);
18826 vector long long vec_vaddudm (vector long long, vector long long);
18827 vector long long vec_vaddudm (vector bool long long, vector long long);
18828 vector long long vec_vaddudm (vector long long, vector bool long long);
18829 vector unsigned long long vec_vaddudm (vector unsigned long long,
18830 vector unsigned long long);
18831 vector unsigned long long vec_vaddudm (vector bool unsigned long long,
18832 vector unsigned long long);
18833 vector unsigned long long vec_vaddudm (vector unsigned long long,
18834 vector bool unsigned long long);
18836 vector long long vec_vbpermq (vector signed char, vector signed char);
18837 vector long long vec_vbpermq (vector unsigned char, vector unsigned char);
18839 vector unsigned char vec_bperm (vector unsigned char, vector unsigned char);
18840 vector unsigned char vec_bperm (vector unsigned long long,
18841 vector unsigned char);
18842 vector unsigned long long vec_bperm (vector unsigned __int128,
18843 vector unsigned char);
18845 vector long long vec_cntlz (vector long long);
18846 vector unsigned long long vec_cntlz (vector unsigned long long);
18847 vector int vec_cntlz (vector int);
18848 vector unsigned int vec_cntlz (vector int);
18849 vector short vec_cntlz (vector short);
18850 vector unsigned short vec_cntlz (vector unsigned short);
18851 vector signed char vec_cntlz (vector signed char);
18852 vector unsigned char vec_cntlz (vector unsigned char);
18854 vector long long vec_vclz (vector long long);
18855 vector unsigned long long vec_vclz (vector unsigned long long);
18856 vector int vec_vclz (vector int);
18857 vector unsigned int vec_vclz (vector int);
18858 vector short vec_vclz (vector short);
18859 vector unsigned short vec_vclz (vector unsigned short);
18860 vector signed char vec_vclz (vector signed char);
18861 vector unsigned char vec_vclz (vector unsigned char);
18863 vector signed char vec_vclzb (vector signed char);
18864 vector unsigned char vec_vclzb (vector unsigned char);
18866 vector long long vec_vclzd (vector long long);
18867 vector unsigned long long vec_vclzd (vector unsigned long long);
18869 vector short vec_vclzh (vector short);
18870 vector unsigned short vec_vclzh (vector unsigned short);
18872 vector int vec_vclzw (vector int);
18873 vector unsigned int vec_vclzw (vector int);
18875 vector signed char vec_vgbbd (vector signed char);
18876 vector unsigned char vec_vgbbd (vector unsigned char);
18878 vector long long vec_vmaxsd (vector long long, vector long long);
18880 vector unsigned long long vec_vmaxud (vector unsigned long long,
18881 unsigned vector long long);
18883 vector long long vec_vminsd (vector long long, vector long long);
18885 vector unsigned long long vec_vminud (vector long long,
18888 vector int vec_vpksdss (vector long long, vector long long);
18889 vector unsigned int vec_vpksdss (vector long long, vector long long);
18891 vector unsigned int vec_vpkudus (vector unsigned long long,
18892 vector unsigned long long);
18894 vector int vec_vpkudum (vector long long, vector long long);
18895 vector unsigned int vec_vpkudum (vector unsigned long long,
18896 vector unsigned long long);
18897 vector bool int vec_vpkudum (vector bool long long, vector bool long long);
18899 vector long long vec_vpopcnt (vector long long);
18900 vector unsigned long long vec_vpopcnt (vector unsigned long long);
18901 vector int vec_vpopcnt (vector int);
18902 vector unsigned int vec_vpopcnt (vector int);
18903 vector short vec_vpopcnt (vector short);
18904 vector unsigned short vec_vpopcnt (vector unsigned short);
18905 vector signed char vec_vpopcnt (vector signed char);
18906 vector unsigned char vec_vpopcnt (vector unsigned char);
18908 vector signed char vec_vpopcntb (vector signed char);
18909 vector unsigned char vec_vpopcntb (vector unsigned char);
18911 vector long long vec_vpopcntd (vector long long);
18912 vector unsigned long long vec_vpopcntd (vector unsigned long long);
18914 vector short vec_vpopcnth (vector short);
18915 vector unsigned short vec_vpopcnth (vector unsigned short);
18917 vector int vec_vpopcntw (vector int);
18918 vector unsigned int vec_vpopcntw (vector int);
18920 vector long long vec_vrld (vector long long, vector unsigned long long);
18921 vector unsigned long long vec_vrld (vector unsigned long long,
18922 vector unsigned long long);
18924 vector long long vec_vsld (vector long long, vector unsigned long long);
18925 vector long long vec_vsld (vector unsigned long long,
18926 vector unsigned long long);
18928 vector long long vec_vsrad (vector long long, vector unsigned long long);
18929 vector unsigned long long vec_vsrad (vector unsigned long long,
18930 vector unsigned long long);
18932 vector long long vec_vsrd (vector long long, vector unsigned long long);
18933 vector unsigned long long char vec_vsrd (vector unsigned long long,
18934 vector unsigned long long);
18936 vector long long vec_vsubudm (vector long long, vector long long);
18937 vector long long vec_vsubudm (vector bool long long, vector long long);
18938 vector long long vec_vsubudm (vector long long, vector bool long long);
18939 vector unsigned long long vec_vsubudm (vector unsigned long long,
18940 vector unsigned long long);
18941 vector unsigned long long vec_vsubudm (vector bool long long,
18942 vector unsigned long long);
18943 vector unsigned long long vec_vsubudm (vector unsigned long long,
18944 vector bool long long);
18946 vector long long vec_vupkhsw (vector int);
18947 vector unsigned long long vec_vupkhsw (vector unsigned int);
18949 vector long long vec_vupklsw (vector int);
18950 vector unsigned long long vec_vupklsw (vector int);
18953 If the ISA 2.07 additions to the vector/scalar (power8-vector)
18954 instruction set are available, the following additional functions are
18955 available for 64-bit targets. New vector types
18956 (@var{vector __int128_t} and @var{vector __uint128_t}) are available
18957 to hold the @var{__int128_t} and @var{__uint128_t} types to use these
18960 The normal vector extract, and set operations work on
18961 @var{vector __int128_t} and @var{vector __uint128_t} types,
18962 but the index value must be 0.
18965 vector __int128_t vec_vaddcuq (vector __int128_t, vector __int128_t);
18966 vector __uint128_t vec_vaddcuq (vector __uint128_t, vector __uint128_t);
18968 vector __int128_t vec_vadduqm (vector __int128_t, vector __int128_t);
18969 vector __uint128_t vec_vadduqm (vector __uint128_t, vector __uint128_t);
18971 vector __int128_t vec_vaddecuq (vector __int128_t, vector __int128_t,
18972 vector __int128_t);
18973 vector __uint128_t vec_vaddecuq (vector __uint128_t, vector __uint128_t,
18974 vector __uint128_t);
18976 vector __int128_t vec_vaddeuqm (vector __int128_t, vector __int128_t,
18977 vector __int128_t);
18978 vector __uint128_t vec_vaddeuqm (vector __uint128_t, vector __uint128_t,
18979 vector __uint128_t);
18981 vector __int128_t vec_vsubecuq (vector __int128_t, vector __int128_t,
18982 vector __int128_t);
18983 vector __uint128_t vec_vsubecuq (vector __uint128_t, vector __uint128_t,
18984 vector __uint128_t);
18986 vector __int128_t vec_vsubeuqm (vector __int128_t, vector __int128_t,
18987 vector __int128_t);
18988 vector __uint128_t vec_vsubeuqm (vector __uint128_t, vector __uint128_t,
18989 vector __uint128_t);
18991 vector __int128_t vec_vsubcuq (vector __int128_t, vector __int128_t);
18992 vector __uint128_t vec_vsubcuq (vector __uint128_t, vector __uint128_t);
18994 __int128_t vec_vsubuqm (__int128_t, __int128_t);
18995 __uint128_t vec_vsubuqm (__uint128_t, __uint128_t);
18997 vector __int128_t __builtin_bcdadd (vector __int128_t, vector __int128_t);
18998 int __builtin_bcdadd_lt (vector __int128_t, vector __int128_t);
18999 int __builtin_bcdadd_eq (vector __int128_t, vector __int128_t);
19000 int __builtin_bcdadd_gt (vector __int128_t, vector __int128_t);
19001 int __builtin_bcdadd_ov (vector __int128_t, vector __int128_t);
19002 vector __int128_t bcdsub (vector __int128_t, vector __int128_t);
19003 int __builtin_bcdsub_lt (vector __int128_t, vector __int128_t);
19004 int __builtin_bcdsub_eq (vector __int128_t, vector __int128_t);
19005 int __builtin_bcdsub_gt (vector __int128_t, vector __int128_t);
19006 int __builtin_bcdsub_ov (vector __int128_t, vector __int128_t);
19009 If the ISA 3.0 instruction set additions (@option{-mcpu=power9})
19013 vector unsigned long long vec_bperm (vector unsigned long long,
19014 vector unsigned char);
19016 vector bool char vec_cmpne (vector bool char, vector bool char);
19017 vector bool char vec_cmpne (vector signed char, vector signed char);
19018 vector bool char vec_cmpne (vector unsigned char, vector unsigned char);
19019 vector bool int vec_cmpne (vector bool int, vector bool int);
19020 vector bool int vec_cmpne (vector signed int, vector signed int);
19021 vector bool int vec_cmpne (vector unsigned int, vector unsigned int);
19022 vector bool long long vec_cmpne (vector bool long long, vector bool long long);
19023 vector bool long long vec_cmpne (vector signed long long,
19024 vector signed long long);
19025 vector bool long long vec_cmpne (vector unsigned long long,
19026 vector unsigned long long);
19027 vector bool short vec_cmpne (vector bool short, vector bool short);
19028 vector bool short vec_cmpne (vector signed short, vector signed short);
19029 vector bool short vec_cmpne (vector unsigned short, vector unsigned short);
19030 vector bool long long vec_cmpne (vector double, vector double);
19031 vector bool int vec_cmpne (vector float, vector float);
19033 vector float vec_extract_fp32_from_shorth (vector unsigned short);
19034 vector float vec_extract_fp32_from_shortl (vector unsigned short);
19036 vector long long vec_vctz (vector long long);
19037 vector unsigned long long vec_vctz (vector unsigned long long);
19038 vector int vec_vctz (vector int);
19039 vector unsigned int vec_vctz (vector int);
19040 vector short vec_vctz (vector short);
19041 vector unsigned short vec_vctz (vector unsigned short);
19042 vector signed char vec_vctz (vector signed char);
19043 vector unsigned char vec_vctz (vector unsigned char);
19045 vector signed char vec_vctzb (vector signed char);
19046 vector unsigned char vec_vctzb (vector unsigned char);
19048 vector long long vec_vctzd (vector long long);
19049 vector unsigned long long vec_vctzd (vector unsigned long long);
19051 vector short vec_vctzh (vector short);
19052 vector unsigned short vec_vctzh (vector unsigned short);
19054 vector int vec_vctzw (vector int);
19055 vector unsigned int vec_vctzw (vector int);
19057 long long vec_vextract4b (const vector signed char, const int);
19058 long long vec_vextract4b (const vector unsigned char, const int);
19060 vector signed char vec_insert4b (vector int, vector signed char, const int);
19061 vector unsigned char vec_insert4b (vector unsigned int, vector unsigned char,
19063 vector signed char vec_insert4b (long long, vector signed char, const int);
19064 vector unsigned char vec_insert4b (long long, vector unsigned char, const int);
19066 vector unsigned int vec_parity_lsbb (vector signed int);
19067 vector unsigned int vec_parity_lsbb (vector unsigned int);
19068 vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
19069 vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
19070 vector unsigned long long vec_parity_lsbb (vector signed long long);
19071 vector unsigned long long vec_parity_lsbb (vector unsigned long long);
19073 vector int vec_vprtyb (vector int);
19074 vector unsigned int vec_vprtyb (vector unsigned int);
19075 vector long long vec_vprtyb (vector long long);
19076 vector unsigned long long vec_vprtyb (vector unsigned long long);
19078 vector int vec_vprtybw (vector int);
19079 vector unsigned int vec_vprtybw (vector unsigned int);
19081 vector long long vec_vprtybd (vector long long);
19082 vector unsigned long long vec_vprtybd (vector unsigned long long);
19085 On 64-bit targets, if the ISA 3.0 additions (@option{-mcpu=power9})
19089 vector long vec_vprtyb (vector long);
19090 vector unsigned long vec_vprtyb (vector unsigned long);
19091 vector __int128_t vec_vprtyb (vector __int128_t);
19092 vector __uint128_t vec_vprtyb (vector __uint128_t);
19094 vector long vec_vprtybd (vector long);
19095 vector unsigned long vec_vprtybd (vector unsigned long);
19097 vector __int128_t vec_vprtybq (vector __int128_t);
19098 vector __uint128_t vec_vprtybd (vector __uint128_t);
19101 The following built-in vector functions are available for the PowerPC family
19102 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
19104 __vector unsigned char
19105 vec_slv (__vector unsigned char src, __vector unsigned char shift_distance);
19106 __vector unsigned char
19107 vec_srv (__vector unsigned char src, __vector unsigned char shift_distance);
19110 The @code{vec_slv} and @code{vec_srv} functions operate on
19111 all of the bytes of their @code{src} and @code{shift_distance}
19112 arguments in parallel. The behavior of the @code{vec_slv} is as if
19113 there existed a temporary array of 17 unsigned characters
19114 @code{slv_array} within which elements 0 through 15 are the same as
19115 the entries in the @code{src} array and element 16 equals 0. The
19116 result returned from the @code{vec_slv} function is a
19117 @code{__vector} of 16 unsigned characters within which element
19118 @code{i} is computed using the C expression
19119 @code{0xff & (*((unsigned short *)(slv_array + i)) << (0x07 &
19120 shift_distance[i]))},
19121 with this resulting value coerced to the @code{unsigned char} type.
19122 The behavior of the @code{vec_srv} is as if
19123 there existed a temporary array of 17 unsigned characters
19124 @code{srv_array} within which element 0 equals zero and
19125 elements 1 through 16 equal the elements 0 through 15 of
19126 the @code{src} array. The
19127 result returned from the @code{vec_srv} function is a
19128 @code{__vector} of 16 unsigned characters within which element
19129 @code{i} is computed using the C expression
19130 @code{0xff & (*((unsigned short *)(srv_array + i)) >>
19131 (0x07 & shift_distance[i]))},
19132 with this resulting value coerced to the @code{unsigned char} type.
19134 The following built-in functions are available for the PowerPC family
19135 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
19137 __vector unsigned char
19138 vec_absd (__vector unsigned char arg1, __vector unsigned char arg2);
19139 __vector unsigned short
19140 vec_absd (__vector unsigned short arg1, __vector unsigned short arg2);
19141 __vector unsigned int
19142 vec_absd (__vector unsigned int arg1, __vector unsigned int arg2);
19144 __vector unsigned char
19145 vec_absdb (__vector unsigned char arg1, __vector unsigned char arg2);
19146 __vector unsigned short
19147 vec_absdh (__vector unsigned short arg1, __vector unsigned short arg2);
19148 __vector unsigned int
19149 vec_absdw (__vector unsigned int arg1, __vector unsigned int arg2);
19152 The @code{vec_absd}, @code{vec_absdb}, @code{vec_absdh}, and
19153 @code{vec_absdw} built-in functions each computes the absolute
19154 differences of the pairs of vector elements supplied in its two vector
19155 arguments, placing the absolute differences into the corresponding
19156 elements of the vector result.
19158 The following built-in functions are available for the PowerPC family
19159 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
19161 __vector unsigned int
19162 vec_extract_exp (__vector float source);
19163 __vector unsigned long long int
19164 vec_extract_exp (__vector double source);
19166 __vector unsigned int
19167 vec_extract_sig (__vector float source);
19168 __vector unsigned long long int
19169 vec_extract_sig (__vector double source);
19172 vec_insert_exp (__vector unsigned int significands,
19173 __vector unsigned int exponents);
19175 vec_insert_exp (__vector unsigned float significands,
19176 __vector unsigned int exponents);
19178 vec_insert_exp (__vector unsigned long long int significands,
19179 __vector unsigned long long int exponents);
19181 vec_insert_exp (__vector unsigned double significands,
19182 __vector unsigned long long int exponents);
19184 __vector bool int vec_test_data_class (__vector float source,
19185 const int condition);
19186 __vector bool long long int vec_test_data_class (__vector double source,
19187 const int condition);
19190 The @code{vec_extract_sig} and @code{vec_extract_exp} built-in
19191 functions return vectors representing the significands and biased
19192 exponent values of their @code{source} arguments respectively.
19193 Within the result vector returned by @code{vec_extract_sig}, the
19194 @code{0x800000} bit of each vector element returned when the
19195 function's @code{source} argument is of type @code{float} is set to 1
19196 if the corresponding floating point value is in normalized form.
19197 Otherwise, this bit is set to 0. When the @code{source} argument is
19198 of type @code{double}, the @code{0x10000000000000} bit within each of
19199 the result vector's elements is set according to the same rules.
19200 Note that the sign of the significand is not represented in the result
19201 returned from the @code{vec_extract_sig} function. To extract the
19203 @code{vec_cpsgn} function, which returns a new vector within which all
19204 of the sign bits of its second argument vector are overwritten with the
19205 sign bits copied from the coresponding elements of its first argument
19206 vector, and all other (non-sign) bits of the second argument vector
19207 are copied unchanged into the result vector.
19209 The @code{vec_insert_exp} built-in functions return a vector of
19210 single- or double-precision floating
19211 point values constructed by assembling the values of their
19212 @code{significands} and @code{exponents} arguments into the
19213 corresponding elements of the returned vector.
19215 element of the result is copied from the most significant bit of the
19216 corresponding entry within the @code{significands} argument.
19217 Note that the relevant
19218 bits of the @code{significands} argument are the same, for both integer
19219 and floating point types.
19221 significand and exponent components of each element of the result are
19222 composed of the least significant bits of the corresponding
19223 @code{significands} element and the least significant bits of the
19224 corresponding @code{exponents} element.
19226 The @code{vec_test_data_class} built-in function returns a vector
19227 representing the results of testing the @code{source} vector for the
19228 condition selected by the @code{condition} argument. The
19229 @code{condition} argument must be a compile-time constant integer with
19230 value not exceeding 127. The
19231 @code{condition} argument is encoded as a bitmask with each bit
19232 enabling the testing of a different condition, as characterized by the
19236 0x20 Test for +Infinity
19237 0x10 Test for -Infinity
19238 0x08 Test for +Zero
19239 0x04 Test for -Zero
19240 0x02 Test for +Denormal
19241 0x01 Test for -Denormal
19244 If any of the enabled test conditions is true, the corresponding entry
19245 in the result vector is -1. Otherwise (all of the enabled test
19246 conditions are false), the corresponding entry of the result vector is 0.
19248 The following built-in functions are available for the PowerPC family
19249 of processors, starting with ISA 3.0 or later (@option{-mcpu=power9}):
19251 vector unsigned int vec_rlmi (vector unsigned int, vector unsigned int,
19252 vector unsigned int);
19253 vector unsigned long long vec_rlmi (vector unsigned long long,
19254 vector unsigned long long,
19255 vector unsigned long long);
19256 vector unsigned int vec_rlnm (vector unsigned int, vector unsigned int,
19257 vector unsigned int);
19258 vector unsigned long long vec_rlnm (vector unsigned long long,
19259 vector unsigned long long,
19260 vector unsigned long long);
19261 vector unsigned int vec_vrlnm (vector unsigned int, vector unsigned int);
19262 vector unsigned long long vec_vrlnm (vector unsigned long long,
19263 vector unsigned long long);
19266 The result of @code{vec_rlmi} is obtained by rotating each element of
19267 the first argument vector left and inserting it under mask into the
19268 second argument vector. The third argument vector contains the mask
19269 beginning in bits 11:15, the mask end in bits 19:23, and the shift
19270 count in bits 27:31, of each element.
19272 The result of @code{vec_rlnm} is obtained by rotating each element of
19273 the first argument vector left and ANDing it with a mask specified by
19274 the second and third argument vectors. The second argument vector
19275 contains the shift count for each element in the low-order byte. The
19276 third argument vector contains the mask end for each element in the
19277 low-order byte, with the mask begin in the next higher byte.
19279 The result of @code{vec_vrlnm} is obtained by rotating each element
19280 of the first argument vector left and ANDing it with a mask. The
19281 second argument vector contains the mask beginning in bits 11:15,
19282 the mask end in bits 19:23, and the shift count in bits 27:31,
19285 If the ISA 3.0 instruction set additions (@option{-mcpu=power9})
19288 vector signed bool char vec_revb (vector signed char);
19289 vector signed char vec_revb (vector signed char);
19290 vector unsigned char vec_revb (vector unsigned char);
19291 vector bool short vec_revb (vector bool short);
19292 vector short vec_revb (vector short);
19293 vector unsigned short vec_revb (vector unsigned short);
19294 vector bool int vec_revb (vector bool int);
19295 vector int vec_revb (vector int);
19296 vector unsigned int vec_revb (vector unsigned int);
19297 vector float vec_revb (vector float);
19298 vector bool long long vec_revb (vector bool long long);
19299 vector long long vec_revb (vector long long);
19300 vector unsigned long long vec_revb (vector unsigned long long);
19301 vector double vec_revb (vector double);
19304 On 64-bit targets, if the ISA 3.0 additions (@option{-mcpu=power9})
19307 vector long vec_revb (vector long);
19308 vector unsigned long vec_revb (vector unsigned long);
19309 vector __int128_t vec_revb (vector __int128_t);
19310 vector __uint128_t vec_revb (vector __uint128_t);
19313 The @code{vec_revb} built-in function reverses the bytes on an element
19314 by element basis. A vector of @code{vector unsigned char} or
19315 @code{vector signed char} reverses the bytes in the whole word.
19317 If the cryptographic instructions are enabled (@option{-mcrypto} or
19318 @option{-mcpu=power8}), the following builtins are enabled.
19321 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long);
19323 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long,
19324 vector unsigned long long);
19326 vector unsigned long long __builtin_crypto_vcipherlast
19327 (vector unsigned long long,
19328 vector unsigned long long);
19330 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long,
19331 vector unsigned long long);
19333 vector unsigned long long __builtin_crypto_vncipherlast
19334 (vector unsigned long long,
19335 vector unsigned long long);
19337 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char,
19338 vector unsigned char,
19339 vector unsigned char);
19341 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short,
19342 vector unsigned short,
19343 vector unsigned short);
19345 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int,
19346 vector unsigned int,
19347 vector unsigned int);
19349 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long,
19350 vector unsigned long long,
19351 vector unsigned long long);
19353 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char,
19354 vector unsigned char);
19356 vector unsigned short __builtin_crypto_vpmsumb (vector unsigned short,
19357 vector unsigned short);
19359 vector unsigned int __builtin_crypto_vpmsumb (vector unsigned int,
19360 vector unsigned int);
19362 vector unsigned long long __builtin_crypto_vpmsumb (vector unsigned long long,
19363 vector unsigned long long);
19365 vector unsigned long long __builtin_crypto_vshasigmad
19366 (vector unsigned long long, int, int);
19368 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int,
19372 The second argument to @var{__builtin_crypto_vshasigmad} and
19373 @var{__builtin_crypto_vshasigmaw} must be a constant
19374 integer that is 0 or 1. The third argument to these built-in functions
19375 must be a constant integer in the range of 0 to 15.
19377 If the ISA 3.0 instruction set additions
19378 are enabled (@option{-mcpu=power9}), the following additional
19379 functions are available for both 32-bit and 64-bit targets.
19381 vector short vec_xl (int, vector short *);
19382 vector short vec_xl (int, short *);
19383 vector unsigned short vec_xl (int, vector unsigned short *);
19384 vector unsigned short vec_xl (int, unsigned short *);
19385 vector char vec_xl (int, vector char *);
19386 vector char vec_xl (int, char *);
19387 vector unsigned char vec_xl (int, vector unsigned char *);
19388 vector unsigned char vec_xl (int, unsigned char *);
19390 void vec_xst (vector short, int, vector short *);
19391 void vec_xst (vector short, int, short *);
19392 void vec_xst (vector unsigned short, int, vector unsigned short *);
19393 void vec_xst (vector unsigned short, int, unsigned short *);
19394 void vec_xst (vector char, int, vector char *);
19395 void vec_xst (vector char, int, char *);
19396 void vec_xst (vector unsigned char, int, vector unsigned char *);
19397 void vec_xst (vector unsigned char, int, unsigned char *);
19399 @node PowerPC Hardware Transactional Memory Built-in Functions
19400 @subsection PowerPC Hardware Transactional Memory Built-in Functions
19401 GCC provides two interfaces for accessing the Hardware Transactional
19402 Memory (HTM) instructions available on some of the PowerPC family
19403 of processors (eg, POWER8). The two interfaces come in a low level
19404 interface, consisting of built-in functions specific to PowerPC and a
19405 higher level interface consisting of inline functions that are common
19406 between PowerPC and S/390.
19408 @subsubsection PowerPC HTM Low Level Built-in Functions
19410 The following low level built-in functions are available with
19411 @option{-mhtm} or @option{-mcpu=CPU} where CPU is `power8' or later.
19412 They all generate the machine instruction that is part of the name.
19414 The HTM builtins (with the exception of @code{__builtin_tbegin}) return
19415 the full 4-bit condition register value set by their associated hardware
19416 instruction. The header file @code{htmintrin.h} defines some macros that can
19417 be used to decipher the return value. The @code{__builtin_tbegin} builtin
19418 returns a simple true or false value depending on whether a transaction was
19419 successfully started or not. The arguments of the builtins match exactly the
19420 type and order of the associated hardware instruction's operands, except for
19421 the @code{__builtin_tcheck} builtin, which does not take any input arguments.
19422 Refer to the ISA manual for a description of each instruction's operands.
19425 unsigned int __builtin_tbegin (unsigned int)
19426 unsigned int __builtin_tend (unsigned int)
19428 unsigned int __builtin_tabort (unsigned int)
19429 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int)
19430 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int)
19431 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int)
19432 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int)
19434 unsigned int __builtin_tcheck (void)
19435 unsigned int __builtin_treclaim (unsigned int)
19436 unsigned int __builtin_trechkpt (void)
19437 unsigned int __builtin_tsr (unsigned int)
19440 In addition to the above HTM built-ins, we have added built-ins for
19441 some common extended mnemonics of the HTM instructions:
19444 unsigned int __builtin_tendall (void)
19445 unsigned int __builtin_tresume (void)
19446 unsigned int __builtin_tsuspend (void)
19449 Note that the semantics of the above HTM builtins are required to mimic
19450 the locking semantics used for critical sections. Builtins that are used
19451 to create a new transaction or restart a suspended transaction must have
19452 lock acquisition like semantics while those builtins that end or suspend a
19453 transaction must have lock release like semantics. Specifically, this must
19454 mimic lock semantics as specified by C++11, for example: Lock acquisition is
19455 as-if an execution of __atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE)
19456 that returns 0, and lock release is as-if an execution of
19457 __atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an
19458 implicit implementation-defined lock used for all transactions. The HTM
19459 instructions associated with with the builtins inherently provide the
19460 correct acquisition and release hardware barriers required. However,
19461 the compiler must also be prohibited from moving loads and stores across
19462 the builtins in a way that would violate their semantics. This has been
19463 accomplished by adding memory barriers to the associated HTM instructions
19464 (which is a conservative approach to provide acquire and release semantics).
19465 Earlier versions of the compiler did not treat the HTM instructions as
19466 memory barriers. A @code{__TM_FENCE__} macro has been added, which can
19467 be used to determine whether the current compiler treats HTM instructions
19468 as memory barriers or not. This allows the user to explicitly add memory
19469 barriers to their code when using an older version of the compiler.
19471 The following set of built-in functions are available to gain access
19472 to the HTM specific special purpose registers.
19475 unsigned long __builtin_get_texasr (void)
19476 unsigned long __builtin_get_texasru (void)
19477 unsigned long __builtin_get_tfhar (void)
19478 unsigned long __builtin_get_tfiar (void)
19480 void __builtin_set_texasr (unsigned long);
19481 void __builtin_set_texasru (unsigned long);
19482 void __builtin_set_tfhar (unsigned long);
19483 void __builtin_set_tfiar (unsigned long);
19486 Example usage of these low level built-in functions may look like:
19489 #include <htmintrin.h>
19491 int num_retries = 10;
19495 if (__builtin_tbegin (0))
19497 /* Transaction State Initiated. */
19498 if (is_locked (lock))
19499 __builtin_tabort (0);
19500 ... transaction code...
19501 __builtin_tend (0);
19506 /* Transaction State Failed. Use locks if the transaction
19507 failure is "persistent" or we've tried too many times. */
19508 if (num_retries-- <= 0
19509 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ()))
19511 acquire_lock (lock);
19512 ... non transactional fallback path...
19513 release_lock (lock);
19520 One final built-in function has been added that returns the value of
19521 the 2-bit Transaction State field of the Machine Status Register (MSR)
19522 as stored in @code{CR0}.
19525 unsigned long __builtin_ttest (void)
19528 This built-in can be used to determine the current transaction state
19529 using the following code example:
19532 #include <htmintrin.h>
19534 unsigned char tx_state = _HTM_STATE (__builtin_ttest ());
19536 if (tx_state == _HTM_TRANSACTIONAL)
19538 /* Code to use in transactional state. */
19540 else if (tx_state == _HTM_NONTRANSACTIONAL)
19542 /* Code to use in non-transactional state. */
19544 else if (tx_state == _HTM_SUSPENDED)
19546 /* Code to use in transaction suspended state. */
19550 @subsubsection PowerPC HTM High Level Inline Functions
19552 The following high level HTM interface is made available by including
19553 @code{<htmxlintrin.h>} and using @option{-mhtm} or @option{-mcpu=CPU}
19554 where CPU is `power8' or later. This interface is common between PowerPC
19555 and S/390, allowing users to write one HTM source implementation that
19556 can be compiled and executed on either system.
19559 long __TM_simple_begin (void)
19560 long __TM_begin (void* const TM_buff)
19561 long __TM_end (void)
19562 void __TM_abort (void)
19563 void __TM_named_abort (unsigned char const code)
19564 void __TM_resume (void)
19565 void __TM_suspend (void)
19567 long __TM_is_user_abort (void* const TM_buff)
19568 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code)
19569 long __TM_is_illegal (void* const TM_buff)
19570 long __TM_is_footprint_exceeded (void* const TM_buff)
19571 long __TM_nesting_depth (void* const TM_buff)
19572 long __TM_is_nested_too_deep(void* const TM_buff)
19573 long __TM_is_conflict(void* const TM_buff)
19574 long __TM_is_failure_persistent(void* const TM_buff)
19575 long __TM_failure_address(void* const TM_buff)
19576 long long __TM_failure_code(void* const TM_buff)
19579 Using these common set of HTM inline functions, we can create
19580 a more portable version of the HTM example in the previous
19581 section that will work on either PowerPC or S/390:
19584 #include <htmxlintrin.h>
19586 int num_retries = 10;
19587 TM_buff_type TM_buff;
19591 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED)
19593 /* Transaction State Initiated. */
19594 if (is_locked (lock))
19596 ... transaction code...
19602 /* Transaction State Failed. Use locks if the transaction
19603 failure is "persistent" or we've tried too many times. */
19604 if (num_retries-- <= 0
19605 || __TM_is_failure_persistent (TM_buff))
19607 acquire_lock (lock);
19608 ... non transactional fallback path...
19609 release_lock (lock);
19616 @node PowerPC Atomic Memory Operation Functions
19617 @subsection PowerPC Atomic Memory Operation Functions
19618 ISA 3.0 of the PowerPC added new atomic memory operation (amo)
19619 instructions. GCC provides support for these instructions in 64-bit
19620 environments. All of the functions are declared in the include file
19623 The functions supported are:
19628 uint32_t amo_lwat_add (uint32_t *, uint32_t);
19629 uint32_t amo_lwat_xor (uint32_t *, uint32_t);
19630 uint32_t amo_lwat_ior (uint32_t *, uint32_t);
19631 uint32_t amo_lwat_and (uint32_t *, uint32_t);
19632 uint32_t amo_lwat_umax (uint32_t *, uint32_t);
19633 uint32_t amo_lwat_umin (uint32_t *, uint32_t);
19634 uint32_t amo_lwat_swap (uint32_t *, uint32_t);
19636 int32_t amo_lwat_sadd (int32_t *, int32_t);
19637 int32_t amo_lwat_smax (int32_t *, int32_t);
19638 int32_t amo_lwat_smin (int32_t *, int32_t);
19639 int32_t amo_lwat_sswap (int32_t *, int32_t);
19641 uint64_t amo_ldat_add (uint64_t *, uint64_t);
19642 uint64_t amo_ldat_xor (uint64_t *, uint64_t);
19643 uint64_t amo_ldat_ior (uint64_t *, uint64_t);
19644 uint64_t amo_ldat_and (uint64_t *, uint64_t);
19645 uint64_t amo_ldat_umax (uint64_t *, uint64_t);
19646 uint64_t amo_ldat_umin (uint64_t *, uint64_t);
19647 uint64_t amo_ldat_swap (uint64_t *, uint64_t);
19649 int64_t amo_ldat_sadd (int64_t *, int64_t);
19650 int64_t amo_ldat_smax (int64_t *, int64_t);
19651 int64_t amo_ldat_smin (int64_t *, int64_t);
19652 int64_t amo_ldat_sswap (int64_t *, int64_t);
19654 void amo_stwat_add (uint32_t *, uint32_t);
19655 void amo_stwat_xor (uint32_t *, uint32_t);
19656 void amo_stwat_ior (uint32_t *, uint32_t);
19657 void amo_stwat_and (uint32_t *, uint32_t);
19658 void amo_stwat_umax (uint32_t *, uint32_t);
19659 void amo_stwat_umin (uint32_t *, uint32_t);
19661 void amo_stwat_sadd (int32_t *, int32_t);
19662 void amo_stwat_smax (int32_t *, int32_t);
19663 void amo_stwat_smin (int32_t *, int32_t);
19665 void amo_stdat_add (uint64_t *, uint64_t);
19666 void amo_stdat_xor (uint64_t *, uint64_t);
19667 void amo_stdat_ior (uint64_t *, uint64_t);
19668 void amo_stdat_and (uint64_t *, uint64_t);
19669 void amo_stdat_umax (uint64_t *, uint64_t);
19670 void amo_stdat_umin (uint64_t *, uint64_t);
19672 void amo_stdat_sadd (int64_t *, int64_t);
19673 void amo_stdat_smax (int64_t *, int64_t);
19674 void amo_stdat_smin (int64_t *, int64_t);
19677 @node RX Built-in Functions
19678 @subsection RX Built-in Functions
19679 GCC supports some of the RX instructions which cannot be expressed in
19680 the C programming language via the use of built-in functions. The
19681 following functions are supported:
19683 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
19684 Generates the @code{brk} machine instruction.
19687 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
19688 Generates the @code{clrpsw} machine instruction to clear the specified
19689 bit in the processor status word.
19692 @deftypefn {Built-in Function} void __builtin_rx_int (int)
19693 Generates the @code{int} machine instruction to generate an interrupt
19694 with the specified value.
19697 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
19698 Generates the @code{machi} machine instruction to add the result of
19699 multiplying the top 16 bits of the two arguments into the
19703 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
19704 Generates the @code{maclo} machine instruction to add the result of
19705 multiplying the bottom 16 bits of the two arguments into the
19709 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
19710 Generates the @code{mulhi} machine instruction to place the result of
19711 multiplying the top 16 bits of the two arguments into the
19715 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
19716 Generates the @code{mullo} machine instruction to place the result of
19717 multiplying the bottom 16 bits of the two arguments into the
19721 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
19722 Generates the @code{mvfachi} machine instruction to read the top
19723 32 bits of the accumulator.
19726 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
19727 Generates the @code{mvfacmi} machine instruction to read the middle
19728 32 bits of the accumulator.
19731 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
19732 Generates the @code{mvfc} machine instruction which reads the control
19733 register specified in its argument and returns its value.
19736 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
19737 Generates the @code{mvtachi} machine instruction to set the top
19738 32 bits of the accumulator.
19741 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
19742 Generates the @code{mvtaclo} machine instruction to set the bottom
19743 32 bits of the accumulator.
19746 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
19747 Generates the @code{mvtc} machine instruction which sets control
19748 register number @code{reg} to @code{val}.
19751 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
19752 Generates the @code{mvtipl} machine instruction set the interrupt
19756 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
19757 Generates the @code{racw} machine instruction to round the accumulator
19758 according to the specified mode.
19761 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
19762 Generates the @code{revw} machine instruction which swaps the bytes in
19763 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
19764 and also bits 16--23 occupy bits 24--31 and vice versa.
19767 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
19768 Generates the @code{rmpa} machine instruction which initiates a
19769 repeated multiply and accumulate sequence.
19772 @deftypefn {Built-in Function} void __builtin_rx_round (float)
19773 Generates the @code{round} machine instruction which returns the
19774 floating-point argument rounded according to the current rounding mode
19775 set in the floating-point status word register.
19778 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
19779 Generates the @code{sat} machine instruction which returns the
19780 saturated value of the argument.
19783 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
19784 Generates the @code{setpsw} machine instruction to set the specified
19785 bit in the processor status word.
19788 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
19789 Generates the @code{wait} machine instruction.
19792 @node S/390 System z Built-in Functions
19793 @subsection S/390 System z Built-in Functions
19794 @deftypefn {Built-in Function} int __builtin_tbegin (void*)
19795 Generates the @code{tbegin} machine instruction starting a
19796 non-constrained hardware transaction. If the parameter is non-NULL the
19797 memory area is used to store the transaction diagnostic buffer and
19798 will be passed as first operand to @code{tbegin}. This buffer can be
19799 defined using the @code{struct __htm_tdb} C struct defined in
19800 @code{htmintrin.h} and must reside on a double-word boundary. The
19801 second tbegin operand is set to @code{0xff0c}. This enables
19802 save/restore of all GPRs and disables aborts for FPR and AR
19803 manipulations inside the transaction body. The condition code set by
19804 the tbegin instruction is returned as integer value. The tbegin
19805 instruction by definition overwrites the content of all FPRs. The
19806 compiler will generate code which saves and restores the FPRs. For
19807 soft-float code it is recommended to used the @code{*_nofloat}
19808 variant. In order to prevent a TDB from being written it is required
19809 to pass a constant zero value as parameter. Passing a zero value
19810 through a variable is not sufficient. Although modifications of
19811 access registers inside the transaction will not trigger an
19812 transaction abort it is not supported to actually modify them. Access
19813 registers do not get saved when entering a transaction. They will have
19814 undefined state when reaching the abort code.
19817 Macros for the possible return codes of tbegin are defined in the
19818 @code{htmintrin.h} header file:
19821 @item _HTM_TBEGIN_STARTED
19822 @code{tbegin} has been executed as part of normal processing. The
19823 transaction body is supposed to be executed.
19824 @item _HTM_TBEGIN_INDETERMINATE
19825 The transaction was aborted due to an indeterminate condition which
19826 might be persistent.
19827 @item _HTM_TBEGIN_TRANSIENT
19828 The transaction aborted due to a transient failure. The transaction
19829 should be re-executed in that case.
19830 @item _HTM_TBEGIN_PERSISTENT
19831 The transaction aborted due to a persistent failure. Re-execution
19832 under same circumstances will not be productive.
19835 @defmac _HTM_FIRST_USER_ABORT_CODE
19836 The @code{_HTM_FIRST_USER_ABORT_CODE} defined in @code{htmintrin.h}
19837 specifies the first abort code which can be used for
19838 @code{__builtin_tabort}. Values below this threshold are reserved for
19842 @deftp {Data type} {struct __htm_tdb}
19843 The @code{struct __htm_tdb} defined in @code{htmintrin.h} describes
19844 the structure of the transaction diagnostic block as specified in the
19845 Principles of Operation manual chapter 5-91.
19848 @deftypefn {Built-in Function} int __builtin_tbegin_nofloat (void*)
19849 Same as @code{__builtin_tbegin} but without FPR saves and restores.
19850 Using this variant in code making use of FPRs will leave the FPRs in
19851 undefined state when entering the transaction abort handler code.
19854 @deftypefn {Built-in Function} int __builtin_tbegin_retry (void*, int)
19855 In addition to @code{__builtin_tbegin} a loop for transient failures
19856 is generated. If tbegin returns a condition code of 2 the transaction
19857 will be retried as often as specified in the second argument. The
19858 perform processor assist instruction is used to tell the CPU about the
19859 number of fails so far.
19862 @deftypefn {Built-in Function} int __builtin_tbegin_retry_nofloat (void*, int)
19863 Same as @code{__builtin_tbegin_retry} but without FPR saves and
19864 restores. Using this variant in code making use of FPRs will leave
19865 the FPRs in undefined state when entering the transaction abort
19869 @deftypefn {Built-in Function} void __builtin_tbeginc (void)
19870 Generates the @code{tbeginc} machine instruction starting a constrained
19871 hardware transaction. The second operand is set to @code{0xff08}.
19874 @deftypefn {Built-in Function} int __builtin_tend (void)
19875 Generates the @code{tend} machine instruction finishing a transaction
19876 and making the changes visible to other threads. The condition code
19877 generated by tend is returned as integer value.
19880 @deftypefn {Built-in Function} void __builtin_tabort (int)
19881 Generates the @code{tabort} machine instruction with the specified
19882 abort code. Abort codes from 0 through 255 are reserved and will
19883 result in an error message.
19886 @deftypefn {Built-in Function} void __builtin_tx_assist (int)
19887 Generates the @code{ppa rX,rY,1} machine instruction. Where the
19888 integer parameter is loaded into rX and a value of zero is loaded into
19889 rY. The integer parameter specifies the number of times the
19890 transaction repeatedly aborted.
19893 @deftypefn {Built-in Function} int __builtin_tx_nesting_depth (void)
19894 Generates the @code{etnd} machine instruction. The current nesting
19895 depth is returned as integer value. For a nesting depth of 0 the code
19896 is not executed as part of an transaction.
19899 @deftypefn {Built-in Function} void __builtin_non_tx_store (uint64_t *, uint64_t)
19901 Generates the @code{ntstg} machine instruction. The second argument
19902 is written to the first arguments location. The store operation will
19903 not be rolled-back in case of an transaction abort.
19906 @node SH Built-in Functions
19907 @subsection SH Built-in Functions
19908 The following built-in functions are supported on the SH1, SH2, SH3 and SH4
19909 families of processors:
19911 @deftypefn {Built-in Function} {void} __builtin_set_thread_pointer (void *@var{ptr})
19912 Sets the @samp{GBR} register to the specified value @var{ptr}. This is usually
19913 used by system code that manages threads and execution contexts. The compiler
19914 normally does not generate code that modifies the contents of @samp{GBR} and
19915 thus the value is preserved across function calls. Changing the @samp{GBR}
19916 value in user code must be done with caution, since the compiler might use
19917 @samp{GBR} in order to access thread local variables.
19921 @deftypefn {Built-in Function} {void *} __builtin_thread_pointer (void)
19922 Returns the value that is currently set in the @samp{GBR} register.
19923 Memory loads and stores that use the thread pointer as a base address are
19924 turned into @samp{GBR} based displacement loads and stores, if possible.
19932 int get_tcb_value (void)
19934 // Generate @samp{mov.l @@(8,gbr),r0} instruction
19935 return ((my_tcb*)__builtin_thread_pointer ())->c;
19941 @deftypefn {Built-in Function} {unsigned int} __builtin_sh_get_fpscr (void)
19942 Returns the value that is currently set in the @samp{FPSCR} register.
19945 @deftypefn {Built-in Function} {void} __builtin_sh_set_fpscr (unsigned int @var{val})
19946 Sets the @samp{FPSCR} register to the specified value @var{val}, while
19947 preserving the current values of the FR, SZ and PR bits.
19950 @node SPARC VIS Built-in Functions
19951 @subsection SPARC VIS Built-in Functions
19953 GCC supports SIMD operations on the SPARC using both the generic vector
19954 extensions (@pxref{Vector Extensions}) as well as built-in functions for
19955 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
19956 switch, the VIS extension is exposed as the following built-in functions:
19959 typedef int v1si __attribute__ ((vector_size (4)));
19960 typedef int v2si __attribute__ ((vector_size (8)));
19961 typedef short v4hi __attribute__ ((vector_size (8)));
19962 typedef short v2hi __attribute__ ((vector_size (4)));
19963 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
19964 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
19966 void __builtin_vis_write_gsr (int64_t);
19967 int64_t __builtin_vis_read_gsr (void);
19969 void * __builtin_vis_alignaddr (void *, long);
19970 void * __builtin_vis_alignaddrl (void *, long);
19971 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
19972 v2si __builtin_vis_faligndatav2si (v2si, v2si);
19973 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
19974 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
19976 v4hi __builtin_vis_fexpand (v4qi);
19978 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
19979 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
19980 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
19981 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
19982 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
19983 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
19984 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
19986 v4qi __builtin_vis_fpack16 (v4hi);
19987 v8qi __builtin_vis_fpack32 (v2si, v8qi);
19988 v2hi __builtin_vis_fpackfix (v2si);
19989 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
19991 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
19993 long __builtin_vis_edge8 (void *, void *);
19994 long __builtin_vis_edge8l (void *, void *);
19995 long __builtin_vis_edge16 (void *, void *);
19996 long __builtin_vis_edge16l (void *, void *);
19997 long __builtin_vis_edge32 (void *, void *);
19998 long __builtin_vis_edge32l (void *, void *);
20000 long __builtin_vis_fcmple16 (v4hi, v4hi);
20001 long __builtin_vis_fcmple32 (v2si, v2si);
20002 long __builtin_vis_fcmpne16 (v4hi, v4hi);
20003 long __builtin_vis_fcmpne32 (v2si, v2si);
20004 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
20005 long __builtin_vis_fcmpgt32 (v2si, v2si);
20006 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
20007 long __builtin_vis_fcmpeq32 (v2si, v2si);
20009 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
20010 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
20011 v2si __builtin_vis_fpadd32 (v2si, v2si);
20012 v1si __builtin_vis_fpadd32s (v1si, v1si);
20013 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
20014 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
20015 v2si __builtin_vis_fpsub32 (v2si, v2si);
20016 v1si __builtin_vis_fpsub32s (v1si, v1si);
20018 long __builtin_vis_array8 (long, long);
20019 long __builtin_vis_array16 (long, long);
20020 long __builtin_vis_array32 (long, long);
20023 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
20024 functions also become available:
20027 long __builtin_vis_bmask (long, long);
20028 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
20029 v2si __builtin_vis_bshufflev2si (v2si, v2si);
20030 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
20031 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
20033 long __builtin_vis_edge8n (void *, void *);
20034 long __builtin_vis_edge8ln (void *, void *);
20035 long __builtin_vis_edge16n (void *, void *);
20036 long __builtin_vis_edge16ln (void *, void *);
20037 long __builtin_vis_edge32n (void *, void *);
20038 long __builtin_vis_edge32ln (void *, void *);
20041 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
20042 functions also become available:
20045 void __builtin_vis_cmask8 (long);
20046 void __builtin_vis_cmask16 (long);
20047 void __builtin_vis_cmask32 (long);
20049 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
20051 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
20052 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
20053 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
20054 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
20055 v2si __builtin_vis_fsll16 (v2si, v2si);
20056 v2si __builtin_vis_fslas16 (v2si, v2si);
20057 v2si __builtin_vis_fsrl16 (v2si, v2si);
20058 v2si __builtin_vis_fsra16 (v2si, v2si);
20060 long __builtin_vis_pdistn (v8qi, v8qi);
20062 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
20064 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
20065 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
20067 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
20068 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
20069 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
20070 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
20071 v2si __builtin_vis_fpadds32 (v2si, v2si);
20072 v1si __builtin_vis_fpadds32s (v1si, v1si);
20073 v2si __builtin_vis_fpsubs32 (v2si, v2si);
20074 v1si __builtin_vis_fpsubs32s (v1si, v1si);
20076 long __builtin_vis_fucmple8 (v8qi, v8qi);
20077 long __builtin_vis_fucmpne8 (v8qi, v8qi);
20078 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
20079 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
20081 float __builtin_vis_fhadds (float, float);
20082 double __builtin_vis_fhaddd (double, double);
20083 float __builtin_vis_fhsubs (float, float);
20084 double __builtin_vis_fhsubd (double, double);
20085 float __builtin_vis_fnhadds (float, float);
20086 double __builtin_vis_fnhaddd (double, double);
20088 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
20089 int64_t __builtin_vis_xmulx (int64_t, int64_t);
20090 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
20093 When you use the @option{-mvis4} switch, the VIS version 4.0 built-in
20094 functions also become available:
20097 v8qi __builtin_vis_fpadd8 (v8qi, v8qi);
20098 v8qi __builtin_vis_fpadds8 (v8qi, v8qi);
20099 v8qi __builtin_vis_fpaddus8 (v8qi, v8qi);
20100 v4hi __builtin_vis_fpaddus16 (v4hi, v4hi);
20102 v8qi __builtin_vis_fpsub8 (v8qi, v8qi);
20103 v8qi __builtin_vis_fpsubs8 (v8qi, v8qi);
20104 v8qi __builtin_vis_fpsubus8 (v8qi, v8qi);
20105 v4hi __builtin_vis_fpsubus16 (v4hi, v4hi);
20107 long __builtin_vis_fpcmple8 (v8qi, v8qi);
20108 long __builtin_vis_fpcmpgt8 (v8qi, v8qi);
20109 long __builtin_vis_fpcmpule16 (v4hi, v4hi);
20110 long __builtin_vis_fpcmpugt16 (v4hi, v4hi);
20111 long __builtin_vis_fpcmpule32 (v2si, v2si);
20112 long __builtin_vis_fpcmpugt32 (v2si, v2si);
20114 v8qi __builtin_vis_fpmax8 (v8qi, v8qi);
20115 v4hi __builtin_vis_fpmax16 (v4hi, v4hi);
20116 v2si __builtin_vis_fpmax32 (v2si, v2si);
20118 v8qi __builtin_vis_fpmaxu8 (v8qi, v8qi);
20119 v4hi __builtin_vis_fpmaxu16 (v4hi, v4hi);
20120 v2si __builtin_vis_fpmaxu32 (v2si, v2si);
20123 v8qi __builtin_vis_fpmin8 (v8qi, v8qi);
20124 v4hi __builtin_vis_fpmin16 (v4hi, v4hi);
20125 v2si __builtin_vis_fpmin32 (v2si, v2si);
20127 v8qi __builtin_vis_fpminu8 (v8qi, v8qi);
20128 v4hi __builtin_vis_fpminu16 (v4hi, v4hi);
20129 v2si __builtin_vis_fpminu32 (v2si, v2si);
20132 When you use the @option{-mvis4b} switch, the VIS version 4.0B
20133 built-in functions also become available:
20136 v8qi __builtin_vis_dictunpack8 (double, int);
20137 v4hi __builtin_vis_dictunpack16 (double, int);
20138 v2si __builtin_vis_dictunpack32 (double, int);
20140 long __builtin_vis_fpcmple8shl (v8qi, v8qi, int);
20141 long __builtin_vis_fpcmpgt8shl (v8qi, v8qi, int);
20142 long __builtin_vis_fpcmpeq8shl (v8qi, v8qi, int);
20143 long __builtin_vis_fpcmpne8shl (v8qi, v8qi, int);
20145 long __builtin_vis_fpcmple16shl (v4hi, v4hi, int);
20146 long __builtin_vis_fpcmpgt16shl (v4hi, v4hi, int);
20147 long __builtin_vis_fpcmpeq16shl (v4hi, v4hi, int);
20148 long __builtin_vis_fpcmpne16shl (v4hi, v4hi, int);
20150 long __builtin_vis_fpcmple32shl (v2si, v2si, int);
20151 long __builtin_vis_fpcmpgt32shl (v2si, v2si, int);
20152 long __builtin_vis_fpcmpeq32shl (v2si, v2si, int);
20153 long __builtin_vis_fpcmpne32shl (v2si, v2si, int);
20155 long __builtin_vis_fpcmpule8shl (v8qi, v8qi, int);
20156 long __builtin_vis_fpcmpugt8shl (v8qi, v8qi, int);
20157 long __builtin_vis_fpcmpule16shl (v4hi, v4hi, int);
20158 long __builtin_vis_fpcmpugt16shl (v4hi, v4hi, int);
20159 long __builtin_vis_fpcmpule32shl (v2si, v2si, int);
20160 long __builtin_vis_fpcmpugt32shl (v2si, v2si, int);
20162 long __builtin_vis_fpcmpde8shl (v8qi, v8qi, int);
20163 long __builtin_vis_fpcmpde16shl (v4hi, v4hi, int);
20164 long __builtin_vis_fpcmpde32shl (v2si, v2si, int);
20166 long __builtin_vis_fpcmpur8shl (v8qi, v8qi, int);
20167 long __builtin_vis_fpcmpur16shl (v4hi, v4hi, int);
20168 long __builtin_vis_fpcmpur32shl (v2si, v2si, int);
20171 @node SPU Built-in Functions
20172 @subsection SPU Built-in Functions
20174 GCC provides extensions for the SPU processor as described in the
20175 Sony/Toshiba/IBM SPU Language Extensions Specification. GCC's
20176 implementation differs in several ways.
20181 The optional extension of specifying vector constants in parentheses is
20185 A vector initializer requires no cast if the vector constant is of the
20186 same type as the variable it is initializing.
20189 If @code{signed} or @code{unsigned} is omitted, the signedness of the
20190 vector type is the default signedness of the base type. The default
20191 varies depending on the operating system, so a portable program should
20192 always specify the signedness.
20195 By default, the keyword @code{__vector} is added. The macro
20196 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
20200 GCC allows using a @code{typedef} name as the type specifier for a
20204 For C, overloaded functions are implemented with macros so the following
20208 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
20212 Since @code{spu_add} is a macro, the vector constant in the example
20213 is treated as four separate arguments. Wrap the entire argument in
20214 parentheses for this to work.
20217 The extended version of @code{__builtin_expect} is not supported.
20221 @emph{Note:} Only the interface described in the aforementioned
20222 specification is supported. Internally, GCC uses built-in functions to
20223 implement the required functionality, but these are not supported and
20224 are subject to change without notice.
20226 @node TI C6X Built-in Functions
20227 @subsection TI C6X Built-in Functions
20229 GCC provides intrinsics to access certain instructions of the TI C6X
20230 processors. These intrinsics, listed below, are available after
20231 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
20232 to C6X instructions.
20236 int _sadd (int, int)
20237 int _ssub (int, int)
20238 int _sadd2 (int, int)
20239 int _ssub2 (int, int)
20240 long long _mpy2 (int, int)
20241 long long _smpy2 (int, int)
20242 int _add4 (int, int)
20243 int _sub4 (int, int)
20244 int _saddu4 (int, int)
20246 int _smpy (int, int)
20247 int _smpyh (int, int)
20248 int _smpyhl (int, int)
20249 int _smpylh (int, int)
20251 int _sshl (int, int)
20252 int _subc (int, int)
20254 int _avg2 (int, int)
20255 int _avgu4 (int, int)
20257 int _clrr (int, int)
20258 int _extr (int, int)
20259 int _extru (int, int)
20265 @node TILE-Gx Built-in Functions
20266 @subsection TILE-Gx Built-in Functions
20268 GCC provides intrinsics to access every instruction of the TILE-Gx
20269 processor. The intrinsics are of the form:
20273 unsigned long long __insn_@var{op} (...)
20277 Where @var{op} is the name of the instruction. Refer to the ISA manual
20278 for the complete list of instructions.
20280 GCC also provides intrinsics to directly access the network registers.
20281 The intrinsics are:
20285 unsigned long long __tile_idn0_receive (void)
20286 unsigned long long __tile_idn1_receive (void)
20287 unsigned long long __tile_udn0_receive (void)
20288 unsigned long long __tile_udn1_receive (void)
20289 unsigned long long __tile_udn2_receive (void)
20290 unsigned long long __tile_udn3_receive (void)
20291 void __tile_idn_send (unsigned long long)
20292 void __tile_udn_send (unsigned long long)
20296 The intrinsic @code{void __tile_network_barrier (void)} is used to
20297 guarantee that no network operations before it are reordered with
20300 @node TILEPro Built-in Functions
20301 @subsection TILEPro Built-in Functions
20303 GCC provides intrinsics to access every instruction of the TILEPro
20304 processor. The intrinsics are of the form:
20308 unsigned __insn_@var{op} (...)
20313 where @var{op} is the name of the instruction. Refer to the ISA manual
20314 for the complete list of instructions.
20316 GCC also provides intrinsics to directly access the network registers.
20317 The intrinsics are:
20321 unsigned __tile_idn0_receive (void)
20322 unsigned __tile_idn1_receive (void)
20323 unsigned __tile_sn_receive (void)
20324 unsigned __tile_udn0_receive (void)
20325 unsigned __tile_udn1_receive (void)
20326 unsigned __tile_udn2_receive (void)
20327 unsigned __tile_udn3_receive (void)
20328 void __tile_idn_send (unsigned)
20329 void __tile_sn_send (unsigned)
20330 void __tile_udn_send (unsigned)
20334 The intrinsic @code{void __tile_network_barrier (void)} is used to
20335 guarantee that no network operations before it are reordered with
20338 @node x86 Built-in Functions
20339 @subsection x86 Built-in Functions
20341 These built-in functions are available for the x86-32 and x86-64 family
20342 of computers, depending on the command-line switches used.
20344 If you specify command-line switches such as @option{-msse},
20345 the compiler could use the extended instruction sets even if the built-ins
20346 are not used explicitly in the program. For this reason, applications
20347 that perform run-time CPU detection must compile separate files for each
20348 supported architecture, using the appropriate flags. In particular,
20349 the file containing the CPU detection code should be compiled without
20352 The following machine modes are available for use with MMX built-in functions
20353 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
20354 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
20355 vector of eight 8-bit integers. Some of the built-in functions operate on
20356 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
20358 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
20359 of two 32-bit floating-point values.
20361 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
20362 floating-point values. Some instructions use a vector of four 32-bit
20363 integers, these use @code{V4SI}. Finally, some instructions operate on an
20364 entire vector register, interpreting it as a 128-bit integer, these use mode
20367 The x86-32 and x86-64 family of processors use additional built-in
20368 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
20369 floating point and @code{TC} 128-bit complex floating-point values.
20371 The following floating-point built-in functions are always available. All
20372 of them implement the function that is part of the name.
20375 __float128 __builtin_fabsq (__float128)
20376 __float128 __builtin_copysignq (__float128, __float128)
20379 The following built-in functions are always available.
20382 @item __float128 __builtin_infq (void)
20383 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
20384 @findex __builtin_infq
20386 @item __float128 __builtin_huge_valq (void)
20387 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
20388 @findex __builtin_huge_valq
20390 @item __float128 __builtin_nanq (void)
20391 Similar to @code{__builtin_nan}, except the return type is @code{__float128}.
20392 @findex __builtin_nanq
20394 @item __float128 __builtin_nansq (void)
20395 Similar to @code{__builtin_nans}, except the return type is @code{__float128}.
20396 @findex __builtin_nansq
20399 The following built-in function is always available.
20402 @item void __builtin_ia32_pause (void)
20403 Generates the @code{pause} machine instruction with a compiler memory
20407 The following built-in functions are always available and can be used to
20408 check the target platform type.
20410 @deftypefn {Built-in Function} void __builtin_cpu_init (void)
20411 This function runs the CPU detection code to check the type of CPU and the
20412 features supported. This built-in function needs to be invoked along with the built-in functions
20413 to check CPU type and features, @code{__builtin_cpu_is} and
20414 @code{__builtin_cpu_supports}, only when used in a function that is
20415 executed before any constructors are called. The CPU detection code is
20416 automatically executed in a very high priority constructor.
20418 For example, this function has to be used in @code{ifunc} resolvers that
20419 check for CPU type using the built-in functions @code{__builtin_cpu_is}
20420 and @code{__builtin_cpu_supports}, or in constructors on targets that
20421 don't support constructor priority.
20424 static void (*resolve_memcpy (void)) (void)
20426 // ifunc resolvers fire before constructors, explicitly call the init
20428 __builtin_cpu_init ();
20429 if (__builtin_cpu_supports ("ssse3"))
20430 return ssse3_memcpy; // super fast memcpy with ssse3 instructions.
20432 return default_memcpy;
20435 void *memcpy (void *, const void *, size_t)
20436 __attribute__ ((ifunc ("resolve_memcpy")));
20441 @deftypefn {Built-in Function} int __builtin_cpu_is (const char *@var{cpuname})
20442 This function returns a positive integer if the run-time CPU
20443 is of type @var{cpuname}
20444 and returns @code{0} otherwise. The following CPU names can be detected:
20460 Intel Core i7 Nehalem CPU.
20463 Intel Core i7 Westmere CPU.
20466 Intel Core i7 Sandy Bridge CPU.
20472 AMD Family 10h CPU.
20475 AMD Family 10h Barcelona CPU.
20478 AMD Family 10h Shanghai CPU.
20481 AMD Family 10h Istanbul CPU.
20484 AMD Family 14h CPU.
20487 AMD Family 15h CPU.
20490 AMD Family 15h Bulldozer version 1.
20493 AMD Family 15h Bulldozer version 2.
20496 AMD Family 15h Bulldozer version 3.
20499 AMD Family 15h Bulldozer version 4.
20502 AMD Family 16h CPU.
20505 AMD Family 17h CPU.
20508 AMD Family 17h Zen version 1.
20511 Here is an example:
20513 if (__builtin_cpu_is ("corei7"))
20515 do_corei7 (); // Core i7 specific implementation.
20519 do_generic (); // Generic implementation.
20524 @deftypefn {Built-in Function} int __builtin_cpu_supports (const char *@var{feature})
20525 This function returns a positive integer if the run-time CPU
20526 supports @var{feature}
20527 and returns @code{0} otherwise. The following features can be detected:
20535 POPCNT instruction.
20543 SSSE3 instructions.
20545 SSE4.1 instructions.
20547 SSE4.2 instructions.
20553 AVX512F instructions.
20556 Here is an example:
20558 if (__builtin_cpu_supports ("popcnt"))
20560 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc");
20564 count = generic_countbits (n); //generic implementation.
20570 The following built-in functions are made available by @option{-mmmx}.
20571 All of them generate the machine instruction that is part of the name.
20574 v8qi __builtin_ia32_paddb (v8qi, v8qi)
20575 v4hi __builtin_ia32_paddw (v4hi, v4hi)
20576 v2si __builtin_ia32_paddd (v2si, v2si)
20577 v8qi __builtin_ia32_psubb (v8qi, v8qi)
20578 v4hi __builtin_ia32_psubw (v4hi, v4hi)
20579 v2si __builtin_ia32_psubd (v2si, v2si)
20580 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
20581 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
20582 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
20583 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
20584 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
20585 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
20586 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
20587 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
20588 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
20589 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
20590 di __builtin_ia32_pand (di, di)
20591 di __builtin_ia32_pandn (di,di)
20592 di __builtin_ia32_por (di, di)
20593 di __builtin_ia32_pxor (di, di)
20594 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
20595 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
20596 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
20597 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
20598 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
20599 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
20600 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
20601 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
20602 v2si __builtin_ia32_punpckhdq (v2si, v2si)
20603 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
20604 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
20605 v2si __builtin_ia32_punpckldq (v2si, v2si)
20606 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
20607 v4hi __builtin_ia32_packssdw (v2si, v2si)
20608 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
20610 v4hi __builtin_ia32_psllw (v4hi, v4hi)
20611 v2si __builtin_ia32_pslld (v2si, v2si)
20612 v1di __builtin_ia32_psllq (v1di, v1di)
20613 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
20614 v2si __builtin_ia32_psrld (v2si, v2si)
20615 v1di __builtin_ia32_psrlq (v1di, v1di)
20616 v4hi __builtin_ia32_psraw (v4hi, v4hi)
20617 v2si __builtin_ia32_psrad (v2si, v2si)
20618 v4hi __builtin_ia32_psllwi (v4hi, int)
20619 v2si __builtin_ia32_pslldi (v2si, int)
20620 v1di __builtin_ia32_psllqi (v1di, int)
20621 v4hi __builtin_ia32_psrlwi (v4hi, int)
20622 v2si __builtin_ia32_psrldi (v2si, int)
20623 v1di __builtin_ia32_psrlqi (v1di, int)
20624 v4hi __builtin_ia32_psrawi (v4hi, int)
20625 v2si __builtin_ia32_psradi (v2si, int)
20629 The following built-in functions are made available either with
20630 @option{-msse}, or with @option{-m3dnowa}. All of them generate
20631 the machine instruction that is part of the name.
20634 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
20635 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
20636 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
20637 v1di __builtin_ia32_psadbw (v8qi, v8qi)
20638 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
20639 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
20640 v8qi __builtin_ia32_pminub (v8qi, v8qi)
20641 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
20642 int __builtin_ia32_pmovmskb (v8qi)
20643 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
20644 void __builtin_ia32_movntq (di *, di)
20645 void __builtin_ia32_sfence (void)
20648 The following built-in functions are available when @option{-msse} is used.
20649 All of them generate the machine instruction that is part of the name.
20652 int __builtin_ia32_comieq (v4sf, v4sf)
20653 int __builtin_ia32_comineq (v4sf, v4sf)
20654 int __builtin_ia32_comilt (v4sf, v4sf)
20655 int __builtin_ia32_comile (v4sf, v4sf)
20656 int __builtin_ia32_comigt (v4sf, v4sf)
20657 int __builtin_ia32_comige (v4sf, v4sf)
20658 int __builtin_ia32_ucomieq (v4sf, v4sf)
20659 int __builtin_ia32_ucomineq (v4sf, v4sf)
20660 int __builtin_ia32_ucomilt (v4sf, v4sf)
20661 int __builtin_ia32_ucomile (v4sf, v4sf)
20662 int __builtin_ia32_ucomigt (v4sf, v4sf)
20663 int __builtin_ia32_ucomige (v4sf, v4sf)
20664 v4sf __builtin_ia32_addps (v4sf, v4sf)
20665 v4sf __builtin_ia32_subps (v4sf, v4sf)
20666 v4sf __builtin_ia32_mulps (v4sf, v4sf)
20667 v4sf __builtin_ia32_divps (v4sf, v4sf)
20668 v4sf __builtin_ia32_addss (v4sf, v4sf)
20669 v4sf __builtin_ia32_subss (v4sf, v4sf)
20670 v4sf __builtin_ia32_mulss (v4sf, v4sf)
20671 v4sf __builtin_ia32_divss (v4sf, v4sf)
20672 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf)
20673 v4sf __builtin_ia32_cmpltps (v4sf, v4sf)
20674 v4sf __builtin_ia32_cmpleps (v4sf, v4sf)
20675 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf)
20676 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf)
20677 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf)
20678 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf)
20679 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf)
20680 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf)
20681 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf)
20682 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf)
20683 v4sf __builtin_ia32_cmpordps (v4sf, v4sf)
20684 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf)
20685 v4sf __builtin_ia32_cmpltss (v4sf, v4sf)
20686 v4sf __builtin_ia32_cmpless (v4sf, v4sf)
20687 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf)
20688 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf)
20689 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf)
20690 v4sf __builtin_ia32_cmpnless (v4sf, v4sf)
20691 v4sf __builtin_ia32_cmpordss (v4sf, v4sf)
20692 v4sf __builtin_ia32_maxps (v4sf, v4sf)
20693 v4sf __builtin_ia32_maxss (v4sf, v4sf)
20694 v4sf __builtin_ia32_minps (v4sf, v4sf)
20695 v4sf __builtin_ia32_minss (v4sf, v4sf)
20696 v4sf __builtin_ia32_andps (v4sf, v4sf)
20697 v4sf __builtin_ia32_andnps (v4sf, v4sf)
20698 v4sf __builtin_ia32_orps (v4sf, v4sf)
20699 v4sf __builtin_ia32_xorps (v4sf, v4sf)
20700 v4sf __builtin_ia32_movss (v4sf, v4sf)
20701 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
20702 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
20703 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
20704 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
20705 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
20706 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
20707 v2si __builtin_ia32_cvtps2pi (v4sf)
20708 int __builtin_ia32_cvtss2si (v4sf)
20709 v2si __builtin_ia32_cvttps2pi (v4sf)
20710 int __builtin_ia32_cvttss2si (v4sf)
20711 v4sf __builtin_ia32_rcpps (v4sf)
20712 v4sf __builtin_ia32_rsqrtps (v4sf)
20713 v4sf __builtin_ia32_sqrtps (v4sf)
20714 v4sf __builtin_ia32_rcpss (v4sf)
20715 v4sf __builtin_ia32_rsqrtss (v4sf)
20716 v4sf __builtin_ia32_sqrtss (v4sf)
20717 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
20718 void __builtin_ia32_movntps (float *, v4sf)
20719 int __builtin_ia32_movmskps (v4sf)
20722 The following built-in functions are available when @option{-msse} is used.
20725 @item v4sf __builtin_ia32_loadups (float *)
20726 Generates the @code{movups} machine instruction as a load from memory.
20727 @item void __builtin_ia32_storeups (float *, v4sf)
20728 Generates the @code{movups} machine instruction as a store to memory.
20729 @item v4sf __builtin_ia32_loadss (float *)
20730 Generates the @code{movss} machine instruction as a load from memory.
20731 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
20732 Generates the @code{movhps} machine instruction as a load from memory.
20733 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
20734 Generates the @code{movlps} machine instruction as a load from memory
20735 @item void __builtin_ia32_storehps (v2sf *, v4sf)
20736 Generates the @code{movhps} machine instruction as a store to memory.
20737 @item void __builtin_ia32_storelps (v2sf *, v4sf)
20738 Generates the @code{movlps} machine instruction as a store to memory.
20741 The following built-in functions are available when @option{-msse2} is used.
20742 All of them generate the machine instruction that is part of the name.
20745 int __builtin_ia32_comisdeq (v2df, v2df)
20746 int __builtin_ia32_comisdlt (v2df, v2df)
20747 int __builtin_ia32_comisdle (v2df, v2df)
20748 int __builtin_ia32_comisdgt (v2df, v2df)
20749 int __builtin_ia32_comisdge (v2df, v2df)
20750 int __builtin_ia32_comisdneq (v2df, v2df)
20751 int __builtin_ia32_ucomisdeq (v2df, v2df)
20752 int __builtin_ia32_ucomisdlt (v2df, v2df)
20753 int __builtin_ia32_ucomisdle (v2df, v2df)
20754 int __builtin_ia32_ucomisdgt (v2df, v2df)
20755 int __builtin_ia32_ucomisdge (v2df, v2df)
20756 int __builtin_ia32_ucomisdneq (v2df, v2df)
20757 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
20758 v2df __builtin_ia32_cmpltpd (v2df, v2df)
20759 v2df __builtin_ia32_cmplepd (v2df, v2df)
20760 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
20761 v2df __builtin_ia32_cmpgepd (v2df, v2df)
20762 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
20763 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
20764 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
20765 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
20766 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
20767 v2df __builtin_ia32_cmpngepd (v2df, v2df)
20768 v2df __builtin_ia32_cmpordpd (v2df, v2df)
20769 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
20770 v2df __builtin_ia32_cmpltsd (v2df, v2df)
20771 v2df __builtin_ia32_cmplesd (v2df, v2df)
20772 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
20773 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
20774 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
20775 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
20776 v2df __builtin_ia32_cmpordsd (v2df, v2df)
20777 v2di __builtin_ia32_paddq (v2di, v2di)
20778 v2di __builtin_ia32_psubq (v2di, v2di)
20779 v2df __builtin_ia32_addpd (v2df, v2df)
20780 v2df __builtin_ia32_subpd (v2df, v2df)
20781 v2df __builtin_ia32_mulpd (v2df, v2df)
20782 v2df __builtin_ia32_divpd (v2df, v2df)
20783 v2df __builtin_ia32_addsd (v2df, v2df)
20784 v2df __builtin_ia32_subsd (v2df, v2df)
20785 v2df __builtin_ia32_mulsd (v2df, v2df)
20786 v2df __builtin_ia32_divsd (v2df, v2df)
20787 v2df __builtin_ia32_minpd (v2df, v2df)
20788 v2df __builtin_ia32_maxpd (v2df, v2df)
20789 v2df __builtin_ia32_minsd (v2df, v2df)
20790 v2df __builtin_ia32_maxsd (v2df, v2df)
20791 v2df __builtin_ia32_andpd (v2df, v2df)
20792 v2df __builtin_ia32_andnpd (v2df, v2df)
20793 v2df __builtin_ia32_orpd (v2df, v2df)
20794 v2df __builtin_ia32_xorpd (v2df, v2df)
20795 v2df __builtin_ia32_movsd (v2df, v2df)
20796 v2df __builtin_ia32_unpckhpd (v2df, v2df)
20797 v2df __builtin_ia32_unpcklpd (v2df, v2df)
20798 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
20799 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
20800 v4si __builtin_ia32_paddd128 (v4si, v4si)
20801 v2di __builtin_ia32_paddq128 (v2di, v2di)
20802 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
20803 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
20804 v4si __builtin_ia32_psubd128 (v4si, v4si)
20805 v2di __builtin_ia32_psubq128 (v2di, v2di)
20806 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
20807 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
20808 v2di __builtin_ia32_pand128 (v2di, v2di)
20809 v2di __builtin_ia32_pandn128 (v2di, v2di)
20810 v2di __builtin_ia32_por128 (v2di, v2di)
20811 v2di __builtin_ia32_pxor128 (v2di, v2di)
20812 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
20813 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
20814 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
20815 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
20816 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
20817 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
20818 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
20819 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
20820 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
20821 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
20822 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
20823 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
20824 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
20825 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
20826 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
20827 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
20828 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
20829 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
20830 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
20831 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
20832 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
20833 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
20834 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
20835 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
20836 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
20837 v2df __builtin_ia32_loadupd (double *)
20838 void __builtin_ia32_storeupd (double *, v2df)
20839 v2df __builtin_ia32_loadhpd (v2df, double const *)
20840 v2df __builtin_ia32_loadlpd (v2df, double const *)
20841 int __builtin_ia32_movmskpd (v2df)
20842 int __builtin_ia32_pmovmskb128 (v16qi)
20843 void __builtin_ia32_movnti (int *, int)
20844 void __builtin_ia32_movnti64 (long long int *, long long int)
20845 void __builtin_ia32_movntpd (double *, v2df)
20846 void __builtin_ia32_movntdq (v2df *, v2df)
20847 v4si __builtin_ia32_pshufd (v4si, int)
20848 v8hi __builtin_ia32_pshuflw (v8hi, int)
20849 v8hi __builtin_ia32_pshufhw (v8hi, int)
20850 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
20851 v2df __builtin_ia32_sqrtpd (v2df)
20852 v2df __builtin_ia32_sqrtsd (v2df)
20853 v2df __builtin_ia32_shufpd (v2df, v2df, int)
20854 v2df __builtin_ia32_cvtdq2pd (v4si)
20855 v4sf __builtin_ia32_cvtdq2ps (v4si)
20856 v4si __builtin_ia32_cvtpd2dq (v2df)
20857 v2si __builtin_ia32_cvtpd2pi (v2df)
20858 v4sf __builtin_ia32_cvtpd2ps (v2df)
20859 v4si __builtin_ia32_cvttpd2dq (v2df)
20860 v2si __builtin_ia32_cvttpd2pi (v2df)
20861 v2df __builtin_ia32_cvtpi2pd (v2si)
20862 int __builtin_ia32_cvtsd2si (v2df)
20863 int __builtin_ia32_cvttsd2si (v2df)
20864 long long __builtin_ia32_cvtsd2si64 (v2df)
20865 long long __builtin_ia32_cvttsd2si64 (v2df)
20866 v4si __builtin_ia32_cvtps2dq (v4sf)
20867 v2df __builtin_ia32_cvtps2pd (v4sf)
20868 v4si __builtin_ia32_cvttps2dq (v4sf)
20869 v2df __builtin_ia32_cvtsi2sd (v2df, int)
20870 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
20871 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
20872 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
20873 void __builtin_ia32_clflush (const void *)
20874 void __builtin_ia32_lfence (void)
20875 void __builtin_ia32_mfence (void)
20876 v16qi __builtin_ia32_loaddqu (const char *)
20877 void __builtin_ia32_storedqu (char *, v16qi)
20878 v1di __builtin_ia32_pmuludq (v2si, v2si)
20879 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
20880 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
20881 v4si __builtin_ia32_pslld128 (v4si, v4si)
20882 v2di __builtin_ia32_psllq128 (v2di, v2di)
20883 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
20884 v4si __builtin_ia32_psrld128 (v4si, v4si)
20885 v2di __builtin_ia32_psrlq128 (v2di, v2di)
20886 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
20887 v4si __builtin_ia32_psrad128 (v4si, v4si)
20888 v2di __builtin_ia32_pslldqi128 (v2di, int)
20889 v8hi __builtin_ia32_psllwi128 (v8hi, int)
20890 v4si __builtin_ia32_pslldi128 (v4si, int)
20891 v2di __builtin_ia32_psllqi128 (v2di, int)
20892 v2di __builtin_ia32_psrldqi128 (v2di, int)
20893 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
20894 v4si __builtin_ia32_psrldi128 (v4si, int)
20895 v2di __builtin_ia32_psrlqi128 (v2di, int)
20896 v8hi __builtin_ia32_psrawi128 (v8hi, int)
20897 v4si __builtin_ia32_psradi128 (v4si, int)
20898 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
20899 v2di __builtin_ia32_movq128 (v2di)
20902 The following built-in functions are available when @option{-msse3} is used.
20903 All of them generate the machine instruction that is part of the name.
20906 v2df __builtin_ia32_addsubpd (v2df, v2df)
20907 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
20908 v2df __builtin_ia32_haddpd (v2df, v2df)
20909 v4sf __builtin_ia32_haddps (v4sf, v4sf)
20910 v2df __builtin_ia32_hsubpd (v2df, v2df)
20911 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
20912 v16qi __builtin_ia32_lddqu (char const *)
20913 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
20914 v4sf __builtin_ia32_movshdup (v4sf)
20915 v4sf __builtin_ia32_movsldup (v4sf)
20916 void __builtin_ia32_mwait (unsigned int, unsigned int)
20919 The following built-in functions are available when @option{-mssse3} is used.
20920 All of them generate the machine instruction that is part of the name.
20923 v2si __builtin_ia32_phaddd (v2si, v2si)
20924 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
20925 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
20926 v2si __builtin_ia32_phsubd (v2si, v2si)
20927 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
20928 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
20929 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
20930 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
20931 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
20932 v8qi __builtin_ia32_psignb (v8qi, v8qi)
20933 v2si __builtin_ia32_psignd (v2si, v2si)
20934 v4hi __builtin_ia32_psignw (v4hi, v4hi)
20935 v1di __builtin_ia32_palignr (v1di, v1di, int)
20936 v8qi __builtin_ia32_pabsb (v8qi)
20937 v2si __builtin_ia32_pabsd (v2si)
20938 v4hi __builtin_ia32_pabsw (v4hi)
20941 The following built-in functions are available when @option{-mssse3} is used.
20942 All of them generate the machine instruction that is part of the name.
20945 v4si __builtin_ia32_phaddd128 (v4si, v4si)
20946 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
20947 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
20948 v4si __builtin_ia32_phsubd128 (v4si, v4si)
20949 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
20950 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
20951 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
20952 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
20953 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
20954 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
20955 v4si __builtin_ia32_psignd128 (v4si, v4si)
20956 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
20957 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
20958 v16qi __builtin_ia32_pabsb128 (v16qi)
20959 v4si __builtin_ia32_pabsd128 (v4si)
20960 v8hi __builtin_ia32_pabsw128 (v8hi)
20963 The following built-in functions are available when @option{-msse4.1} is
20964 used. All of them generate the machine instruction that is part of the
20968 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
20969 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
20970 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
20971 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
20972 v2df __builtin_ia32_dppd (v2df, v2df, const int)
20973 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
20974 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
20975 v2di __builtin_ia32_movntdqa (v2di *);
20976 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
20977 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
20978 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
20979 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
20980 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
20981 v8hi __builtin_ia32_phminposuw128 (v8hi)
20982 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
20983 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
20984 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
20985 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
20986 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
20987 v4si __builtin_ia32_pminsd128 (v4si, v4si)
20988 v4si __builtin_ia32_pminud128 (v4si, v4si)
20989 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
20990 v4si __builtin_ia32_pmovsxbd128 (v16qi)
20991 v2di __builtin_ia32_pmovsxbq128 (v16qi)
20992 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
20993 v2di __builtin_ia32_pmovsxdq128 (v4si)
20994 v4si __builtin_ia32_pmovsxwd128 (v8hi)
20995 v2di __builtin_ia32_pmovsxwq128 (v8hi)
20996 v4si __builtin_ia32_pmovzxbd128 (v16qi)
20997 v2di __builtin_ia32_pmovzxbq128 (v16qi)
20998 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
20999 v2di __builtin_ia32_pmovzxdq128 (v4si)
21000 v4si __builtin_ia32_pmovzxwd128 (v8hi)
21001 v2di __builtin_ia32_pmovzxwq128 (v8hi)
21002 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
21003 v4si __builtin_ia32_pmulld128 (v4si, v4si)
21004 int __builtin_ia32_ptestc128 (v2di, v2di)
21005 int __builtin_ia32_ptestnzc128 (v2di, v2di)
21006 int __builtin_ia32_ptestz128 (v2di, v2di)
21007 v2df __builtin_ia32_roundpd (v2df, const int)
21008 v4sf __builtin_ia32_roundps (v4sf, const int)
21009 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
21010 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
21013 The following built-in functions are available when @option{-msse4.1} is
21017 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
21018 Generates the @code{insertps} machine instruction.
21019 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
21020 Generates the @code{pextrb} machine instruction.
21021 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
21022 Generates the @code{pinsrb} machine instruction.
21023 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
21024 Generates the @code{pinsrd} machine instruction.
21025 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
21026 Generates the @code{pinsrq} machine instruction in 64bit mode.
21029 The following built-in functions are changed to generate new SSE4.1
21030 instructions when @option{-msse4.1} is used.
21033 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
21034 Generates the @code{extractps} machine instruction.
21035 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
21036 Generates the @code{pextrd} machine instruction.
21037 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
21038 Generates the @code{pextrq} machine instruction in 64bit mode.
21041 The following built-in functions are available when @option{-msse4.2} is
21042 used. All of them generate the machine instruction that is part of the
21046 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
21047 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
21048 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
21049 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
21050 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
21051 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
21052 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
21053 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
21054 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
21055 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
21056 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
21057 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
21058 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
21059 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
21060 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
21063 The following built-in functions are available when @option{-msse4.2} is
21067 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
21068 Generates the @code{crc32b} machine instruction.
21069 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
21070 Generates the @code{crc32w} machine instruction.
21071 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
21072 Generates the @code{crc32l} machine instruction.
21073 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
21074 Generates the @code{crc32q} machine instruction.
21077 The following built-in functions are changed to generate new SSE4.2
21078 instructions when @option{-msse4.2} is used.
21081 @item int __builtin_popcount (unsigned int)
21082 Generates the @code{popcntl} machine instruction.
21083 @item int __builtin_popcountl (unsigned long)
21084 Generates the @code{popcntl} or @code{popcntq} machine instruction,
21085 depending on the size of @code{unsigned long}.
21086 @item int __builtin_popcountll (unsigned long long)
21087 Generates the @code{popcntq} machine instruction.
21090 The following built-in functions are available when @option{-mavx} is
21091 used. All of them generate the machine instruction that is part of the
21095 v4df __builtin_ia32_addpd256 (v4df,v4df)
21096 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
21097 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
21098 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
21099 v4df __builtin_ia32_andnpd256 (v4df,v4df)
21100 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
21101 v4df __builtin_ia32_andpd256 (v4df,v4df)
21102 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
21103 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
21104 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
21105 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
21106 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
21107 v2df __builtin_ia32_cmppd (v2df,v2df,int)
21108 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
21109 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
21110 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
21111 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
21112 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
21113 v4df __builtin_ia32_cvtdq2pd256 (v4si)
21114 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
21115 v4si __builtin_ia32_cvtpd2dq256 (v4df)
21116 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
21117 v8si __builtin_ia32_cvtps2dq256 (v8sf)
21118 v4df __builtin_ia32_cvtps2pd256 (v4sf)
21119 v4si __builtin_ia32_cvttpd2dq256 (v4df)
21120 v8si __builtin_ia32_cvttps2dq256 (v8sf)
21121 v4df __builtin_ia32_divpd256 (v4df,v4df)
21122 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
21123 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
21124 v4df __builtin_ia32_haddpd256 (v4df,v4df)
21125 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
21126 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
21127 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
21128 v32qi __builtin_ia32_lddqu256 (pcchar)
21129 v32qi __builtin_ia32_loaddqu256 (pcchar)
21130 v4df __builtin_ia32_loadupd256 (pcdouble)
21131 v8sf __builtin_ia32_loadups256 (pcfloat)
21132 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
21133 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
21134 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
21135 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
21136 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
21137 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
21138 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
21139 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
21140 v4df __builtin_ia32_maxpd256 (v4df,v4df)
21141 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
21142 v4df __builtin_ia32_minpd256 (v4df,v4df)
21143 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
21144 v4df __builtin_ia32_movddup256 (v4df)
21145 int __builtin_ia32_movmskpd256 (v4df)
21146 int __builtin_ia32_movmskps256 (v8sf)
21147 v8sf __builtin_ia32_movshdup256 (v8sf)
21148 v8sf __builtin_ia32_movsldup256 (v8sf)
21149 v4df __builtin_ia32_mulpd256 (v4df,v4df)
21150 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
21151 v4df __builtin_ia32_orpd256 (v4df,v4df)
21152 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
21153 v2df __builtin_ia32_pd_pd256 (v4df)
21154 v4df __builtin_ia32_pd256_pd (v2df)
21155 v4sf __builtin_ia32_ps_ps256 (v8sf)
21156 v8sf __builtin_ia32_ps256_ps (v4sf)
21157 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
21158 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
21159 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
21160 v8sf __builtin_ia32_rcpps256 (v8sf)
21161 v4df __builtin_ia32_roundpd256 (v4df,int)
21162 v8sf __builtin_ia32_roundps256 (v8sf,int)
21163 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
21164 v8sf __builtin_ia32_rsqrtps256 (v8sf)
21165 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
21166 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
21167 v4si __builtin_ia32_si_si256 (v8si)
21168 v8si __builtin_ia32_si256_si (v4si)
21169 v4df __builtin_ia32_sqrtpd256 (v4df)
21170 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
21171 v8sf __builtin_ia32_sqrtps256 (v8sf)
21172 void __builtin_ia32_storedqu256 (pchar,v32qi)
21173 void __builtin_ia32_storeupd256 (pdouble,v4df)
21174 void __builtin_ia32_storeups256 (pfloat,v8sf)
21175 v4df __builtin_ia32_subpd256 (v4df,v4df)
21176 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
21177 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
21178 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
21179 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
21180 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
21181 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
21182 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
21183 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
21184 v4sf __builtin_ia32_vbroadcastss (pcfloat)
21185 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
21186 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
21187 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
21188 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
21189 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
21190 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
21191 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
21192 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
21193 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
21194 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
21195 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
21196 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
21197 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
21198 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
21199 v2df __builtin_ia32_vpermilpd (v2df,int)
21200 v4df __builtin_ia32_vpermilpd256 (v4df,int)
21201 v4sf __builtin_ia32_vpermilps (v4sf,int)
21202 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
21203 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
21204 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
21205 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
21206 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
21207 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
21208 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
21209 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
21210 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
21211 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
21212 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
21213 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
21214 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
21215 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
21216 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
21217 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
21218 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
21219 void __builtin_ia32_vzeroall (void)
21220 void __builtin_ia32_vzeroupper (void)
21221 v4df __builtin_ia32_xorpd256 (v4df,v4df)
21222 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
21225 The following built-in functions are available when @option{-mavx2} is
21226 used. All of them generate the machine instruction that is part of the
21230 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int)
21231 v32qi __builtin_ia32_pabsb256 (v32qi)
21232 v16hi __builtin_ia32_pabsw256 (v16hi)
21233 v8si __builtin_ia32_pabsd256 (v8si)
21234 v16hi __builtin_ia32_packssdw256 (v8si,v8si)
21235 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
21236 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
21237 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
21238 v32qi __builtin_ia32_paddb256 (v32qi,v32qi)
21239 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
21240 v8si __builtin_ia32_paddd256 (v8si,v8si)
21241 v4di __builtin_ia32_paddq256 (v4di,v4di)
21242 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
21243 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
21244 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
21245 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
21246 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
21247 v4di __builtin_ia32_andsi256 (v4di,v4di)
21248 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
21249 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi)
21250 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
21251 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
21252 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
21253 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
21254 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
21255 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
21256 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
21257 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
21258 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
21259 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
21260 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
21261 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
21262 v8si __builtin_ia32_phaddd256 (v8si,v8si)
21263 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
21264 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
21265 v8si __builtin_ia32_phsubd256 (v8si,v8si)
21266 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
21267 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
21268 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
21269 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
21270 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
21271 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
21272 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
21273 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
21274 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
21275 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
21276 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
21277 v8si __builtin_ia32_pminsd256 (v8si,v8si)
21278 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
21279 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
21280 v8si __builtin_ia32_pminud256 (v8si,v8si)
21281 int __builtin_ia32_pmovmskb256 (v32qi)
21282 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
21283 v8si __builtin_ia32_pmovsxbd256 (v16qi)
21284 v4di __builtin_ia32_pmovsxbq256 (v16qi)
21285 v8si __builtin_ia32_pmovsxwd256 (v8hi)
21286 v4di __builtin_ia32_pmovsxwq256 (v8hi)
21287 v4di __builtin_ia32_pmovsxdq256 (v4si)
21288 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
21289 v8si __builtin_ia32_pmovzxbd256 (v16qi)
21290 v4di __builtin_ia32_pmovzxbq256 (v16qi)
21291 v8si __builtin_ia32_pmovzxwd256 (v8hi)
21292 v4di __builtin_ia32_pmovzxwq256 (v8hi)
21293 v4di __builtin_ia32_pmovzxdq256 (v4si)
21294 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
21295 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
21296 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
21297 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
21298 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
21299 v8si __builtin_ia32_pmulld256 (v8si,v8si)
21300 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
21301 v4di __builtin_ia32_por256 (v4di,v4di)
21302 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
21303 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
21304 v8si __builtin_ia32_pshufd256 (v8si,int)
21305 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
21306 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
21307 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
21308 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
21309 v8si __builtin_ia32_psignd256 (v8si,v8si)
21310 v4di __builtin_ia32_pslldqi256 (v4di,int)
21311 v16hi __builtin_ia32_psllwi256 (16hi,int)
21312 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
21313 v8si __builtin_ia32_pslldi256 (v8si,int)
21314 v8si __builtin_ia32_pslld256(v8si,v4si)
21315 v4di __builtin_ia32_psllqi256 (v4di,int)
21316 v4di __builtin_ia32_psllq256(v4di,v2di)
21317 v16hi __builtin_ia32_psrawi256 (v16hi,int)
21318 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
21319 v8si __builtin_ia32_psradi256 (v8si,int)
21320 v8si __builtin_ia32_psrad256 (v8si,v4si)
21321 v4di __builtin_ia32_psrldqi256 (v4di, int)
21322 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
21323 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
21324 v8si __builtin_ia32_psrldi256 (v8si,int)
21325 v8si __builtin_ia32_psrld256 (v8si,v4si)
21326 v4di __builtin_ia32_psrlqi256 (v4di,int)
21327 v4di __builtin_ia32_psrlq256(v4di,v2di)
21328 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
21329 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
21330 v8si __builtin_ia32_psubd256 (v8si,v8si)
21331 v4di __builtin_ia32_psubq256 (v4di,v4di)
21332 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
21333 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
21334 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
21335 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
21336 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
21337 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
21338 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
21339 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
21340 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
21341 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
21342 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
21343 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
21344 v4di __builtin_ia32_pxor256 (v4di,v4di)
21345 v4di __builtin_ia32_movntdqa256 (pv4di)
21346 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
21347 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
21348 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
21349 v4di __builtin_ia32_vbroadcastsi256 (v2di)
21350 v4si __builtin_ia32_pblendd128 (v4si,v4si)
21351 v8si __builtin_ia32_pblendd256 (v8si,v8si)
21352 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
21353 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
21354 v8si __builtin_ia32_pbroadcastd256 (v4si)
21355 v4di __builtin_ia32_pbroadcastq256 (v2di)
21356 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
21357 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
21358 v4si __builtin_ia32_pbroadcastd128 (v4si)
21359 v2di __builtin_ia32_pbroadcastq128 (v2di)
21360 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
21361 v4df __builtin_ia32_permdf256 (v4df,int)
21362 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
21363 v4di __builtin_ia32_permdi256 (v4di,int)
21364 v4di __builtin_ia32_permti256 (v4di,v4di,int)
21365 v4di __builtin_ia32_extract128i256 (v4di,int)
21366 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
21367 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
21368 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
21369 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
21370 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
21371 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
21372 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
21373 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
21374 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
21375 v8si __builtin_ia32_psllv8si (v8si,v8si)
21376 v4si __builtin_ia32_psllv4si (v4si,v4si)
21377 v4di __builtin_ia32_psllv4di (v4di,v4di)
21378 v2di __builtin_ia32_psllv2di (v2di,v2di)
21379 v8si __builtin_ia32_psrav8si (v8si,v8si)
21380 v4si __builtin_ia32_psrav4si (v4si,v4si)
21381 v8si __builtin_ia32_psrlv8si (v8si,v8si)
21382 v4si __builtin_ia32_psrlv4si (v4si,v4si)
21383 v4di __builtin_ia32_psrlv4di (v4di,v4di)
21384 v2di __builtin_ia32_psrlv2di (v2di,v2di)
21385 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
21386 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
21387 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
21388 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
21389 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
21390 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
21391 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
21392 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
21393 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
21394 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
21395 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
21396 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
21397 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
21398 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
21399 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
21400 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
21403 The following built-in functions are available when @option{-maes} is
21404 used. All of them generate the machine instruction that is part of the
21408 v2di __builtin_ia32_aesenc128 (v2di, v2di)
21409 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
21410 v2di __builtin_ia32_aesdec128 (v2di, v2di)
21411 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
21412 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
21413 v2di __builtin_ia32_aesimc128 (v2di)
21416 The following built-in function is available when @option{-mpclmul} is
21420 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
21421 Generates the @code{pclmulqdq} machine instruction.
21424 The following built-in function is available when @option{-mfsgsbase} is
21425 used. All of them generate the machine instruction that is part of the
21429 unsigned int __builtin_ia32_rdfsbase32 (void)
21430 unsigned long long __builtin_ia32_rdfsbase64 (void)
21431 unsigned int __builtin_ia32_rdgsbase32 (void)
21432 unsigned long long __builtin_ia32_rdgsbase64 (void)
21433 void _writefsbase_u32 (unsigned int)
21434 void _writefsbase_u64 (unsigned long long)
21435 void _writegsbase_u32 (unsigned int)
21436 void _writegsbase_u64 (unsigned long long)
21439 The following built-in function is available when @option{-mrdrnd} is
21440 used. All of them generate the machine instruction that is part of the
21444 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
21445 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
21446 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
21449 The following built-in functions are available when @option{-msse4a} is used.
21450 All of them generate the machine instruction that is part of the name.
21453 void __builtin_ia32_movntsd (double *, v2df)
21454 void __builtin_ia32_movntss (float *, v4sf)
21455 v2di __builtin_ia32_extrq (v2di, v16qi)
21456 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
21457 v2di __builtin_ia32_insertq (v2di, v2di)
21458 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
21461 The following built-in functions are available when @option{-mxop} is used.
21463 v2df __builtin_ia32_vfrczpd (v2df)
21464 v4sf __builtin_ia32_vfrczps (v4sf)
21465 v2df __builtin_ia32_vfrczsd (v2df)
21466 v4sf __builtin_ia32_vfrczss (v4sf)
21467 v4df __builtin_ia32_vfrczpd256 (v4df)
21468 v8sf __builtin_ia32_vfrczps256 (v8sf)
21469 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
21470 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
21471 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
21472 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
21473 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
21474 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
21475 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
21476 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
21477 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
21478 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
21479 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
21480 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
21481 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
21482 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
21483 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
21484 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
21485 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
21486 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
21487 v4si __builtin_ia32_vpcomequd (v4si, v4si)
21488 v2di __builtin_ia32_vpcomequq (v2di, v2di)
21489 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
21490 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
21491 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
21492 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
21493 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
21494 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
21495 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
21496 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
21497 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
21498 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
21499 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
21500 v4si __builtin_ia32_vpcomged (v4si, v4si)
21501 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
21502 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
21503 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
21504 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
21505 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
21506 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
21507 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
21508 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
21509 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
21510 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
21511 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
21512 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
21513 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
21514 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
21515 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
21516 v4si __builtin_ia32_vpcomled (v4si, v4si)
21517 v2di __builtin_ia32_vpcomleq (v2di, v2di)
21518 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
21519 v4si __builtin_ia32_vpcomleud (v4si, v4si)
21520 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
21521 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
21522 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
21523 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
21524 v4si __builtin_ia32_vpcomltd (v4si, v4si)
21525 v2di __builtin_ia32_vpcomltq (v2di, v2di)
21526 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
21527 v4si __builtin_ia32_vpcomltud (v4si, v4si)
21528 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
21529 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
21530 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
21531 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
21532 v4si __builtin_ia32_vpcomned (v4si, v4si)
21533 v2di __builtin_ia32_vpcomneq (v2di, v2di)
21534 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
21535 v4si __builtin_ia32_vpcomneud (v4si, v4si)
21536 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
21537 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
21538 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
21539 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
21540 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
21541 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
21542 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
21543 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
21544 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
21545 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
21546 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
21547 v4si __builtin_ia32_vphaddbd (v16qi)
21548 v2di __builtin_ia32_vphaddbq (v16qi)
21549 v8hi __builtin_ia32_vphaddbw (v16qi)
21550 v2di __builtin_ia32_vphadddq (v4si)
21551 v4si __builtin_ia32_vphaddubd (v16qi)
21552 v2di __builtin_ia32_vphaddubq (v16qi)
21553 v8hi __builtin_ia32_vphaddubw (v16qi)
21554 v2di __builtin_ia32_vphaddudq (v4si)
21555 v4si __builtin_ia32_vphadduwd (v8hi)
21556 v2di __builtin_ia32_vphadduwq (v8hi)
21557 v4si __builtin_ia32_vphaddwd (v8hi)
21558 v2di __builtin_ia32_vphaddwq (v8hi)
21559 v8hi __builtin_ia32_vphsubbw (v16qi)
21560 v2di __builtin_ia32_vphsubdq (v4si)
21561 v4si __builtin_ia32_vphsubwd (v8hi)
21562 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
21563 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
21564 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
21565 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
21566 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
21567 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
21568 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
21569 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
21570 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
21571 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
21572 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
21573 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
21574 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
21575 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
21576 v4si __builtin_ia32_vprotd (v4si, v4si)
21577 v2di __builtin_ia32_vprotq (v2di, v2di)
21578 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
21579 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
21580 v4si __builtin_ia32_vpshad (v4si, v4si)
21581 v2di __builtin_ia32_vpshaq (v2di, v2di)
21582 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
21583 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
21584 v4si __builtin_ia32_vpshld (v4si, v4si)
21585 v2di __builtin_ia32_vpshlq (v2di, v2di)
21586 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
21589 The following built-in functions are available when @option{-mfma4} is used.
21590 All of them generate the machine instruction that is part of the name.
21593 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df)
21594 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf)
21595 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df)
21596 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf)
21597 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df)
21598 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf)
21599 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df)
21600 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf)
21601 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df)
21602 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf)
21603 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df)
21604 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf)
21605 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df)
21606 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf)
21607 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df)
21608 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf)
21609 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df)
21610 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf)
21611 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df)
21612 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf)
21613 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df)
21614 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf)
21615 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df)
21616 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf)
21617 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df)
21618 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf)
21619 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df)
21620 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf)
21621 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df)
21622 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf)
21623 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df)
21624 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf)
21628 The following built-in functions are available when @option{-mlwp} is used.
21631 void __builtin_ia32_llwpcb16 (void *);
21632 void __builtin_ia32_llwpcb32 (void *);
21633 void __builtin_ia32_llwpcb64 (void *);
21634 void * __builtin_ia32_llwpcb16 (void);
21635 void * __builtin_ia32_llwpcb32 (void);
21636 void * __builtin_ia32_llwpcb64 (void);
21637 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
21638 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
21639 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
21640 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
21641 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
21642 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
21645 The following built-in functions are available when @option{-mbmi} is used.
21646 All of them generate the machine instruction that is part of the name.
21648 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
21649 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
21652 The following built-in functions are available when @option{-mbmi2} is used.
21653 All of them generate the machine instruction that is part of the name.
21655 unsigned int _bzhi_u32 (unsigned int, unsigned int)
21656 unsigned int _pdep_u32 (unsigned int, unsigned int)
21657 unsigned int _pext_u32 (unsigned int, unsigned int)
21658 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
21659 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
21660 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
21663 The following built-in functions are available when @option{-mlzcnt} is used.
21664 All of them generate the machine instruction that is part of the name.
21666 unsigned short __builtin_ia32_lzcnt_u16(unsigned short);
21667 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
21668 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
21671 The following built-in functions are available when @option{-mfxsr} is used.
21672 All of them generate the machine instruction that is part of the name.
21674 void __builtin_ia32_fxsave (void *)
21675 void __builtin_ia32_fxrstor (void *)
21676 void __builtin_ia32_fxsave64 (void *)
21677 void __builtin_ia32_fxrstor64 (void *)
21680 The following built-in functions are available when @option{-mxsave} is used.
21681 All of them generate the machine instruction that is part of the name.
21683 void __builtin_ia32_xsave (void *, long long)
21684 void __builtin_ia32_xrstor (void *, long long)
21685 void __builtin_ia32_xsave64 (void *, long long)
21686 void __builtin_ia32_xrstor64 (void *, long long)
21689 The following built-in functions are available when @option{-mxsaveopt} is used.
21690 All of them generate the machine instruction that is part of the name.
21692 void __builtin_ia32_xsaveopt (void *, long long)
21693 void __builtin_ia32_xsaveopt64 (void *, long long)
21696 The following built-in functions are available when @option{-mtbm} is used.
21697 Both of them generate the immediate form of the bextr machine instruction.
21699 unsigned int __builtin_ia32_bextri_u32 (unsigned int,
21700 const unsigned int);
21701 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long,
21702 const unsigned long long);
21706 The following built-in functions are available when @option{-m3dnow} is used.
21707 All of them generate the machine instruction that is part of the name.
21710 void __builtin_ia32_femms (void)
21711 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
21712 v2si __builtin_ia32_pf2id (v2sf)
21713 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
21714 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
21715 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
21716 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
21717 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
21718 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
21719 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
21720 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
21721 v2sf __builtin_ia32_pfrcp (v2sf)
21722 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
21723 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
21724 v2sf __builtin_ia32_pfrsqrt (v2sf)
21725 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
21726 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
21727 v2sf __builtin_ia32_pi2fd (v2si)
21728 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
21731 The following built-in functions are available when @option{-m3dnowa} is used.
21732 All of them generate the machine instruction that is part of the name.
21735 v2si __builtin_ia32_pf2iw (v2sf)
21736 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
21737 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
21738 v2sf __builtin_ia32_pi2fw (v2si)
21739 v2sf __builtin_ia32_pswapdsf (v2sf)
21740 v2si __builtin_ia32_pswapdsi (v2si)
21743 The following built-in functions are available when @option{-mrtm} is used
21744 They are used for restricted transactional memory. These are the internal
21745 low level functions. Normally the functions in
21746 @ref{x86 transactional memory intrinsics} should be used instead.
21749 int __builtin_ia32_xbegin ()
21750 void __builtin_ia32_xend ()
21751 void __builtin_ia32_xabort (status)
21752 int __builtin_ia32_xtest ()
21755 The following built-in functions are available when @option{-mmwaitx} is used.
21756 All of them generate the machine instruction that is part of the name.
21758 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int)
21759 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int)
21762 The following built-in functions are available when @option{-mclzero} is used.
21763 All of them generate the machine instruction that is part of the name.
21765 void __builtin_i32_clzero (void *)
21768 The following built-in functions are available when @option{-mpku} is used.
21769 They generate reads and writes to PKRU.
21771 void __builtin_ia32_wrpkru (unsigned int)
21772 unsigned int __builtin_ia32_rdpkru ()
21775 The following built-in functions are available when @option{-mcet} is used.
21776 They are used to support Intel Control-flow Enforcment Technology (CET).
21777 Each built-in function generates the machine instruction that is part of the
21780 unsigned int __builtin_ia32_rdsspd (unsigned int)
21781 unsigned long long __builtin_ia32_rdsspq (unsigned long long)
21782 void __builtin_ia32_incsspd (unsigned int)
21783 void __builtin_ia32_incsspq (unsigned long long)
21784 void __builtin_ia32_saveprevssp(void);
21785 void __builtin_ia32_rstorssp(void *);
21786 void __builtin_ia32_wrssd(unsigned int, void *);
21787 void __builtin_ia32_wrssq(unsigned long long, void *);
21788 void __builtin_ia32_wrussd(unsigned int, void *);
21789 void __builtin_ia32_wrussq(unsigned long long, void *);
21790 void __builtin_ia32_setssbsy(void);
21791 void __builtin_ia32_clrssbsy(void *);
21794 @node x86 transactional memory intrinsics
21795 @subsection x86 Transactional Memory Intrinsics
21797 These hardware transactional memory intrinsics for x86 allow you to use
21798 memory transactions with RTM (Restricted Transactional Memory).
21799 This support is enabled with the @option{-mrtm} option.
21800 For using HLE (Hardware Lock Elision) see
21801 @ref{x86 specific memory model extensions for transactional memory} instead.
21803 A memory transaction commits all changes to memory in an atomic way,
21804 as visible to other threads. If the transaction fails it is rolled back
21805 and all side effects discarded.
21807 Generally there is no guarantee that a memory transaction ever succeeds
21808 and suitable fallback code always needs to be supplied.
21810 @deftypefn {RTM Function} {unsigned} _xbegin ()
21811 Start a RTM (Restricted Transactional Memory) transaction.
21812 Returns @code{_XBEGIN_STARTED} when the transaction
21813 started successfully (note this is not 0, so the constant has to be
21814 explicitly tested).
21816 If the transaction aborts, all side-effects
21817 are undone and an abort code encoded as a bit mask is returned.
21818 The following macros are defined:
21821 @item _XABORT_EXPLICIT
21822 Transaction was explicitly aborted with @code{_xabort}. The parameter passed
21823 to @code{_xabort} is available with @code{_XABORT_CODE(status)}.
21824 @item _XABORT_RETRY
21825 Transaction retry is possible.
21826 @item _XABORT_CONFLICT
21827 Transaction abort due to a memory conflict with another thread.
21828 @item _XABORT_CAPACITY
21829 Transaction abort due to the transaction using too much memory.
21830 @item _XABORT_DEBUG
21831 Transaction abort due to a debug trap.
21832 @item _XABORT_NESTED
21833 Transaction abort in an inner nested transaction.
21836 There is no guarantee
21837 any transaction ever succeeds, so there always needs to be a valid
21841 @deftypefn {RTM Function} {void} _xend ()
21842 Commit the current transaction. When no transaction is active this faults.
21843 All memory side-effects of the transaction become visible
21844 to other threads in an atomic manner.
21847 @deftypefn {RTM Function} {int} _xtest ()
21848 Return a nonzero value if a transaction is currently active, otherwise 0.
21851 @deftypefn {RTM Function} {void} _xabort (status)
21852 Abort the current transaction. When no transaction is active this is a no-op.
21853 The @var{status} is an 8-bit constant; its value is encoded in the return
21854 value from @code{_xbegin}.
21857 Here is an example showing handling for @code{_XABORT_RETRY}
21858 and a fallback path for other failures:
21861 #include <immintrin.h>
21863 int n_tries, max_tries;
21864 unsigned status = _XABORT_EXPLICIT;
21867 for (n_tries = 0; n_tries < max_tries; n_tries++)
21869 status = _xbegin ();
21870 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY))
21873 if (status == _XBEGIN_STARTED)
21875 ... transaction code...
21880 ... non-transactional fallback path...
21885 Note that, in most cases, the transactional and non-transactional code
21886 must synchronize together to ensure consistency.
21888 @node Target Format Checks
21889 @section Format Checks Specific to Particular Target Machines
21891 For some target machines, GCC supports additional options to the
21893 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
21896 * Solaris Format Checks::
21897 * Darwin Format Checks::
21900 @node Solaris Format Checks
21901 @subsection Solaris Format Checks
21903 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
21904 check. @code{cmn_err} accepts a subset of the standard @code{printf}
21905 conversions, and the two-argument @code{%b} conversion for displaying
21906 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
21908 @node Darwin Format Checks
21909 @subsection Darwin Format Checks
21911 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
21912 attribute context. Declarations made with such attribution are parsed for correct syntax
21913 and format argument types. However, parsing of the format string itself is currently undefined
21914 and is not carried out by this version of the compiler.
21916 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
21917 also be used as format arguments. Note that the relevant headers are only likely to be
21918 available on Darwin (OSX) installations. On such installations, the XCode and system
21919 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
21920 associated functions.
21923 @section Pragmas Accepted by GCC
21925 @cindex @code{#pragma}
21927 GCC supports several types of pragmas, primarily in order to compile
21928 code originally written for other compilers. Note that in general
21929 we do not recommend the use of pragmas; @xref{Function Attributes},
21930 for further explanation.
21933 * AArch64 Pragmas::
21937 * RS/6000 and PowerPC Pragmas::
21940 * Solaris Pragmas::
21941 * Symbol-Renaming Pragmas::
21942 * Structure-Layout Pragmas::
21944 * Diagnostic Pragmas::
21945 * Visibility Pragmas::
21946 * Push/Pop Macro Pragmas::
21947 * Function Specific Option Pragmas::
21948 * Loop-Specific Pragmas::
21951 @node AArch64 Pragmas
21952 @subsection AArch64 Pragmas
21954 The pragmas defined by the AArch64 target correspond to the AArch64
21955 target function attributes. They can be specified as below:
21957 #pragma GCC target("string")
21960 where @code{@var{string}} can be any string accepted as an AArch64 target
21961 attribute. @xref{AArch64 Function Attributes}, for more details
21962 on the permissible values of @code{string}.
21965 @subsection ARM Pragmas
21967 The ARM target defines pragmas for controlling the default addition of
21968 @code{long_call} and @code{short_call} attributes to functions.
21969 @xref{Function Attributes}, for information about the effects of these
21974 @cindex pragma, long_calls
21975 Set all subsequent functions to have the @code{long_call} attribute.
21977 @item no_long_calls
21978 @cindex pragma, no_long_calls
21979 Set all subsequent functions to have the @code{short_call} attribute.
21981 @item long_calls_off
21982 @cindex pragma, long_calls_off
21983 Do not affect the @code{long_call} or @code{short_call} attributes of
21984 subsequent functions.
21988 @subsection M32C Pragmas
21991 @item GCC memregs @var{number}
21992 @cindex pragma, memregs
21993 Overrides the command-line option @code{-memregs=} for the current
21994 file. Use with care! This pragma must be before any function in the
21995 file, and mixing different memregs values in different objects may
21996 make them incompatible. This pragma is useful when a
21997 performance-critical function uses a memreg for temporary values,
21998 as it may allow you to reduce the number of memregs used.
22000 @item ADDRESS @var{name} @var{address}
22001 @cindex pragma, address
22002 For any declared symbols matching @var{name}, this does three things
22003 to that symbol: it forces the symbol to be located at the given
22004 address (a number), it forces the symbol to be volatile, and it
22005 changes the symbol's scope to be static. This pragma exists for
22006 compatibility with other compilers, but note that the common
22007 @code{1234H} numeric syntax is not supported (use @code{0x1234}
22011 #pragma ADDRESS port3 0x103
22018 @subsection MeP Pragmas
22022 @item custom io_volatile (on|off)
22023 @cindex pragma, custom io_volatile
22024 Overrides the command-line option @code{-mio-volatile} for the current
22025 file. Note that for compatibility with future GCC releases, this
22026 option should only be used once before any @code{io} variables in each
22029 @item GCC coprocessor available @var{registers}
22030 @cindex pragma, coprocessor available
22031 Specifies which coprocessor registers are available to the register
22032 allocator. @var{registers} may be a single register, register range
22033 separated by ellipses, or comma-separated list of those. Example:
22036 #pragma GCC coprocessor available $c0...$c10, $c28
22039 @item GCC coprocessor call_saved @var{registers}
22040 @cindex pragma, coprocessor call_saved
22041 Specifies which coprocessor registers are to be saved and restored by
22042 any function using them. @var{registers} may be a single register,
22043 register range separated by ellipses, or comma-separated list of
22047 #pragma GCC coprocessor call_saved $c4...$c6, $c31
22050 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
22051 @cindex pragma, coprocessor subclass
22052 Creates and defines a register class. These register classes can be
22053 used by inline @code{asm} constructs. @var{registers} may be a single
22054 register, register range separated by ellipses, or comma-separated
22055 list of those. Example:
22058 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
22060 asm ("cpfoo %0" : "=B" (x));
22063 @item GCC disinterrupt @var{name} , @var{name} @dots{}
22064 @cindex pragma, disinterrupt
22065 For the named functions, the compiler adds code to disable interrupts
22066 for the duration of those functions. If any functions so named
22067 are not encountered in the source, a warning is emitted that the pragma is
22068 not used. Examples:
22071 #pragma disinterrupt foo
22072 #pragma disinterrupt bar, grill
22073 int foo () @{ @dots{} @}
22076 @item GCC call @var{name} , @var{name} @dots{}
22077 @cindex pragma, call
22078 For the named functions, the compiler always uses a register-indirect
22079 call model when calling the named functions. Examples:
22088 @node RS/6000 and PowerPC Pragmas
22089 @subsection RS/6000 and PowerPC Pragmas
22091 The RS/6000 and PowerPC targets define one pragma for controlling
22092 whether or not the @code{longcall} attribute is added to function
22093 declarations by default. This pragma overrides the @option{-mlongcall}
22094 option, but not the @code{longcall} and @code{shortcall} attributes.
22095 @xref{RS/6000 and PowerPC Options}, for more information about when long
22096 calls are and are not necessary.
22100 @cindex pragma, longcall
22101 Apply the @code{longcall} attribute to all subsequent function
22105 Do not apply the @code{longcall} attribute to subsequent function
22109 @c Describe h8300 pragmas here.
22110 @c Describe sh pragmas here.
22111 @c Describe v850 pragmas here.
22113 @node S/390 Pragmas
22114 @subsection S/390 Pragmas
22116 The pragmas defined by the S/390 target correspond to the S/390
22117 target function attributes and some the additional options:
22124 Note that options of the pragma, unlike options of the target
22125 attribute, do change the value of preprocessor macros like
22126 @code{__VEC__}. They can be specified as below:
22129 #pragma GCC target("string[,string]...")
22130 #pragma GCC target("string"[,"string"]...)
22133 @node Darwin Pragmas
22134 @subsection Darwin Pragmas
22136 The following pragmas are available for all architectures running the
22137 Darwin operating system. These are useful for compatibility with other
22141 @item mark @var{tokens}@dots{}
22142 @cindex pragma, mark
22143 This pragma is accepted, but has no effect.
22145 @item options align=@var{alignment}
22146 @cindex pragma, options align
22147 This pragma sets the alignment of fields in structures. The values of
22148 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
22149 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
22150 properly; to restore the previous setting, use @code{reset} for the
22153 @item segment @var{tokens}@dots{}
22154 @cindex pragma, segment
22155 This pragma is accepted, but has no effect.
22157 @item unused (@var{var} [, @var{var}]@dots{})
22158 @cindex pragma, unused
22159 This pragma declares variables to be possibly unused. GCC does not
22160 produce warnings for the listed variables. The effect is similar to
22161 that of the @code{unused} attribute, except that this pragma may appear
22162 anywhere within the variables' scopes.
22165 @node Solaris Pragmas
22166 @subsection Solaris Pragmas
22168 The Solaris target supports @code{#pragma redefine_extname}
22169 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
22170 @code{#pragma} directives for compatibility with the system compiler.
22173 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
22174 @cindex pragma, align
22176 Increase the minimum alignment of each @var{variable} to @var{alignment}.
22177 This is the same as GCC's @code{aligned} attribute @pxref{Variable
22178 Attributes}). Macro expansion occurs on the arguments to this pragma
22179 when compiling C and Objective-C@. It does not currently occur when
22180 compiling C++, but this is a bug which may be fixed in a future
22183 @item fini (@var{function} [, @var{function}]...)
22184 @cindex pragma, fini
22186 This pragma causes each listed @var{function} to be called after
22187 main, or during shared module unloading, by adding a call to the
22188 @code{.fini} section.
22190 @item init (@var{function} [, @var{function}]...)
22191 @cindex pragma, init
22193 This pragma causes each listed @var{function} to be called during
22194 initialization (before @code{main}) or during shared module loading, by
22195 adding a call to the @code{.init} section.
22199 @node Symbol-Renaming Pragmas
22200 @subsection Symbol-Renaming Pragmas
22202 GCC supports a @code{#pragma} directive that changes the name used in
22203 assembly for a given declaration. While this pragma is supported on all
22204 platforms, it is intended primarily to provide compatibility with the
22205 Solaris system headers. This effect can also be achieved using the asm
22206 labels extension (@pxref{Asm Labels}).
22209 @item redefine_extname @var{oldname} @var{newname}
22210 @cindex pragma, redefine_extname
22212 This pragma gives the C function @var{oldname} the assembly symbol
22213 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
22214 is defined if this pragma is available (currently on all platforms).
22217 This pragma and the asm labels extension interact in a complicated
22218 manner. Here are some corner cases you may want to be aware of:
22221 @item This pragma silently applies only to declarations with external
22222 linkage. Asm labels do not have this restriction.
22224 @item In C++, this pragma silently applies only to declarations with
22225 ``C'' linkage. Again, asm labels do not have this restriction.
22227 @item If either of the ways of changing the assembly name of a
22228 declaration are applied to a declaration whose assembly name has
22229 already been determined (either by a previous use of one of these
22230 features, or because the compiler needed the assembly name in order to
22231 generate code), and the new name is different, a warning issues and
22232 the name does not change.
22234 @item The @var{oldname} used by @code{#pragma redefine_extname} is
22235 always the C-language name.
22238 @node Structure-Layout Pragmas
22239 @subsection Structure-Layout Pragmas
22241 For compatibility with Microsoft Windows compilers, GCC supports a
22242 set of @code{#pragma} directives that change the maximum alignment of
22243 members of structures (other than zero-width bit-fields), unions, and
22244 classes subsequently defined. The @var{n} value below always is required
22245 to be a small power of two and specifies the new alignment in bytes.
22248 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
22249 @item @code{#pragma pack()} sets the alignment to the one that was in
22250 effect when compilation started (see also command-line option
22251 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
22252 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
22253 setting on an internal stack and then optionally sets the new alignment.
22254 @item @code{#pragma pack(pop)} restores the alignment setting to the one
22255 saved at the top of the internal stack (and removes that stack entry).
22256 Note that @code{#pragma pack([@var{n}])} does not influence this internal
22257 stack; thus it is possible to have @code{#pragma pack(push)} followed by
22258 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
22259 @code{#pragma pack(pop)}.
22262 Some targets, e.g.@: x86 and PowerPC, support the @code{#pragma ms_struct}
22263 directive which lays out structures and unions subsequently defined as the
22264 documented @code{__attribute__ ((ms_struct))}.
22267 @item @code{#pragma ms_struct on} turns on the Microsoft layout.
22268 @item @code{#pragma ms_struct off} turns off the Microsoft layout.
22269 @item @code{#pragma ms_struct reset} goes back to the default layout.
22272 Most targets also support the @code{#pragma scalar_storage_order} directive
22273 which lays out structures and unions subsequently defined as the documented
22274 @code{__attribute__ ((scalar_storage_order))}.
22277 @item @code{#pragma scalar_storage_order big-endian} sets the storage order
22278 of the scalar fields to big-endian.
22279 @item @code{#pragma scalar_storage_order little-endian} sets the storage order
22280 of the scalar fields to little-endian.
22281 @item @code{#pragma scalar_storage_order default} goes back to the endianness
22282 that was in effect when compilation started (see also command-line option
22283 @option{-fsso-struct=@var{endianness}} @pxref{C Dialect Options}).
22287 @subsection Weak Pragmas
22289 For compatibility with SVR4, GCC supports a set of @code{#pragma}
22290 directives for declaring symbols to be weak, and defining weak
22294 @item #pragma weak @var{symbol}
22295 @cindex pragma, weak
22296 This pragma declares @var{symbol} to be weak, as if the declaration
22297 had the attribute of the same name. The pragma may appear before
22298 or after the declaration of @var{symbol}. It is not an error for
22299 @var{symbol} to never be defined at all.
22301 @item #pragma weak @var{symbol1} = @var{symbol2}
22302 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
22303 It is an error if @var{symbol2} is not defined in the current
22307 @node Diagnostic Pragmas
22308 @subsection Diagnostic Pragmas
22310 GCC allows the user to selectively enable or disable certain types of
22311 diagnostics, and change the kind of the diagnostic. For example, a
22312 project's policy might require that all sources compile with
22313 @option{-Werror} but certain files might have exceptions allowing
22314 specific types of warnings. Or, a project might selectively enable
22315 diagnostics and treat them as errors depending on which preprocessor
22316 macros are defined.
22319 @item #pragma GCC diagnostic @var{kind} @var{option}
22320 @cindex pragma, diagnostic
22322 Modifies the disposition of a diagnostic. Note that not all
22323 diagnostics are modifiable; at the moment only warnings (normally
22324 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
22325 Use @option{-fdiagnostics-show-option} to determine which diagnostics
22326 are controllable and which option controls them.
22328 @var{kind} is @samp{error} to treat this diagnostic as an error,
22329 @samp{warning} to treat it like a warning (even if @option{-Werror} is
22330 in effect), or @samp{ignored} if the diagnostic is to be ignored.
22331 @var{option} is a double quoted string that matches the command-line
22335 #pragma GCC diagnostic warning "-Wformat"
22336 #pragma GCC diagnostic error "-Wformat"
22337 #pragma GCC diagnostic ignored "-Wformat"
22340 Note that these pragmas override any command-line options. GCC keeps
22341 track of the location of each pragma, and issues diagnostics according
22342 to the state as of that point in the source file. Thus, pragmas occurring
22343 after a line do not affect diagnostics caused by that line.
22345 @item #pragma GCC diagnostic push
22346 @itemx #pragma GCC diagnostic pop
22348 Causes GCC to remember the state of the diagnostics as of each
22349 @code{push}, and restore to that point at each @code{pop}. If a
22350 @code{pop} has no matching @code{push}, the command-line options are
22354 #pragma GCC diagnostic error "-Wuninitialized"
22355 foo(a); /* error is given for this one */
22356 #pragma GCC diagnostic push
22357 #pragma GCC diagnostic ignored "-Wuninitialized"
22358 foo(b); /* no diagnostic for this one */
22359 #pragma GCC diagnostic pop
22360 foo(c); /* error is given for this one */
22361 #pragma GCC diagnostic pop
22362 foo(d); /* depends on command-line options */
22367 GCC also offers a simple mechanism for printing messages during
22371 @item #pragma message @var{string}
22372 @cindex pragma, diagnostic
22374 Prints @var{string} as a compiler message on compilation. The message
22375 is informational only, and is neither a compilation warning nor an error.
22378 #pragma message "Compiling " __FILE__ "..."
22381 @var{string} may be parenthesized, and is printed with location
22382 information. For example,
22385 #define DO_PRAGMA(x) _Pragma (#x)
22386 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
22388 TODO(Remember to fix this)
22392 prints @samp{/tmp/file.c:4: note: #pragma message:
22393 TODO - Remember to fix this}.
22397 @node Visibility Pragmas
22398 @subsection Visibility Pragmas
22401 @item #pragma GCC visibility push(@var{visibility})
22402 @itemx #pragma GCC visibility pop
22403 @cindex pragma, visibility
22405 This pragma allows the user to set the visibility for multiple
22406 declarations without having to give each a visibility attribute
22407 (@pxref{Function Attributes}).
22409 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
22410 declarations. Class members and template specializations are not
22411 affected; if you want to override the visibility for a particular
22412 member or instantiation, you must use an attribute.
22417 @node Push/Pop Macro Pragmas
22418 @subsection Push/Pop Macro Pragmas
22420 For compatibility with Microsoft Windows compilers, GCC supports
22421 @samp{#pragma push_macro(@var{"macro_name"})}
22422 and @samp{#pragma pop_macro(@var{"macro_name"})}.
22425 @item #pragma push_macro(@var{"macro_name"})
22426 @cindex pragma, push_macro
22427 This pragma saves the value of the macro named as @var{macro_name} to
22428 the top of the stack for this macro.
22430 @item #pragma pop_macro(@var{"macro_name"})
22431 @cindex pragma, pop_macro
22432 This pragma sets the value of the macro named as @var{macro_name} to
22433 the value on top of the stack for this macro. If the stack for
22434 @var{macro_name} is empty, the value of the macro remains unchanged.
22441 #pragma push_macro("X")
22444 #pragma pop_macro("X")
22449 In this example, the definition of X as 1 is saved by @code{#pragma
22450 push_macro} and restored by @code{#pragma pop_macro}.
22452 @node Function Specific Option Pragmas
22453 @subsection Function Specific Option Pragmas
22456 @item #pragma GCC target (@var{"string"}...)
22457 @cindex pragma GCC target
22459 This pragma allows you to set target specific options for functions
22460 defined later in the source file. One or more strings can be
22461 specified. Each function that is defined after this point is as
22462 if @code{attribute((target("STRING")))} was specified for that
22463 function. The parenthesis around the options is optional.
22464 @xref{Function Attributes}, for more information about the
22465 @code{target} attribute and the attribute syntax.
22467 The @code{#pragma GCC target} pragma is presently implemented for
22468 x86, ARM, AArch64, PowerPC, S/390, and Nios II targets only.
22470 @item #pragma GCC optimize (@var{"string"}...)
22471 @cindex pragma GCC optimize
22473 This pragma allows you to set global optimization options for functions
22474 defined later in the source file. One or more strings can be
22475 specified. Each function that is defined after this point is as
22476 if @code{attribute((optimize("STRING")))} was specified for that
22477 function. The parenthesis around the options is optional.
22478 @xref{Function Attributes}, for more information about the
22479 @code{optimize} attribute and the attribute syntax.
22481 @item #pragma GCC push_options
22482 @itemx #pragma GCC pop_options
22483 @cindex pragma GCC push_options
22484 @cindex pragma GCC pop_options
22486 These pragmas maintain a stack of the current target and optimization
22487 options. It is intended for include files where you temporarily want
22488 to switch to using a different @samp{#pragma GCC target} or
22489 @samp{#pragma GCC optimize} and then to pop back to the previous
22492 @item #pragma GCC reset_options
22493 @cindex pragma GCC reset_options
22495 This pragma clears the current @code{#pragma GCC target} and
22496 @code{#pragma GCC optimize} to use the default switches as specified
22497 on the command line.
22501 @node Loop-Specific Pragmas
22502 @subsection Loop-Specific Pragmas
22505 @item #pragma GCC ivdep
22506 @cindex pragma GCC ivdep
22508 With this pragma, the programmer asserts that there are no loop-carried
22509 dependencies which would prevent consecutive iterations of
22510 the following loop from executing concurrently with SIMD
22511 (single instruction multiple data) instructions.
22513 For example, the compiler can only unconditionally vectorize the following
22514 loop with the pragma:
22517 void foo (int n, int *a, int *b, int *c)
22521 for (i = 0; i < n; ++i)
22522 a[i] = b[i] + c[i];
22527 In this example, using the @code{restrict} qualifier had the same
22528 effect. In the following example, that would not be possible. Assume
22529 @math{k < -m} or @math{k >= m}. Only with the pragma, the compiler knows
22530 that it can unconditionally vectorize the following loop:
22533 void ignore_vec_dep (int *a, int k, int c, int m)
22536 for (int i = 0; i < m; i++)
22537 a[i] = a[i + k] * c;
22541 @item #pragma GCC unroll @var{n}
22542 @cindex pragma GCC unroll @var{n}
22544 You can use this pragma to control how many times a loop should be unrolled.
22545 It must be placed immediately before a @code{for}, @code{while} or @code{do}
22546 loop or a @code{#pragma GCC ivdep}, and applies only to the loop that follows.
22547 @var{n} is an integer constant expression specifying the unrolling factor.
22548 The values of @math{0} and @math{1} block any unrolling of the loop.
22552 @node Unnamed Fields
22553 @section Unnamed Structure and Union Fields
22554 @cindex @code{struct}
22555 @cindex @code{union}
22557 As permitted by ISO C11 and for compatibility with other compilers,
22558 GCC allows you to define
22559 a structure or union that contains, as fields, structures and unions
22560 without names. For example:
22574 In this example, you are able to access members of the unnamed
22575 union with code like @samp{foo.b}. Note that only unnamed structs and
22576 unions are allowed, you may not have, for example, an unnamed
22579 You must never create such structures that cause ambiguous field definitions.
22580 For example, in this structure:
22592 it is ambiguous which @code{a} is being referred to with @samp{foo.a}.
22593 The compiler gives errors for such constructs.
22595 @opindex fms-extensions
22596 Unless @option{-fms-extensions} is used, the unnamed field must be a
22597 structure or union definition without a tag (for example, @samp{struct
22598 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
22599 also be a definition with a tag such as @samp{struct foo @{ int a;
22600 @};}, a reference to a previously defined structure or union such as
22601 @samp{struct foo;}, or a reference to a @code{typedef} name for a
22602 previously defined structure or union type.
22604 @opindex fplan9-extensions
22605 The option @option{-fplan9-extensions} enables
22606 @option{-fms-extensions} as well as two other extensions. First, a
22607 pointer to a structure is automatically converted to a pointer to an
22608 anonymous field for assignments and function calls. For example:
22611 struct s1 @{ int a; @};
22612 struct s2 @{ struct s1; @};
22613 extern void f1 (struct s1 *);
22614 void f2 (struct s2 *p) @{ f1 (p); @}
22618 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
22619 converted into a pointer to the anonymous field.
22621 Second, when the type of an anonymous field is a @code{typedef} for a
22622 @code{struct} or @code{union}, code may refer to the field using the
22623 name of the @code{typedef}.
22626 typedef struct @{ int a; @} s1;
22627 struct s2 @{ s1; @};
22628 s1 f1 (struct s2 *p) @{ return p->s1; @}
22631 These usages are only permitted when they are not ambiguous.
22634 @section Thread-Local Storage
22635 @cindex Thread-Local Storage
22636 @cindex @acronym{TLS}
22637 @cindex @code{__thread}
22639 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
22640 are allocated such that there is one instance of the variable per extant
22641 thread. The runtime model GCC uses to implement this originates
22642 in the IA-64 processor-specific ABI, but has since been migrated
22643 to other processors as well. It requires significant support from
22644 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
22645 system libraries (@file{libc.so} and @file{libpthread.so}), so it
22646 is not available everywhere.
22648 At the user level, the extension is visible with a new storage
22649 class keyword: @code{__thread}. For example:
22653 extern __thread struct state s;
22654 static __thread char *p;
22657 The @code{__thread} specifier may be used alone, with the @code{extern}
22658 or @code{static} specifiers, but with no other storage class specifier.
22659 When used with @code{extern} or @code{static}, @code{__thread} must appear
22660 immediately after the other storage class specifier.
22662 The @code{__thread} specifier may be applied to any global, file-scoped
22663 static, function-scoped static, or static data member of a class. It may
22664 not be applied to block-scoped automatic or non-static data member.
22666 When the address-of operator is applied to a thread-local variable, it is
22667 evaluated at run time and returns the address of the current thread's
22668 instance of that variable. An address so obtained may be used by any
22669 thread. When a thread terminates, any pointers to thread-local variables
22670 in that thread become invalid.
22672 No static initialization may refer to the address of a thread-local variable.
22674 In C++, if an initializer is present for a thread-local variable, it must
22675 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
22678 See @uref{https://www.akkadia.org/drepper/tls.pdf,
22679 ELF Handling For Thread-Local Storage} for a detailed explanation of
22680 the four thread-local storage addressing models, and how the runtime
22681 is expected to function.
22684 * C99 Thread-Local Edits::
22685 * C++98 Thread-Local Edits::
22688 @node C99 Thread-Local Edits
22689 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
22691 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
22692 that document the exact semantics of the language extension.
22696 @cite{5.1.2 Execution environments}
22698 Add new text after paragraph 1
22701 Within either execution environment, a @dfn{thread} is a flow of
22702 control within a program. It is implementation defined whether
22703 or not there may be more than one thread associated with a program.
22704 It is implementation defined how threads beyond the first are
22705 created, the name and type of the function called at thread
22706 startup, and how threads may be terminated. However, objects
22707 with thread storage duration shall be initialized before thread
22712 @cite{6.2.4 Storage durations of objects}
22714 Add new text before paragraph 3
22717 An object whose identifier is declared with the storage-class
22718 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
22719 Its lifetime is the entire execution of the thread, and its
22720 stored value is initialized only once, prior to thread startup.
22724 @cite{6.4.1 Keywords}
22726 Add @code{__thread}.
22729 @cite{6.7.1 Storage-class specifiers}
22731 Add @code{__thread} to the list of storage class specifiers in
22734 Change paragraph 2 to
22737 With the exception of @code{__thread}, at most one storage-class
22738 specifier may be given [@dots{}]. The @code{__thread} specifier may
22739 be used alone, or immediately following @code{extern} or
22743 Add new text after paragraph 6
22746 The declaration of an identifier for a variable that has
22747 block scope that specifies @code{__thread} shall also
22748 specify either @code{extern} or @code{static}.
22750 The @code{__thread} specifier shall be used only with
22755 @node C++98 Thread-Local Edits
22756 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
22758 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
22759 that document the exact semantics of the language extension.
22763 @b{[intro.execution]}
22765 New text after paragraph 4
22768 A @dfn{thread} is a flow of control within the abstract machine.
22769 It is implementation defined whether or not there may be more than
22773 New text after paragraph 7
22776 It is unspecified whether additional action must be taken to
22777 ensure when and whether side effects are visible to other threads.
22783 Add @code{__thread}.
22786 @b{[basic.start.main]}
22788 Add after paragraph 5
22791 The thread that begins execution at the @code{main} function is called
22792 the @dfn{main thread}. It is implementation defined how functions
22793 beginning threads other than the main thread are designated or typed.
22794 A function so designated, as well as the @code{main} function, is called
22795 a @dfn{thread startup function}. It is implementation defined what
22796 happens if a thread startup function returns. It is implementation
22797 defined what happens to other threads when any thread calls @code{exit}.
22801 @b{[basic.start.init]}
22803 Add after paragraph 4
22806 The storage for an object of thread storage duration shall be
22807 statically initialized before the first statement of the thread startup
22808 function. An object of thread storage duration shall not require
22809 dynamic initialization.
22813 @b{[basic.start.term]}
22815 Add after paragraph 3
22818 The type of an object with thread storage duration shall not have a
22819 non-trivial destructor, nor shall it be an array type whose elements
22820 (directly or indirectly) have non-trivial destructors.
22826 Add ``thread storage duration'' to the list in paragraph 1.
22831 Thread, static, and automatic storage durations are associated with
22832 objects introduced by declarations [@dots{}].
22835 Add @code{__thread} to the list of specifiers in paragraph 3.
22838 @b{[basic.stc.thread]}
22840 New section before @b{[basic.stc.static]}
22843 The keyword @code{__thread} applied to a non-local object gives the
22844 object thread storage duration.
22846 A local variable or class data member declared both @code{static}
22847 and @code{__thread} gives the variable or member thread storage
22852 @b{[basic.stc.static]}
22857 All objects that have neither thread storage duration, dynamic
22858 storage duration nor are local [@dots{}].
22864 Add @code{__thread} to the list in paragraph 1.
22869 With the exception of @code{__thread}, at most one
22870 @var{storage-class-specifier} shall appear in a given
22871 @var{decl-specifier-seq}. The @code{__thread} specifier may
22872 be used alone, or immediately following the @code{extern} or
22873 @code{static} specifiers. [@dots{}]
22876 Add after paragraph 5
22879 The @code{__thread} specifier can be applied only to the names of objects
22880 and to anonymous unions.
22886 Add after paragraph 6
22889 Non-@code{static} members shall not be @code{__thread}.
22893 @node Binary constants
22894 @section Binary Constants using the @samp{0b} Prefix
22895 @cindex Binary constants using the @samp{0b} prefix
22897 Integer constants can be written as binary constants, consisting of a
22898 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
22899 @samp{0B}. This is particularly useful in environments that operate a
22900 lot on the bit level (like microcontrollers).
22902 The following statements are identical:
22911 The type of these constants follows the same rules as for octal or
22912 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
22915 @node C++ Extensions
22916 @chapter Extensions to the C++ Language
22917 @cindex extensions, C++ language
22918 @cindex C++ language extensions
22920 The GNU compiler provides these extensions to the C++ language (and you
22921 can also use most of the C language extensions in your C++ programs). If you
22922 want to write code that checks whether these features are available, you can
22923 test for the GNU compiler the same way as for C programs: check for a
22924 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
22925 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
22926 Predefined Macros,cpp,The GNU C Preprocessor}).
22929 * C++ Volatiles:: What constitutes an access to a volatile object.
22930 * Restricted Pointers:: C99 restricted pointers and references.
22931 * Vague Linkage:: Where G++ puts inlines, vtables and such.
22932 * C++ Interface:: You can use a single C++ header file for both
22933 declarations and definitions.
22934 * Template Instantiation:: Methods for ensuring that exactly one copy of
22935 each needed template instantiation is emitted.
22936 * Bound member functions:: You can extract a function pointer to the
22937 method denoted by a @samp{->*} or @samp{.*} expression.
22938 * C++ Attributes:: Variable, function, and type attributes for C++ only.
22939 * Function Multiversioning:: Declaring multiple function versions.
22940 * Type Traits:: Compiler support for type traits.
22941 * C++ Concepts:: Improved support for generic programming.
22942 * Deprecated Features:: Things will disappear from G++.
22943 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
22946 @node C++ Volatiles
22947 @section When is a Volatile C++ Object Accessed?
22948 @cindex accessing volatiles
22949 @cindex volatile read
22950 @cindex volatile write
22951 @cindex volatile access
22953 The C++ standard differs from the C standard in its treatment of
22954 volatile objects. It fails to specify what constitutes a volatile
22955 access, except to say that C++ should behave in a similar manner to C
22956 with respect to volatiles, where possible. However, the different
22957 lvalueness of expressions between C and C++ complicate the behavior.
22958 G++ behaves the same as GCC for volatile access, @xref{C
22959 Extensions,,Volatiles}, for a description of GCC's behavior.
22961 The C and C++ language specifications differ when an object is
22962 accessed in a void context:
22965 volatile int *src = @var{somevalue};
22969 The C++ standard specifies that such expressions do not undergo lvalue
22970 to rvalue conversion, and that the type of the dereferenced object may
22971 be incomplete. The C++ standard does not specify explicitly that it
22972 is lvalue to rvalue conversion that is responsible for causing an
22973 access. There is reason to believe that it is, because otherwise
22974 certain simple expressions become undefined. However, because it
22975 would surprise most programmers, G++ treats dereferencing a pointer to
22976 volatile object of complete type as GCC would do for an equivalent
22977 type in C@. When the object has incomplete type, G++ issues a
22978 warning; if you wish to force an error, you must force a conversion to
22979 rvalue with, for instance, a static cast.
22981 When using a reference to volatile, G++ does not treat equivalent
22982 expressions as accesses to volatiles, but instead issues a warning that
22983 no volatile is accessed. The rationale for this is that otherwise it
22984 becomes difficult to determine where volatile access occur, and not
22985 possible to ignore the return value from functions returning volatile
22986 references. Again, if you wish to force a read, cast the reference to
22989 G++ implements the same behavior as GCC does when assigning to a
22990 volatile object---there is no reread of the assigned-to object, the
22991 assigned rvalue is reused. Note that in C++ assignment expressions
22992 are lvalues, and if used as an lvalue, the volatile object is
22993 referred to. For instance, @var{vref} refers to @var{vobj}, as
22994 expected, in the following example:
22998 volatile int &vref = vobj = @var{something};
23001 @node Restricted Pointers
23002 @section Restricting Pointer Aliasing
23003 @cindex restricted pointers
23004 @cindex restricted references
23005 @cindex restricted this pointer
23007 As with the C front end, G++ understands the C99 feature of restricted pointers,
23008 specified with the @code{__restrict__}, or @code{__restrict} type
23009 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
23010 language flag, @code{restrict} is not a keyword in C++.
23012 In addition to allowing restricted pointers, you can specify restricted
23013 references, which indicate that the reference is not aliased in the local
23017 void fn (int *__restrict__ rptr, int &__restrict__ rref)
23024 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
23025 @var{rref} refers to a (different) unaliased integer.
23027 You may also specify whether a member function's @var{this} pointer is
23028 unaliased by using @code{__restrict__} as a member function qualifier.
23031 void T::fn () __restrict__
23038 Within the body of @code{T::fn}, @var{this} has the effective
23039 definition @code{T *__restrict__ const this}. Notice that the
23040 interpretation of a @code{__restrict__} member function qualifier is
23041 different to that of @code{const} or @code{volatile} qualifier, in that it
23042 is applied to the pointer rather than the object. This is consistent with
23043 other compilers that implement restricted pointers.
23045 As with all outermost parameter qualifiers, @code{__restrict__} is
23046 ignored in function definition matching. This means you only need to
23047 specify @code{__restrict__} in a function definition, rather than
23048 in a function prototype as well.
23050 @node Vague Linkage
23051 @section Vague Linkage
23052 @cindex vague linkage
23054 There are several constructs in C++ that require space in the object
23055 file but are not clearly tied to a single translation unit. We say that
23056 these constructs have ``vague linkage''. Typically such constructs are
23057 emitted wherever they are needed, though sometimes we can be more
23061 @item Inline Functions
23062 Inline functions are typically defined in a header file which can be
23063 included in many different compilations. Hopefully they can usually be
23064 inlined, but sometimes an out-of-line copy is necessary, if the address
23065 of the function is taken or if inlining fails. In general, we emit an
23066 out-of-line copy in all translation units where one is needed. As an
23067 exception, we only emit inline virtual functions with the vtable, since
23068 it always requires a copy.
23070 Local static variables and string constants used in an inline function
23071 are also considered to have vague linkage, since they must be shared
23072 between all inlined and out-of-line instances of the function.
23076 C++ virtual functions are implemented in most compilers using a lookup
23077 table, known as a vtable. The vtable contains pointers to the virtual
23078 functions provided by a class, and each object of the class contains a
23079 pointer to its vtable (or vtables, in some multiple-inheritance
23080 situations). If the class declares any non-inline, non-pure virtual
23081 functions, the first one is chosen as the ``key method'' for the class,
23082 and the vtable is only emitted in the translation unit where the key
23085 @emph{Note:} If the chosen key method is later defined as inline, the
23086 vtable is still emitted in every translation unit that defines it.
23087 Make sure that any inline virtuals are declared inline in the class
23088 body, even if they are not defined there.
23090 @item @code{type_info} objects
23091 @cindex @code{type_info}
23093 C++ requires information about types to be written out in order to
23094 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
23095 For polymorphic classes (classes with virtual functions), the @samp{type_info}
23096 object is written out along with the vtable so that @samp{dynamic_cast}
23097 can determine the dynamic type of a class object at run time. For all
23098 other types, we write out the @samp{type_info} object when it is used: when
23099 applying @samp{typeid} to an expression, throwing an object, or
23100 referring to a type in a catch clause or exception specification.
23102 @item Template Instantiations
23103 Most everything in this section also applies to template instantiations,
23104 but there are other options as well.
23105 @xref{Template Instantiation,,Where's the Template?}.
23109 When used with GNU ld version 2.8 or later on an ELF system such as
23110 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
23111 these constructs will be discarded at link time. This is known as
23114 On targets that don't support COMDAT, but do support weak symbols, GCC
23115 uses them. This way one copy overrides all the others, but
23116 the unused copies still take up space in the executable.
23118 For targets that do not support either COMDAT or weak symbols,
23119 most entities with vague linkage are emitted as local symbols to
23120 avoid duplicate definition errors from the linker. This does not happen
23121 for local statics in inlines, however, as having multiple copies
23122 almost certainly breaks things.
23124 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
23125 another way to control placement of these constructs.
23127 @node C++ Interface
23128 @section C++ Interface and Implementation Pragmas
23130 @cindex interface and implementation headers, C++
23131 @cindex C++ interface and implementation headers
23132 @cindex pragmas, interface and implementation
23134 @code{#pragma interface} and @code{#pragma implementation} provide the
23135 user with a way of explicitly directing the compiler to emit entities
23136 with vague linkage (and debugging information) in a particular
23139 @emph{Note:} These @code{#pragma}s have been superceded as of GCC 2.7.2
23140 by COMDAT support and the ``key method'' heuristic
23141 mentioned in @ref{Vague Linkage}. Using them can actually cause your
23142 program to grow due to unnecessary out-of-line copies of inline
23146 @item #pragma interface
23147 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
23148 @kindex #pragma interface
23149 Use this directive in @emph{header files} that define object classes, to save
23150 space in most of the object files that use those classes. Normally,
23151 local copies of certain information (backup copies of inline member
23152 functions, debugging information, and the internal tables that implement
23153 virtual functions) must be kept in each object file that includes class
23154 definitions. You can use this pragma to avoid such duplication. When a
23155 header file containing @samp{#pragma interface} is included in a
23156 compilation, this auxiliary information is not generated (unless
23157 the main input source file itself uses @samp{#pragma implementation}).
23158 Instead, the object files contain references to be resolved at link
23161 The second form of this directive is useful for the case where you have
23162 multiple headers with the same name in different directories. If you
23163 use this form, you must specify the same string to @samp{#pragma
23166 @item #pragma implementation
23167 @itemx #pragma implementation "@var{objects}.h"
23168 @kindex #pragma implementation
23169 Use this pragma in a @emph{main input file}, when you want full output from
23170 included header files to be generated (and made globally visible). The
23171 included header file, in turn, should use @samp{#pragma interface}.
23172 Backup copies of inline member functions, debugging information, and the
23173 internal tables used to implement virtual functions are all generated in
23174 implementation files.
23176 @cindex implied @code{#pragma implementation}
23177 @cindex @code{#pragma implementation}, implied
23178 @cindex naming convention, implementation headers
23179 If you use @samp{#pragma implementation} with no argument, it applies to
23180 an include file with the same basename@footnote{A file's @dfn{basename}
23181 is the name stripped of all leading path information and of trailing
23182 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
23183 file. For example, in @file{allclass.cc}, giving just
23184 @samp{#pragma implementation}
23185 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
23187 Use the string argument if you want a single implementation file to
23188 include code from multiple header files. (You must also use
23189 @samp{#include} to include the header file; @samp{#pragma
23190 implementation} only specifies how to use the file---it doesn't actually
23193 There is no way to split up the contents of a single header file into
23194 multiple implementation files.
23197 @cindex inlining and C++ pragmas
23198 @cindex C++ pragmas, effect on inlining
23199 @cindex pragmas in C++, effect on inlining
23200 @samp{#pragma implementation} and @samp{#pragma interface} also have an
23201 effect on function inlining.
23203 If you define a class in a header file marked with @samp{#pragma
23204 interface}, the effect on an inline function defined in that class is
23205 similar to an explicit @code{extern} declaration---the compiler emits
23206 no code at all to define an independent version of the function. Its
23207 definition is used only for inlining with its callers.
23209 @opindex fno-implement-inlines
23210 Conversely, when you include the same header file in a main source file
23211 that declares it as @samp{#pragma implementation}, the compiler emits
23212 code for the function itself; this defines a version of the function
23213 that can be found via pointers (or by callers compiled without
23214 inlining). If all calls to the function can be inlined, you can avoid
23215 emitting the function by compiling with @option{-fno-implement-inlines}.
23216 If any calls are not inlined, you will get linker errors.
23218 @node Template Instantiation
23219 @section Where's the Template?
23220 @cindex template instantiation
23222 C++ templates were the first language feature to require more
23223 intelligence from the environment than was traditionally found on a UNIX
23224 system. Somehow the compiler and linker have to make sure that each
23225 template instance occurs exactly once in the executable if it is needed,
23226 and not at all otherwise. There are two basic approaches to this
23227 problem, which are referred to as the Borland model and the Cfront model.
23230 @item Borland model
23231 Borland C++ solved the template instantiation problem by adding the code
23232 equivalent of common blocks to their linker; the compiler emits template
23233 instances in each translation unit that uses them, and the linker
23234 collapses them together. The advantage of this model is that the linker
23235 only has to consider the object files themselves; there is no external
23236 complexity to worry about. The disadvantage is that compilation time
23237 is increased because the template code is being compiled repeatedly.
23238 Code written for this model tends to include definitions of all
23239 templates in the header file, since they must be seen to be
23243 The AT&T C++ translator, Cfront, solved the template instantiation
23244 problem by creating the notion of a template repository, an
23245 automatically maintained place where template instances are stored. A
23246 more modern version of the repository works as follows: As individual
23247 object files are built, the compiler places any template definitions and
23248 instantiations encountered in the repository. At link time, the link
23249 wrapper adds in the objects in the repository and compiles any needed
23250 instances that were not previously emitted. The advantages of this
23251 model are more optimal compilation speed and the ability to use the
23252 system linker; to implement the Borland model a compiler vendor also
23253 needs to replace the linker. The disadvantages are vastly increased
23254 complexity, and thus potential for error; for some code this can be
23255 just as transparent, but in practice it can been very difficult to build
23256 multiple programs in one directory and one program in multiple
23257 directories. Code written for this model tends to separate definitions
23258 of non-inline member templates into a separate file, which should be
23259 compiled separately.
23262 G++ implements the Borland model on targets where the linker supports it,
23263 including ELF targets (such as GNU/Linux), Mac OS X and Microsoft Windows.
23264 Otherwise G++ implements neither automatic model.
23266 You have the following options for dealing with template instantiations:
23270 Do nothing. Code written for the Borland model works fine, but
23271 each translation unit contains instances of each of the templates it
23272 uses. The duplicate instances will be discarded by the linker, but in
23273 a large program, this can lead to an unacceptable amount of code
23274 duplication in object files or shared libraries.
23276 Duplicate instances of a template can be avoided by defining an explicit
23277 instantiation in one object file, and preventing the compiler from doing
23278 implicit instantiations in any other object files by using an explicit
23279 instantiation declaration, using the @code{extern template} syntax:
23282 extern template int max (int, int);
23285 This syntax is defined in the C++ 2011 standard, but has been supported by
23286 G++ and other compilers since well before 2011.
23288 Explicit instantiations can be used for the largest or most frequently
23289 duplicated instances, without having to know exactly which other instances
23290 are used in the rest of the program. You can scatter the explicit
23291 instantiations throughout your program, perhaps putting them in the
23292 translation units where the instances are used or the translation units
23293 that define the templates themselves; you can put all of the explicit
23294 instantiations you need into one big file; or you can create small files
23301 template class Foo<int>;
23302 template ostream& operator <<
23303 (ostream&, const Foo<int>&);
23307 for each of the instances you need, and create a template instantiation
23308 library from those.
23310 This is the simplest option, but also offers flexibility and
23311 fine-grained control when necessary. It is also the most portable
23312 alternative and programs using this approach will work with most modern
23317 Compile your template-using code with @option{-frepo}. The compiler
23318 generates files with the extension @samp{.rpo} listing all of the
23319 template instantiations used in the corresponding object files that
23320 could be instantiated there; the link wrapper, @samp{collect2},
23321 then updates the @samp{.rpo} files to tell the compiler where to place
23322 those instantiations and rebuild any affected object files. The
23323 link-time overhead is negligible after the first pass, as the compiler
23324 continues to place the instantiations in the same files.
23326 This can be a suitable option for application code written for the Borland
23327 model, as it usually just works. Code written for the Cfront model
23328 needs to be modified so that the template definitions are available at
23329 one or more points of instantiation; usually this is as simple as adding
23330 @code{#include <tmethods.cc>} to the end of each template header.
23332 For library code, if you want the library to provide all of the template
23333 instantiations it needs, just try to link all of its object files
23334 together; the link will fail, but cause the instantiations to be
23335 generated as a side effect. Be warned, however, that this may cause
23336 conflicts if multiple libraries try to provide the same instantiations.
23337 For greater control, use explicit instantiation as described in the next
23341 @opindex fno-implicit-templates
23342 Compile your code with @option{-fno-implicit-templates} to disable the
23343 implicit generation of template instances, and explicitly instantiate
23344 all the ones you use. This approach requires more knowledge of exactly
23345 which instances you need than do the others, but it's less
23346 mysterious and allows greater control if you want to ensure that only
23347 the intended instances are used.
23349 If you are using Cfront-model code, you can probably get away with not
23350 using @option{-fno-implicit-templates} when compiling files that don't
23351 @samp{#include} the member template definitions.
23353 If you use one big file to do the instantiations, you may want to
23354 compile it without @option{-fno-implicit-templates} so you get all of the
23355 instances required by your explicit instantiations (but not by any
23356 other files) without having to specify them as well.
23358 In addition to forward declaration of explicit instantiations
23359 (with @code{extern}), G++ has extended the template instantiation
23360 syntax to support instantiation of the compiler support data for a
23361 template class (i.e.@: the vtable) without instantiating any of its
23362 members (with @code{inline}), and instantiation of only the static data
23363 members of a template class, without the support data or member
23364 functions (with @code{static}):
23367 inline template class Foo<int>;
23368 static template class Foo<int>;
23372 @node Bound member functions
23373 @section Extracting the Function Pointer from a Bound Pointer to Member Function
23375 @cindex pointer to member function
23376 @cindex bound pointer to member function
23378 In C++, pointer to member functions (PMFs) are implemented using a wide
23379 pointer of sorts to handle all the possible call mechanisms; the PMF
23380 needs to store information about how to adjust the @samp{this} pointer,
23381 and if the function pointed to is virtual, where to find the vtable, and
23382 where in the vtable to look for the member function. If you are using
23383 PMFs in an inner loop, you should really reconsider that decision. If
23384 that is not an option, you can extract the pointer to the function that
23385 would be called for a given object/PMF pair and call it directly inside
23386 the inner loop, to save a bit of time.
23388 Note that you still pay the penalty for the call through a
23389 function pointer; on most modern architectures, such a call defeats the
23390 branch prediction features of the CPU@. This is also true of normal
23391 virtual function calls.
23393 The syntax for this extension is
23397 extern int (A::*fp)();
23398 typedef int (*fptr)(A *);
23400 fptr p = (fptr)(a.*fp);
23403 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
23404 no object is needed to obtain the address of the function. They can be
23405 converted to function pointers directly:
23408 fptr p1 = (fptr)(&A::foo);
23411 @opindex Wno-pmf-conversions
23412 You must specify @option{-Wno-pmf-conversions} to use this extension.
23414 @node C++ Attributes
23415 @section C++-Specific Variable, Function, and Type Attributes
23417 Some attributes only make sense for C++ programs.
23420 @item abi_tag ("@var{tag}", ...)
23421 @cindex @code{abi_tag} function attribute
23422 @cindex @code{abi_tag} variable attribute
23423 @cindex @code{abi_tag} type attribute
23424 The @code{abi_tag} attribute can be applied to a function, variable, or class
23425 declaration. It modifies the mangled name of the entity to
23426 incorporate the tag name, in order to distinguish the function or
23427 class from an earlier version with a different ABI; perhaps the class
23428 has changed size, or the function has a different return type that is
23429 not encoded in the mangled name.
23431 The attribute can also be applied to an inline namespace, but does not
23432 affect the mangled name of the namespace; in this case it is only used
23433 for @option{-Wabi-tag} warnings and automatic tagging of functions and
23434 variables. Tagging inline namespaces is generally preferable to
23435 tagging individual declarations, but the latter is sometimes
23436 necessary, such as when only certain members of a class need to be
23439 The argument can be a list of strings of arbitrary length. The
23440 strings are sorted on output, so the order of the list is
23443 A redeclaration of an entity must not add new ABI tags,
23444 since doing so would change the mangled name.
23446 The ABI tags apply to a name, so all instantiations and
23447 specializations of a template have the same tags. The attribute will
23448 be ignored if applied to an explicit specialization or instantiation.
23450 The @option{-Wabi-tag} flag enables a warning about a class which does
23451 not have all the ABI tags used by its subobjects and virtual functions; for users with code
23452 that needs to coexist with an earlier ABI, using this option can help
23453 to find all affected types that need to be tagged.
23455 When a type involving an ABI tag is used as the type of a variable or
23456 return type of a function where that tag is not already present in the
23457 signature of the function, the tag is automatically applied to the
23458 variable or function. @option{-Wabi-tag} also warns about this
23459 situation; this warning can be avoided by explicitly tagging the
23460 variable or function or moving it into a tagged inline namespace.
23462 @item init_priority (@var{priority})
23463 @cindex @code{init_priority} variable attribute
23465 In Standard C++, objects defined at namespace scope are guaranteed to be
23466 initialized in an order in strict accordance with that of their definitions
23467 @emph{in a given translation unit}. No guarantee is made for initializations
23468 across translation units. However, GNU C++ allows users to control the
23469 order of initialization of objects defined at namespace scope with the
23470 @code{init_priority} attribute by specifying a relative @var{priority},
23471 a constant integral expression currently bounded between 101 and 65535
23472 inclusive. Lower numbers indicate a higher priority.
23474 In the following example, @code{A} would normally be created before
23475 @code{B}, but the @code{init_priority} attribute reverses that order:
23478 Some_Class A __attribute__ ((init_priority (2000)));
23479 Some_Class B __attribute__ ((init_priority (543)));
23483 Note that the particular values of @var{priority} do not matter; only their
23487 @cindex @code{warn_unused} type attribute
23489 For C++ types with non-trivial constructors and/or destructors it is
23490 impossible for the compiler to determine whether a variable of this
23491 type is truly unused if it is not referenced. This type attribute
23492 informs the compiler that variables of this type should be warned
23493 about if they appear to be unused, just like variables of fundamental
23496 This attribute is appropriate for types which just represent a value,
23497 such as @code{std::string}; it is not appropriate for types which
23498 control a resource, such as @code{std::lock_guard}.
23500 This attribute is also accepted in C, but it is unnecessary because C
23501 does not have constructors or destructors.
23505 @node Function Multiversioning
23506 @section Function Multiversioning
23507 @cindex function versions
23509 With the GNU C++ front end, for x86 targets, you may specify multiple
23510 versions of a function, where each function is specialized for a
23511 specific target feature. At runtime, the appropriate version of the
23512 function is automatically executed depending on the characteristics of
23513 the execution platform. Here is an example.
23516 __attribute__ ((target ("default")))
23519 // The default version of foo.
23523 __attribute__ ((target ("sse4.2")))
23526 // foo version for SSE4.2
23530 __attribute__ ((target ("arch=atom")))
23533 // foo version for the Intel ATOM processor
23537 __attribute__ ((target ("arch=amdfam10")))
23540 // foo version for the AMD Family 0x10 processors.
23547 assert ((*p) () == foo ());
23552 In the above example, four versions of function foo are created. The
23553 first version of foo with the target attribute "default" is the default
23554 version. This version gets executed when no other target specific
23555 version qualifies for execution on a particular platform. A new version
23556 of foo is created by using the same function signature but with a
23557 different target string. Function foo is called or a pointer to it is
23558 taken just like a regular function. GCC takes care of doing the
23559 dispatching to call the right version at runtime. Refer to the
23560 @uref{http://gcc.gnu.org/wiki/FunctionMultiVersioning, GCC wiki on
23561 Function Multiversioning} for more details.
23564 @section Type Traits
23566 The C++ front end implements syntactic extensions that allow
23567 compile-time determination of
23568 various characteristics of a type (or of a
23572 @item __has_nothrow_assign (type)
23573 If @code{type} is const qualified or is a reference type then the trait is
23574 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
23575 is true, else if @code{type} is a cv class or union type with copy assignment
23576 operators that are known not to throw an exception then the trait is true,
23577 else it is false. Requires: @code{type} shall be a complete type,
23578 (possibly cv-qualified) @code{void}, or an array of unknown bound.
23580 @item __has_nothrow_copy (type)
23581 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
23582 @code{type} is a cv class or union type with copy constructors that
23583 are known not to throw an exception then the trait is true, else it is false.
23584 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
23585 @code{void}, or an array of unknown bound.
23587 @item __has_nothrow_constructor (type)
23588 If @code{__has_trivial_constructor (type)} is true then the trait is
23589 true, else if @code{type} is a cv class or union type (or array
23590 thereof) with a default constructor that is known not to throw an
23591 exception then the trait is true, else it is false. Requires:
23592 @code{type} shall be a complete type, (possibly cv-qualified)
23593 @code{void}, or an array of unknown bound.
23595 @item __has_trivial_assign (type)
23596 If @code{type} is const qualified or is a reference type then the trait is
23597 false. Otherwise if @code{__is_pod (type)} is true then the trait is
23598 true, else if @code{type} is a cv class or union type with a trivial
23599 copy assignment ([class.copy]) then the trait is true, else it is
23600 false. Requires: @code{type} shall be a complete type, (possibly
23601 cv-qualified) @code{void}, or an array of unknown bound.
23603 @item __has_trivial_copy (type)
23604 If @code{__is_pod (type)} is true or @code{type} is a reference type
23605 then the trait is true, else if @code{type} is a cv class or union type
23606 with a trivial copy constructor ([class.copy]) then the trait
23607 is true, else it is false. Requires: @code{type} shall be a complete
23608 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23610 @item __has_trivial_constructor (type)
23611 If @code{__is_pod (type)} is true then the trait is true, else if
23612 @code{type} is a cv class or union type (or array thereof) with a
23613 trivial default constructor ([class.ctor]) then the trait is true,
23614 else it is false. Requires: @code{type} shall be a complete
23615 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23617 @item __has_trivial_destructor (type)
23618 If @code{__is_pod (type)} is true or @code{type} is a reference type then
23619 the trait is true, else if @code{type} is a cv class or union type (or
23620 array thereof) with a trivial destructor ([class.dtor]) then the trait
23621 is true, else it is false. Requires: @code{type} shall be a complete
23622 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23624 @item __has_virtual_destructor (type)
23625 If @code{type} is a class type with a virtual destructor
23626 ([class.dtor]) then the trait is true, else it is false. Requires:
23627 @code{type} shall be a complete type, (possibly cv-qualified)
23628 @code{void}, or an array of unknown bound.
23630 @item __is_abstract (type)
23631 If @code{type} is an abstract class ([class.abstract]) then the trait
23632 is true, else it is false. Requires: @code{type} shall be a complete
23633 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23635 @item __is_base_of (base_type, derived_type)
23636 If @code{base_type} is a base class of @code{derived_type}
23637 ([class.derived]) then the trait is true, otherwise it is false.
23638 Top-level cv qualifications of @code{base_type} and
23639 @code{derived_type} are ignored. For the purposes of this trait, a
23640 class type is considered is own base. Requires: if @code{__is_class
23641 (base_type)} and @code{__is_class (derived_type)} are true and
23642 @code{base_type} and @code{derived_type} are not the same type
23643 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
23644 type. A diagnostic is produced if this requirement is not met.
23646 @item __is_class (type)
23647 If @code{type} is a cv class type, and not a union type
23648 ([basic.compound]) the trait is true, else it is false.
23650 @item __is_empty (type)
23651 If @code{__is_class (type)} is false then the trait is false.
23652 Otherwise @code{type} is considered empty if and only if: @code{type}
23653 has no non-static data members, or all non-static data members, if
23654 any, are bit-fields of length 0, and @code{type} has no virtual
23655 members, and @code{type} has no virtual base classes, and @code{type}
23656 has no base classes @code{base_type} for which
23657 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
23658 be a complete type, (possibly cv-qualified) @code{void}, or an array
23661 @item __is_enum (type)
23662 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
23663 true, else it is false.
23665 @item __is_literal_type (type)
23666 If @code{type} is a literal type ([basic.types]) the trait is
23667 true, else it is false. Requires: @code{type} shall be a complete type,
23668 (possibly cv-qualified) @code{void}, or an array of unknown bound.
23670 @item __is_pod (type)
23671 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
23672 else it is false. Requires: @code{type} shall be a complete type,
23673 (possibly cv-qualified) @code{void}, or an array of unknown bound.
23675 @item __is_polymorphic (type)
23676 If @code{type} is a polymorphic class ([class.virtual]) then the trait
23677 is true, else it is false. Requires: @code{type} shall be a complete
23678 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23680 @item __is_standard_layout (type)
23681 If @code{type} is a standard-layout type ([basic.types]) the trait is
23682 true, else it is false. Requires: @code{type} shall be a complete
23683 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23685 @item __is_trivial (type)
23686 If @code{type} is a trivial type ([basic.types]) the trait is
23687 true, else it is false. Requires: @code{type} shall be a complete
23688 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
23690 @item __is_union (type)
23691 If @code{type} is a cv union type ([basic.compound]) the trait is
23692 true, else it is false.
23694 @item __underlying_type (type)
23695 The underlying type of @code{type}. Requires: @code{type} shall be
23696 an enumeration type ([dcl.enum]).
23698 @item __integer_pack (length)
23699 When used as the pattern of a pack expansion within a template
23700 definition, expands to a template argument pack containing integers
23701 from @code{0} to @code{length-1}. This is provided for efficient
23702 implementation of @code{std::make_integer_sequence}.
23708 @section C++ Concepts
23710 C++ concepts provide much-improved support for generic programming. In
23711 particular, they allow the specification of constraints on template arguments.
23712 The constraints are used to extend the usual overloading and partial
23713 specialization capabilities of the language, allowing generic data structures
23714 and algorithms to be ``refined'' based on their properties rather than their
23717 The following keywords are reserved for concepts.
23721 States an expression as an assumption, and if possible, verifies that the
23722 assumption is valid. For example, @code{assume(n > 0)}.
23725 Introduces an axiom definition. Axioms introduce requirements on values.
23728 Introduces a universally quantified object in an axiom. For example,
23729 @code{forall (int n) n + 0 == n}).
23732 Introduces a concept definition. Concepts are sets of syntactic and semantic
23733 requirements on types and their values.
23736 Introduces constraints on template arguments or requirements for a member
23737 function of a class template.
23741 The front end also exposes a number of internal mechanism that can be used
23742 to simplify the writing of type traits. Note that some of these traits are
23743 likely to be removed in the future.
23746 @item __is_same (type1, type2)
23747 A binary type trait: true whenever the type arguments are the same.
23752 @node Deprecated Features
23753 @section Deprecated Features
23755 In the past, the GNU C++ compiler was extended to experiment with new
23756 features, at a time when the C++ language was still evolving. Now that
23757 the C++ standard is complete, some of those features are superseded by
23758 superior alternatives. Using the old features might cause a warning in
23759 some cases that the feature will be dropped in the future. In other
23760 cases, the feature might be gone already.
23762 While the list below is not exhaustive, it documents some of the options
23763 that are now deprecated:
23766 @item -fexternal-templates
23767 @itemx -falt-external-templates
23768 These are two of the many ways for G++ to implement template
23769 instantiation. @xref{Template Instantiation}. The C++ standard clearly
23770 defines how template definitions have to be organized across
23771 implementation units. G++ has an implicit instantiation mechanism that
23772 should work just fine for standard-conforming code.
23774 @item -fstrict-prototype
23775 @itemx -fno-strict-prototype
23776 Previously it was possible to use an empty prototype parameter list to
23777 indicate an unspecified number of parameters (like C), rather than no
23778 parameters, as C++ demands. This feature has been removed, except where
23779 it is required for backwards compatibility. @xref{Backwards Compatibility}.
23782 G++ allows a virtual function returning @samp{void *} to be overridden
23783 by one returning a different pointer type. This extension to the
23784 covariant return type rules is now deprecated and will be removed from a
23787 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
23788 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
23789 and are now removed from G++. Code using these operators should be
23790 modified to use @code{std::min} and @code{std::max} instead.
23792 The named return value extension has been deprecated, and is now
23795 The use of initializer lists with new expressions has been deprecated,
23796 and is now removed from G++.
23798 Floating and complex non-type template parameters have been deprecated,
23799 and are now removed from G++.
23801 The implicit typename extension has been deprecated and is now
23804 The use of default arguments in function pointers, function typedefs
23805 and other places where they are not permitted by the standard is
23806 deprecated and will be removed from a future version of G++.
23808 G++ allows floating-point literals to appear in integral constant expressions,
23809 e.g.@: @samp{ enum E @{ e = int(2.2 * 3.7) @} }
23810 This extension is deprecated and will be removed from a future version.
23812 G++ allows static data members of const floating-point type to be declared
23813 with an initializer in a class definition. The standard only allows
23814 initializers for static members of const integral types and const
23815 enumeration types so this extension has been deprecated and will be removed
23816 from a future version.
23818 @node Backwards Compatibility
23819 @section Backwards Compatibility
23820 @cindex Backwards Compatibility
23821 @cindex ARM [Annotated C++ Reference Manual]
23823 Now that there is a definitive ISO standard C++, G++ has a specification
23824 to adhere to. The C++ language evolved over time, and features that
23825 used to be acceptable in previous drafts of the standard, such as the ARM
23826 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
23827 compilation of C++ written to such drafts, G++ contains some backwards
23828 compatibilities. @emph{All such backwards compatibility features are
23829 liable to disappear in future versions of G++.} They should be considered
23830 deprecated. @xref{Deprecated Features}.
23834 If a variable is declared at for scope, it used to remain in scope until
23835 the end of the scope that contained the for statement (rather than just
23836 within the for scope). G++ retains this, but issues a warning, if such a
23837 variable is accessed outside the for scope.
23839 @item Implicit C language
23840 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
23841 scope to set the language. On such systems, all header files are
23842 implicitly scoped inside a C language scope. Also, an empty prototype
23843 @code{()} is treated as an unspecified number of arguments, rather
23844 than no arguments, as C++ demands.
23847 @c LocalWords: emph deftypefn builtin ARCv2EM SIMD builtins msimd
23848 @c LocalWords: typedef v4si v8hi DMA dma vdiwr vdowr