1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012
3 @c Free Software Foundation, Inc.
5 @c This is part of the GCC manual.
6 @c For copying conditions, see the file gcc.texi.
9 @chapter Extensions to the C Language Family
10 @cindex extensions, C language
11 @cindex C language extensions
14 GNU C provides several language features not found in ISO standard C@.
15 (The @option{-pedantic} option directs GCC to print a warning message if
16 any of these features is used.) To test for the availability of these
17 features in conditional compilation, check for a predefined macro
18 @code{__GNUC__}, which is always defined under GCC@.
20 These extensions are available in C and Objective-C@. Most of them are
21 also available in C++. @xref{C++ Extensions,,Extensions to the
22 C++ Language}, for extensions that apply @emph{only} to C++.
24 Some features that are in ISO C99 but not C90 or C++ are also, as
25 extensions, accepted by GCC in C90 mode and in C++.
28 * Statement Exprs:: Putting statements and declarations inside expressions.
29 * Local Labels:: Labels local to a block.
30 * Labels as Values:: Getting pointers to labels, and computed gotos.
31 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
32 * Constructing Calls:: Dispatching a call to another function.
33 * Typeof:: @code{typeof}: referring to the type of an expression.
34 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
35 * Long Long:: Double-word integers---@code{long long int}.
36 * __int128:: 128-bit integers---@code{__int128}.
37 * Complex:: Data types for complex numbers.
38 * Floating Types:: Additional Floating Types.
39 * Half-Precision:: Half-Precision Floating Point.
40 * Decimal Float:: Decimal Floating Types.
41 * Hex Floats:: Hexadecimal floating-point constants.
42 * Fixed-Point:: Fixed-Point Types.
43 * Named Address Spaces::Named address spaces.
44 * Zero Length:: Zero-length arrays.
45 * Variable Length:: Arrays whose length is computed at run time.
46 * Empty Structures:: Structures with no members.
47 * Variadic Macros:: Macros with a variable number of arguments.
48 * Escaped Newlines:: Slightly looser rules for escaped newlines.
49 * Subscripting:: Any array can be subscripted, even if not an lvalue.
50 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
51 * Initializers:: Non-constant initializers.
52 * Compound Literals:: Compound literals give structures, unions
54 * Designated Inits:: Labeling elements of initializers.
55 * Cast to Union:: Casting to union type from any member of the union.
56 * Case Ranges:: `case 1 ... 9' and such.
57 * Mixed Declarations:: Mixing declarations and code.
58 * Function Attributes:: Declaring that functions have no side effects,
59 or that they can never return.
60 * Attribute Syntax:: Formal syntax for attributes.
61 * Function Prototypes:: Prototype declarations and old-style definitions.
62 * C++ Comments:: C++ comments are recognized.
63 * Dollar Signs:: Dollar sign is allowed in identifiers.
64 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
65 * Variable Attributes:: Specifying attributes of variables.
66 * Type Attributes:: Specifying attributes of types.
67 * Alignment:: Inquiring about the alignment of a type or variable.
68 * Inline:: Defining inline functions (as fast as macros).
69 * Volatiles:: What constitutes an access to a volatile object.
70 * Extended Asm:: Assembler instructions with C expressions as operands.
71 (With them you can define ``built-in'' functions.)
72 * Constraints:: Constraints for asm operands
73 * Asm Labels:: Specifying the assembler name to use for a C symbol.
74 * Explicit Reg Vars:: Defining variables residing in specified registers.
75 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
76 * Incomplete Enums:: @code{enum foo;}, with details to follow.
77 * Function Names:: Printable strings which are the name of the current
79 * Return Address:: Getting the return or frame address of a function.
80 * Vector Extensions:: Using vector instructions through built-in functions.
81 * Offsetof:: Special syntax for implementing @code{offsetof}.
82 * __sync Builtins:: Legacy built-in functions for atomic memory access.
83 * __atomic Builtins:: Atomic built-in functions with memory model.
84 * Object Size Checking:: Built-in functions for limited buffer overflow
86 * Other Builtins:: Other built-in functions.
87 * Target Builtins:: Built-in functions specific to particular targets.
88 * Target Format Checks:: Format checks specific to particular targets.
89 * Pragmas:: Pragmas accepted by GCC.
90 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
91 * Thread-Local:: Per-thread variables.
92 * Binary constants:: Binary constants using the @samp{0b} prefix.
96 @section Statements and Declarations in Expressions
97 @cindex statements inside expressions
98 @cindex declarations inside expressions
99 @cindex expressions containing statements
100 @cindex macros, statements in expressions
102 @c the above section title wrapped and causes an underfull hbox.. i
103 @c changed it from "within" to "in". --mew 4feb93
104 A compound statement enclosed in parentheses may appear as an expression
105 in GNU C@. This allows you to use loops, switches, and local variables
106 within an expression.
108 Recall that a compound statement is a sequence of statements surrounded
109 by braces; in this construct, parentheses go around the braces. For
113 (@{ int y = foo (); int z;
120 is a valid (though slightly more complex than necessary) expression
121 for the absolute value of @code{foo ()}.
123 The last thing in the compound statement should be an expression
124 followed by a semicolon; the value of this subexpression serves as the
125 value of the entire construct. (If you use some other kind of statement
126 last within the braces, the construct has type @code{void}, and thus
127 effectively no value.)
129 This feature is especially useful in making macro definitions ``safe'' (so
130 that they evaluate each operand exactly once). For example, the
131 ``maximum'' function is commonly defined as a macro in standard C as
135 #define max(a,b) ((a) > (b) ? (a) : (b))
139 @cindex side effects, macro argument
140 But this definition computes either @var{a} or @var{b} twice, with bad
141 results if the operand has side effects. In GNU C, if you know the
142 type of the operands (here taken as @code{int}), you can define
143 the macro safely as follows:
146 #define maxint(a,b) \
147 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
150 Embedded statements are not allowed in constant expressions, such as
151 the value of an enumeration constant, the width of a bit-field, or
152 the initial value of a static variable.
154 If you don't know the type of the operand, you can still do this, but you
155 must use @code{typeof} (@pxref{Typeof}).
157 In G++, the result value of a statement expression undergoes array and
158 function pointer decay, and is returned by value to the enclosing
159 expression. For instance, if @code{A} is a class, then
168 will construct a temporary @code{A} object to hold the result of the
169 statement expression, and that will be used to invoke @code{Foo}.
170 Therefore the @code{this} pointer observed by @code{Foo} will not be the
173 Any temporaries created within a statement within a statement expression
174 will be destroyed at the statement's end. This makes statement
175 expressions inside macros slightly different from function calls. In
176 the latter case temporaries introduced during argument evaluation will
177 be destroyed at the end of the statement that includes the function
178 call. In the statement expression case they will be destroyed during
179 the statement expression. For instance,
182 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
183 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
193 will have different places where temporaries are destroyed. For the
194 @code{macro} case, the temporary @code{X} will be destroyed just after
195 the initialization of @code{b}. In the @code{function} case that
196 temporary will be destroyed when the function returns.
198 These considerations mean that it is probably a bad idea to use
199 statement-expressions of this form in header files that are designed to
200 work with C++. (Note that some versions of the GNU C Library contained
201 header files using statement-expression that lead to precisely this
204 Jumping into a statement expression with @code{goto} or using a
205 @code{switch} statement outside the statement expression with a
206 @code{case} or @code{default} label inside the statement expression is
207 not permitted. Jumping into a statement expression with a computed
208 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
209 Jumping out of a statement expression is permitted, but if the
210 statement expression is part of a larger expression then it is
211 unspecified which other subexpressions of that expression have been
212 evaluated except where the language definition requires certain
213 subexpressions to be evaluated before or after the statement
214 expression. In any case, as with a function call the evaluation of a
215 statement expression is not interleaved with the evaluation of other
216 parts of the containing expression. For example,
219 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
223 will call @code{foo} and @code{bar1} and will not call @code{baz} but
224 may or may not call @code{bar2}. If @code{bar2} is called, it will be
225 called after @code{foo} and before @code{bar1}
228 @section Locally Declared Labels
230 @cindex macros, local labels
232 GCC allows you to declare @dfn{local labels} in any nested block
233 scope. A local label is just like an ordinary label, but you can
234 only reference it (with a @code{goto} statement, or by taking its
235 address) within the block in which it was declared.
237 A local label declaration looks like this:
240 __label__ @var{label};
247 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
250 Local label declarations must come at the beginning of the block,
251 before any ordinary declarations or statements.
253 The label declaration defines the label @emph{name}, but does not define
254 the label itself. You must do this in the usual way, with
255 @code{@var{label}:}, within the statements of the statement expression.
257 The local label feature is useful for complex macros. If a macro
258 contains nested loops, a @code{goto} can be useful for breaking out of
259 them. However, an ordinary label whose scope is the whole function
260 cannot be used: if the macro can be expanded several times in one
261 function, the label will be multiply defined in that function. A
262 local label avoids this problem. For example:
265 #define SEARCH(value, array, target) \
268 typeof (target) _SEARCH_target = (target); \
269 typeof (*(array)) *_SEARCH_array = (array); \
272 for (i = 0; i < max; i++) \
273 for (j = 0; j < max; j++) \
274 if (_SEARCH_array[i][j] == _SEARCH_target) \
275 @{ (value) = i; goto found; @} \
281 This could also be written using a statement-expression:
284 #define SEARCH(array, target) \
287 typeof (target) _SEARCH_target = (target); \
288 typeof (*(array)) *_SEARCH_array = (array); \
291 for (i = 0; i < max; i++) \
292 for (j = 0; j < max; j++) \
293 if (_SEARCH_array[i][j] == _SEARCH_target) \
294 @{ value = i; goto found; @} \
301 Local label declarations also make the labels they declare visible to
302 nested functions, if there are any. @xref{Nested Functions}, for details.
304 @node Labels as Values
305 @section Labels as Values
306 @cindex labels as values
307 @cindex computed gotos
308 @cindex goto with computed label
309 @cindex address of a label
311 You can get the address of a label defined in the current function
312 (or a containing function) with the unary operator @samp{&&}. The
313 value has type @code{void *}. This value is a constant and can be used
314 wherever a constant of that type is valid. For example:
322 To use these values, you need to be able to jump to one. This is done
323 with the computed goto statement@footnote{The analogous feature in
324 Fortran is called an assigned goto, but that name seems inappropriate in
325 C, where one can do more than simply store label addresses in label
326 variables.}, @code{goto *@var{exp};}. For example,
333 Any expression of type @code{void *} is allowed.
335 One way of using these constants is in initializing a static array that
336 will serve as a jump table:
339 static void *array[] = @{ &&foo, &&bar, &&hack @};
342 Then you can select a label with indexing, like this:
349 Note that this does not check whether the subscript is in bounds---array
350 indexing in C never does that.
352 Such an array of label values serves a purpose much like that of the
353 @code{switch} statement. The @code{switch} statement is cleaner, so
354 use that rather than an array unless the problem does not fit a
355 @code{switch} statement very well.
357 Another use of label values is in an interpreter for threaded code.
358 The labels within the interpreter function can be stored in the
359 threaded code for super-fast dispatching.
361 You may not use this mechanism to jump to code in a different function.
362 If you do that, totally unpredictable things will happen. The best way to
363 avoid this is to store the label address only in automatic variables and
364 never pass it as an argument.
366 An alternate way to write the above example is
369 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
371 goto *(&&foo + array[i]);
375 This is more friendly to code living in shared libraries, as it reduces
376 the number of dynamic relocations that are needed, and by consequence,
377 allows the data to be read-only.
379 The @code{&&foo} expressions for the same label might have different
380 values if the containing function is inlined or cloned. If a program
381 relies on them being always the same,
382 @code{__attribute__((__noinline__,__noclone__))} should be used to
383 prevent inlining and cloning. If @code{&&foo} is used in a static
384 variable initializer, inlining and cloning is forbidden.
386 @node Nested Functions
387 @section Nested Functions
388 @cindex nested functions
389 @cindex downward funargs
392 A @dfn{nested function} is a function defined inside another function.
393 (Nested functions are not supported for GNU C++.) The nested function's
394 name is local to the block where it is defined. For example, here we
395 define a nested function named @code{square}, and call it twice:
399 foo (double a, double b)
401 double square (double z) @{ return z * z; @}
403 return square (a) + square (b);
408 The nested function can access all the variables of the containing
409 function that are visible at the point of its definition. This is
410 called @dfn{lexical scoping}. For example, here we show a nested
411 function which uses an inherited variable named @code{offset}:
415 bar (int *array, int offset, int size)
417 int access (int *array, int index)
418 @{ return array[index + offset]; @}
421 for (i = 0; i < size; i++)
422 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
427 Nested function definitions are permitted within functions in the places
428 where variable definitions are allowed; that is, in any block, mixed
429 with the other declarations and statements in the block.
431 It is possible to call the nested function from outside the scope of its
432 name by storing its address or passing the address to another function:
435 hack (int *array, int size)
437 void store (int index, int value)
438 @{ array[index] = value; @}
440 intermediate (store, size);
444 Here, the function @code{intermediate} receives the address of
445 @code{store} as an argument. If @code{intermediate} calls @code{store},
446 the arguments given to @code{store} are used to store into @code{array}.
447 But this technique works only so long as the containing function
448 (@code{hack}, in this example) does not exit.
450 If you try to call the nested function through its address after the
451 containing function has exited, all hell will break loose. If you try
452 to call it after a containing scope level has exited, and if it refers
453 to some of the variables that are no longer in scope, you may be lucky,
454 but it's not wise to take the risk. If, however, the nested function
455 does not refer to anything that has gone out of scope, you should be
458 GCC implements taking the address of a nested function using a technique
459 called @dfn{trampolines}. This technique was described in
460 @cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
461 C++ Conference Proceedings, October 17-21, 1988).
463 A nested function can jump to a label inherited from a containing
464 function, provided the label was explicitly declared in the containing
465 function (@pxref{Local Labels}). Such a jump returns instantly to the
466 containing function, exiting the nested function which did the
467 @code{goto} and any intermediate functions as well. Here is an example:
471 bar (int *array, int offset, int size)
474 int access (int *array, int index)
478 return array[index + offset];
482 for (i = 0; i < size; i++)
483 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
487 /* @r{Control comes here from @code{access}
488 if it detects an error.} */
495 A nested function always has no linkage. Declaring one with
496 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
497 before its definition, use @code{auto} (which is otherwise meaningless
498 for function declarations).
501 bar (int *array, int offset, int size)
504 auto int access (int *, int);
506 int access (int *array, int index)
510 return array[index + offset];
516 @node Constructing Calls
517 @section Constructing Function Calls
518 @cindex constructing calls
519 @cindex forwarding calls
521 Using the built-in functions described below, you can record
522 the arguments a function received, and call another function
523 with the same arguments, without knowing the number or types
526 You can also record the return value of that function call,
527 and later return that value, without knowing what data type
528 the function tried to return (as long as your caller expects
531 However, these built-in functions may interact badly with some
532 sophisticated features or other extensions of the language. It
533 is, therefore, not recommended to use them outside very simple
534 functions acting as mere forwarders for their arguments.
536 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
537 This built-in function returns a pointer to data
538 describing how to perform a call with the same arguments as were passed
539 to the current function.
541 The function saves the arg pointer register, structure value address,
542 and all registers that might be used to pass arguments to a function
543 into a block of memory allocated on the stack. Then it returns the
544 address of that block.
547 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
548 This built-in function invokes @var{function}
549 with a copy of the parameters described by @var{arguments}
552 The value of @var{arguments} should be the value returned by
553 @code{__builtin_apply_args}. The argument @var{size} specifies the size
554 of the stack argument data, in bytes.
556 This function returns a pointer to data describing
557 how to return whatever value was returned by @var{function}. The data
558 is saved in a block of memory allocated on the stack.
560 It is not always simple to compute the proper value for @var{size}. The
561 value is used by @code{__builtin_apply} to compute the amount of data
562 that should be pushed on the stack and copied from the incoming argument
566 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
567 This built-in function returns the value described by @var{result} from
568 the containing function. You should specify, for @var{result}, a value
569 returned by @code{__builtin_apply}.
572 @deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
573 This built-in function represents all anonymous arguments of an inline
574 function. It can be used only in inline functions which will be always
575 inlined, never compiled as a separate function, such as those using
576 @code{__attribute__ ((__always_inline__))} or
577 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
578 It must be only passed as last argument to some other function
579 with variable arguments. This is useful for writing small wrapper
580 inlines for variable argument functions, when using preprocessor
581 macros is undesirable. For example:
583 extern int myprintf (FILE *f, const char *format, ...);
584 extern inline __attribute__ ((__gnu_inline__)) int
585 myprintf (FILE *f, const char *format, ...)
587 int r = fprintf (f, "myprintf: ");
590 int s = fprintf (f, format, __builtin_va_arg_pack ());
598 @deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
599 This built-in function returns the number of anonymous arguments of
600 an inline function. It can be used only in inline functions which
601 will be always inlined, never compiled as a separate function, such
602 as those using @code{__attribute__ ((__always_inline__))} or
603 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
604 For example following will do link or runtime checking of open
605 arguments for optimized code:
608 extern inline __attribute__((__gnu_inline__)) int
609 myopen (const char *path, int oflag, ...)
611 if (__builtin_va_arg_pack_len () > 1)
612 warn_open_too_many_arguments ();
614 if (__builtin_constant_p (oflag))
616 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
618 warn_open_missing_mode ();
619 return __open_2 (path, oflag);
621 return open (path, oflag, __builtin_va_arg_pack ());
624 if (__builtin_va_arg_pack_len () < 1)
625 return __open_2 (path, oflag);
627 return open (path, oflag, __builtin_va_arg_pack ());
634 @section Referring to a Type with @code{typeof}
637 @cindex macros, types of arguments
639 Another way to refer to the type of an expression is with @code{typeof}.
640 The syntax of using of this keyword looks like @code{sizeof}, but the
641 construct acts semantically like a type name defined with @code{typedef}.
643 There are two ways of writing the argument to @code{typeof}: with an
644 expression or with a type. Here is an example with an expression:
651 This assumes that @code{x} is an array of pointers to functions;
652 the type described is that of the values of the functions.
654 Here is an example with a typename as the argument:
661 Here the type described is that of pointers to @code{int}.
663 If you are writing a header file that must work when included in ISO C
664 programs, write @code{__typeof__} instead of @code{typeof}.
665 @xref{Alternate Keywords}.
667 A @code{typeof}-construct can be used anywhere a typedef name could be
668 used. For example, you can use it in a declaration, in a cast, or inside
669 of @code{sizeof} or @code{typeof}.
671 The operand of @code{typeof} is evaluated for its side effects if and
672 only if it is an expression of variably modified type or the name of
675 @code{typeof} is often useful in conjunction with the
676 statements-within-expressions feature. Here is how the two together can
677 be used to define a safe ``maximum'' macro that operates on any
678 arithmetic type and evaluates each of its arguments exactly once:
682 (@{ typeof (a) _a = (a); \
683 typeof (b) _b = (b); \
684 _a > _b ? _a : _b; @})
687 @cindex underscores in variables in macros
688 @cindex @samp{_} in variables in macros
689 @cindex local variables in macros
690 @cindex variables, local, in macros
691 @cindex macros, local variables in
693 The reason for using names that start with underscores for the local
694 variables is to avoid conflicts with variable names that occur within the
695 expressions that are substituted for @code{a} and @code{b}. Eventually we
696 hope to design a new form of declaration syntax that allows you to declare
697 variables whose scopes start only after their initializers; this will be a
698 more reliable way to prevent such conflicts.
701 Some more examples of the use of @code{typeof}:
705 This declares @code{y} with the type of what @code{x} points to.
712 This declares @code{y} as an array of such values.
719 This declares @code{y} as an array of pointers to characters:
722 typeof (typeof (char *)[4]) y;
726 It is equivalent to the following traditional C declaration:
732 To see the meaning of the declaration using @code{typeof}, and why it
733 might be a useful way to write, rewrite it with these macros:
736 #define pointer(T) typeof(T *)
737 #define array(T, N) typeof(T [N])
741 Now the declaration can be rewritten this way:
744 array (pointer (char), 4) y;
748 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
749 pointers to @code{char}.
752 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
753 a more limited extension which permitted one to write
756 typedef @var{T} = @var{expr};
760 with the effect of declaring @var{T} to have the type of the expression
761 @var{expr}. This extension does not work with GCC 3 (versions between
762 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
763 relies on it should be rewritten to use @code{typeof}:
766 typedef typeof(@var{expr}) @var{T};
770 This will work with all versions of GCC@.
773 @section Conditionals with Omitted Operands
774 @cindex conditional expressions, extensions
775 @cindex omitted middle-operands
776 @cindex middle-operands, omitted
777 @cindex extensions, @code{?:}
778 @cindex @code{?:} extensions
780 The middle operand in a conditional expression may be omitted. Then
781 if the first operand is nonzero, its value is the value of the conditional
784 Therefore, the expression
791 has the value of @code{x} if that is nonzero; otherwise, the value of
794 This example is perfectly equivalent to
800 @cindex side effect in @code{?:}
801 @cindex @code{?:} side effect
803 In this simple case, the ability to omit the middle operand is not
804 especially useful. When it becomes useful is when the first operand does,
805 or may (if it is a macro argument), contain a side effect. Then repeating
806 the operand in the middle would perform the side effect twice. Omitting
807 the middle operand uses the value already computed without the undesirable
808 effects of recomputing it.
811 @section 128-bits integers
812 @cindex @code{__int128} data types
814 As an extension the integer scalar type @code{__int128} is supported for
815 targets having an integer mode wide enough to hold 128-bit.
816 Simply write @code{__int128} for a signed 128-bit integer, or
817 @code{unsigned __int128} for an unsigned 128-bit integer. There is no
818 support in GCC to express an integer constant of type @code{__int128}
819 for targets having @code{long long} integer with less then 128 bit width.
822 @section Double-Word Integers
823 @cindex @code{long long} data types
824 @cindex double-word arithmetic
825 @cindex multiprecision arithmetic
826 @cindex @code{LL} integer suffix
827 @cindex @code{ULL} integer suffix
829 ISO C99 supports data types for integers that are at least 64 bits wide,
830 and as an extension GCC supports them in C90 mode and in C++.
831 Simply write @code{long long int} for a signed integer, or
832 @code{unsigned long long int} for an unsigned integer. To make an
833 integer constant of type @code{long long int}, add the suffix @samp{LL}
834 to the integer. To make an integer constant of type @code{unsigned long
835 long int}, add the suffix @samp{ULL} to the integer.
837 You can use these types in arithmetic like any other integer types.
838 Addition, subtraction, and bitwise boolean operations on these types
839 are open-coded on all types of machines. Multiplication is open-coded
840 if the machine supports fullword-to-doubleword a widening multiply
841 instruction. Division and shifts are open-coded only on machines that
842 provide special support. The operations that are not open-coded use
843 special library routines that come with GCC@.
845 There may be pitfalls when you use @code{long long} types for function
846 arguments, unless you declare function prototypes. If a function
847 expects type @code{int} for its argument, and you pass a value of type
848 @code{long long int}, confusion will result because the caller and the
849 subroutine will disagree about the number of bytes for the argument.
850 Likewise, if the function expects @code{long long int} and you pass
851 @code{int}. The best way to avoid such problems is to use prototypes.
854 @section Complex Numbers
855 @cindex complex numbers
856 @cindex @code{_Complex} keyword
857 @cindex @code{__complex__} keyword
859 ISO C99 supports complex floating data types, and as an extension GCC
860 supports them in C90 mode and in C++, and supports complex integer data
861 types which are not part of ISO C99. You can declare complex types
862 using the keyword @code{_Complex}. As an extension, the older GNU
863 keyword @code{__complex__} is also supported.
865 For example, @samp{_Complex double x;} declares @code{x} as a
866 variable whose real part and imaginary part are both of type
867 @code{double}. @samp{_Complex short int y;} declares @code{y} to
868 have real and imaginary parts of type @code{short int}; this is not
869 likely to be useful, but it shows that the set of complex types is
872 To write a constant with a complex data type, use the suffix @samp{i} or
873 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
874 has type @code{_Complex float} and @code{3i} has type
875 @code{_Complex int}. Such a constant always has a pure imaginary
876 value, but you can form any complex value you like by adding one to a
877 real constant. This is a GNU extension; if you have an ISO C99
878 conforming C library (such as GNU libc), and want to construct complex
879 constants of floating type, you should include @code{<complex.h>} and
880 use the macros @code{I} or @code{_Complex_I} instead.
882 @cindex @code{__real__} keyword
883 @cindex @code{__imag__} keyword
884 To extract the real part of a complex-valued expression @var{exp}, write
885 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
886 extract the imaginary part. This is a GNU extension; for values of
887 floating type, you should use the ISO C99 functions @code{crealf},
888 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
889 @code{cimagl}, declared in @code{<complex.h>} and also provided as
890 built-in functions by GCC@.
892 @cindex complex conjugation
893 The operator @samp{~} performs complex conjugation when used on a value
894 with a complex type. This is a GNU extension; for values of
895 floating type, you should use the ISO C99 functions @code{conjf},
896 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
897 provided as built-in functions by GCC@.
899 GCC can allocate complex automatic variables in a noncontiguous
900 fashion; it's even possible for the real part to be in a register while
901 the imaginary part is on the stack (or vice-versa). Only the DWARF2
902 debug info format can represent this, so use of DWARF2 is recommended.
903 If you are using the stabs debug info format, GCC describes a noncontiguous
904 complex variable as if it were two separate variables of noncomplex type.
905 If the variable's actual name is @code{foo}, the two fictitious
906 variables are named @code{foo$real} and @code{foo$imag}. You can
907 examine and set these two fictitious variables with your debugger.
910 @section Additional Floating Types
911 @cindex additional floating types
912 @cindex @code{__float80} data type
913 @cindex @code{__float128} data type
914 @cindex @code{w} floating point suffix
915 @cindex @code{q} floating point suffix
916 @cindex @code{W} floating point suffix
917 @cindex @code{Q} floating point suffix
919 As an extension, the GNU C compiler supports additional floating
920 types, @code{__float80} and @code{__float128} to support 80bit
921 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
922 Support for additional types includes the arithmetic operators:
923 add, subtract, multiply, divide; unary arithmetic operators;
924 relational operators; equality operators; and conversions to and from
925 integer and other floating types. Use a suffix @samp{w} or @samp{W}
926 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
927 for @code{_float128}. You can declare complex types using the
928 corresponding internal complex type, @code{XCmode} for @code{__float80}
929 type and @code{TCmode} for @code{__float128} type:
932 typedef _Complex float __attribute__((mode(TC))) _Complex128;
933 typedef _Complex float __attribute__((mode(XC))) _Complex80;
936 Not all targets support additional floating point types. @code{__float80}
937 and @code{__float128} types are supported on i386, x86_64 and ia64 targets.
938 The @code{__float128} type is supported on hppa HP-UX targets.
941 @section Half-Precision Floating Point
942 @cindex half-precision floating point
943 @cindex @code{__fp16} data type
945 On ARM targets, GCC supports half-precision (16-bit) floating point via
946 the @code{__fp16} type. You must enable this type explicitly
947 with the @option{-mfp16-format} command-line option in order to use it.
949 ARM supports two incompatible representations for half-precision
950 floating-point values. You must choose one of the representations and
951 use it consistently in your program.
953 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
954 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
955 There are 11 bits of significand precision, approximately 3
958 Specifying @option{-mfp16-format=alternative} selects the ARM
959 alternative format. This representation is similar to the IEEE
960 format, but does not support infinities or NaNs. Instead, the range
961 of exponents is extended, so that this format can represent normalized
962 values in the range of @math{2^{-14}} to 131008.
964 The @code{__fp16} type is a storage format only. For purposes
965 of arithmetic and other operations, @code{__fp16} values in C or C++
966 expressions are automatically promoted to @code{float}. In addition,
967 you cannot declare a function with a return value or parameters
968 of type @code{__fp16}.
970 Note that conversions from @code{double} to @code{__fp16}
971 involve an intermediate conversion to @code{float}. Because
972 of rounding, this can sometimes produce a different result than a
975 ARM provides hardware support for conversions between
976 @code{__fp16} and @code{float} values
977 as an extension to VFP and NEON (Advanced SIMD). GCC generates
978 code using these hardware instructions if you compile with
979 options to select an FPU that provides them;
980 for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
981 in addition to the @option{-mfp16-format} option to select
982 a half-precision format.
984 Language-level support for the @code{__fp16} data type is
985 independent of whether GCC generates code using hardware floating-point
986 instructions. In cases where hardware support is not specified, GCC
987 implements conversions between @code{__fp16} and @code{float} values
991 @section Decimal Floating Types
992 @cindex decimal floating types
993 @cindex @code{_Decimal32} data type
994 @cindex @code{_Decimal64} data type
995 @cindex @code{_Decimal128} data type
996 @cindex @code{df} integer suffix
997 @cindex @code{dd} integer suffix
998 @cindex @code{dl} integer suffix
999 @cindex @code{DF} integer suffix
1000 @cindex @code{DD} integer suffix
1001 @cindex @code{DL} integer suffix
1003 As an extension, the GNU C compiler supports decimal floating types as
1004 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
1005 floating types in GCC will evolve as the draft technical report changes.
1006 Calling conventions for any target might also change. Not all targets
1007 support decimal floating types.
1009 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
1010 @code{_Decimal128}. They use a radix of ten, unlike the floating types
1011 @code{float}, @code{double}, and @code{long double} whose radix is not
1012 specified by the C standard but is usually two.
1014 Support for decimal floating types includes the arithmetic operators
1015 add, subtract, multiply, divide; unary arithmetic operators;
1016 relational operators; equality operators; and conversions to and from
1017 integer and other floating types. Use a suffix @samp{df} or
1018 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1019 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1022 GCC support of decimal float as specified by the draft technical report
1027 When the value of a decimal floating type cannot be represented in the
1028 integer type to which it is being converted, the result is undefined
1029 rather than the result value specified by the draft technical report.
1032 GCC does not provide the C library functionality associated with
1033 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1034 @file{wchar.h}, which must come from a separate C library implementation.
1035 Because of this the GNU C compiler does not define macro
1036 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1037 the technical report.
1040 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1041 are supported by the DWARF2 debug information format.
1047 ISO C99 supports floating-point numbers written not only in the usual
1048 decimal notation, such as @code{1.55e1}, but also numbers such as
1049 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1050 supports this in C90 mode (except in some cases when strictly
1051 conforming) and in C++. In that format the
1052 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1053 mandatory. The exponent is a decimal number that indicates the power of
1054 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1061 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1062 is the same as @code{1.55e1}.
1064 Unlike for floating-point numbers in the decimal notation the exponent
1065 is always required in the hexadecimal notation. Otherwise the compiler
1066 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1067 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1068 extension for floating-point constants of type @code{float}.
1071 @section Fixed-Point Types
1072 @cindex fixed-point types
1073 @cindex @code{_Fract} data type
1074 @cindex @code{_Accum} data type
1075 @cindex @code{_Sat} data type
1076 @cindex @code{hr} fixed-suffix
1077 @cindex @code{r} fixed-suffix
1078 @cindex @code{lr} fixed-suffix
1079 @cindex @code{llr} fixed-suffix
1080 @cindex @code{uhr} fixed-suffix
1081 @cindex @code{ur} fixed-suffix
1082 @cindex @code{ulr} fixed-suffix
1083 @cindex @code{ullr} fixed-suffix
1084 @cindex @code{hk} fixed-suffix
1085 @cindex @code{k} fixed-suffix
1086 @cindex @code{lk} fixed-suffix
1087 @cindex @code{llk} fixed-suffix
1088 @cindex @code{uhk} fixed-suffix
1089 @cindex @code{uk} fixed-suffix
1090 @cindex @code{ulk} fixed-suffix
1091 @cindex @code{ullk} fixed-suffix
1092 @cindex @code{HR} fixed-suffix
1093 @cindex @code{R} fixed-suffix
1094 @cindex @code{LR} fixed-suffix
1095 @cindex @code{LLR} fixed-suffix
1096 @cindex @code{UHR} fixed-suffix
1097 @cindex @code{UR} fixed-suffix
1098 @cindex @code{ULR} fixed-suffix
1099 @cindex @code{ULLR} fixed-suffix
1100 @cindex @code{HK} fixed-suffix
1101 @cindex @code{K} fixed-suffix
1102 @cindex @code{LK} fixed-suffix
1103 @cindex @code{LLK} fixed-suffix
1104 @cindex @code{UHK} fixed-suffix
1105 @cindex @code{UK} fixed-suffix
1106 @cindex @code{ULK} fixed-suffix
1107 @cindex @code{ULLK} fixed-suffix
1109 As an extension, the GNU C compiler supports fixed-point types as
1110 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1111 types in GCC will evolve as the draft technical report changes.
1112 Calling conventions for any target might also change. Not all targets
1113 support fixed-point types.
1115 The fixed-point types are
1116 @code{short _Fract},
1119 @code{long long _Fract},
1120 @code{unsigned short _Fract},
1121 @code{unsigned _Fract},
1122 @code{unsigned long _Fract},
1123 @code{unsigned long long _Fract},
1124 @code{_Sat short _Fract},
1126 @code{_Sat long _Fract},
1127 @code{_Sat long long _Fract},
1128 @code{_Sat unsigned short _Fract},
1129 @code{_Sat unsigned _Fract},
1130 @code{_Sat unsigned long _Fract},
1131 @code{_Sat unsigned long long _Fract},
1132 @code{short _Accum},
1135 @code{long long _Accum},
1136 @code{unsigned short _Accum},
1137 @code{unsigned _Accum},
1138 @code{unsigned long _Accum},
1139 @code{unsigned long long _Accum},
1140 @code{_Sat short _Accum},
1142 @code{_Sat long _Accum},
1143 @code{_Sat long long _Accum},
1144 @code{_Sat unsigned short _Accum},
1145 @code{_Sat unsigned _Accum},
1146 @code{_Sat unsigned long _Accum},
1147 @code{_Sat unsigned long long _Accum}.
1149 Fixed-point data values contain fractional and optional integral parts.
1150 The format of fixed-point data varies and depends on the target machine.
1152 Support for fixed-point types includes:
1155 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1157 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1159 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1161 binary shift operators (@code{<<}, @code{>>})
1163 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1165 equality operators (@code{==}, @code{!=})
1167 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1168 @code{<<=}, @code{>>=})
1170 conversions to and from integer, floating-point, or fixed-point types
1173 Use a suffix in a fixed-point literal constant:
1175 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1176 @code{_Sat short _Fract}
1177 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1178 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1179 @code{_Sat long _Fract}
1180 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1181 @code{_Sat long long _Fract}
1182 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1183 @code{_Sat unsigned short _Fract}
1184 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1185 @code{_Sat unsigned _Fract}
1186 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1187 @code{_Sat unsigned long _Fract}
1188 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1189 and @code{_Sat unsigned long long _Fract}
1190 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1191 @code{_Sat short _Accum}
1192 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1193 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1194 @code{_Sat long _Accum}
1195 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1196 @code{_Sat long long _Accum}
1197 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1198 @code{_Sat unsigned short _Accum}
1199 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1200 @code{_Sat unsigned _Accum}
1201 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1202 @code{_Sat unsigned long _Accum}
1203 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1204 and @code{_Sat unsigned long long _Accum}
1207 GCC support of fixed-point types as specified by the draft technical report
1212 Pragmas to control overflow and rounding behaviors are not implemented.
1215 Fixed-point types are supported by the DWARF2 debug information format.
1217 @node Named Address Spaces
1218 @section Named Address Spaces
1219 @cindex Named Address Spaces
1221 As an extension, the GNU C compiler supports named address spaces as
1222 defined in the N1275 draft of ISO/IEC DTR 18037. Support for named
1223 address spaces in GCC will evolve as the draft technical report
1224 changes. Calling conventions for any target might also change. At
1225 present, only the AVR, SPU, M32C, and RL78 targets support address
1226 spaces other than the generic address space.
1228 Address space identifiers may be used exactly like any other C type
1229 qualifier (e.g., @code{const} or @code{volatile}). See the N1275
1230 document for more details.
1232 @anchor{AVR Named Address Spaces}
1233 @subsection AVR Named Address Spaces
1235 On the AVR target, there are several address spaces that can be used
1236 in order to put read-only data into the flash memory and access that
1237 data by means of the special instructions @code{LPM} or @code{ELPM}
1238 needed to read from flash.
1240 Per default, any data including read-only data is located in RAM
1241 (the generic address space) so that non-generic address spaces are
1242 needed to locate read-only data in flash memory
1243 @emph{and} to generate the right instructions to access this data
1244 without using (inline) assembler code.
1248 @cindex @code{__flash} AVR Named Address Spaces
1249 The @code{__flash} qualifier will locate data in the
1250 @code{.progmem.data} section. Data will be read using the @code{LPM}
1251 instruction. Pointers to this address space are 16 bits wide.
1258 @cindex @code{__flash1} AVR Named Address Spaces
1259 @cindex @code{__flash2} AVR Named Address Spaces
1260 @cindex @code{__flash3} AVR Named Address Spaces
1261 @cindex @code{__flash4} AVR Named Address Spaces
1262 @cindex @code{__flash5} AVR Named Address Spaces
1263 These are 16-bit address spaces locating data in section
1264 @code{.progmem@var{N}.data} where @var{N} refers to
1265 address space @code{__flash@var{N}}.
1266 The compiler will set the @code{RAMPZ} segment register approptiately
1267 before reading data by means of the @code{ELPM} instruction.
1269 On devices with less 64@tie{}kiB flash segments as indicated by the address
1270 space, the compiler will cut down the segment number to a number the
1271 device actually supports. Counting starts at@tie{}@code{0}
1272 for space @code{__flash}. For example, if you access address space
1273 @code{__flash3} on an ATmega128 device with two 64@tie{}kiB flash segments,
1274 the compiler will generate a read from @code{__flash1}, i.e.@: it
1275 will load @code{RAMPZ} with@tie{}@code{1} before reading.
1278 @cindex @code{__memx} AVR Named Address Spaces
1279 This is a 24-bit address space that linearizes flash and RAM:
1280 If the high bit of the address is set, data is read from
1281 RAM using the lower two bytes as RAM address.
1282 If the high bit of the address is clear, data is read from flash
1283 with @code{RAMPZ} set according to the high byte of the address.
1285 Objects in this address space will be located in @code{.progmem.data}.
1291 char my_read (const __flash char ** p)
1293 /* p is a pointer to RAM that points to a pointer to flash.
1294 The first indirection of p will read that flash pointer
1295 from RAM and the second indirection reads a char from this
1301 /* Locate array[] in flash memory */
1302 const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};
1308 /* Return 17 by reading from flash memory */
1309 return array[array[i]];
1313 For each named address space supported by avr-gcc there is an equally
1314 named but uppercase built-in macro defined.
1315 The purpose is to facilitate testing if respective address space
1316 support is available or not:
1320 const __flash int var = 1;
1327 #include <avr/pgmspace.h> /* From avr-libc */
1329 const int var PROGMEM = 1;
1333 return (int) pgm_read_word (&i);
1335 #endif /* __FLASH */
1338 Notice that attribute @ref{AVR Variable Attributes,@code{progmem}}
1339 locates data in flash but
1340 accesses to these data will read from generic address space, i.e.@:
1342 so that you need special accessors like @code{pgm_read_byte}
1343 from @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc}}.
1345 @b{Limitations and caveats}
1349 Reading across the 64@tie{}KiB section boundary of
1350 the @code{__flash} or @code{__flash@var{N}} address spaces
1351 will show undefined behaviour. The only address space that
1352 supports reading across the 64@tie{}KiB flash segment boundaries is
1356 If you use one if the @code{__flash@var{N}} address spaces
1357 you will have to arrange your linker skript to locate the
1358 @code{.progmem@var{N}.data} sections according to your needs.
1361 Any data or pointers to the non-generic address spaces must
1362 be qualified as @code{const}, i.e.@: as read-only data.
1363 This still applies if the data in one of these address
1364 spaces like software version number or calibration lookup table are intended to
1365 be changed after load time by, say, a boot loader. In this case
1366 the right qualification is @code{const} @code{volatile} so that the compiler
1367 must not optimize away known values or insert them
1368 as immediates into operands of instructions.
1371 Code like the following is not yet supported because of missing
1372 support in avr-binutils,
1373 see @w{@uref{http://sourceware.org/PR13503,PR13503}}.
1375 extern const __memx char foo;
1376 const __memx void *pfoo = &foo;
1378 The code will throw an assembler warning and the high byte of
1379 @code{pfoo} will be initialized with@tie{}@code{0}, i.e.@: the
1380 initialization will be as if @code{foo} was located in the first
1381 64@tie{}KiB chunk of flash.
1385 @subsection M32C Named Address Spaces
1386 @cindex @code{__far} M32C Named Address Spaces
1388 On the M32C target, with the R8C and M16C cpu variants, variables
1389 qualified with @code{__far} are accessed using 32-bit addresses in
1390 order to access memory beyond the first 64@tie{}Ki bytes. If
1391 @code{__far} is used with the M32CM or M32C cpu variants, it has no
1394 @subsection RL78 Named Address Spaces
1395 @cindex @code{__far} RL78 Named Address Spaces
1397 On the RL78 target, variables qualified with @code{__far} are accessed
1398 with 32-bit pointers (20-bit addresses) rather than the default 16-bit
1399 addresses. Non-far variables are assumed to appear in the topmost
1400 64@tie{}KiB of the address space.
1402 @subsection SPU Named Address Spaces
1403 @cindex @code{__ea} SPU Named Address Spaces
1405 On the SPU target variables may be declared as
1406 belonging to another address space by qualifying the type with the
1407 @code{__ea} address space identifier:
1413 When the variable @code{i} is accessed, the compiler will generate
1414 special code to access this variable. It may use runtime library
1415 support, or generate special machine instructions to access that address
1419 @section Arrays of Length Zero
1420 @cindex arrays of length zero
1421 @cindex zero-length arrays
1422 @cindex length-zero arrays
1423 @cindex flexible array members
1425 Zero-length arrays are allowed in GNU C@. They are very useful as the
1426 last element of a structure which is really a header for a variable-length
1435 struct line *thisline = (struct line *)
1436 malloc (sizeof (struct line) + this_length);
1437 thisline->length = this_length;
1440 In ISO C90, you would have to give @code{contents} a length of 1, which
1441 means either you waste space or complicate the argument to @code{malloc}.
1443 In ISO C99, you would use a @dfn{flexible array member}, which is
1444 slightly different in syntax and semantics:
1448 Flexible array members are written as @code{contents[]} without
1452 Flexible array members have incomplete type, and so the @code{sizeof}
1453 operator may not be applied. As a quirk of the original implementation
1454 of zero-length arrays, @code{sizeof} evaluates to zero.
1457 Flexible array members may only appear as the last member of a
1458 @code{struct} that is otherwise non-empty.
1461 A structure containing a flexible array member, or a union containing
1462 such a structure (possibly recursively), may not be a member of a
1463 structure or an element of an array. (However, these uses are
1464 permitted by GCC as extensions.)
1467 GCC versions before 3.0 allowed zero-length arrays to be statically
1468 initialized, as if they were flexible arrays. In addition to those
1469 cases that were useful, it also allowed initializations in situations
1470 that would corrupt later data. Non-empty initialization of zero-length
1471 arrays is now treated like any case where there are more initializer
1472 elements than the array holds, in that a suitable warning about "excess
1473 elements in array" is given, and the excess elements (all of them, in
1474 this case) are ignored.
1476 Instead GCC allows static initialization of flexible array members.
1477 This is equivalent to defining a new structure containing the original
1478 structure followed by an array of sufficient size to contain the data.
1479 I.e.@: in the following, @code{f1} is constructed as if it were declared
1485 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1488 struct f1 f1; int data[3];
1489 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1493 The convenience of this extension is that @code{f1} has the desired
1494 type, eliminating the need to consistently refer to @code{f2.f1}.
1496 This has symmetry with normal static arrays, in that an array of
1497 unknown size is also written with @code{[]}.
1499 Of course, this extension only makes sense if the extra data comes at
1500 the end of a top-level object, as otherwise we would be overwriting
1501 data at subsequent offsets. To avoid undue complication and confusion
1502 with initialization of deeply nested arrays, we simply disallow any
1503 non-empty initialization except when the structure is the top-level
1504 object. For example:
1507 struct foo @{ int x; int y[]; @};
1508 struct bar @{ struct foo z; @};
1510 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1511 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1512 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1513 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1516 @node Empty Structures
1517 @section Structures With No Members
1518 @cindex empty structures
1519 @cindex zero-size structures
1521 GCC permits a C structure to have no members:
1528 The structure will have size zero. In C++, empty structures are part
1529 of the language. G++ treats empty structures as if they had a single
1530 member of type @code{char}.
1532 @node Variable Length
1533 @section Arrays of Variable Length
1534 @cindex variable-length arrays
1535 @cindex arrays of variable length
1538 Variable-length automatic arrays are allowed in ISO C99, and as an
1539 extension GCC accepts them in C90 mode and in C++. These arrays are
1540 declared like any other automatic arrays, but with a length that is not
1541 a constant expression. The storage is allocated at the point of
1542 declaration and deallocated when the brace-level is exited. For
1547 concat_fopen (char *s1, char *s2, char *mode)
1549 char str[strlen (s1) + strlen (s2) + 1];
1552 return fopen (str, mode);
1556 @cindex scope of a variable length array
1557 @cindex variable-length array scope
1558 @cindex deallocating variable length arrays
1559 Jumping or breaking out of the scope of the array name deallocates the
1560 storage. Jumping into the scope is not allowed; you get an error
1563 @cindex @code{alloca} vs variable-length arrays
1564 You can use the function @code{alloca} to get an effect much like
1565 variable-length arrays. The function @code{alloca} is available in
1566 many other C implementations (but not in all). On the other hand,
1567 variable-length arrays are more elegant.
1569 There are other differences between these two methods. Space allocated
1570 with @code{alloca} exists until the containing @emph{function} returns.
1571 The space for a variable-length array is deallocated as soon as the array
1572 name's scope ends. (If you use both variable-length arrays and
1573 @code{alloca} in the same function, deallocation of a variable-length array
1574 will also deallocate anything more recently allocated with @code{alloca}.)
1576 You can also use variable-length arrays as arguments to functions:
1580 tester (int len, char data[len][len])
1586 The length of an array is computed once when the storage is allocated
1587 and is remembered for the scope of the array in case you access it with
1590 If you want to pass the array first and the length afterward, you can
1591 use a forward declaration in the parameter list---another GNU extension.
1595 tester (int len; char data[len][len], int len)
1601 @cindex parameter forward declaration
1602 The @samp{int len} before the semicolon is a @dfn{parameter forward
1603 declaration}, and it serves the purpose of making the name @code{len}
1604 known when the declaration of @code{data} is parsed.
1606 You can write any number of such parameter forward declarations in the
1607 parameter list. They can be separated by commas or semicolons, but the
1608 last one must end with a semicolon, which is followed by the ``real''
1609 parameter declarations. Each forward declaration must match a ``real''
1610 declaration in parameter name and data type. ISO C99 does not support
1611 parameter forward declarations.
1613 @node Variadic Macros
1614 @section Macros with a Variable Number of Arguments.
1615 @cindex variable number of arguments
1616 @cindex macro with variable arguments
1617 @cindex rest argument (in macro)
1618 @cindex variadic macros
1620 In the ISO C standard of 1999, a macro can be declared to accept a
1621 variable number of arguments much as a function can. The syntax for
1622 defining the macro is similar to that of a function. Here is an
1626 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1629 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1630 such a macro, it represents the zero or more tokens until the closing
1631 parenthesis that ends the invocation, including any commas. This set of
1632 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1633 wherever it appears. See the CPP manual for more information.
1635 GCC has long supported variadic macros, and used a different syntax that
1636 allowed you to give a name to the variable arguments just like any other
1637 argument. Here is an example:
1640 #define debug(format, args...) fprintf (stderr, format, args)
1643 This is in all ways equivalent to the ISO C example above, but arguably
1644 more readable and descriptive.
1646 GNU CPP has two further variadic macro extensions, and permits them to
1647 be used with either of the above forms of macro definition.
1649 In standard C, you are not allowed to leave the variable argument out
1650 entirely; but you are allowed to pass an empty argument. For example,
1651 this invocation is invalid in ISO C, because there is no comma after
1658 GNU CPP permits you to completely omit the variable arguments in this
1659 way. In the above examples, the compiler would complain, though since
1660 the expansion of the macro still has the extra comma after the format
1663 To help solve this problem, CPP behaves specially for variable arguments
1664 used with the token paste operator, @samp{##}. If instead you write
1667 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1670 and if the variable arguments are omitted or empty, the @samp{##}
1671 operator causes the preprocessor to remove the comma before it. If you
1672 do provide some variable arguments in your macro invocation, GNU CPP
1673 does not complain about the paste operation and instead places the
1674 variable arguments after the comma. Just like any other pasted macro
1675 argument, these arguments are not macro expanded.
1677 @node Escaped Newlines
1678 @section Slightly Looser Rules for Escaped Newlines
1679 @cindex escaped newlines
1680 @cindex newlines (escaped)
1682 Recently, the preprocessor has relaxed its treatment of escaped
1683 newlines. Previously, the newline had to immediately follow a
1684 backslash. The current implementation allows whitespace in the form
1685 of spaces, horizontal and vertical tabs, and form feeds between the
1686 backslash and the subsequent newline. The preprocessor issues a
1687 warning, but treats it as a valid escaped newline and combines the two
1688 lines to form a single logical line. This works within comments and
1689 tokens, as well as between tokens. Comments are @emph{not} treated as
1690 whitespace for the purposes of this relaxation, since they have not
1691 yet been replaced with spaces.
1694 @section Non-Lvalue Arrays May Have Subscripts
1695 @cindex subscripting
1696 @cindex arrays, non-lvalue
1698 @cindex subscripting and function values
1699 In ISO C99, arrays that are not lvalues still decay to pointers, and
1700 may be subscripted, although they may not be modified or used after
1701 the next sequence point and the unary @samp{&} operator may not be
1702 applied to them. As an extension, GCC allows such arrays to be
1703 subscripted in C90 mode, though otherwise they do not decay to
1704 pointers outside C99 mode. For example,
1705 this is valid in GNU C though not valid in C90:
1709 struct foo @{int a[4];@};
1715 return f().a[index];
1721 @section Arithmetic on @code{void}- and Function-Pointers
1722 @cindex void pointers, arithmetic
1723 @cindex void, size of pointer to
1724 @cindex function pointers, arithmetic
1725 @cindex function, size of pointer to
1727 In GNU C, addition and subtraction operations are supported on pointers to
1728 @code{void} and on pointers to functions. This is done by treating the
1729 size of a @code{void} or of a function as 1.
1731 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1732 and on function types, and returns 1.
1734 @opindex Wpointer-arith
1735 The option @option{-Wpointer-arith} requests a warning if these extensions
1739 @section Non-Constant Initializers
1740 @cindex initializers, non-constant
1741 @cindex non-constant initializers
1743 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1744 automatic variable are not required to be constant expressions in GNU C@.
1745 Here is an example of an initializer with run-time varying elements:
1748 foo (float f, float g)
1750 float beat_freqs[2] = @{ f-g, f+g @};
1755 @node Compound Literals
1756 @section Compound Literals
1757 @cindex constructor expressions
1758 @cindex initializations in expressions
1759 @cindex structures, constructor expression
1760 @cindex expressions, constructor
1761 @cindex compound literals
1762 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1764 ISO C99 supports compound literals. A compound literal looks like
1765 a cast containing an initializer. Its value is an object of the
1766 type specified in the cast, containing the elements specified in
1767 the initializer; it is an lvalue. As an extension, GCC supports
1768 compound literals in C90 mode and in C++.
1770 Usually, the specified type is a structure. Assume that
1771 @code{struct foo} and @code{structure} are declared as shown:
1774 struct foo @{int a; char b[2];@} structure;
1778 Here is an example of constructing a @code{struct foo} with a compound literal:
1781 structure = ((struct foo) @{x + y, 'a', 0@});
1785 This is equivalent to writing the following:
1789 struct foo temp = @{x + y, 'a', 0@};
1794 You can also construct an array. If all the elements of the compound literal
1795 are (made up of) simple constant expressions, suitable for use in
1796 initializers of objects of static storage duration, then the compound
1797 literal can be coerced to a pointer to its first element and used in
1798 such an initializer, as shown here:
1801 char **foo = (char *[]) @{ "x", "y", "z" @};
1804 Compound literals for scalar types and union types are
1805 also allowed, but then the compound literal is equivalent
1808 As a GNU extension, GCC allows initialization of objects with static storage
1809 duration by compound literals (which is not possible in ISO C99, because
1810 the initializer is not a constant).
1811 It is handled as if the object was initialized only with the bracket
1812 enclosed list if the types of the compound literal and the object match.
1813 The initializer list of the compound literal must be constant.
1814 If the object being initialized has array type of unknown size, the size is
1815 determined by compound literal size.
1818 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1819 static int y[] = (int []) @{1, 2, 3@};
1820 static int z[] = (int [3]) @{1@};
1824 The above lines are equivalent to the following:
1826 static struct foo x = @{1, 'a', 'b'@};
1827 static int y[] = @{1, 2, 3@};
1828 static int z[] = @{1, 0, 0@};
1831 @node Designated Inits
1832 @section Designated Initializers
1833 @cindex initializers with labeled elements
1834 @cindex labeled elements in initializers
1835 @cindex case labels in initializers
1836 @cindex designated initializers
1838 Standard C90 requires the elements of an initializer to appear in a fixed
1839 order, the same as the order of the elements in the array or structure
1842 In ISO C99 you can give the elements in any order, specifying the array
1843 indices or structure field names they apply to, and GNU C allows this as
1844 an extension in C90 mode as well. This extension is not
1845 implemented in GNU C++.
1847 To specify an array index, write
1848 @samp{[@var{index}] =} before the element value. For example,
1851 int a[6] = @{ [4] = 29, [2] = 15 @};
1858 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1862 The index values must be constant expressions, even if the array being
1863 initialized is automatic.
1865 An alternative syntax for this which has been obsolete since GCC 2.5 but
1866 GCC still accepts is to write @samp{[@var{index}]} before the element
1867 value, with no @samp{=}.
1869 To initialize a range of elements to the same value, write
1870 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1871 extension. For example,
1874 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1878 If the value in it has side-effects, the side-effects will happen only once,
1879 not for each initialized field by the range initializer.
1882 Note that the length of the array is the highest value specified
1885 In a structure initializer, specify the name of a field to initialize
1886 with @samp{.@var{fieldname} =} before the element value. For example,
1887 given the following structure,
1890 struct point @{ int x, y; @};
1894 the following initialization
1897 struct point p = @{ .y = yvalue, .x = xvalue @};
1904 struct point p = @{ xvalue, yvalue @};
1907 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1908 @samp{@var{fieldname}:}, as shown here:
1911 struct point p = @{ y: yvalue, x: xvalue @};
1915 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1916 @dfn{designator}. You can also use a designator (or the obsolete colon
1917 syntax) when initializing a union, to specify which element of the union
1918 should be used. For example,
1921 union foo @{ int i; double d; @};
1923 union foo f = @{ .d = 4 @};
1927 will convert 4 to a @code{double} to store it in the union using
1928 the second element. By contrast, casting 4 to type @code{union foo}
1929 would store it into the union as the integer @code{i}, since it is
1930 an integer. (@xref{Cast to Union}.)
1932 You can combine this technique of naming elements with ordinary C
1933 initialization of successive elements. Each initializer element that
1934 does not have a designator applies to the next consecutive element of the
1935 array or structure. For example,
1938 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1945 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1948 Labeling the elements of an array initializer is especially useful
1949 when the indices are characters or belong to an @code{enum} type.
1954 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1955 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1958 @cindex designator lists
1959 You can also write a series of @samp{.@var{fieldname}} and
1960 @samp{[@var{index}]} designators before an @samp{=} to specify a
1961 nested subobject to initialize; the list is taken relative to the
1962 subobject corresponding to the closest surrounding brace pair. For
1963 example, with the @samp{struct point} declaration above:
1966 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1970 If the same field is initialized multiple times, it will have value from
1971 the last initialization. If any such overridden initialization has
1972 side-effect, it is unspecified whether the side-effect happens or not.
1973 Currently, GCC will discard them and issue a warning.
1976 @section Case Ranges
1978 @cindex ranges in case statements
1980 You can specify a range of consecutive values in a single @code{case} label,
1984 case @var{low} ... @var{high}:
1988 This has the same effect as the proper number of individual @code{case}
1989 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1991 This feature is especially useful for ranges of ASCII character codes:
1997 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1998 it may be parsed wrong when you use it with integer values. For example,
2013 @section Cast to a Union Type
2014 @cindex cast to a union
2015 @cindex union, casting to a
2017 A cast to union type is similar to other casts, except that the type
2018 specified is a union type. You can specify the type either with
2019 @code{union @var{tag}} or with a typedef name. A cast to union is actually
2020 a constructor though, not a cast, and hence does not yield an lvalue like
2021 normal casts. (@xref{Compound Literals}.)
2023 The types that may be cast to the union type are those of the members
2024 of the union. Thus, given the following union and variables:
2027 union foo @{ int i; double d; @};
2033 both @code{x} and @code{y} can be cast to type @code{union foo}.
2035 Using the cast as the right-hand side of an assignment to a variable of
2036 union type is equivalent to storing in a member of the union:
2041 u = (union foo) x @equiv{} u.i = x
2042 u = (union foo) y @equiv{} u.d = y
2045 You can also use the union cast as a function argument:
2048 void hack (union foo);
2050 hack ((union foo) x);
2053 @node Mixed Declarations
2054 @section Mixed Declarations and Code
2055 @cindex mixed declarations and code
2056 @cindex declarations, mixed with code
2057 @cindex code, mixed with declarations
2059 ISO C99 and ISO C++ allow declarations and code to be freely mixed
2060 within compound statements. As an extension, GCC also allows this in
2061 C90 mode. For example, you could do:
2070 Each identifier is visible from where it is declared until the end of
2071 the enclosing block.
2073 @node Function Attributes
2074 @section Declaring Attributes of Functions
2075 @cindex function attributes
2076 @cindex declaring attributes of functions
2077 @cindex functions that never return
2078 @cindex functions that return more than once
2079 @cindex functions that have no side effects
2080 @cindex functions in arbitrary sections
2081 @cindex functions that behave like malloc
2082 @cindex @code{volatile} applied to function
2083 @cindex @code{const} applied to function
2084 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
2085 @cindex functions with non-null pointer arguments
2086 @cindex functions that are passed arguments in registers on the 386
2087 @cindex functions that pop the argument stack on the 386
2088 @cindex functions that do not pop the argument stack on the 386
2089 @cindex functions that have different compilation options on the 386
2090 @cindex functions that have different optimization options
2091 @cindex functions that are dynamically resolved
2093 In GNU C, you declare certain things about functions called in your program
2094 which help the compiler optimize function calls and check your code more
2097 The keyword @code{__attribute__} allows you to specify special
2098 attributes when making a declaration. This keyword is followed by an
2099 attribute specification inside double parentheses. The following
2100 attributes are currently defined for functions on all targets:
2101 @code{aligned}, @code{alloc_size}, @code{noreturn},
2102 @code{returns_twice}, @code{noinline}, @code{noclone},
2103 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
2104 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
2105 @code{no_instrument_function}, @code{no_split_stack},
2106 @code{section}, @code{constructor},
2107 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
2108 @code{weak}, @code{malloc}, @code{alias}, @code{ifunc},
2109 @code{warn_unused_result}, @code{nonnull}, @code{gnu_inline},
2110 @code{externally_visible}, @code{hot}, @code{cold}, @code{artificial},
2111 @code{error} and @code{warning}. Several other attributes are defined
2112 for functions on particular target systems. Other attributes,
2113 including @code{section} are supported for variables declarations
2114 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
2116 GCC plugins may provide their own attributes.
2118 You may also specify attributes with @samp{__} preceding and following
2119 each keyword. This allows you to use them in header files without
2120 being concerned about a possible macro of the same name. For example,
2121 you may use @code{__noreturn__} instead of @code{noreturn}.
2123 @xref{Attribute Syntax}, for details of the exact syntax for using
2127 @c Keep this table alphabetized by attribute name. Treat _ as space.
2129 @item alias ("@var{target}")
2130 @cindex @code{alias} attribute
2131 The @code{alias} attribute causes the declaration to be emitted as an
2132 alias for another symbol, which must be specified. For instance,
2135 void __f () @{ /* @r{Do something.} */; @}
2136 void f () __attribute__ ((weak, alias ("__f")));
2139 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
2140 mangled name for the target must be used. It is an error if @samp{__f}
2141 is not defined in the same translation unit.
2143 Not all target machines support this attribute.
2145 @item aligned (@var{alignment})
2146 @cindex @code{aligned} attribute
2147 This attribute specifies a minimum alignment for the function,
2150 You cannot use this attribute to decrease the alignment of a function,
2151 only to increase it. However, when you explicitly specify a function
2152 alignment this will override the effect of the
2153 @option{-falign-functions} (@pxref{Optimize Options}) option for this
2156 Note that the effectiveness of @code{aligned} attributes may be
2157 limited by inherent limitations in your linker. On many systems, the
2158 linker is only able to arrange for functions to be aligned up to a
2159 certain maximum alignment. (For some linkers, the maximum supported
2160 alignment may be very very small.) See your linker documentation for
2161 further information.
2163 The @code{aligned} attribute can also be used for variables and fields
2164 (@pxref{Variable Attributes}.)
2167 @cindex @code{alloc_size} attribute
2168 The @code{alloc_size} attribute is used to tell the compiler that the
2169 function return value points to memory, where the size is given by
2170 one or two of the functions parameters. GCC uses this
2171 information to improve the correctness of @code{__builtin_object_size}.
2173 The function parameter(s) denoting the allocated size are specified by
2174 one or two integer arguments supplied to the attribute. The allocated size
2175 is either the value of the single function argument specified or the product
2176 of the two function arguments specified. Argument numbering starts at
2182 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
2183 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
2186 declares that my_calloc will return memory of the size given by
2187 the product of parameter 1 and 2 and that my_realloc will return memory
2188 of the size given by parameter 2.
2191 @cindex @code{always_inline} function attribute
2192 Generally, functions are not inlined unless optimization is specified.
2193 For functions declared inline, this attribute inlines the function even
2194 if no optimization level was specified.
2197 @cindex @code{gnu_inline} function attribute
2198 This attribute should be used with a function which is also declared
2199 with the @code{inline} keyword. It directs GCC to treat the function
2200 as if it were defined in gnu90 mode even when compiling in C99 or
2203 If the function is declared @code{extern}, then this definition of the
2204 function is used only for inlining. In no case is the function
2205 compiled as a standalone function, not even if you take its address
2206 explicitly. Such an address becomes an external reference, as if you
2207 had only declared the function, and had not defined it. This has
2208 almost the effect of a macro. The way to use this is to put a
2209 function definition in a header file with this attribute, and put
2210 another copy of the function, without @code{extern}, in a library
2211 file. The definition in the header file will cause most calls to the
2212 function to be inlined. If any uses of the function remain, they will
2213 refer to the single copy in the library. Note that the two
2214 definitions of the functions need not be precisely the same, although
2215 if they do not have the same effect your program may behave oddly.
2217 In C, if the function is neither @code{extern} nor @code{static}, then
2218 the function is compiled as a standalone function, as well as being
2219 inlined where possible.
2221 This is how GCC traditionally handled functions declared
2222 @code{inline}. Since ISO C99 specifies a different semantics for
2223 @code{inline}, this function attribute is provided as a transition
2224 measure and as a useful feature in its own right. This attribute is
2225 available in GCC 4.1.3 and later. It is available if either of the
2226 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2227 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2228 Function is As Fast As a Macro}.
2230 In C++, this attribute does not depend on @code{extern} in any way,
2231 but it still requires the @code{inline} keyword to enable its special
2235 @cindex @code{artificial} function attribute
2236 This attribute is useful for small inline wrappers which if possible
2237 should appear during debugging as a unit, depending on the debug
2238 info format it will either mean marking the function as artificial
2239 or using the caller location for all instructions within the inlined
2243 @cindex interrupt handler functions
2244 When added to an interrupt handler with the M32C port, causes the
2245 prologue and epilogue to use bank switching to preserve the registers
2246 rather than saving them on the stack.
2249 @cindex @code{flatten} function attribute
2250 Generally, inlining into a function is limited. For a function marked with
2251 this attribute, every call inside this function will be inlined, if possible.
2252 Whether the function itself is considered for inlining depends on its size and
2253 the current inlining parameters.
2255 @item error ("@var{message}")
2256 @cindex @code{error} function attribute
2257 If this attribute is used on a function declaration and a call to such a function
2258 is not eliminated through dead code elimination or other optimizations, an error
2259 which will include @var{message} will be diagnosed. This is useful
2260 for compile time checking, especially together with @code{__builtin_constant_p}
2261 and inline functions where checking the inline function arguments is not
2262 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2263 While it is possible to leave the function undefined and thus invoke
2264 a link failure, when using this attribute the problem will be diagnosed
2265 earlier and with exact location of the call even in presence of inline
2266 functions or when not emitting debugging information.
2268 @item warning ("@var{message}")
2269 @cindex @code{warning} function attribute
2270 If this attribute is used on a function declaration and a call to such a function
2271 is not eliminated through dead code elimination or other optimizations, a warning
2272 which will include @var{message} will be diagnosed. This is useful
2273 for compile time checking, especially together with @code{__builtin_constant_p}
2274 and inline functions. While it is possible to define the function with
2275 a message in @code{.gnu.warning*} section, when using this attribute the problem
2276 will be diagnosed earlier and with exact location of the call even in presence
2277 of inline functions or when not emitting debugging information.
2280 @cindex functions that do pop the argument stack on the 386
2282 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2283 assume that the calling function will pop off the stack space used to
2284 pass arguments. This is
2285 useful to override the effects of the @option{-mrtd} switch.
2288 @cindex @code{const} function attribute
2289 Many functions do not examine any values except their arguments, and
2290 have no effects except the return value. Basically this is just slightly
2291 more strict class than the @code{pure} attribute below, since function is not
2292 allowed to read global memory.
2294 @cindex pointer arguments
2295 Note that a function that has pointer arguments and examines the data
2296 pointed to must @emph{not} be declared @code{const}. Likewise, a
2297 function that calls a non-@code{const} function usually must not be
2298 @code{const}. It does not make sense for a @code{const} function to
2301 The attribute @code{const} is not implemented in GCC versions earlier
2302 than 2.5. An alternative way to declare that a function has no side
2303 effects, which works in the current version and in some older versions,
2307 typedef int intfn ();
2309 extern const intfn square;
2312 This approach does not work in GNU C++ from 2.6.0 on, since the language
2313 specifies that the @samp{const} must be attached to the return value.
2317 @itemx constructor (@var{priority})
2318 @itemx destructor (@var{priority})
2319 @cindex @code{constructor} function attribute
2320 @cindex @code{destructor} function attribute
2321 The @code{constructor} attribute causes the function to be called
2322 automatically before execution enters @code{main ()}. Similarly, the
2323 @code{destructor} attribute causes the function to be called
2324 automatically after @code{main ()} has completed or @code{exit ()} has
2325 been called. Functions with these attributes are useful for
2326 initializing data that will be used implicitly during the execution of
2329 You may provide an optional integer priority to control the order in
2330 which constructor and destructor functions are run. A constructor
2331 with a smaller priority number runs before a constructor with a larger
2332 priority number; the opposite relationship holds for destructors. So,
2333 if you have a constructor that allocates a resource and a destructor
2334 that deallocates the same resource, both functions typically have the
2335 same priority. The priorities for constructor and destructor
2336 functions are the same as those specified for namespace-scope C++
2337 objects (@pxref{C++ Attributes}).
2339 These attributes are not currently implemented for Objective-C@.
2342 @itemx deprecated (@var{msg})
2343 @cindex @code{deprecated} attribute.
2344 The @code{deprecated} attribute results in a warning if the function
2345 is used anywhere in the source file. This is useful when identifying
2346 functions that are expected to be removed in a future version of a
2347 program. The warning also includes the location of the declaration
2348 of the deprecated function, to enable users to easily find further
2349 information about why the function is deprecated, or what they should
2350 do instead. Note that the warnings only occurs for uses:
2353 int old_fn () __attribute__ ((deprecated));
2355 int (*fn_ptr)() = old_fn;
2358 results in a warning on line 3 but not line 2. The optional msg
2359 argument, which must be a string, will be printed in the warning if
2362 The @code{deprecated} attribute can also be used for variables and
2363 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2366 @cindex @code{disinterrupt} attribute
2367 On Epiphany and MeP targets, this attribute causes the compiler to emit
2368 instructions to disable interrupts for the duration of the given
2372 @cindex @code{__declspec(dllexport)}
2373 On Microsoft Windows targets and Symbian OS targets the
2374 @code{dllexport} attribute causes the compiler to provide a global
2375 pointer to a pointer in a DLL, so that it can be referenced with the
2376 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2377 name is formed by combining @code{_imp__} and the function or variable
2380 You can use @code{__declspec(dllexport)} as a synonym for
2381 @code{__attribute__ ((dllexport))} for compatibility with other
2384 On systems that support the @code{visibility} attribute, this
2385 attribute also implies ``default'' visibility. It is an error to
2386 explicitly specify any other visibility.
2388 In previous versions of GCC, the @code{dllexport} attribute was ignored
2389 for inlined functions, unless the @option{-fkeep-inline-functions} flag
2390 had been used. The default behaviour now is to emit all dllexported
2391 inline functions; however, this can cause object file-size bloat, in
2392 which case the old behaviour can be restored by using
2393 @option{-fno-keep-inline-dllexport}.
2395 The attribute is also ignored for undefined symbols.
2397 When applied to C++ classes, the attribute marks defined non-inlined
2398 member functions and static data members as exports. Static consts
2399 initialized in-class are not marked unless they are also defined
2402 For Microsoft Windows targets there are alternative methods for
2403 including the symbol in the DLL's export table such as using a
2404 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2405 the @option{--export-all} linker flag.
2408 @cindex @code{__declspec(dllimport)}
2409 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2410 attribute causes the compiler to reference a function or variable via
2411 a global pointer to a pointer that is set up by the DLL exporting the
2412 symbol. The attribute implies @code{extern}. On Microsoft Windows
2413 targets, the pointer name is formed by combining @code{_imp__} and the
2414 function or variable name.
2416 You can use @code{__declspec(dllimport)} as a synonym for
2417 @code{__attribute__ ((dllimport))} for compatibility with other
2420 On systems that support the @code{visibility} attribute, this
2421 attribute also implies ``default'' visibility. It is an error to
2422 explicitly specify any other visibility.
2424 Currently, the attribute is ignored for inlined functions. If the
2425 attribute is applied to a symbol @emph{definition}, an error is reported.
2426 If a symbol previously declared @code{dllimport} is later defined, the
2427 attribute is ignored in subsequent references, and a warning is emitted.
2428 The attribute is also overridden by a subsequent declaration as
2431 When applied to C++ classes, the attribute marks non-inlined
2432 member functions and static data members as imports. However, the
2433 attribute is ignored for virtual methods to allow creation of vtables
2436 On the SH Symbian OS target the @code{dllimport} attribute also has
2437 another affect---it can cause the vtable and run-time type information
2438 for a class to be exported. This happens when the class has a
2439 dllimport'ed constructor or a non-inline, non-pure virtual function
2440 and, for either of those two conditions, the class also has an inline
2441 constructor or destructor and has a key function that is defined in
2442 the current translation unit.
2444 For Microsoft Windows based targets the use of the @code{dllimport}
2445 attribute on functions is not necessary, but provides a small
2446 performance benefit by eliminating a thunk in the DLL@. The use of the
2447 @code{dllimport} attribute on imported variables was required on older
2448 versions of the GNU linker, but can now be avoided by passing the
2449 @option{--enable-auto-import} switch to the GNU linker. As with
2450 functions, using the attribute for a variable eliminates a thunk in
2453 One drawback to using this attribute is that a pointer to a
2454 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2455 address. However, a pointer to a @emph{function} with the
2456 @code{dllimport} attribute can be used as a constant initializer; in
2457 this case, the address of a stub function in the import lib is
2458 referenced. On Microsoft Windows targets, the attribute can be disabled
2459 for functions by setting the @option{-mnop-fun-dllimport} flag.
2462 @cindex eight bit data on the H8/300, H8/300H, and H8S
2463 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2464 variable should be placed into the eight bit data section.
2465 The compiler will generate more efficient code for certain operations
2466 on data in the eight bit data area. Note the eight bit data area is limited to
2469 You must use GAS and GLD from GNU binutils version 2.7 or later for
2470 this attribute to work correctly.
2472 @item exception_handler
2473 @cindex exception handler functions on the Blackfin processor
2474 Use this attribute on the Blackfin to indicate that the specified function
2475 is an exception handler. The compiler will generate function entry and
2476 exit sequences suitable for use in an exception handler when this
2477 attribute is present.
2479 @item externally_visible
2480 @cindex @code{externally_visible} attribute.
2481 This attribute, attached to a global variable or function, nullifies
2482 the effect of the @option{-fwhole-program} command-line option, so the
2483 object remains visible outside the current compilation unit. If @option{-fwhole-program} is used together with @option{-flto} and @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
2486 @cindex functions which handle memory bank switching
2487 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2488 use a calling convention that takes care of switching memory banks when
2489 entering and leaving a function. This calling convention is also the
2490 default when using the @option{-mlong-calls} option.
2492 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2493 to call and return from a function.
2495 On 68HC11 the compiler will generate a sequence of instructions
2496 to invoke a board-specific routine to switch the memory bank and call the
2497 real function. The board-specific routine simulates a @code{call}.
2498 At the end of a function, it will jump to a board-specific routine
2499 instead of using @code{rts}. The board-specific return routine simulates
2502 On MeP targets this causes the compiler to use a calling convention
2503 which assumes the called function is too far away for the built-in
2506 @item fast_interrupt
2507 @cindex interrupt handler functions
2508 Use this attribute on the M32C and RX ports to indicate that the specified
2509 function is a fast interrupt handler. This is just like the
2510 @code{interrupt} attribute, except that @code{freit} is used to return
2511 instead of @code{reit}.
2514 @cindex functions that pop the argument stack on the 386
2515 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2516 pass the first argument (if of integral type) in the register ECX and
2517 the second argument (if of integral type) in the register EDX@. Subsequent
2518 and other typed arguments are passed on the stack. The called function will
2519 pop the arguments off the stack. If the number of arguments is variable all
2520 arguments are pushed on the stack.
2523 @cindex functions that pop the argument stack on the 386
2524 On the Intel 386, the @code{thiscall} attribute causes the compiler to
2525 pass the first argument (if of integral type) in the register ECX.
2526 Subsequent and other typed arguments are passed on the stack. The called
2527 function will pop the arguments off the stack.
2528 If the number of arguments is variable all arguments are pushed on the
2530 The @code{thiscall} attribute is intended for C++ non-static member functions.
2531 As gcc extension this calling convention can be used for C-functions
2532 and for static member methods.
2534 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2535 @cindex @code{format} function attribute
2537 The @code{format} attribute specifies that a function takes @code{printf},
2538 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2539 should be type-checked against a format string. For example, the
2544 my_printf (void *my_object, const char *my_format, ...)
2545 __attribute__ ((format (printf, 2, 3)));
2549 causes the compiler to check the arguments in calls to @code{my_printf}
2550 for consistency with the @code{printf} style format string argument
2553 The parameter @var{archetype} determines how the format string is
2554 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2555 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2556 @code{strfmon}. (You can also use @code{__printf__},
2557 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2558 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2559 @code{ms_strftime} are also present.
2560 @var{archtype} values such as @code{printf} refer to the formats accepted
2561 by the system's C run-time library, while @code{gnu_} values always refer
2562 to the formats accepted by the GNU C Library. On Microsoft Windows
2563 targets, @code{ms_} values refer to the formats accepted by the
2564 @file{msvcrt.dll} library.
2565 The parameter @var{string-index}
2566 specifies which argument is the format string argument (starting
2567 from 1), while @var{first-to-check} is the number of the first
2568 argument to check against the format string. For functions
2569 where the arguments are not available to be checked (such as
2570 @code{vprintf}), specify the third parameter as zero. In this case the
2571 compiler only checks the format string for consistency. For
2572 @code{strftime} formats, the third parameter is required to be zero.
2573 Since non-static C++ methods have an implicit @code{this} argument, the
2574 arguments of such methods should be counted from two, not one, when
2575 giving values for @var{string-index} and @var{first-to-check}.
2577 In the example above, the format string (@code{my_format}) is the second
2578 argument of the function @code{my_print}, and the arguments to check
2579 start with the third argument, so the correct parameters for the format
2580 attribute are 2 and 3.
2582 @opindex ffreestanding
2583 @opindex fno-builtin
2584 The @code{format} attribute allows you to identify your own functions
2585 which take format strings as arguments, so that GCC can check the
2586 calls to these functions for errors. The compiler always (unless
2587 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2588 for the standard library functions @code{printf}, @code{fprintf},
2589 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2590 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2591 warnings are requested (using @option{-Wformat}), so there is no need to
2592 modify the header file @file{stdio.h}. In C99 mode, the functions
2593 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2594 @code{vsscanf} are also checked. Except in strictly conforming C
2595 standard modes, the X/Open function @code{strfmon} is also checked as
2596 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2597 @xref{C Dialect Options,,Options Controlling C Dialect}.
2599 For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
2600 recognized in the same context. Declarations including these format attributes
2601 will be parsed for correct syntax, however the result of checking of such format
2602 strings is not yet defined, and will not be carried out by this version of the
2605 The target may also provide additional types of format checks.
2606 @xref{Target Format Checks,,Format Checks Specific to Particular
2609 @item format_arg (@var{string-index})
2610 @cindex @code{format_arg} function attribute
2611 @opindex Wformat-nonliteral
2612 The @code{format_arg} attribute specifies that a function takes a format
2613 string for a @code{printf}, @code{scanf}, @code{strftime} or
2614 @code{strfmon} style function and modifies it (for example, to translate
2615 it into another language), so the result can be passed to a
2616 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2617 function (with the remaining arguments to the format function the same
2618 as they would have been for the unmodified string). For example, the
2623 my_dgettext (char *my_domain, const char *my_format)
2624 __attribute__ ((format_arg (2)));
2628 causes the compiler to check the arguments in calls to a @code{printf},
2629 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2630 format string argument is a call to the @code{my_dgettext} function, for
2631 consistency with the format string argument @code{my_format}. If the
2632 @code{format_arg} attribute had not been specified, all the compiler
2633 could tell in such calls to format functions would be that the format
2634 string argument is not constant; this would generate a warning when
2635 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2636 without the attribute.
2638 The parameter @var{string-index} specifies which argument is the format
2639 string argument (starting from one). Since non-static C++ methods have
2640 an implicit @code{this} argument, the arguments of such methods should
2641 be counted from two.
2643 The @code{format-arg} attribute allows you to identify your own
2644 functions which modify format strings, so that GCC can check the
2645 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2646 type function whose operands are a call to one of your own function.
2647 The compiler always treats @code{gettext}, @code{dgettext}, and
2648 @code{dcgettext} in this manner except when strict ISO C support is
2649 requested by @option{-ansi} or an appropriate @option{-std} option, or
2650 @option{-ffreestanding} or @option{-fno-builtin}
2651 is used. @xref{C Dialect Options,,Options
2652 Controlling C Dialect}.
2654 For Objective-C dialects, the @code{format-arg} attribute may refer to an
2655 @code{NSString} reference for compatibility with the @code{format} attribute
2658 The target may also allow additional types in @code{format-arg} attributes.
2659 @xref{Target Format Checks,,Format Checks Specific to Particular
2662 @item function_vector
2663 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2664 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2665 function should be called through the function vector. Calling a
2666 function through the function vector will reduce code size, however;
2667 the function vector has a limited size (maximum 128 entries on the H8/300
2668 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2670 In SH2A target, this attribute declares a function to be called using the
2671 TBR relative addressing mode. The argument to this attribute is the entry
2672 number of the same function in a vector table containing all the TBR
2673 relative addressable functions. For the successful jump, register TBR
2674 should contain the start address of this TBR relative vector table.
2675 In the startup routine of the user application, user needs to care of this
2676 TBR register initialization. The TBR relative vector table can have at
2677 max 256 function entries. The jumps to these functions will be generated
2678 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2679 You must use GAS and GLD from GNU binutils version 2.7 or later for
2680 this attribute to work correctly.
2682 Please refer the example of M16C target, to see the use of this
2683 attribute while declaring a function,
2685 In an application, for a function being called once, this attribute will
2686 save at least 8 bytes of code; and if other successive calls are being
2687 made to the same function, it will save 2 bytes of code per each of these
2690 On M16C/M32C targets, the @code{function_vector} attribute declares a
2691 special page subroutine call function. Use of this attribute reduces
2692 the code size by 2 bytes for each call generated to the
2693 subroutine. The argument to the attribute is the vector number entry
2694 from the special page vector table which contains the 16 low-order
2695 bits of the subroutine's entry address. Each vector table has special
2696 page number (18 to 255) which are used in @code{jsrs} instruction.
2697 Jump addresses of the routines are generated by adding 0x0F0000 (in
2698 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2699 byte addresses set in the vector table. Therefore you need to ensure
2700 that all the special page vector routines should get mapped within the
2701 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2704 In the following example 2 bytes will be saved for each call to
2705 function @code{foo}.
2708 void foo (void) __attribute__((function_vector(0x18)));
2719 If functions are defined in one file and are called in another file,
2720 then be sure to write this declaration in both files.
2722 This attribute is ignored for R8C target.
2725 @cindex interrupt handler functions
2726 Use this attribute on the ARM, AVR, CR16, Epiphany, M32C, M32R/D, m68k, MeP, MIPS,
2727 RL78, RX and Xstormy16 ports to indicate that the specified function is an
2728 interrupt handler. The compiler will generate function entry and exit
2729 sequences suitable for use in an interrupt handler when this attribute
2732 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, MicroBlaze,
2733 and SH processors can be specified via the @code{interrupt_handler} attribute.
2735 Note, on the AVR, interrupts will be enabled inside the function.
2737 Note, for the ARM, you can specify the kind of interrupt to be handled by
2738 adding an optional parameter to the interrupt attribute like this:
2741 void f () __attribute__ ((interrupt ("IRQ")));
2744 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2746 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2747 may be called with a word aligned stack pointer.
2749 On MIPS targets, you can use the following attributes to modify the behavior
2750 of an interrupt handler:
2752 @item use_shadow_register_set
2753 @cindex @code{use_shadow_register_set} attribute
2754 Assume that the handler uses a shadow register set, instead of
2755 the main general-purpose registers.
2757 @item keep_interrupts_masked
2758 @cindex @code{keep_interrupts_masked} attribute
2759 Keep interrupts masked for the whole function. Without this attribute,
2760 GCC tries to reenable interrupts for as much of the function as it can.
2762 @item use_debug_exception_return
2763 @cindex @code{use_debug_exception_return} attribute
2764 Return using the @code{deret} instruction. Interrupt handlers that don't
2765 have this attribute return using @code{eret} instead.
2768 You can use any combination of these attributes, as shown below:
2770 void __attribute__ ((interrupt)) v0 ();
2771 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
2772 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
2773 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
2774 void __attribute__ ((interrupt, use_shadow_register_set,
2775 keep_interrupts_masked)) v4 ();
2776 void __attribute__ ((interrupt, use_shadow_register_set,
2777 use_debug_exception_return)) v5 ();
2778 void __attribute__ ((interrupt, keep_interrupts_masked,
2779 use_debug_exception_return)) v6 ();
2780 void __attribute__ ((interrupt, use_shadow_register_set,
2781 keep_interrupts_masked,
2782 use_debug_exception_return)) v7 ();
2785 On RL78, use @code{brk_interrupt} instead of @code{interrupt} for
2786 handlers intended to be used with the @code{BRK} opcode (i.e. those
2787 that must end with @code{RETB} instead of @code{RETI}).
2789 @item ifunc ("@var{resolver}")
2790 @cindex @code{ifunc} attribute
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 can 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's declaration is be a
2801 function returning pointer to void function returning void:
2804 void *my_memcpy (void *dst, const void *src, size_t len)
2809 static void (*resolve_memcpy (void)) (void)
2811 return my_memcpy; // we'll just always select this routine
2815 The exported header file declaring the function the user calls would
2819 extern void *memcpy (void *, const void *, size_t);
2822 allowing the user to call this as a regular function, unaware of the
2823 implementation. Finally, the indirect function needs to be defined in
2824 the same translation unit as the resolver function:
2827 void *memcpy (void *, const void *, size_t)
2828 __attribute__ ((ifunc ("resolve_memcpy")));
2831 Indirect functions cannot be weak, and require a recent binutils (at
2832 least version 2.20.1), and GNU C library (at least version 2.11.1).
2834 @item interrupt_handler
2835 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2836 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2837 indicate that the specified function is an interrupt handler. The compiler
2838 will generate function entry and exit sequences suitable for use in an
2839 interrupt handler when this attribute is present.
2841 @item interrupt_thread
2842 @cindex interrupt thread functions on fido
2843 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2844 that the specified function is an interrupt handler that is designed
2845 to run as a thread. The compiler omits generate prologue/epilogue
2846 sequences and replaces the return instruction with a @code{sleep}
2847 instruction. This attribute is available only on fido.
2850 @cindex interrupt service routines on ARM
2851 Use this attribute on ARM to write Interrupt Service Routines. This is an
2852 alias to the @code{interrupt} attribute above.
2855 @cindex User stack pointer in interrupts on the Blackfin
2856 When used together with @code{interrupt_handler}, @code{exception_handler}
2857 or @code{nmi_handler}, code will be generated to load the stack pointer
2858 from the USP register in the function prologue.
2861 @cindex @code{l1_text} function attribute
2862 This attribute specifies a function to be placed into L1 Instruction
2863 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2864 With @option{-mfdpic}, function calls with a such function as the callee
2865 or caller will use inlined PLT.
2868 @cindex @code{l2} function attribute
2869 On the Blackfin, this attribute specifies a function to be placed into L2
2870 SRAM. The function will be put into a specific section named
2871 @code{.l1.text}. With @option{-mfdpic}, callers of such functions will use
2875 @cindex @code{leaf} function attribute
2876 Calls to external functions with this attribute must return to the current
2877 compilation unit only by return or by exception handling. In particular, leaf
2878 functions are not allowed to call callback function passed to it from the current
2879 compilation unit or directly call functions exported by the unit or longjmp
2880 into the unit. Leaf function might still call functions from other compilation
2881 units and thus they are not necessarily leaf in the sense that they contain no
2882 function calls at all.
2884 The attribute is intended for library functions to improve dataflow analysis.
2885 The compiler takes the hint that any data not escaping the current compilation unit can
2886 not be used or modified by the leaf function. For example, the @code{sin} function
2887 is a leaf function, but @code{qsort} is not.
2889 Note that leaf functions might invoke signals and signal handlers might be
2890 defined in the current compilation unit and use static variables. The only
2891 compliant way to write such a signal handler is to declare such variables
2894 The attribute has no effect on functions defined within the current compilation
2895 unit. This is to allow easy merging of multiple compilation units into one,
2896 for example, by using the link time optimization. For this reason the
2897 attribute is not allowed on types to annotate indirect calls.
2899 @item long_call/short_call
2900 @cindex indirect calls on ARM
2901 This attribute specifies how a particular function is called on
2902 ARM and Epiphany. Both attributes override the
2903 @option{-mlong-calls} (@pxref{ARM Options})
2904 command-line switch and @code{#pragma long_calls} settings. The
2905 @code{long_call} attribute indicates that the function might be far
2906 away from the call site and require a different (more expensive)
2907 calling sequence. The @code{short_call} attribute always places
2908 the offset to the function from the call site into the @samp{BL}
2909 instruction directly.
2911 @item longcall/shortcall
2912 @cindex functions called via pointer on the RS/6000 and PowerPC
2913 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2914 indicates that the function might be far away from the call site and
2915 require a different (more expensive) calling sequence. The
2916 @code{shortcall} attribute indicates that the function is always close
2917 enough for the shorter calling sequence to be used. These attributes
2918 override both the @option{-mlongcall} switch and, on the RS/6000 and
2919 PowerPC, the @code{#pragma longcall} setting.
2921 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2922 calls are necessary.
2924 @item long_call/near/far
2925 @cindex indirect calls on MIPS
2926 These attributes specify how a particular function is called on MIPS@.
2927 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2928 command-line switch. The @code{long_call} and @code{far} attributes are
2929 synonyms, and cause the compiler to always call
2930 the function by first loading its address into a register, and then using
2931 the contents of that register. The @code{near} attribute has the opposite
2932 effect; it specifies that non-PIC calls should be made using the more
2933 efficient @code{jal} instruction.
2936 @cindex @code{malloc} attribute
2937 The @code{malloc} attribute is used to tell the compiler that a function
2938 may be treated as if any non-@code{NULL} pointer it returns cannot
2939 alias any other pointer valid when the function returns and that the memory
2940 has undefined content.
2941 This will often improve optimization.
2942 Standard functions with this property include @code{malloc} and
2943 @code{calloc}. @code{realloc}-like functions do not have this
2944 property as the memory pointed to does not have undefined content.
2946 @item mips16/nomips16
2947 @cindex @code{mips16} attribute
2948 @cindex @code{nomips16} attribute
2950 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2951 function attributes to locally select or turn off MIPS16 code generation.
2952 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2953 while MIPS16 code generation is disabled for functions with the
2954 @code{nomips16} attribute. These attributes override the
2955 @option{-mips16} and @option{-mno-mips16} options on the command line
2956 (@pxref{MIPS Options}).
2958 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2959 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2960 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2961 may interact badly with some GCC extensions such as @code{__builtin_apply}
2962 (@pxref{Constructing Calls}).
2964 @item model (@var{model-name})
2965 @cindex function addressability on the M32R/D
2966 @cindex variable addressability on the IA-64
2968 On the M32R/D, use this attribute to set the addressability of an
2969 object, and of the code generated for a function. The identifier
2970 @var{model-name} is one of @code{small}, @code{medium}, or
2971 @code{large}, representing each of the code models.
2973 Small model objects live in the lower 16MB of memory (so that their
2974 addresses can be loaded with the @code{ld24} instruction), and are
2975 callable with the @code{bl} instruction.
2977 Medium model objects may live anywhere in the 32-bit address space (the
2978 compiler will generate @code{seth/add3} instructions to load their addresses),
2979 and are callable with the @code{bl} instruction.
2981 Large model objects may live anywhere in the 32-bit address space (the
2982 compiler will generate @code{seth/add3} instructions to load their addresses),
2983 and may not be reachable with the @code{bl} instruction (the compiler will
2984 generate the much slower @code{seth/add3/jl} instruction sequence).
2986 On IA-64, use this attribute to set the addressability of an object.
2987 At present, the only supported identifier for @var{model-name} is
2988 @code{small}, indicating addressability via ``small'' (22-bit)
2989 addresses (so that their addresses can be loaded with the @code{addl}
2990 instruction). Caveat: such addressing is by definition not position
2991 independent and hence this attribute must not be used for objects
2992 defined by shared libraries.
2994 @item ms_abi/sysv_abi
2995 @cindex @code{ms_abi} attribute
2996 @cindex @code{sysv_abi} attribute
2998 On 32-bit and 64-bit (i?86|x86_64)-*-* targets, you can use an ABI attribute
2999 to indicate which calling convention should be used for a function. The
3000 @code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
3001 while the @code{sysv_abi} attribute tells the compiler to use the ABI
3002 used on GNU/Linux and other systems. The default is to use the Microsoft ABI
3003 when targeting Windows. On all other systems, the default is the x86/AMD ABI.
3005 Note, the @code{ms_abi} attribute for Windows 64-bit targets currently
3006 requires the @option{-maccumulate-outgoing-args} option.
3008 @item callee_pop_aggregate_return (@var{number})
3009 @cindex @code{callee_pop_aggregate_return} attribute
3011 On 32-bit i?86-*-* targets, you can control by those attribute for
3012 aggregate return in memory, if the caller is responsible to pop the hidden
3013 pointer together with the rest of the arguments - @var{number} equal to
3014 zero -, or if the callee is responsible to pop hidden pointer - @var{number}
3015 equal to one. The default i386 ABI assumes that the callee pops the
3016 stack for hidden pointer.
3018 Note, that on 32-bit i386 Windows targets the compiler assumes that the
3019 caller pops the stack for hidden pointer.
3021 @item ms_hook_prologue
3022 @cindex @code{ms_hook_prologue} attribute
3024 On 32 bit i[34567]86-*-* targets and 64 bit x86_64-*-* targets, you can use
3025 this function attribute to make gcc generate the "hot-patching" function
3026 prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
3030 @cindex function without a prologue/epilogue code
3031 Use this attribute on the ARM, AVR, MCORE, RX and SPU ports to indicate that
3032 the specified function does not need prologue/epilogue sequences generated by
3033 the compiler. It is up to the programmer to provide these sequences. The
3034 only statements that can be safely included in naked functions are
3035 @code{asm} statements that do not have operands. All other statements,
3036 including declarations of local variables, @code{if} statements, and so
3037 forth, should be avoided. Naked functions should be used to implement the
3038 body of an assembly function, while allowing the compiler to construct
3039 the requisite function declaration for the assembler.
3042 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
3043 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
3044 use the normal calling convention based on @code{jsr} and @code{rts}.
3045 This attribute can be used to cancel the effect of the @option{-mlong-calls}
3048 On MeP targets this attribute causes the compiler to assume the called
3049 function is close enough to use the normal calling convention,
3050 overriding the @code{-mtf} command line option.
3053 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
3054 Use this attribute together with @code{interrupt_handler},
3055 @code{exception_handler} or @code{nmi_handler} to indicate that the function
3056 entry code should enable nested interrupts or exceptions.
3059 @cindex NMI handler functions on the Blackfin processor
3060 Use this attribute on the Blackfin to indicate that the specified function
3061 is an NMI handler. The compiler will generate function entry and
3062 exit sequences suitable for use in an NMI handler when this
3063 attribute is present.
3065 @item no_instrument_function
3066 @cindex @code{no_instrument_function} function attribute
3067 @opindex finstrument-functions
3068 If @option{-finstrument-functions} is given, profiling function calls will
3069 be generated at entry and exit of most user-compiled functions.
3070 Functions with this attribute will not be so instrumented.
3072 @item no_split_stack
3073 @cindex @code{no_split_stack} function attribute
3074 @opindex fsplit-stack
3075 If @option{-fsplit-stack} is given, functions will have a small
3076 prologue which decides whether to split the stack. Functions with the
3077 @code{no_split_stack} attribute will not have that prologue, and thus
3078 may run with only a small amount of stack space available.
3081 @cindex @code{noinline} function attribute
3082 This function attribute prevents a function from being considered for
3084 @c Don't enumerate the optimizations by name here; we try to be
3085 @c future-compatible with this mechanism.
3086 If the function does not have side-effects, there are optimizations
3087 other than inlining that causes function calls to be optimized away,
3088 although the function call is live. To keep such calls from being
3093 (@pxref{Extended Asm}) in the called function, to serve as a special
3097 @cindex @code{noclone} function attribute
3098 This function attribute prevents a function from being considered for
3099 cloning - a mechanism which produces specialized copies of functions
3100 and which is (currently) performed by interprocedural constant
3103 @item nonnull (@var{arg-index}, @dots{})
3104 @cindex @code{nonnull} function attribute
3105 The @code{nonnull} attribute specifies that some function parameters should
3106 be non-null pointers. For instance, the declaration:
3110 my_memcpy (void *dest, const void *src, size_t len)
3111 __attribute__((nonnull (1, 2)));
3115 causes the compiler to check that, in calls to @code{my_memcpy},
3116 arguments @var{dest} and @var{src} are non-null. If the compiler
3117 determines that a null pointer is passed in an argument slot marked
3118 as non-null, and the @option{-Wnonnull} option is enabled, a warning
3119 is issued. The compiler may also choose to make optimizations based
3120 on the knowledge that certain function arguments will not be null.
3122 If no argument index list is given to the @code{nonnull} attribute,
3123 all pointer arguments are marked as non-null. To illustrate, the
3124 following declaration is equivalent to the previous example:
3128 my_memcpy (void *dest, const void *src, size_t len)
3129 __attribute__((nonnull));
3133 @cindex @code{noreturn} function attribute
3134 A few standard library functions, such as @code{abort} and @code{exit},
3135 cannot return. GCC knows this automatically. Some programs define
3136 their own functions that never return. You can declare them
3137 @code{noreturn} to tell the compiler this fact. For example,
3141 void fatal () __attribute__ ((noreturn));
3144 fatal (/* @r{@dots{}} */)
3146 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
3152 The @code{noreturn} keyword tells the compiler to assume that
3153 @code{fatal} cannot return. It can then optimize without regard to what
3154 would happen if @code{fatal} ever did return. This makes slightly
3155 better code. More importantly, it helps avoid spurious warnings of
3156 uninitialized variables.
3158 The @code{noreturn} keyword does not affect the exceptional path when that
3159 applies: a @code{noreturn}-marked function may still return to the caller
3160 by throwing an exception or calling @code{longjmp}.
3162 Do not assume that registers saved by the calling function are
3163 restored before calling the @code{noreturn} function.
3165 It does not make sense for a @code{noreturn} function to have a return
3166 type other than @code{void}.
3168 The attribute @code{noreturn} is not implemented in GCC versions
3169 earlier than 2.5. An alternative way to declare that a function does
3170 not return, which works in the current version and in some older
3171 versions, is as follows:
3174 typedef void voidfn ();
3176 volatile voidfn fatal;
3179 This approach does not work in GNU C++.
3182 @cindex @code{nothrow} function attribute
3183 The @code{nothrow} attribute is used to inform the compiler that a
3184 function cannot throw an exception. For example, most functions in
3185 the standard C library can be guaranteed not to throw an exception
3186 with the notable exceptions of @code{qsort} and @code{bsearch} that
3187 take function pointer arguments. The @code{nothrow} attribute is not
3188 implemented in GCC versions earlier than 3.3.
3191 @cindex @code{optimize} function attribute
3192 The @code{optimize} attribute is used to specify that a function is to
3193 be compiled with different optimization options than specified on the
3194 command line. Arguments can either be numbers or strings. Numbers
3195 are assumed to be an optimization level. Strings that begin with
3196 @code{O} are assumed to be an optimization option, while other options
3197 are assumed to be used with a @code{-f} prefix. You can also use the
3198 @samp{#pragma GCC optimize} pragma to set the optimization options
3199 that affect more than one function.
3200 @xref{Function Specific Option Pragmas}, for details about the
3201 @samp{#pragma GCC optimize} pragma.
3203 This can be used for instance to have frequently executed functions
3204 compiled with more aggressive optimization options that produce faster
3205 and larger code, while other functions can be called with less
3208 @item OS_main/OS_task
3209 @cindex @code{OS_main} AVR function attribute
3210 @cindex @code{OS_task} AVR function attribute
3211 On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
3212 do not save/restore any call-saved register in their prologue/epilogue.
3214 The @code{OS_main} attribute can be used when there @emph{is
3215 guarantee} that interrupts are disabled at the time when the function
3216 is entered. This will save resources when the stack pointer has to be
3217 changed to set up a frame for local variables.
3219 The @code{OS_task} attribute can be used when there is @emph{no
3220 guarantee} that interrupts are disabled at that time when the function
3221 is entered like for, e@.g@. task functions in a multi-threading operating
3222 system. In that case, changing the stack pointer register will be
3223 guarded by save/clear/restore of the global interrupt enable flag.
3225 The differences to the @code{naked} function attribute are:
3227 @item @code{naked} functions do not have a return instruction whereas
3228 @code{OS_main} and @code{OS_task} functions will have a @code{RET} or
3229 @code{RETI} return instruction.
3230 @item @code{naked} functions do not set up a frame for local variables
3231 or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
3236 @cindex @code{pcs} function attribute
3238 The @code{pcs} attribute can be used to control the calling convention
3239 used for a function on ARM. The attribute takes an argument that specifies
3240 the calling convention to use.
3242 When compiling using the AAPCS ABI (or a variant of that) then valid
3243 values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}. In
3244 order to use a variant other than @code{"aapcs"} then the compiler must
3245 be permitted to use the appropriate co-processor registers (i.e., the
3246 VFP registers must be available in order to use @code{"aapcs-vfp"}).
3250 /* Argument passed in r0, and result returned in r0+r1. */
3251 double f2d (float) __attribute__((pcs("aapcs")));
3254 Variadic functions always use the @code{"aapcs"} calling convention and
3255 the compiler will reject attempts to specify an alternative.
3258 @cindex @code{pure} function attribute
3259 Many functions have no effects except the return value and their
3260 return value depends only on the parameters and/or global variables.
3261 Such a function can be subject
3262 to common subexpression elimination and loop optimization just as an
3263 arithmetic operator would be. These functions should be declared
3264 with the attribute @code{pure}. For example,
3267 int square (int) __attribute__ ((pure));
3271 says that the hypothetical function @code{square} is safe to call
3272 fewer times than the program says.
3274 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
3275 Interesting non-pure functions are functions with infinite loops or those
3276 depending on volatile memory or other system resource, that may change between
3277 two consecutive calls (such as @code{feof} in a multithreading environment).
3279 The attribute @code{pure} is not implemented in GCC versions earlier
3283 @cindex @code{hot} function attribute
3284 The @code{hot} attribute is used to inform the compiler that a function is a
3285 hot spot of the compiled program. The function is optimized more aggressively
3286 and on many target it is placed into special subsection of the text section so
3287 all hot functions appears close together improving locality.
3289 When profile feedback is available, via @option{-fprofile-use}, hot functions
3290 are automatically detected and this attribute is ignored.
3292 The @code{hot} attribute is not implemented in GCC versions earlier
3296 @cindex @code{cold} function attribute
3297 The @code{cold} attribute is used to inform the compiler that a function is
3298 unlikely executed. The function is optimized for size rather than speed and on
3299 many targets it is placed into special subsection of the text section so all
3300 cold functions appears close together improving code locality of non-cold parts
3301 of program. The paths leading to call of cold functions within code are marked
3302 as unlikely by the branch prediction mechanism. It is thus useful to mark
3303 functions used to handle unlikely conditions, such as @code{perror}, as cold to
3304 improve optimization of hot functions that do call marked functions in rare
3307 When profile feedback is available, via @option{-fprofile-use}, hot functions
3308 are automatically detected and this attribute is ignored.
3310 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
3312 @item regparm (@var{number})
3313 @cindex @code{regparm} attribute
3314 @cindex functions that are passed arguments in registers on the 386
3315 On the Intel 386, the @code{regparm} attribute causes the compiler to
3316 pass arguments number one to @var{number} if they are of integral type
3317 in registers EAX, EDX, and ECX instead of on the stack. Functions that
3318 take a variable number of arguments will continue to be passed all of their
3319 arguments on the stack.
3321 Beware that on some ELF systems this attribute is unsuitable for
3322 global functions in shared libraries with lazy binding (which is the
3323 default). Lazy binding will send the first call via resolving code in
3324 the loader, which might assume EAX, EDX and ECX can be clobbered, as
3325 per the standard calling conventions. Solaris 8 is affected by this.
3326 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
3327 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
3328 disabled with the linker or the loader if desired, to avoid the
3332 @cindex @code{sseregparm} attribute
3333 On the Intel 386 with SSE support, the @code{sseregparm} attribute
3334 causes the compiler to pass up to 3 floating point arguments in
3335 SSE registers instead of on the stack. Functions that take a
3336 variable number of arguments will continue to pass all of their
3337 floating point arguments on the stack.
3339 @item force_align_arg_pointer
3340 @cindex @code{force_align_arg_pointer} attribute
3341 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
3342 applied to individual function definitions, generating an alternate
3343 prologue and epilogue that realigns the runtime stack if necessary.
3344 This supports mixing legacy codes that run with a 4-byte aligned stack
3345 with modern codes that keep a 16-byte stack for SSE compatibility.
3348 @cindex @code{resbank} attribute
3349 On the SH2A target, this attribute enables the high-speed register
3350 saving and restoration using a register bank for @code{interrupt_handler}
3351 routines. Saving to the bank is performed automatically after the CPU
3352 accepts an interrupt that uses a register bank.
3354 The nineteen 32-bit registers comprising general register R0 to R14,
3355 control register GBR, and system registers MACH, MACL, and PR and the
3356 vector table address offset are saved into a register bank. Register
3357 banks are stacked in first-in last-out (FILO) sequence. Restoration
3358 from the bank is executed by issuing a RESBANK instruction.
3361 @cindex @code{returns_twice} attribute
3362 The @code{returns_twice} attribute tells the compiler that a function may
3363 return more than one time. The compiler will ensure that all registers
3364 are dead before calling such a function and will emit a warning about
3365 the variables that may be clobbered after the second return from the
3366 function. Examples of such functions are @code{setjmp} and @code{vfork}.
3367 The @code{longjmp}-like counterpart of such function, if any, might need
3368 to be marked with the @code{noreturn} attribute.
3371 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
3372 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
3373 all registers except the stack pointer should be saved in the prologue
3374 regardless of whether they are used or not.
3376 @item save_volatiles
3377 @cindex save volatile registers on the MicroBlaze
3378 Use this attribute on the MicroBlaze to indicate that the function is
3379 an interrupt handler. All volatile registers (in addition to non-volatile
3380 registers) will be saved in the function prologue. If the function is a leaf
3381 function, only volatiles used by the function are saved. A normal function
3382 return is generated instead of a return from interrupt.
3384 @item section ("@var{section-name}")
3385 @cindex @code{section} function attribute
3386 Normally, the compiler places the code it generates in the @code{text} section.
3387 Sometimes, however, you need additional sections, or you need certain
3388 particular functions to appear in special sections. The @code{section}
3389 attribute specifies that a function lives in a particular section.
3390 For example, the declaration:
3393 extern void foobar (void) __attribute__ ((section ("bar")));
3397 puts the function @code{foobar} in the @code{bar} section.
3399 Some file formats do not support arbitrary sections so the @code{section}
3400 attribute is not available on all platforms.
3401 If you need to map the entire contents of a module to a particular
3402 section, consider using the facilities of the linker instead.
3405 @cindex @code{sentinel} function attribute
3406 This function attribute ensures that a parameter in a function call is
3407 an explicit @code{NULL}. The attribute is only valid on variadic
3408 functions. By default, the sentinel is located at position zero, the
3409 last parameter of the function call. If an optional integer position
3410 argument P is supplied to the attribute, the sentinel must be located at
3411 position P counting backwards from the end of the argument list.
3414 __attribute__ ((sentinel))
3416 __attribute__ ((sentinel(0)))
3419 The attribute is automatically set with a position of 0 for the built-in
3420 functions @code{execl} and @code{execlp}. The built-in function
3421 @code{execle} has the attribute set with a position of 1.
3423 A valid @code{NULL} in this context is defined as zero with any pointer
3424 type. If your system defines the @code{NULL} macro with an integer type
3425 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3426 with a copy that redefines NULL appropriately.
3428 The warnings for missing or incorrect sentinels are enabled with
3432 See long_call/short_call.
3435 See longcall/shortcall.
3438 @cindex signal handler functions on the AVR processors
3439 Use this attribute on the AVR to indicate that the specified
3440 function is a signal handler. The compiler will generate function
3441 entry and exit sequences suitable for use in a signal handler when this
3442 attribute is present. Interrupts will be disabled inside the function.
3445 Use this attribute on the SH to indicate an @code{interrupt_handler}
3446 function should switch to an alternate stack. It expects a string
3447 argument that names a global variable holding the address of the
3452 void f () __attribute__ ((interrupt_handler,
3453 sp_switch ("alt_stack")));
3457 @cindex functions that pop the argument stack on the 386
3458 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3459 assume that the called function will pop off the stack space used to
3460 pass arguments, unless it takes a variable number of arguments.
3462 @item syscall_linkage
3463 @cindex @code{syscall_linkage} attribute
3464 This attribute is used to modify the IA64 calling convention by marking
3465 all input registers as live at all function exits. This makes it possible
3466 to restart a system call after an interrupt without having to save/restore
3467 the input registers. This also prevents kernel data from leaking into
3471 @cindex @code{target} function attribute
3472 The @code{target} attribute is used to specify that a function is to
3473 be compiled with different target options than specified on the
3474 command line. This can be used for instance to have functions
3475 compiled with a different ISA (instruction set architecture) than the
3476 default. You can also use the @samp{#pragma GCC target} pragma to set
3477 more than one function to be compiled with specific target options.
3478 @xref{Function Specific Option Pragmas}, for details about the
3479 @samp{#pragma GCC target} pragma.
3481 For instance on a 386, you could compile one function with
3482 @code{target("sse4.1,arch=core2")} and another with
3483 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
3484 compiling the first function with @option{-msse4.1} and
3485 @option{-march=core2} options, and the second function with
3486 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3487 user to make sure that a function is only invoked on a machine that
3488 supports the particular ISA it was compiled for (for example by using
3489 @code{cpuid} on 386 to determine what feature bits and architecture
3493 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3494 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3497 On the 386, the following options are allowed:
3502 @cindex @code{target("abm")} attribute
3503 Enable/disable the generation of the advanced bit instructions.
3507 @cindex @code{target("aes")} attribute
3508 Enable/disable the generation of the AES instructions.
3512 @cindex @code{target("mmx")} attribute
3513 Enable/disable the generation of the MMX instructions.
3517 @cindex @code{target("pclmul")} attribute
3518 Enable/disable the generation of the PCLMUL instructions.
3522 @cindex @code{target("popcnt")} attribute
3523 Enable/disable the generation of the POPCNT instruction.
3527 @cindex @code{target("sse")} attribute
3528 Enable/disable the generation of the SSE instructions.
3532 @cindex @code{target("sse2")} attribute
3533 Enable/disable the generation of the SSE2 instructions.
3537 @cindex @code{target("sse3")} attribute
3538 Enable/disable the generation of the SSE3 instructions.
3542 @cindex @code{target("sse4")} attribute
3543 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3548 @cindex @code{target("sse4.1")} attribute
3549 Enable/disable the generation of the sse4.1 instructions.
3553 @cindex @code{target("sse4.2")} attribute
3554 Enable/disable the generation of the sse4.2 instructions.
3558 @cindex @code{target("sse4a")} attribute
3559 Enable/disable the generation of the SSE4A instructions.
3563 @cindex @code{target("fma4")} attribute
3564 Enable/disable the generation of the FMA4 instructions.
3568 @cindex @code{target("xop")} attribute
3569 Enable/disable the generation of the XOP instructions.
3573 @cindex @code{target("lwp")} attribute
3574 Enable/disable the generation of the LWP instructions.
3578 @cindex @code{target("ssse3")} attribute
3579 Enable/disable the generation of the SSSE3 instructions.
3583 @cindex @code{target("cld")} attribute
3584 Enable/disable the generation of the CLD before string moves.
3586 @item fancy-math-387
3587 @itemx no-fancy-math-387
3588 @cindex @code{target("fancy-math-387")} attribute
3589 Enable/disable the generation of the @code{sin}, @code{cos}, and
3590 @code{sqrt} instructions on the 387 floating point unit.
3593 @itemx no-fused-madd
3594 @cindex @code{target("fused-madd")} attribute
3595 Enable/disable the generation of the fused multiply/add instructions.
3599 @cindex @code{target("ieee-fp")} attribute
3600 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3602 @item inline-all-stringops
3603 @itemx no-inline-all-stringops
3604 @cindex @code{target("inline-all-stringops")} attribute
3605 Enable/disable inlining of string operations.
3607 @item inline-stringops-dynamically
3608 @itemx no-inline-stringops-dynamically
3609 @cindex @code{target("inline-stringops-dynamically")} attribute
3610 Enable/disable the generation of the inline code to do small string
3611 operations and calling the library routines for large operations.
3613 @item align-stringops
3614 @itemx no-align-stringops
3615 @cindex @code{target("align-stringops")} attribute
3616 Do/do not align destination of inlined string operations.
3620 @cindex @code{target("recip")} attribute
3621 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3622 instructions followed an additional Newton-Raphson step instead of
3623 doing a floating point division.
3625 @item arch=@var{ARCH}
3626 @cindex @code{target("arch=@var{ARCH}")} attribute
3627 Specify the architecture to generate code for in compiling the function.
3629 @item tune=@var{TUNE}
3630 @cindex @code{target("tune=@var{TUNE}")} attribute
3631 Specify the architecture to tune for in compiling the function.
3633 @item fpmath=@var{FPMATH}
3634 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3635 Specify which floating point unit to use. The
3636 @code{target("fpmath=sse,387")} option must be specified as
3637 @code{target("fpmath=sse+387")} because the comma would separate
3641 On the PowerPC, the following options are allowed:
3646 @cindex @code{target("altivec")} attribute
3647 Generate code that uses (does not use) AltiVec instructions. In
3648 32-bit code, you cannot enable Altivec instructions unless
3649 @option{-mabi=altivec} was used on the command line.
3653 @cindex @code{target("cmpb")} attribute
3654 Generate code that uses (does not use) the compare bytes instruction
3655 implemented on the POWER6 processor and other processors that support
3656 the PowerPC V2.05 architecture.
3660 @cindex @code{target("dlmzb")} attribute
3661 Generate code that uses (does not use) the string-search @samp{dlmzb}
3662 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
3663 generated by default when targetting those processors.
3667 @cindex @code{target("fprnd")} attribute
3668 Generate code that uses (does not use) the FP round to integer
3669 instructions implemented on the POWER5+ processor and other processors
3670 that support the PowerPC V2.03 architecture.
3674 @cindex @code{target("hard-dfp")} attribute
3675 Generate code that uses (does not use) the decimal floating point
3676 instructions implemented on some POWER processors.
3680 @cindex @code{target("isel")} attribute
3681 Generate code that uses (does not use) ISEL instruction.
3685 @cindex @code{target("mfcrf")} attribute
3686 Generate code that uses (does not use) the move from condition
3687 register field instruction implemented on the POWER4 processor and
3688 other processors that support the PowerPC V2.01 architecture.
3692 @cindex @code{target("mfpgpr")} attribute
3693 Generate code that uses (does not use) the FP move to/from general
3694 purpose register instructions implemented on the POWER6X processor and
3695 other processors that support the extended PowerPC V2.05 architecture.
3699 @cindex @code{target("mulhw")} attribute
3700 Generate code that uses (does not use) the half-word multiply and
3701 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
3702 These instructions are generated by default when targetting those
3707 @cindex @code{target("multiple")} attribute
3708 Generate code that uses (does not use) the load multiple word
3709 instructions and the store multiple word instructions.
3713 @cindex @code{target("update")} attribute
3714 Generate code that uses (does not use) the load or store instructions
3715 that update the base register to the address of the calculated memory
3720 @cindex @code{target("popcntb")} attribute
3721 Generate code that uses (does not use) the popcount and double
3722 precision FP reciprocal estimate instruction implemented on the POWER5
3723 processor and other processors that support the PowerPC V2.02
3728 @cindex @code{target("popcntd")} attribute
3729 Generate code that uses (does not use) the popcount instruction
3730 implemented on the POWER7 processor and other processors that support
3731 the PowerPC V2.06 architecture.
3733 @item powerpc-gfxopt
3734 @itemx no-powerpc-gfxopt
3735 @cindex @code{target("powerpc-gfxopt")} attribute
3736 Generate code that uses (does not use) the optional PowerPC
3737 architecture instructions in the Graphics group, including
3738 floating-point select.
3741 @itemx no-powerpc-gpopt
3742 @cindex @code{target("powerpc-gpopt")} attribute
3743 Generate code that uses (does not use) the optional PowerPC
3744 architecture instructions in the General Purpose group, including
3745 floating-point square root.
3747 @item recip-precision
3748 @itemx no-recip-precision
3749 @cindex @code{target("recip-precision")} attribute
3750 Assume (do not assume) that the reciprocal estimate instructions
3751 provide higher precision estimates than is mandated by the powerpc
3756 @cindex @code{target("string")} attribute
3757 Generate code that uses (does not use) the load string instructions
3758 and the store string word instructions to save multiple registers and
3759 do small block moves.
3763 @cindex @code{target("vsx")} attribute
3764 Generate code that uses (does not use) vector/scalar (VSX)
3765 instructions, and also enable the use of built-in functions that allow
3766 more direct access to the VSX instruction set. In 32-bit code, you
3767 cannot enable VSX or Altivec instructions unless
3768 @option{-mabi=altivec} was used on the command line.
3772 @cindex @code{target("friz")} attribute
3773 Generate (do not generate) the @code{friz} instruction when the
3774 @option{-funsafe-math-optimizations} option is used to optimize
3775 rounding a floating point value to 64-bit integer and back to floating
3776 point. The @code{friz} instruction does not return the same value if
3777 the floating point number is too large to fit in an integer.
3779 @item avoid-indexed-addresses
3780 @itemx no-avoid-indexed-addresses
3781 @cindex @code{target("avoid-indexed-addresses")} attribute
3782 Generate code that tries to avoid (not avoid) the use of indexed load
3783 or store instructions.
3787 @cindex @code{target("paired")} attribute
3788 Generate code that uses (does not use) the generation of PAIRED simd
3793 @cindex @code{target("longcall")} attribute
3794 Generate code that assumes (does not assume) that all calls are far
3795 away so that a longer more expensive calling sequence is required.
3798 @cindex @code{target("cpu=@var{CPU}")} attribute
3799 Specify the architecture to generate code for when compiling the
3800 function. If you select the @code{target("cpu=power7")} attribute when
3801 generating 32-bit code, VSX and Altivec instructions are not generated
3802 unless you use the @option{-mabi=altivec} option on the command line.
3804 @item tune=@var{TUNE}
3805 @cindex @code{target("tune=@var{TUNE}")} attribute
3806 Specify the architecture to tune for when compiling the function. If
3807 you do not specify the @code{target("tune=@var{TUNE}")} attribute and
3808 you do specify the @code{target("cpu=@var{CPU}")} attribute,
3809 compilation will tune for the @var{CPU} architecture, and not the
3810 default tuning specified on the command line.
3813 On the 386/x86_64 and PowerPC backends, you can use either multiple
3814 strings to specify multiple options, or you can separate the option
3815 with a comma (@code{,}).
3817 On the 386/x86_64 and PowerPC backends, the inliner will not inline a
3818 function that has different target options than the caller, unless the
3819 callee has a subset of the target options of the caller. For example
3820 a function declared with @code{target("sse3")} can inline a function
3821 with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
3823 The @code{target} attribute is not implemented in GCC versions earlier
3824 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. It is
3825 not currently implemented for other backends.
3828 @cindex tiny data section on the H8/300H and H8S
3829 Use this attribute on the H8/300H and H8S to indicate that the specified
3830 variable should be placed into the tiny data section.
3831 The compiler will generate more efficient code for loads and stores
3832 on data in the tiny data section. Note the tiny data area is limited to
3833 slightly under 32kbytes of data.
3836 Use this attribute on the SH for an @code{interrupt_handler} to return using
3837 @code{trapa} instead of @code{rte}. This attribute expects an integer
3838 argument specifying the trap number to be used.
3841 @cindex @code{unused} attribute.
3842 This attribute, attached to a function, means that the function is meant
3843 to be possibly unused. GCC will not produce a warning for this
3847 @cindex @code{used} attribute.
3848 This attribute, attached to a function, means that code must be emitted
3849 for the function even if it appears that the function is not referenced.
3850 This is useful, for example, when the function is referenced only in
3853 When applied to a member function of a C++ class template, the
3854 attribute also means that the function will be instantiated if the
3855 class itself is instantiated.
3858 @cindex @code{version_id} attribute
3859 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3860 symbol to contain a version string, thus allowing for function level
3861 versioning. HP-UX system header files may use version level functioning
3862 for some system calls.
3865 extern int foo () __attribute__((version_id ("20040821")));
3868 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3870 @item visibility ("@var{visibility_type}")
3871 @cindex @code{visibility} attribute
3872 This attribute affects the linkage of the declaration to which it is attached.
3873 There are four supported @var{visibility_type} values: default,
3874 hidden, protected or internal visibility.
3877 void __attribute__ ((visibility ("protected")))
3878 f () @{ /* @r{Do something.} */; @}
3879 int i __attribute__ ((visibility ("hidden")));
3882 The possible values of @var{visibility_type} correspond to the
3883 visibility settings in the ELF gABI.
3886 @c keep this list of visibilities in alphabetical order.
3889 Default visibility is the normal case for the object file format.
3890 This value is available for the visibility attribute to override other
3891 options that may change the assumed visibility of entities.
3893 On ELF, default visibility means that the declaration is visible to other
3894 modules and, in shared libraries, means that the declared entity may be
3897 On Darwin, default visibility means that the declaration is visible to
3900 Default visibility corresponds to ``external linkage'' in the language.
3903 Hidden visibility indicates that the entity declared will have a new
3904 form of linkage, which we'll call ``hidden linkage''. Two
3905 declarations of an object with hidden linkage refer to the same object
3906 if they are in the same shared object.
3909 Internal visibility is like hidden visibility, but with additional
3910 processor specific semantics. Unless otherwise specified by the
3911 psABI, GCC defines internal visibility to mean that a function is
3912 @emph{never} called from another module. Compare this with hidden
3913 functions which, while they cannot be referenced directly by other
3914 modules, can be referenced indirectly via function pointers. By
3915 indicating that a function cannot be called from outside the module,
3916 GCC may for instance omit the load of a PIC register since it is known
3917 that the calling function loaded the correct value.
3920 Protected visibility is like default visibility except that it
3921 indicates that references within the defining module will bind to the
3922 definition in that module. That is, the declared entity cannot be
3923 overridden by another module.
3927 All visibilities are supported on many, but not all, ELF targets
3928 (supported when the assembler supports the @samp{.visibility}
3929 pseudo-op). Default visibility is supported everywhere. Hidden
3930 visibility is supported on Darwin targets.
3932 The visibility attribute should be applied only to declarations which
3933 would otherwise have external linkage. The attribute should be applied
3934 consistently, so that the same entity should not be declared with
3935 different settings of the attribute.
3937 In C++, the visibility attribute applies to types as well as functions
3938 and objects, because in C++ types have linkage. A class must not have
3939 greater visibility than its non-static data member types and bases,
3940 and class members default to the visibility of their class. Also, a
3941 declaration without explicit visibility is limited to the visibility
3944 In C++, you can mark member functions and static member variables of a
3945 class with the visibility attribute. This is useful if you know a
3946 particular method or static member variable should only be used from
3947 one shared object; then you can mark it hidden while the rest of the
3948 class has default visibility. Care must be taken to avoid breaking
3949 the One Definition Rule; for example, it is usually not useful to mark
3950 an inline method as hidden without marking the whole class as hidden.
3952 A C++ namespace declaration can also have the visibility attribute.
3953 This attribute applies only to the particular namespace body, not to
3954 other definitions of the same namespace; it is equivalent to using
3955 @samp{#pragma GCC visibility} before and after the namespace
3956 definition (@pxref{Visibility Pragmas}).
3958 In C++, if a template argument has limited visibility, this
3959 restriction is implicitly propagated to the template instantiation.
3960 Otherwise, template instantiations and specializations default to the
3961 visibility of their template.
3963 If both the template and enclosing class have explicit visibility, the
3964 visibility from the template is used.
3967 @cindex @code{vliw} attribute
3968 On MeP, the @code{vliw} attribute tells the compiler to emit
3969 instructions in VLIW mode instead of core mode. Note that this
3970 attribute is not allowed unless a VLIW coprocessor has been configured
3971 and enabled through command line options.
3973 @item warn_unused_result
3974 @cindex @code{warn_unused_result} attribute
3975 The @code{warn_unused_result} attribute causes a warning to be emitted
3976 if a caller of the function with this attribute does not use its
3977 return value. This is useful for functions where not checking
3978 the result is either a security problem or always a bug, such as
3982 int fn () __attribute__ ((warn_unused_result));
3985 if (fn () < 0) return -1;
3991 results in warning on line 5.
3994 @cindex @code{weak} attribute
3995 The @code{weak} attribute causes the declaration to be emitted as a weak
3996 symbol rather than a global. This is primarily useful in defining
3997 library functions which can be overridden in user code, though it can
3998 also be used with non-function declarations. Weak symbols are supported
3999 for ELF targets, and also for a.out targets when using the GNU assembler
4003 @itemx weakref ("@var{target}")
4004 @cindex @code{weakref} attribute
4005 The @code{weakref} attribute marks a declaration as a weak reference.
4006 Without arguments, it should be accompanied by an @code{alias} attribute
4007 naming the target symbol. Optionally, the @var{target} may be given as
4008 an argument to @code{weakref} itself. In either case, @code{weakref}
4009 implicitly marks the declaration as @code{weak}. Without a
4010 @var{target}, given as an argument to @code{weakref} or to @code{alias},
4011 @code{weakref} is equivalent to @code{weak}.
4014 static int x() __attribute__ ((weakref ("y")));
4015 /* is equivalent to... */
4016 static int x() __attribute__ ((weak, weakref, alias ("y")));
4018 static int x() __attribute__ ((weakref));
4019 static int x() __attribute__ ((alias ("y")));
4022 A weak reference is an alias that does not by itself require a
4023 definition to be given for the target symbol. If the target symbol is
4024 only referenced through weak references, then it becomes a @code{weak}
4025 undefined symbol. If it is directly referenced, however, then such
4026 strong references prevail, and a definition will be required for the
4027 symbol, not necessarily in the same translation unit.
4029 The effect is equivalent to moving all references to the alias to a
4030 separate translation unit, renaming the alias to the aliased symbol,
4031 declaring it as weak, compiling the two separate translation units and
4032 performing a reloadable link on them.
4034 At present, a declaration to which @code{weakref} is attached can
4035 only be @code{static}.
4039 You can specify multiple attributes in a declaration by separating them
4040 by commas within the double parentheses or by immediately following an
4041 attribute declaration with another attribute declaration.
4043 @cindex @code{#pragma}, reason for not using
4044 @cindex pragma, reason for not using
4045 Some people object to the @code{__attribute__} feature, suggesting that
4046 ISO C's @code{#pragma} should be used instead. At the time
4047 @code{__attribute__} was designed, there were two reasons for not doing
4052 It is impossible to generate @code{#pragma} commands from a macro.
4055 There is no telling what the same @code{#pragma} might mean in another
4059 These two reasons applied to almost any application that might have been
4060 proposed for @code{#pragma}. It was basically a mistake to use
4061 @code{#pragma} for @emph{anything}.
4063 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
4064 to be generated from macros. In addition, a @code{#pragma GCC}
4065 namespace is now in use for GCC-specific pragmas. However, it has been
4066 found convenient to use @code{__attribute__} to achieve a natural
4067 attachment of attributes to their corresponding declarations, whereas
4068 @code{#pragma GCC} is of use for constructs that do not naturally form
4069 part of the grammar. @xref{Other Directives,,Miscellaneous
4070 Preprocessing Directives, cpp, The GNU C Preprocessor}.
4072 @node Attribute Syntax
4073 @section Attribute Syntax
4074 @cindex attribute syntax
4076 This section describes the syntax with which @code{__attribute__} may be
4077 used, and the constructs to which attribute specifiers bind, for the C
4078 language. Some details may vary for C++ and Objective-C@. Because of
4079 infelicities in the grammar for attributes, some forms described here
4080 may not be successfully parsed in all cases.
4082 There are some problems with the semantics of attributes in C++. For
4083 example, there are no manglings for attributes, although they may affect
4084 code generation, so problems may arise when attributed types are used in
4085 conjunction with templates or overloading. Similarly, @code{typeid}
4086 does not distinguish between types with different attributes. Support
4087 for attributes in C++ may be restricted in future to attributes on
4088 declarations only, but not on nested declarators.
4090 @xref{Function Attributes}, for details of the semantics of attributes
4091 applying to functions. @xref{Variable Attributes}, for details of the
4092 semantics of attributes applying to variables. @xref{Type Attributes},
4093 for details of the semantics of attributes applying to structure, union
4094 and enumerated types.
4096 An @dfn{attribute specifier} is of the form
4097 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
4098 is a possibly empty comma-separated sequence of @dfn{attributes}, where
4099 each attribute is one of the following:
4103 Empty. Empty attributes are ignored.
4106 A word (which may be an identifier such as @code{unused}, or a reserved
4107 word such as @code{const}).
4110 A word, followed by, in parentheses, parameters for the attribute.
4111 These parameters take one of the following forms:
4115 An identifier. For example, @code{mode} attributes use this form.
4118 An identifier followed by a comma and a non-empty comma-separated list
4119 of expressions. For example, @code{format} attributes use this form.
4122 A possibly empty comma-separated list of expressions. For example,
4123 @code{format_arg} attributes use this form with the list being a single
4124 integer constant expression, and @code{alias} attributes use this form
4125 with the list being a single string constant.
4129 An @dfn{attribute specifier list} is a sequence of one or more attribute
4130 specifiers, not separated by any other tokens.
4132 In GNU C, an attribute specifier list may appear after the colon following a
4133 label, other than a @code{case} or @code{default} label. The only
4134 attribute it makes sense to use after a label is @code{unused}. This
4135 feature is intended for code generated by programs which contains labels
4136 that may be unused but which is compiled with @option{-Wall}. It would
4137 not normally be appropriate to use in it human-written code, though it
4138 could be useful in cases where the code that jumps to the label is
4139 contained within an @code{#ifdef} conditional. GNU C++ only permits
4140 attributes on labels if the attribute specifier is immediately
4141 followed by a semicolon (i.e., the label applies to an empty
4142 statement). If the semicolon is missing, C++ label attributes are
4143 ambiguous, as it is permissible for a declaration, which could begin
4144 with an attribute list, to be labelled in C++. Declarations cannot be
4145 labelled in C90 or C99, so the ambiguity does not arise there.
4147 An attribute specifier list may appear as part of a @code{struct},
4148 @code{union} or @code{enum} specifier. It may go either immediately
4149 after the @code{struct}, @code{union} or @code{enum} keyword, or after
4150 the closing brace. The former syntax is preferred.
4151 Where attribute specifiers follow the closing brace, they are considered
4152 to relate to the structure, union or enumerated type defined, not to any
4153 enclosing declaration the type specifier appears in, and the type
4154 defined is not complete until after the attribute specifiers.
4155 @c Otherwise, there would be the following problems: a shift/reduce
4156 @c conflict between attributes binding the struct/union/enum and
4157 @c binding to the list of specifiers/qualifiers; and "aligned"
4158 @c attributes could use sizeof for the structure, but the size could be
4159 @c changed later by "packed" attributes.
4161 Otherwise, an attribute specifier appears as part of a declaration,
4162 counting declarations of unnamed parameters and type names, and relates
4163 to that declaration (which may be nested in another declaration, for
4164 example in the case of a parameter declaration), or to a particular declarator
4165 within a declaration. Where an
4166 attribute specifier is applied to a parameter declared as a function or
4167 an array, it should apply to the function or array rather than the
4168 pointer to which the parameter is implicitly converted, but this is not
4169 yet correctly implemented.
4171 Any list of specifiers and qualifiers at the start of a declaration may
4172 contain attribute specifiers, whether or not such a list may in that
4173 context contain storage class specifiers. (Some attributes, however,
4174 are essentially in the nature of storage class specifiers, and only make
4175 sense where storage class specifiers may be used; for example,
4176 @code{section}.) There is one necessary limitation to this syntax: the
4177 first old-style parameter declaration in a function definition cannot
4178 begin with an attribute specifier, because such an attribute applies to
4179 the function instead by syntax described below (which, however, is not
4180 yet implemented in this case). In some other cases, attribute
4181 specifiers are permitted by this grammar but not yet supported by the
4182 compiler. All attribute specifiers in this place relate to the
4183 declaration as a whole. In the obsolescent usage where a type of
4184 @code{int} is implied by the absence of type specifiers, such a list of
4185 specifiers and qualifiers may be an attribute specifier list with no
4186 other specifiers or qualifiers.
4188 At present, the first parameter in a function prototype must have some
4189 type specifier which is not an attribute specifier; this resolves an
4190 ambiguity in the interpretation of @code{void f(int
4191 (__attribute__((foo)) x))}, but is subject to change. At present, if
4192 the parentheses of a function declarator contain only attributes then
4193 those attributes are ignored, rather than yielding an error or warning
4194 or implying a single parameter of type int, but this is subject to
4197 An attribute specifier list may appear immediately before a declarator
4198 (other than the first) in a comma-separated list of declarators in a
4199 declaration of more than one identifier using a single list of
4200 specifiers and qualifiers. Such attribute specifiers apply
4201 only to the identifier before whose declarator they appear. For
4205 __attribute__((noreturn)) void d0 (void),
4206 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
4211 the @code{noreturn} attribute applies to all the functions
4212 declared; the @code{format} attribute only applies to @code{d1}.
4214 An attribute specifier list may appear immediately before the comma,
4215 @code{=} or semicolon terminating the declaration of an identifier other
4216 than a function definition. Such attribute specifiers apply
4217 to the declared object or function. Where an
4218 assembler name for an object or function is specified (@pxref{Asm
4219 Labels}), the attribute must follow the @code{asm}
4222 An attribute specifier list may, in future, be permitted to appear after
4223 the declarator in a function definition (before any old-style parameter
4224 declarations or the function body).
4226 Attribute specifiers may be mixed with type qualifiers appearing inside
4227 the @code{[]} of a parameter array declarator, in the C99 construct by
4228 which such qualifiers are applied to the pointer to which the array is
4229 implicitly converted. Such attribute specifiers apply to the pointer,
4230 not to the array, but at present this is not implemented and they are
4233 An attribute specifier list may appear at the start of a nested
4234 declarator. At present, there are some limitations in this usage: the
4235 attributes correctly apply to the declarator, but for most individual
4236 attributes the semantics this implies are not implemented.
4237 When attribute specifiers follow the @code{*} of a pointer
4238 declarator, they may be mixed with any type qualifiers present.
4239 The following describes the formal semantics of this syntax. It will make the
4240 most sense if you are familiar with the formal specification of
4241 declarators in the ISO C standard.
4243 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
4244 D1}, where @code{T} contains declaration specifiers that specify a type
4245 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
4246 contains an identifier @var{ident}. The type specified for @var{ident}
4247 for derived declarators whose type does not include an attribute
4248 specifier is as in the ISO C standard.
4250 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
4251 and the declaration @code{T D} specifies the type
4252 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4253 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4254 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
4256 If @code{D1} has the form @code{*
4257 @var{type-qualifier-and-attribute-specifier-list} D}, and the
4258 declaration @code{T D} specifies the type
4259 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
4260 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
4261 @var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
4267 void (__attribute__((noreturn)) ****f) (void);
4271 specifies the type ``pointer to pointer to pointer to pointer to
4272 non-returning function returning @code{void}''. As another example,
4275 char *__attribute__((aligned(8))) *f;
4279 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
4280 Note again that this does not work with most attributes; for example,
4281 the usage of @samp{aligned} and @samp{noreturn} attributes given above
4282 is not yet supported.
4284 For compatibility with existing code written for compiler versions that
4285 did not implement attributes on nested declarators, some laxity is
4286 allowed in the placing of attributes. If an attribute that only applies
4287 to types is applied to a declaration, it will be treated as applying to
4288 the type of that declaration. If an attribute that only applies to
4289 declarations is applied to the type of a declaration, it will be treated
4290 as applying to that declaration; and, for compatibility with code
4291 placing the attributes immediately before the identifier declared, such
4292 an attribute applied to a function return type will be treated as
4293 applying to the function type, and such an attribute applied to an array
4294 element type will be treated as applying to the array type. If an
4295 attribute that only applies to function types is applied to a
4296 pointer-to-function type, it will be treated as applying to the pointer
4297 target type; if such an attribute is applied to a function return type
4298 that is not a pointer-to-function type, it will be treated as applying
4299 to the function type.
4301 @node Function Prototypes
4302 @section Prototypes and Old-Style Function Definitions
4303 @cindex function prototype declarations
4304 @cindex old-style function definitions
4305 @cindex promotion of formal parameters
4307 GNU C extends ISO C to allow a function prototype to override a later
4308 old-style non-prototype definition. Consider the following example:
4311 /* @r{Use prototypes unless the compiler is old-fashioned.} */
4318 /* @r{Prototype function declaration.} */
4319 int isroot P((uid_t));
4321 /* @r{Old-style function definition.} */
4323 isroot (x) /* @r{??? lossage here ???} */
4330 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
4331 not allow this example, because subword arguments in old-style
4332 non-prototype definitions are promoted. Therefore in this example the
4333 function definition's argument is really an @code{int}, which does not
4334 match the prototype argument type of @code{short}.
4336 This restriction of ISO C makes it hard to write code that is portable
4337 to traditional C compilers, because the programmer does not know
4338 whether the @code{uid_t} type is @code{short}, @code{int}, or
4339 @code{long}. Therefore, in cases like these GNU C allows a prototype
4340 to override a later old-style definition. More precisely, in GNU C, a
4341 function prototype argument type overrides the argument type specified
4342 by a later old-style definition if the former type is the same as the
4343 latter type before promotion. Thus in GNU C the above example is
4344 equivalent to the following:
4357 GNU C++ does not support old-style function definitions, so this
4358 extension is irrelevant.
4361 @section C++ Style Comments
4363 @cindex C++ comments
4364 @cindex comments, C++ style
4366 In GNU C, you may use C++ style comments, which start with @samp{//} and
4367 continue until the end of the line. Many other C implementations allow
4368 such comments, and they are included in the 1999 C standard. However,
4369 C++ style comments are not recognized if you specify an @option{-std}
4370 option specifying a version of ISO C before C99, or @option{-ansi}
4371 (equivalent to @option{-std=c90}).
4374 @section Dollar Signs in Identifier Names
4376 @cindex dollar signs in identifier names
4377 @cindex identifier names, dollar signs in
4379 In GNU C, you may normally use dollar signs in identifier names.
4380 This is because many traditional C implementations allow such identifiers.
4381 However, dollar signs in identifiers are not supported on a few target
4382 machines, typically because the target assembler does not allow them.
4384 @node Character Escapes
4385 @section The Character @key{ESC} in Constants
4387 You can use the sequence @samp{\e} in a string or character constant to
4388 stand for the ASCII character @key{ESC}.
4390 @node Variable Attributes
4391 @section Specifying Attributes of Variables
4392 @cindex attribute of variables
4393 @cindex variable attributes
4395 The keyword @code{__attribute__} allows you to specify special
4396 attributes of variables or structure fields. This keyword is followed
4397 by an attribute specification inside double parentheses. Some
4398 attributes are currently defined generically for variables.
4399 Other attributes are defined for variables on particular target
4400 systems. Other attributes are available for functions
4401 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
4402 Other front ends might define more attributes
4403 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
4405 You may also specify attributes with @samp{__} preceding and following
4406 each keyword. This allows you to use them in header files without
4407 being concerned about a possible macro of the same name. For example,
4408 you may use @code{__aligned__} instead of @code{aligned}.
4410 @xref{Attribute Syntax}, for details of the exact syntax for using
4414 @cindex @code{aligned} attribute
4415 @item aligned (@var{alignment})
4416 This attribute specifies a minimum alignment for the variable or
4417 structure field, measured in bytes. For example, the declaration:
4420 int x __attribute__ ((aligned (16))) = 0;
4424 causes the compiler to allocate the global variable @code{x} on a
4425 16-byte boundary. On a 68040, this could be used in conjunction with
4426 an @code{asm} expression to access the @code{move16} instruction which
4427 requires 16-byte aligned operands.
4429 You can also specify the alignment of structure fields. For example, to
4430 create a double-word aligned @code{int} pair, you could write:
4433 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
4437 This is an alternative to creating a union with a @code{double} member
4438 that forces the union to be double-word aligned.
4440 As in the preceding examples, you can explicitly specify the alignment
4441 (in bytes) that you wish the compiler to use for a given variable or
4442 structure field. Alternatively, you can leave out the alignment factor
4443 and just ask the compiler to align a variable or field to the
4444 default alignment for the target architecture you are compiling for.
4445 The default alignment is sufficient for all scalar types, but may not be
4446 enough for all vector types on a target which supports vector operations.
4447 The default alignment is fixed for a particular target ABI.
4449 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
4450 which is the largest alignment ever used for any data type on the
4451 target machine you are compiling for. For example, you could write:
4454 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
4457 The compiler automatically sets the alignment for the declared
4458 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
4459 often make copy operations more efficient, because the compiler can
4460 use whatever instructions copy the biggest chunks of memory when
4461 performing copies to or from the variables or fields that you have
4462 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
4463 may change depending on command line options.
4465 When used on a struct, or struct member, the @code{aligned} attribute can
4466 only increase the alignment; in order to decrease it, the @code{packed}
4467 attribute must be specified as well. When used as part of a typedef, the
4468 @code{aligned} attribute can both increase and decrease alignment, and
4469 specifying the @code{packed} attribute will generate a warning.
4471 Note that the effectiveness of @code{aligned} attributes may be limited
4472 by inherent limitations in your linker. On many systems, the linker is
4473 only able to arrange for variables to be aligned up to a certain maximum
4474 alignment. (For some linkers, the maximum supported alignment may
4475 be very very small.) If your linker is only able to align variables
4476 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4477 in an @code{__attribute__} will still only provide you with 8 byte
4478 alignment. See your linker documentation for further information.
4480 The @code{aligned} attribute can also be used for functions
4481 (@pxref{Function Attributes}.)
4483 @item cleanup (@var{cleanup_function})
4484 @cindex @code{cleanup} attribute
4485 The @code{cleanup} attribute runs a function when the variable goes
4486 out of scope. This attribute can only be applied to auto function
4487 scope variables; it may not be applied to parameters or variables
4488 with static storage duration. The function must take one parameter,
4489 a pointer to a type compatible with the variable. The return value
4490 of the function (if any) is ignored.
4492 If @option{-fexceptions} is enabled, then @var{cleanup_function}
4493 will be run during the stack unwinding that happens during the
4494 processing of the exception. Note that the @code{cleanup} attribute
4495 does not allow the exception to be caught, only to perform an action.
4496 It is undefined what happens if @var{cleanup_function} does not
4501 @cindex @code{common} attribute
4502 @cindex @code{nocommon} attribute
4505 The @code{common} attribute requests GCC to place a variable in
4506 ``common'' storage. The @code{nocommon} attribute requests the
4507 opposite---to allocate space for it directly.
4509 These attributes override the default chosen by the
4510 @option{-fno-common} and @option{-fcommon} flags respectively.
4513 @itemx deprecated (@var{msg})
4514 @cindex @code{deprecated} attribute
4515 The @code{deprecated} attribute results in a warning if the variable
4516 is used anywhere in the source file. This is useful when identifying
4517 variables that are expected to be removed in a future version of a
4518 program. The warning also includes the location of the declaration
4519 of the deprecated variable, to enable users to easily find further
4520 information about why the variable is deprecated, or what they should
4521 do instead. Note that the warning only occurs for uses:
4524 extern int old_var __attribute__ ((deprecated));
4526 int new_fn () @{ return old_var; @}
4529 results in a warning on line 3 but not line 2. The optional msg
4530 argument, which must be a string, will be printed in the warning if
4533 The @code{deprecated} attribute can also be used for functions and
4534 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
4536 @item mode (@var{mode})
4537 @cindex @code{mode} attribute
4538 This attribute specifies the data type for the declaration---whichever
4539 type corresponds to the mode @var{mode}. This in effect lets you
4540 request an integer or floating point type according to its width.
4542 You may also specify a mode of @samp{byte} or @samp{__byte__} to
4543 indicate the mode corresponding to a one-byte integer, @samp{word} or
4544 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
4545 or @samp{__pointer__} for the mode used to represent pointers.
4548 @cindex @code{packed} attribute
4549 The @code{packed} attribute specifies that a variable or structure field
4550 should have the smallest possible alignment---one byte for a variable,
4551 and one bit for a field, unless you specify a larger value with the
4552 @code{aligned} attribute.
4554 Here is a structure in which the field @code{x} is packed, so that it
4555 immediately follows @code{a}:
4561 int x[2] __attribute__ ((packed));
4565 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
4566 @code{packed} attribute on bit-fields of type @code{char}. This has
4567 been fixed in GCC 4.4 but the change can lead to differences in the
4568 structure layout. See the documentation of
4569 @option{-Wpacked-bitfield-compat} for more information.
4571 @item section ("@var{section-name}")
4572 @cindex @code{section} variable attribute
4573 Normally, the compiler places the objects it generates in sections like
4574 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
4575 or you need certain particular variables to appear in special sections,
4576 for example to map to special hardware. The @code{section}
4577 attribute specifies that a variable (or function) lives in a particular
4578 section. For example, this small program uses several specific section names:
4581 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
4582 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
4583 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
4584 int init_data __attribute__ ((section ("INITDATA")));
4588 /* @r{Initialize stack pointer} */
4589 init_sp (stack + sizeof (stack));
4591 /* @r{Initialize initialized data} */
4592 memcpy (&init_data, &data, &edata - &data);
4594 /* @r{Turn on the serial ports} */
4601 Use the @code{section} attribute with
4602 @emph{global} variables and not @emph{local} variables,
4603 as shown in the example.
4605 You may use the @code{section} attribute with initialized or
4606 uninitialized global variables but the linker requires
4607 each object be defined once, with the exception that uninitialized
4608 variables tentatively go in the @code{common} (or @code{bss}) section
4609 and can be multiply ``defined''. Using the @code{section} attribute
4610 will change what section the variable goes into and may cause the
4611 linker to issue an error if an uninitialized variable has multiple
4612 definitions. You can force a variable to be initialized with the
4613 @option{-fno-common} flag or the @code{nocommon} attribute.
4615 Some file formats do not support arbitrary sections so the @code{section}
4616 attribute is not available on all platforms.
4617 If you need to map the entire contents of a module to a particular
4618 section, consider using the facilities of the linker instead.
4621 @cindex @code{shared} variable attribute
4622 On Microsoft Windows, in addition to putting variable definitions in a named
4623 section, the section can also be shared among all running copies of an
4624 executable or DLL@. For example, this small program defines shared data
4625 by putting it in a named section @code{shared} and marking the section
4629 int foo __attribute__((section ("shared"), shared)) = 0;
4634 /* @r{Read and write foo. All running
4635 copies see the same value.} */
4641 You may only use the @code{shared} attribute along with @code{section}
4642 attribute with a fully initialized global definition because of the way
4643 linkers work. See @code{section} attribute for more information.
4645 The @code{shared} attribute is only available on Microsoft Windows@.
4647 @item tls_model ("@var{tls_model}")
4648 @cindex @code{tls_model} attribute
4649 The @code{tls_model} attribute sets thread-local storage model
4650 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
4651 overriding @option{-ftls-model=} command-line switch on a per-variable
4653 The @var{tls_model} argument should be one of @code{global-dynamic},
4654 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
4656 Not all targets support this attribute.
4659 This attribute, attached to a variable, means that the variable is meant
4660 to be possibly unused. GCC will not produce a warning for this
4664 This attribute, attached to a variable, means that the variable must be
4665 emitted even if it appears that the variable is not referenced.
4667 When applied to a static data member of a C++ class template, the
4668 attribute also means that the member will be instantiated if the
4669 class itself is instantiated.
4671 @item vector_size (@var{bytes})
4672 This attribute specifies the vector size for the variable, measured in
4673 bytes. For example, the declaration:
4676 int foo __attribute__ ((vector_size (16)));
4680 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
4681 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
4682 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
4684 This attribute is only applicable to integral and float scalars,
4685 although arrays, pointers, and function return values are allowed in
4686 conjunction with this construct.
4688 Aggregates with this attribute are invalid, even if they are of the same
4689 size as a corresponding scalar. For example, the declaration:
4692 struct S @{ int a; @};
4693 struct S __attribute__ ((vector_size (16))) foo;
4697 is invalid even if the size of the structure is the same as the size of
4701 The @code{selectany} attribute causes an initialized global variable to
4702 have link-once semantics. When multiple definitions of the variable are
4703 encountered by the linker, the first is selected and the remainder are
4704 discarded. Following usage by the Microsoft compiler, the linker is told
4705 @emph{not} to warn about size or content differences of the multiple
4708 Although the primary usage of this attribute is for POD types, the
4709 attribute can also be applied to global C++ objects that are initialized
4710 by a constructor. In this case, the static initialization and destruction
4711 code for the object is emitted in each translation defining the object,
4712 but the calls to the constructor and destructor are protected by a
4713 link-once guard variable.
4715 The @code{selectany} attribute is only available on Microsoft Windows
4716 targets. You can use @code{__declspec (selectany)} as a synonym for
4717 @code{__attribute__ ((selectany))} for compatibility with other
4721 The @code{weak} attribute is described in @ref{Function Attributes}.
4724 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4727 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4731 @anchor{AVR Variable Attributes}
4732 @subsection AVR Variable Attributes
4736 @cindex @code{progmem} AVR variable attribute
4737 The @code{progmem} attribute is used on the AVR to place read-only
4738 data in the non-volatile program memory (flash). The @code{progmem}
4739 attribute accomplishes this by putting respective variables into a
4740 section whose name starts with @code{.progmem}.
4742 This attribute works similar to the @code{section} attribute
4743 but adds additional checking. Notice that just like the
4744 @code{section} attribute, @code{progmem} affects the location
4745 of the data but not how this data is accessed.
4747 In order to read data located with the @code{progmem} attribute
4748 (inline) assembler must be used.
4750 /* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc}} */
4751 #include <avr/pgmspace.h>
4753 /* Locate var in flash memory */
4754 const int var[2] PROGMEM = @{ 1, 2 @};
4756 int read_var (int i)
4758 /* Access var[] by accessor macro from avr/pgmspace.h */
4759 return (int) pgm_read_word (& var[i]);
4763 AVR is a Harvard architecture processor and data and read-only data
4764 normally resides in the data memory (RAM).
4766 See also the @ref{AVR Named Address Spaces} section for
4767 an alternate way to locate and access data in flash memory.
4770 @subsection Blackfin Variable Attributes
4772 Three attributes are currently defined for the Blackfin.
4778 @cindex @code{l1_data} variable attribute
4779 @cindex @code{l1_data_A} variable attribute
4780 @cindex @code{l1_data_B} variable attribute
4781 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4782 Variables with @code{l1_data} attribute will be put into the specific section
4783 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4784 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4785 attribute will be put into the specific section named @code{.l1.data.B}.
4788 @cindex @code{l2} variable attribute
4789 Use this attribute on the Blackfin to place the variable into L2 SRAM.
4790 Variables with @code{l2} attribute will be put into the specific section
4791 named @code{.l2.data}.
4794 @subsection M32R/D Variable Attributes
4796 One attribute is currently defined for the M32R/D@.
4799 @item model (@var{model-name})
4800 @cindex variable addressability on the M32R/D
4801 Use this attribute on the M32R/D to set the addressability of an object.
4802 The identifier @var{model-name} is one of @code{small}, @code{medium},
4803 or @code{large}, representing each of the code models.
4805 Small model objects live in the lower 16MB of memory (so that their
4806 addresses can be loaded with the @code{ld24} instruction).
4808 Medium and large model objects may live anywhere in the 32-bit address space
4809 (the compiler will generate @code{seth/add3} instructions to load their
4813 @anchor{MeP Variable Attributes}
4814 @subsection MeP Variable Attributes
4816 The MeP target has a number of addressing modes and busses. The
4817 @code{near} space spans the standard memory space's first 16 megabytes
4818 (24 bits). The @code{far} space spans the entire 32-bit memory space.
4819 The @code{based} space is a 128 byte region in the memory space which
4820 is addressed relative to the @code{$tp} register. The @code{tiny}
4821 space is a 65536 byte region relative to the @code{$gp} register. In
4822 addition to these memory regions, the MeP target has a separate 16-bit
4823 control bus which is specified with @code{cb} attributes.
4828 Any variable with the @code{based} attribute will be assigned to the
4829 @code{.based} section, and will be accessed with relative to the
4830 @code{$tp} register.
4833 Likewise, the @code{tiny} attribute assigned variables to the
4834 @code{.tiny} section, relative to the @code{$gp} register.
4837 Variables with the @code{near} attribute are assumed to have addresses
4838 that fit in a 24-bit addressing mode. This is the default for large
4839 variables (@code{-mtiny=4} is the default) but this attribute can
4840 override @code{-mtiny=} for small variables, or override @code{-ml}.
4843 Variables with the @code{far} attribute are addressed using a full
4844 32-bit address. Since this covers the entire memory space, this
4845 allows modules to make no assumptions about where variables might be
4849 @itemx io (@var{addr})
4850 Variables with the @code{io} attribute are used to address
4851 memory-mapped peripherals. If an address is specified, the variable
4852 is assigned that address, else it is not assigned an address (it is
4853 assumed some other module will assign an address). Example:
4856 int timer_count __attribute__((io(0x123)));
4860 @itemx cb (@var{addr})
4861 Variables with the @code{cb} attribute are used to access the control
4862 bus, using special instructions. @code{addr} indicates the control bus
4866 int cpu_clock __attribute__((cb(0x123)));
4871 @anchor{i386 Variable Attributes}
4872 @subsection i386 Variable Attributes
4874 Two attributes are currently defined for i386 configurations:
4875 @code{ms_struct} and @code{gcc_struct}
4880 @cindex @code{ms_struct} attribute
4881 @cindex @code{gcc_struct} attribute
4883 If @code{packed} is used on a structure, or if bit-fields are used
4884 it may be that the Microsoft ABI packs them differently
4885 than GCC would normally pack them. Particularly when moving packed
4886 data between functions compiled with GCC and the native Microsoft compiler
4887 (either via function call or as data in a file), it may be necessary to access
4890 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4891 compilers to match the native Microsoft compiler.
4893 The Microsoft structure layout algorithm is fairly simple with the exception
4894 of the bitfield packing:
4896 The padding and alignment of members of structures and whether a bit field
4897 can straddle a storage-unit boundary
4900 @item Structure members are stored sequentially in the order in which they are
4901 declared: the first member has the lowest memory address and the last member
4904 @item Every data object has an alignment-requirement. The alignment-requirement
4905 for all data except structures, unions, and arrays is either the size of the
4906 object or the current packing size (specified with either the aligned attribute
4907 or the pack pragma), whichever is less. For structures, unions, and arrays,
4908 the alignment-requirement is the largest alignment-requirement of its members.
4909 Every object is allocated an offset so that:
4911 offset % alignment-requirement == 0
4913 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4914 unit if the integral types are the same size and if the next bit field fits
4915 into the current allocation unit without crossing the boundary imposed by the
4916 common alignment requirements of the bit fields.
4919 Handling of zero-length bitfields:
4921 MSVC interprets zero-length bitfields in the following ways:
4924 @item If a zero-length bitfield is inserted between two bitfields that would
4925 normally be coalesced, the bitfields will not be coalesced.
4932 unsigned long bf_1 : 12;
4934 unsigned long bf_2 : 12;
4938 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4939 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4941 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4942 alignment of the zero-length bitfield is greater than the member that follows it,
4943 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4963 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4964 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4965 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4968 Taking this into account, it is important to note the following:
4971 @item If a zero-length bitfield follows a normal bitfield, the type of the
4972 zero-length bitfield may affect the alignment of the structure as whole. For
4973 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4974 normal bitfield, and is of type short.
4976 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4977 still affect the alignment of the structure:
4987 Here, @code{t4} will take up 4 bytes.
4990 @item Zero-length bitfields following non-bitfield members are ignored:
5001 Here, @code{t5} will take up 2 bytes.
5005 @subsection PowerPC Variable Attributes
5007 Three attributes currently are defined for PowerPC configurations:
5008 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5010 For full documentation of the struct attributes please see the
5011 documentation in @ref{i386 Variable Attributes}.
5013 For documentation of @code{altivec} attribute please see the
5014 documentation in @ref{PowerPC Type Attributes}.
5016 @subsection SPU Variable Attributes
5018 The SPU supports the @code{spu_vector} attribute for variables. For
5019 documentation of this attribute please see the documentation in
5020 @ref{SPU Type Attributes}.
5022 @subsection Xstormy16 Variable Attributes
5024 One attribute is currently defined for xstormy16 configurations:
5029 @cindex @code{below100} attribute
5031 If a variable has the @code{below100} attribute (@code{BELOW100} is
5032 allowed also), GCC will place the variable in the first 0x100 bytes of
5033 memory and use special opcodes to access it. Such variables will be
5034 placed in either the @code{.bss_below100} section or the
5035 @code{.data_below100} section.
5039 @node Type Attributes
5040 @section Specifying Attributes of Types
5041 @cindex attribute of types
5042 @cindex type attributes
5044 The keyword @code{__attribute__} allows you to specify special
5045 attributes of @code{struct} and @code{union} types when you define
5046 such types. This keyword is followed by an attribute specification
5047 inside double parentheses. Seven attributes are currently defined for
5048 types: @code{aligned}, @code{packed}, @code{transparent_union},
5049 @code{unused}, @code{deprecated}, @code{visibility}, and
5050 @code{may_alias}. Other attributes are defined for functions
5051 (@pxref{Function Attributes}) and for variables (@pxref{Variable
5054 You may also specify any one of these attributes with @samp{__}
5055 preceding and following its keyword. This allows you to use these
5056 attributes in header files without being concerned about a possible
5057 macro of the same name. For example, you may use @code{__aligned__}
5058 instead of @code{aligned}.
5060 You may specify type attributes in an enum, struct or union type
5061 declaration or definition, or for other types in a @code{typedef}
5064 For an enum, struct or union type, you may specify attributes either
5065 between the enum, struct or union tag and the name of the type, or
5066 just past the closing curly brace of the @emph{definition}. The
5067 former syntax is preferred.
5069 @xref{Attribute Syntax}, for details of the exact syntax for using
5073 @cindex @code{aligned} attribute
5074 @item aligned (@var{alignment})
5075 This attribute specifies a minimum alignment (in bytes) for variables
5076 of the specified type. For example, the declarations:
5079 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
5080 typedef int more_aligned_int __attribute__ ((aligned (8)));
5084 force the compiler to insure (as far as it can) that each variable whose
5085 type is @code{struct S} or @code{more_aligned_int} will be allocated and
5086 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
5087 variables of type @code{struct S} aligned to 8-byte boundaries allows
5088 the compiler to use the @code{ldd} and @code{std} (doubleword load and
5089 store) instructions when copying one variable of type @code{struct S} to
5090 another, thus improving run-time efficiency.
5092 Note that the alignment of any given @code{struct} or @code{union} type
5093 is required by the ISO C standard to be at least a perfect multiple of
5094 the lowest common multiple of the alignments of all of the members of
5095 the @code{struct} or @code{union} in question. This means that you @emph{can}
5096 effectively adjust the alignment of a @code{struct} or @code{union}
5097 type by attaching an @code{aligned} attribute to any one of the members
5098 of such a type, but the notation illustrated in the example above is a
5099 more obvious, intuitive, and readable way to request the compiler to
5100 adjust the alignment of an entire @code{struct} or @code{union} type.
5102 As in the preceding example, you can explicitly specify the alignment
5103 (in bytes) that you wish the compiler to use for a given @code{struct}
5104 or @code{union} type. Alternatively, you can leave out the alignment factor
5105 and just ask the compiler to align a type to the maximum
5106 useful alignment for the target machine you are compiling for. For
5107 example, you could write:
5110 struct S @{ short f[3]; @} __attribute__ ((aligned));
5113 Whenever you leave out the alignment factor in an @code{aligned}
5114 attribute specification, the compiler automatically sets the alignment
5115 for the type to the largest alignment which is ever used for any data
5116 type on the target machine you are compiling for. Doing this can often
5117 make copy operations more efficient, because the compiler can use
5118 whatever instructions copy the biggest chunks of memory when performing
5119 copies to or from the variables which have types that you have aligned
5122 In the example above, if the size of each @code{short} is 2 bytes, then
5123 the size of the entire @code{struct S} type is 6 bytes. The smallest
5124 power of two which is greater than or equal to that is 8, so the
5125 compiler sets the alignment for the entire @code{struct S} type to 8
5128 Note that although you can ask the compiler to select a time-efficient
5129 alignment for a given type and then declare only individual stand-alone
5130 objects of that type, the compiler's ability to select a time-efficient
5131 alignment is primarily useful only when you plan to create arrays of
5132 variables having the relevant (efficiently aligned) type. If you
5133 declare or use arrays of variables of an efficiently-aligned type, then
5134 it is likely that your program will also be doing pointer arithmetic (or
5135 subscripting, which amounts to the same thing) on pointers to the
5136 relevant type, and the code that the compiler generates for these
5137 pointer arithmetic operations will often be more efficient for
5138 efficiently-aligned types than for other types.
5140 The @code{aligned} attribute can only increase the alignment; but you
5141 can decrease it by specifying @code{packed} as well. See below.
5143 Note that the effectiveness of @code{aligned} attributes may be limited
5144 by inherent limitations in your linker. On many systems, the linker is
5145 only able to arrange for variables to be aligned up to a certain maximum
5146 alignment. (For some linkers, the maximum supported alignment may
5147 be very very small.) If your linker is only able to align variables
5148 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
5149 in an @code{__attribute__} will still only provide you with 8 byte
5150 alignment. See your linker documentation for further information.
5153 This attribute, attached to @code{struct} or @code{union} type
5154 definition, specifies that each member (other than zero-width bitfields)
5155 of the structure or union is placed to minimize the memory required. When
5156 attached to an @code{enum} definition, it indicates that the smallest
5157 integral type should be used.
5159 @opindex fshort-enums
5160 Specifying this attribute for @code{struct} and @code{union} types is
5161 equivalent to specifying the @code{packed} attribute on each of the
5162 structure or union members. Specifying the @option{-fshort-enums}
5163 flag on the line is equivalent to specifying the @code{packed}
5164 attribute on all @code{enum} definitions.
5166 In the following example @code{struct my_packed_struct}'s members are
5167 packed closely together, but the internal layout of its @code{s} member
5168 is not packed---to do that, @code{struct my_unpacked_struct} would need to
5172 struct my_unpacked_struct
5178 struct __attribute__ ((__packed__)) my_packed_struct
5182 struct my_unpacked_struct s;
5186 You may only specify this attribute on the definition of an @code{enum},
5187 @code{struct} or @code{union}, not on a @code{typedef} which does not
5188 also define the enumerated type, structure or union.
5190 @item transparent_union
5191 This attribute, attached to a @code{union} type definition, indicates
5192 that any function parameter having that union type causes calls to that
5193 function to be treated in a special way.
5195 First, the argument corresponding to a transparent union type can be of
5196 any type in the union; no cast is required. Also, if the union contains
5197 a pointer type, the corresponding argument can be a null pointer
5198 constant or a void pointer expression; and if the union contains a void
5199 pointer type, the corresponding argument can be any pointer expression.
5200 If the union member type is a pointer, qualifiers like @code{const} on
5201 the referenced type must be respected, just as with normal pointer
5204 Second, the argument is passed to the function using the calling
5205 conventions of the first member of the transparent union, not the calling
5206 conventions of the union itself. All members of the union must have the
5207 same machine representation; this is necessary for this argument passing
5210 Transparent unions are designed for library functions that have multiple
5211 interfaces for compatibility reasons. For example, suppose the
5212 @code{wait} function must accept either a value of type @code{int *} to
5213 comply with Posix, or a value of type @code{union wait *} to comply with
5214 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
5215 @code{wait} would accept both kinds of arguments, but it would also
5216 accept any other pointer type and this would make argument type checking
5217 less useful. Instead, @code{<sys/wait.h>} might define the interface
5221 typedef union __attribute__ ((__transparent_union__))
5225 @} wait_status_ptr_t;
5227 pid_t wait (wait_status_ptr_t);
5230 This interface allows either @code{int *} or @code{union wait *}
5231 arguments to be passed, using the @code{int *} calling convention.
5232 The program can call @code{wait} with arguments of either type:
5235 int w1 () @{ int w; return wait (&w); @}
5236 int w2 () @{ union wait w; return wait (&w); @}
5239 With this interface, @code{wait}'s implementation might look like this:
5242 pid_t wait (wait_status_ptr_t p)
5244 return waitpid (-1, p.__ip, 0);
5249 When attached to a type (including a @code{union} or a @code{struct}),
5250 this attribute means that variables of that type are meant to appear
5251 possibly unused. GCC will not produce a warning for any variables of
5252 that type, even if the variable appears to do nothing. This is often
5253 the case with lock or thread classes, which are usually defined and then
5254 not referenced, but contain constructors and destructors that have
5255 nontrivial bookkeeping functions.
5258 @itemx deprecated (@var{msg})
5259 The @code{deprecated} attribute results in a warning if the type
5260 is used anywhere in the source file. This is useful when identifying
5261 types that are expected to be removed in a future version of a program.
5262 If possible, the warning also includes the location of the declaration
5263 of the deprecated type, to enable users to easily find further
5264 information about why the type is deprecated, or what they should do
5265 instead. Note that the warnings only occur for uses and then only
5266 if the type is being applied to an identifier that itself is not being
5267 declared as deprecated.
5270 typedef int T1 __attribute__ ((deprecated));
5274 typedef T1 T3 __attribute__ ((deprecated));
5275 T3 z __attribute__ ((deprecated));
5278 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
5279 warning is issued for line 4 because T2 is not explicitly
5280 deprecated. Line 5 has no warning because T3 is explicitly
5281 deprecated. Similarly for line 6. The optional msg
5282 argument, which must be a string, will be printed in the warning if
5285 The @code{deprecated} attribute can also be used for functions and
5286 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
5289 Accesses through pointers to types with this attribute are not subject
5290 to type-based alias analysis, but are instead assumed to be able to alias
5291 any other type of objects. In the context of 6.5/7 an lvalue expression
5292 dereferencing such a pointer is treated like having a character type.
5293 See @option{-fstrict-aliasing} for more information on aliasing issues.
5294 This extension exists to support some vector APIs, in which pointers to
5295 one vector type are permitted to alias pointers to a different vector type.
5297 Note that an object of a type with this attribute does not have any
5303 typedef short __attribute__((__may_alias__)) short_a;
5309 short_a *b = (short_a *) &a;
5313 if (a == 0x12345678)
5320 If you replaced @code{short_a} with @code{short} in the variable
5321 declaration, the above program would abort when compiled with
5322 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
5323 above in recent GCC versions.
5326 In C++, attribute visibility (@pxref{Function Attributes}) can also be
5327 applied to class, struct, union and enum types. Unlike other type
5328 attributes, the attribute must appear between the initial keyword and
5329 the name of the type; it cannot appear after the body of the type.
5331 Note that the type visibility is applied to vague linkage entities
5332 associated with the class (vtable, typeinfo node, etc.). In
5333 particular, if a class is thrown as an exception in one shared object
5334 and caught in another, the class must have default visibility.
5335 Otherwise the two shared objects will be unable to use the same
5336 typeinfo node and exception handling will break.
5340 @subsection ARM Type Attributes
5342 On those ARM targets that support @code{dllimport} (such as Symbian
5343 OS), you can use the @code{notshared} attribute to indicate that the
5344 virtual table and other similar data for a class should not be
5345 exported from a DLL@. For example:
5348 class __declspec(notshared) C @{
5350 __declspec(dllimport) C();
5354 __declspec(dllexport)
5358 In this code, @code{C::C} is exported from the current DLL, but the
5359 virtual table for @code{C} is not exported. (You can use
5360 @code{__attribute__} instead of @code{__declspec} if you prefer, but
5361 most Symbian OS code uses @code{__declspec}.)
5363 @anchor{MeP Type Attributes}
5364 @subsection MeP Type Attributes
5366 Many of the MeP variable attributes may be applied to types as well.
5367 Specifically, the @code{based}, @code{tiny}, @code{near}, and
5368 @code{far} attributes may be applied to either. The @code{io} and
5369 @code{cb} attributes may not be applied to types.
5371 @anchor{i386 Type Attributes}
5372 @subsection i386 Type Attributes
5374 Two attributes are currently defined for i386 configurations:
5375 @code{ms_struct} and @code{gcc_struct}.
5381 @cindex @code{ms_struct}
5382 @cindex @code{gcc_struct}
5384 If @code{packed} is used on a structure, or if bit-fields are used
5385 it may be that the Microsoft ABI packs them differently
5386 than GCC would normally pack them. Particularly when moving packed
5387 data between functions compiled with GCC and the native Microsoft compiler
5388 (either via function call or as data in a file), it may be necessary to access
5391 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
5392 compilers to match the native Microsoft compiler.
5395 To specify multiple attributes, separate them by commas within the
5396 double parentheses: for example, @samp{__attribute__ ((aligned (16),
5399 @anchor{PowerPC Type Attributes}
5400 @subsection PowerPC Type Attributes
5402 Three attributes currently are defined for PowerPC configurations:
5403 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
5405 For full documentation of the @code{ms_struct} and @code{gcc_struct}
5406 attributes please see the documentation in @ref{i386 Type Attributes}.
5408 The @code{altivec} attribute allows one to declare AltiVec vector data
5409 types supported by the AltiVec Programming Interface Manual. The
5410 attribute requires an argument to specify one of three vector types:
5411 @code{vector__}, @code{pixel__} (always followed by unsigned short),
5412 and @code{bool__} (always followed by unsigned).
5415 __attribute__((altivec(vector__)))
5416 __attribute__((altivec(pixel__))) unsigned short
5417 __attribute__((altivec(bool__))) unsigned
5420 These attributes mainly are intended to support the @code{__vector},
5421 @code{__pixel}, and @code{__bool} AltiVec keywords.
5423 @anchor{SPU Type Attributes}
5424 @subsection SPU Type Attributes
5426 The SPU supports the @code{spu_vector} attribute for types. This attribute
5427 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
5428 Language Extensions Specification. It is intended to support the
5429 @code{__vector} keyword.
5432 @section Inquiring on Alignment of Types or Variables
5434 @cindex type alignment
5435 @cindex variable alignment
5437 The keyword @code{__alignof__} allows you to inquire about how an object
5438 is aligned, or the minimum alignment usually required by a type. Its
5439 syntax is just like @code{sizeof}.
5441 For example, if the target machine requires a @code{double} value to be
5442 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
5443 This is true on many RISC machines. On more traditional machine
5444 designs, @code{__alignof__ (double)} is 4 or even 2.
5446 Some machines never actually require alignment; they allow reference to any
5447 data type even at an odd address. For these machines, @code{__alignof__}
5448 reports the smallest alignment that GCC will give the data type, usually as
5449 mandated by the target ABI.
5451 If the operand of @code{__alignof__} is an lvalue rather than a type,
5452 its value is the required alignment for its type, taking into account
5453 any minimum alignment specified with GCC's @code{__attribute__}
5454 extension (@pxref{Variable Attributes}). For example, after this
5458 struct foo @{ int x; char y; @} foo1;
5462 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
5463 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
5465 It is an error to ask for the alignment of an incomplete type.
5469 @section An Inline Function is As Fast As a Macro
5470 @cindex inline functions
5471 @cindex integrating function code
5473 @cindex macros, inline alternative
5475 By declaring a function inline, you can direct GCC to make
5476 calls to that function faster. One way GCC can achieve this is to
5477 integrate that function's code into the code for its callers. This
5478 makes execution faster by eliminating the function-call overhead; in
5479 addition, if any of the actual argument values are constant, their
5480 known values may permit simplifications at compile time so that not
5481 all of the inline function's code needs to be included. The effect on
5482 code size is less predictable; object code may be larger or smaller
5483 with function inlining, depending on the particular case. You can
5484 also direct GCC to try to integrate all ``simple enough'' functions
5485 into their callers with the option @option{-finline-functions}.
5487 GCC implements three different semantics of declaring a function
5488 inline. One is available with @option{-std=gnu89} or
5489 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
5490 on all inline declarations, another when
5491 @option{-std=c99}, @option{-std=c11},
5492 @option{-std=gnu99} or @option{-std=gnu11}
5493 (without @option{-fgnu89-inline}), and the third
5494 is used when compiling C++.
5496 To declare a function inline, use the @code{inline} keyword in its
5497 declaration, like this:
5507 If you are writing a header file to be included in ISO C90 programs, write
5508 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
5510 The three types of inlining behave similarly in two important cases:
5511 when the @code{inline} keyword is used on a @code{static} function,
5512 like the example above, and when a function is first declared without
5513 using the @code{inline} keyword and then is defined with
5514 @code{inline}, like this:
5517 extern int inc (int *a);
5525 In both of these common cases, the program behaves the same as if you
5526 had not used the @code{inline} keyword, except for its speed.
5528 @cindex inline functions, omission of
5529 @opindex fkeep-inline-functions
5530 When a function is both inline and @code{static}, if all calls to the
5531 function are integrated into the caller, and the function's address is
5532 never used, then the function's own assembler code is never referenced.
5533 In this case, GCC does not actually output assembler code for the
5534 function, unless you specify the option @option{-fkeep-inline-functions}.
5535 Some calls cannot be integrated for various reasons (in particular,
5536 calls that precede the function's definition cannot be integrated, and
5537 neither can recursive calls within the definition). If there is a
5538 nonintegrated call, then the function is compiled to assembler code as
5539 usual. The function must also be compiled as usual if the program
5540 refers to its address, because that can't be inlined.
5543 Note that certain usages in a function definition can make it unsuitable
5544 for inline substitution. Among these usages are: use of varargs, use of
5545 alloca, use of variable sized data types (@pxref{Variable Length}),
5546 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
5547 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
5548 will warn when a function marked @code{inline} could not be substituted,
5549 and will give the reason for the failure.
5551 @cindex automatic @code{inline} for C++ member fns
5552 @cindex @code{inline} automatic for C++ member fns
5553 @cindex member fns, automatically @code{inline}
5554 @cindex C++ member fns, automatically @code{inline}
5555 @opindex fno-default-inline
5556 As required by ISO C++, GCC considers member functions defined within
5557 the body of a class to be marked inline even if they are
5558 not explicitly declared with the @code{inline} keyword. You can
5559 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
5560 Options,,Options Controlling C++ Dialect}.
5562 GCC does not inline any functions when not optimizing unless you specify
5563 the @samp{always_inline} attribute for the function, like this:
5566 /* @r{Prototype.} */
5567 inline void foo (const char) __attribute__((always_inline));
5570 The remainder of this section is specific to GNU C90 inlining.
5572 @cindex non-static inline function
5573 When an inline function is not @code{static}, then the compiler must assume
5574 that there may be calls from other source files; since a global symbol can
5575 be defined only once in any program, the function must not be defined in
5576 the other source files, so the calls therein cannot be integrated.
5577 Therefore, a non-@code{static} inline function is always compiled on its
5578 own in the usual fashion.
5580 If you specify both @code{inline} and @code{extern} in the function
5581 definition, then the definition is used only for inlining. In no case
5582 is the function compiled on its own, not even if you refer to its
5583 address explicitly. Such an address becomes an external reference, as
5584 if you had only declared the function, and had not defined it.
5586 This combination of @code{inline} and @code{extern} has almost the
5587 effect of a macro. The way to use it is to put a function definition in
5588 a header file with these keywords, and put another copy of the
5589 definition (lacking @code{inline} and @code{extern}) in a library file.
5590 The definition in the header file will cause most calls to the function
5591 to be inlined. If any uses of the function remain, they will refer to
5592 the single copy in the library.
5595 @section When is a Volatile Object Accessed?
5596 @cindex accessing volatiles
5597 @cindex volatile read
5598 @cindex volatile write
5599 @cindex volatile access
5601 C has the concept of volatile objects. These are normally accessed by
5602 pointers and used for accessing hardware or inter-thread
5603 communication. The standard encourages compilers to refrain from
5604 optimizations concerning accesses to volatile objects, but leaves it
5605 implementation defined as to what constitutes a volatile access. The
5606 minimum requirement is that at a sequence point all previous accesses
5607 to volatile objects have stabilized and no subsequent accesses have
5608 occurred. Thus an implementation is free to reorder and combine
5609 volatile accesses which occur between sequence points, but cannot do
5610 so for accesses across a sequence point. The use of volatile does
5611 not allow you to violate the restriction on updating objects multiple
5612 times between two sequence points.
5614 Accesses to non-volatile objects are not ordered with respect to
5615 volatile accesses. You cannot use a volatile object as a memory
5616 barrier to order a sequence of writes to non-volatile memory. For
5620 int *ptr = @var{something};
5622 *ptr = @var{something};
5626 Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
5627 that the write to @var{*ptr} will have occurred by the time the update
5628 of @var{vobj} has happened. If you need this guarantee, you must use
5629 a stronger memory barrier such as:
5632 int *ptr = @var{something};
5634 *ptr = @var{something};
5635 asm volatile ("" : : : "memory");
5639 A scalar volatile object is read when it is accessed in a void context:
5642 volatile int *src = @var{somevalue};
5646 Such expressions are rvalues, and GCC implements this as a
5647 read of the volatile object being pointed to.
5649 Assignments are also expressions and have an rvalue. However when
5650 assigning to a scalar volatile, the volatile object is not reread,
5651 regardless of whether the assignment expression's rvalue is used or
5652 not. If the assignment's rvalue is used, the value is that assigned
5653 to the volatile object. For instance, there is no read of @var{vobj}
5654 in all the following cases:
5659 vobj = @var{something};
5660 obj = vobj = @var{something};
5661 obj ? vobj = @var{onething} : vobj = @var{anotherthing};
5662 obj = (@var{something}, vobj = @var{anotherthing});
5665 If you need to read the volatile object after an assignment has
5666 occurred, you must use a separate expression with an intervening
5669 As bitfields are not individually addressable, volatile bitfields may
5670 be implicitly read when written to, or when adjacent bitfields are
5671 accessed. Bitfield operations may be optimized such that adjacent
5672 bitfields are only partially accessed, if they straddle a storage unit
5673 boundary. For these reasons it is unwise to use volatile bitfields to
5677 @section Assembler Instructions with C Expression Operands
5678 @cindex extended @code{asm}
5679 @cindex @code{asm} expressions
5680 @cindex assembler instructions
5683 In an assembler instruction using @code{asm}, you can specify the
5684 operands of the instruction using C expressions. This means you need not
5685 guess which registers or memory locations will contain the data you want
5688 You must specify an assembler instruction template much like what
5689 appears in a machine description, plus an operand constraint string for
5692 For example, here is how to use the 68881's @code{fsinx} instruction:
5695 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
5699 Here @code{angle} is the C expression for the input operand while
5700 @code{result} is that of the output operand. Each has @samp{"f"} as its
5701 operand constraint, saying that a floating point register is required.
5702 The @samp{=} in @samp{=f} indicates that the operand is an output; all
5703 output operands' constraints must use @samp{=}. The constraints use the
5704 same language used in the machine description (@pxref{Constraints}).
5706 Each operand is described by an operand-constraint string followed by
5707 the C expression in parentheses. A colon separates the assembler
5708 template from the first output operand and another separates the last
5709 output operand from the first input, if any. Commas separate the
5710 operands within each group. The total number of operands is currently
5711 limited to 30; this limitation may be lifted in some future version of
5714 If there are no output operands but there are input operands, you must
5715 place two consecutive colons surrounding the place where the output
5718 As of GCC version 3.1, it is also possible to specify input and output
5719 operands using symbolic names which can be referenced within the
5720 assembler code. These names are specified inside square brackets
5721 preceding the constraint string, and can be referenced inside the
5722 assembler code using @code{%[@var{name}]} instead of a percentage sign
5723 followed by the operand number. Using named operands the above example
5727 asm ("fsinx %[angle],%[output]"
5728 : [output] "=f" (result)
5729 : [angle] "f" (angle));
5733 Note that the symbolic operand names have no relation whatsoever to
5734 other C identifiers. You may use any name you like, even those of
5735 existing C symbols, but you must ensure that no two operands within the same
5736 assembler construct use the same symbolic name.
5738 Output operand expressions must be lvalues; the compiler can check this.
5739 The input operands need not be lvalues. The compiler cannot check
5740 whether the operands have data types that are reasonable for the
5741 instruction being executed. It does not parse the assembler instruction
5742 template and does not know what it means or even whether it is valid
5743 assembler input. The extended @code{asm} feature is most often used for
5744 machine instructions the compiler itself does not know exist. If
5745 the output expression cannot be directly addressed (for example, it is a
5746 bit-field), your constraint must allow a register. In that case, GCC
5747 will use the register as the output of the @code{asm}, and then store
5748 that register into the output.
5750 The ordinary output operands must be write-only; GCC will assume that
5751 the values in these operands before the instruction are dead and need
5752 not be generated. Extended asm supports input-output or read-write
5753 operands. Use the constraint character @samp{+} to indicate such an
5754 operand and list it with the output operands. You should only use
5755 read-write operands when the constraints for the operand (or the
5756 operand in which only some of the bits are to be changed) allow a
5759 You may, as an alternative, logically split its function into two
5760 separate operands, one input operand and one write-only output
5761 operand. The connection between them is expressed by constraints
5762 which say they need to be in the same location when the instruction
5763 executes. You can use the same C expression for both operands, or
5764 different expressions. For example, here we write the (fictitious)
5765 @samp{combine} instruction with @code{bar} as its read-only source
5766 operand and @code{foo} as its read-write destination:
5769 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
5773 The constraint @samp{"0"} for operand 1 says that it must occupy the
5774 same location as operand 0. A number in constraint is allowed only in
5775 an input operand and it must refer to an output operand.
5777 Only a number in the constraint can guarantee that one operand will be in
5778 the same place as another. The mere fact that @code{foo} is the value
5779 of both operands is not enough to guarantee that they will be in the
5780 same place in the generated assembler code. The following would not
5784 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
5787 Various optimizations or reloading could cause operands 0 and 1 to be in
5788 different registers; GCC knows no reason not to do so. For example, the
5789 compiler might find a copy of the value of @code{foo} in one register and
5790 use it for operand 1, but generate the output operand 0 in a different
5791 register (copying it afterward to @code{foo}'s own address). Of course,
5792 since the register for operand 1 is not even mentioned in the assembler
5793 code, the result will not work, but GCC can't tell that.
5795 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
5796 the operand number for a matching constraint. For example:
5799 asm ("cmoveq %1,%2,%[result]"
5800 : [result] "=r"(result)
5801 : "r" (test), "r"(new), "[result]"(old));
5804 Sometimes you need to make an @code{asm} operand be a specific register,
5805 but there's no matching constraint letter for that register @emph{by
5806 itself}. To force the operand into that register, use a local variable
5807 for the operand and specify the register in the variable declaration.
5808 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
5809 register constraint letter that matches the register:
5812 register int *p1 asm ("r0") = @dots{};
5813 register int *p2 asm ("r1") = @dots{};
5814 register int *result asm ("r0");
5815 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5818 @anchor{Example of asm with clobbered asm reg}
5819 In the above example, beware that a register that is call-clobbered by
5820 the target ABI will be overwritten by any function call in the
5821 assignment, including library calls for arithmetic operators.
5822 Also a register may be clobbered when generating some operations,
5823 like variable shift, memory copy or memory move on x86.
5824 Assuming it is a call-clobbered register, this may happen to @code{r0}
5825 above by the assignment to @code{p2}. If you have to use such a
5826 register, use temporary variables for expressions between the register
5831 register int *p1 asm ("r0") = @dots{};
5832 register int *p2 asm ("r1") = t1;
5833 register int *result asm ("r0");
5834 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5837 Some instructions clobber specific hard registers. To describe this,
5838 write a third colon after the input operands, followed by the names of
5839 the clobbered hard registers (given as strings). Here is a realistic
5840 example for the VAX:
5843 asm volatile ("movc3 %0,%1,%2"
5844 : /* @r{no outputs} */
5845 : "g" (from), "g" (to), "g" (count)
5846 : "r0", "r1", "r2", "r3", "r4", "r5");
5849 You may not write a clobber description in a way that overlaps with an
5850 input or output operand. For example, you may not have an operand
5851 describing a register class with one member if you mention that register
5852 in the clobber list. Variables declared to live in specific registers
5853 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
5854 have no part mentioned in the clobber description.
5855 There is no way for you to specify that an input
5856 operand is modified without also specifying it as an output
5857 operand. Note that if all the output operands you specify are for this
5858 purpose (and hence unused), you will then also need to specify
5859 @code{volatile} for the @code{asm} construct, as described below, to
5860 prevent GCC from deleting the @code{asm} statement as unused.
5862 If you refer to a particular hardware register from the assembler code,
5863 you will probably have to list the register after the third colon to
5864 tell the compiler the register's value is modified. In some assemblers,
5865 the register names begin with @samp{%}; to produce one @samp{%} in the
5866 assembler code, you must write @samp{%%} in the input.
5868 If your assembler instruction can alter the condition code register, add
5869 @samp{cc} to the list of clobbered registers. GCC on some machines
5870 represents the condition codes as a specific hardware register;
5871 @samp{cc} serves to name this register. On other machines, the
5872 condition code is handled differently, and specifying @samp{cc} has no
5873 effect. But it is valid no matter what the machine.
5875 If your assembler instructions access memory in an unpredictable
5876 fashion, add @samp{memory} to the list of clobbered registers. This
5877 will cause GCC to not keep memory values cached in registers across the
5878 assembler instruction and not optimize stores or loads to that memory.
5879 You will also want to add the @code{volatile} keyword if the memory
5880 affected is not listed in the inputs or outputs of the @code{asm}, as
5881 the @samp{memory} clobber does not count as a side-effect of the
5882 @code{asm}. If you know how large the accessed memory is, you can add
5883 it as input or output but if this is not known, you should add
5884 @samp{memory}. As an example, if you access ten bytes of a string, you
5885 can use a memory input like:
5888 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
5891 Note that in the following example the memory input is necessary,
5892 otherwise GCC might optimize the store to @code{x} away:
5899 asm ("magic stuff accessing an 'int' pointed to by '%1'"
5900 "=&d" (r) : "a" (y), "m" (*y));
5905 You can put multiple assembler instructions together in a single
5906 @code{asm} template, separated by the characters normally used in assembly
5907 code for the system. A combination that works in most places is a newline
5908 to break the line, plus a tab character to move to the instruction field
5909 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
5910 assembler allows semicolons as a line-breaking character. Note that some
5911 assembler dialects use semicolons to start a comment.
5912 The input operands are guaranteed not to use any of the clobbered
5913 registers, and neither will the output operands' addresses, so you can
5914 read and write the clobbered registers as many times as you like. Here
5915 is an example of multiple instructions in a template; it assumes the
5916 subroutine @code{_foo} accepts arguments in registers 9 and 10:
5919 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
5921 : "g" (from), "g" (to)
5925 Unless an output operand has the @samp{&} constraint modifier, GCC
5926 may allocate it in the same register as an unrelated input operand, on
5927 the assumption the inputs are consumed before the outputs are produced.
5928 This assumption may be false if the assembler code actually consists of
5929 more than one instruction. In such a case, use @samp{&} for each output
5930 operand that may not overlap an input. @xref{Modifiers}.
5932 If you want to test the condition code produced by an assembler
5933 instruction, you must include a branch and a label in the @code{asm}
5934 construct, as follows:
5937 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5943 This assumes your assembler supports local labels, as the GNU assembler
5944 and most Unix assemblers do.
5946 Speaking of labels, jumps from one @code{asm} to another are not
5947 supported. The compiler's optimizers do not know about these jumps, and
5948 therefore they cannot take account of them when deciding how to
5949 optimize. @xref{Extended asm with goto}.
5951 @cindex macros containing @code{asm}
5952 Usually the most convenient way to use these @code{asm} instructions is to
5953 encapsulate them in macros that look like functions. For example,
5957 (@{ double __value, __arg = (x); \
5958 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5963 Here the variable @code{__arg} is used to make sure that the instruction
5964 operates on a proper @code{double} value, and to accept only those
5965 arguments @code{x} which can convert automatically to a @code{double}.
5967 Another way to make sure the instruction operates on the correct data
5968 type is to use a cast in the @code{asm}. This is different from using a
5969 variable @code{__arg} in that it converts more different types. For
5970 example, if the desired type were @code{int}, casting the argument to
5971 @code{int} would accept a pointer with no complaint, while assigning the
5972 argument to an @code{int} variable named @code{__arg} would warn about
5973 using a pointer unless the caller explicitly casts it.
5975 If an @code{asm} has output operands, GCC assumes for optimization
5976 purposes the instruction has no side effects except to change the output
5977 operands. This does not mean instructions with a side effect cannot be
5978 used, but you must be careful, because the compiler may eliminate them
5979 if the output operands aren't used, or move them out of loops, or
5980 replace two with one if they constitute a common subexpression. Also,
5981 if your instruction does have a side effect on a variable that otherwise
5982 appears not to change, the old value of the variable may be reused later
5983 if it happens to be found in a register.
5985 You can prevent an @code{asm} instruction from being deleted
5986 by writing the keyword @code{volatile} after
5987 the @code{asm}. For example:
5990 #define get_and_set_priority(new) \
5992 asm volatile ("get_and_set_priority %0, %1" \
5993 : "=g" (__old) : "g" (new)); \
5998 The @code{volatile} keyword indicates that the instruction has
5999 important side-effects. GCC will not delete a volatile @code{asm} if
6000 it is reachable. (The instruction can still be deleted if GCC can
6001 prove that control-flow will never reach the location of the
6002 instruction.) Note that even a volatile @code{asm} instruction
6003 can be moved relative to other code, including across jump
6004 instructions. For example, on many targets there is a system
6005 register which can be set to control the rounding mode of
6006 floating point operations. You might try
6007 setting it with a volatile @code{asm}, like this PowerPC example:
6010 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
6015 This will not work reliably, as the compiler may move the addition back
6016 before the volatile @code{asm}. To make it work you need to add an
6017 artificial dependency to the @code{asm} referencing a variable in the code
6018 you don't want moved, for example:
6021 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
6025 Similarly, you can't expect a
6026 sequence of volatile @code{asm} instructions to remain perfectly
6027 consecutive. If you want consecutive output, use a single @code{asm}.
6028 Also, GCC will perform some optimizations across a volatile @code{asm}
6029 instruction; GCC does not ``forget everything'' when it encounters
6030 a volatile @code{asm} instruction the way some other compilers do.
6032 An @code{asm} instruction without any output operands will be treated
6033 identically to a volatile @code{asm} instruction.
6035 It is a natural idea to look for a way to give access to the condition
6036 code left by the assembler instruction. However, when we attempted to
6037 implement this, we found no way to make it work reliably. The problem
6038 is that output operands might need reloading, which would result in
6039 additional following ``store'' instructions. On most machines, these
6040 instructions would alter the condition code before there was time to
6041 test it. This problem doesn't arise for ordinary ``test'' and
6042 ``compare'' instructions because they don't have any output operands.
6044 For reasons similar to those described above, it is not possible to give
6045 an assembler instruction access to the condition code left by previous
6048 @anchor{Extended asm with goto}
6049 As of GCC version 4.5, @code{asm goto} may be used to have the assembly
6050 jump to one or more C labels. In this form, a fifth section after the
6051 clobber list contains a list of all C labels to which the assembly may jump.
6052 Each label operand is implicitly self-named. The @code{asm} is also assumed
6053 to fall through to the next statement.
6055 This form of @code{asm} is restricted to not have outputs. This is due
6056 to a internal restriction in the compiler that control transfer instructions
6057 cannot have outputs. This restriction on @code{asm goto} may be lifted
6058 in some future version of the compiler. In the mean time, @code{asm goto}
6059 may include a memory clobber, and so leave outputs in memory.
6065 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
6066 : : "r"(x), "r"(&y) : "r5", "memory" : error);
6073 In this (inefficient) example, the @code{frob} instruction sets the
6074 carry bit to indicate an error. The @code{jc} instruction detects
6075 this and branches to the @code{error} label. Finally, the output
6076 of the @code{frob} instruction (@code{%r5}) is stored into the memory
6077 for variable @code{y}, which is later read by the @code{return} statement.
6083 asm goto ("mfsr %%r1, 123; jmp %%r1;"
6084 ".pushsection doit_table;"
6085 ".long %l0, %l1, %l2, %l3;"
6087 : : : "r1" : label1, label2, label3, label4);
6088 __builtin_unreachable ();
6103 In this (also inefficient) example, the @code{mfsr} instruction reads
6104 an address from some out-of-band machine register, and the following
6105 @code{jmp} instruction branches to that address. The address read by
6106 the @code{mfsr} instruction is assumed to have been previously set via
6107 some application-specific mechanism to be one of the four values stored
6108 in the @code{doit_table} section. Finally, the @code{asm} is followed
6109 by a call to @code{__builtin_unreachable} to indicate that the @code{asm}
6110 does not in fact fall through.
6113 #define TRACE1(NUM) \
6115 asm goto ("0: nop;" \
6116 ".pushsection trace_table;" \
6119 : : : : trace#NUM); \
6120 if (0) @{ trace#NUM: trace(); @} \
6122 #define TRACE TRACE1(__COUNTER__)
6125 In this example (which in fact inspired the @code{asm goto} feature)
6126 we want on rare occasions to call the @code{trace} function; on other
6127 occasions we'd like to keep the overhead to the absolute minimum.
6128 The normal code path consists of a single @code{nop} instruction.
6129 However, we record the address of this @code{nop} together with the
6130 address of a label that calls the @code{trace} function. This allows
6131 the @code{nop} instruction to be patched at runtime to be an
6132 unconditional branch to the stored label. It is assumed that an
6133 optimizing compiler will move the labeled block out of line, to
6134 optimize the fall through path from the @code{asm}.
6136 If you are writing a header file that should be includable in ISO C
6137 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
6140 @subsection Size of an @code{asm}
6142 Some targets require that GCC track the size of each instruction used in
6143 order to generate correct code. Because the final length of an
6144 @code{asm} is only known by the assembler, GCC must make an estimate as
6145 to how big it will be. The estimate is formed by counting the number of
6146 statements in the pattern of the @code{asm} and multiplying that by the
6147 length of the longest instruction on that processor. Statements in the
6148 @code{asm} are identified by newline characters and whatever statement
6149 separator characters are supported by the assembler; on most processors
6150 this is the `@code{;}' character.
6152 Normally, GCC's estimate is perfectly adequate to ensure that correct
6153 code is generated, but it is possible to confuse the compiler if you use
6154 pseudo instructions or assembler macros that expand into multiple real
6155 instructions or if you use assembler directives that expand to more
6156 space in the object file than would be needed for a single instruction.
6157 If this happens then the assembler will produce a diagnostic saying that
6158 a label is unreachable.
6160 @subsection i386 floating point asm operands
6162 There are several rules on the usage of stack-like regs in
6163 asm_operands insns. These rules apply only to the operands that are
6168 Given a set of input regs that die in an asm_operands, it is
6169 necessary to know which are implicitly popped by the asm, and
6170 which must be explicitly popped by gcc.
6172 An input reg that is implicitly popped by the asm must be
6173 explicitly clobbered, unless it is constrained to match an
6177 For any input reg that is implicitly popped by an asm, it is
6178 necessary to know how to adjust the stack to compensate for the pop.
6179 If any non-popped input is closer to the top of the reg-stack than
6180 the implicitly popped reg, it would not be possible to know what the
6181 stack looked like---it's not clear how the rest of the stack ``slides
6184 All implicitly popped input regs must be closer to the top of
6185 the reg-stack than any input that is not implicitly popped.
6187 It is possible that if an input dies in an insn, reload might
6188 use the input reg for an output reload. Consider this example:
6191 asm ("foo" : "=t" (a) : "f" (b));
6194 This asm says that input B is not popped by the asm, and that
6195 the asm pushes a result onto the reg-stack, i.e., the stack is one
6196 deeper after the asm than it was before. But, it is possible that
6197 reload will think that it can use the same reg for both the input and
6198 the output, if input B dies in this insn.
6200 If any input operand uses the @code{f} constraint, all output reg
6201 constraints must use the @code{&} earlyclobber.
6203 The asm above would be written as
6206 asm ("foo" : "=&t" (a) : "f" (b));
6210 Some operands need to be in particular places on the stack. All
6211 output operands fall in this category---there is no other way to
6212 know which regs the outputs appear in unless the user indicates
6213 this in the constraints.
6215 Output operands must specifically indicate which reg an output
6216 appears in after an asm. @code{=f} is not allowed: the operand
6217 constraints must select a class with a single reg.
6220 Output operands may not be ``inserted'' between existing stack regs.
6221 Since no 387 opcode uses a read/write operand, all output operands
6222 are dead before the asm_operands, and are pushed by the asm_operands.
6223 It makes no sense to push anywhere but the top of the reg-stack.
6225 Output operands must start at the top of the reg-stack: output
6226 operands may not ``skip'' a reg.
6229 Some asm statements may need extra stack space for internal
6230 calculations. This can be guaranteed by clobbering stack registers
6231 unrelated to the inputs and outputs.
6235 Here are a couple of reasonable asms to want to write. This asm
6236 takes one input, which is internally popped, and produces two outputs.
6239 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
6242 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
6243 and replaces them with one output. The user must code the @code{st(1)}
6244 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
6247 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
6253 @section Controlling Names Used in Assembler Code
6254 @cindex assembler names for identifiers
6255 @cindex names used in assembler code
6256 @cindex identifiers, names in assembler code
6258 You can specify the name to be used in the assembler code for a C
6259 function or variable by writing the @code{asm} (or @code{__asm__})
6260 keyword after the declarator as follows:
6263 int foo asm ("myfoo") = 2;
6267 This specifies that the name to be used for the variable @code{foo} in
6268 the assembler code should be @samp{myfoo} rather than the usual
6271 On systems where an underscore is normally prepended to the name of a C
6272 function or variable, this feature allows you to define names for the
6273 linker that do not start with an underscore.
6275 It does not make sense to use this feature with a non-static local
6276 variable since such variables do not have assembler names. If you are
6277 trying to put the variable in a particular register, see @ref{Explicit
6278 Reg Vars}. GCC presently accepts such code with a warning, but will
6279 probably be changed to issue an error, rather than a warning, in the
6282 You cannot use @code{asm} in this way in a function @emph{definition}; but
6283 you can get the same effect by writing a declaration for the function
6284 before its definition and putting @code{asm} there, like this:
6287 extern func () asm ("FUNC");
6294 It is up to you to make sure that the assembler names you choose do not
6295 conflict with any other assembler symbols. Also, you must not use a
6296 register name; that would produce completely invalid assembler code. GCC
6297 does not as yet have the ability to store static variables in registers.
6298 Perhaps that will be added.
6300 @node Explicit Reg Vars
6301 @section Variables in Specified Registers
6302 @cindex explicit register variables
6303 @cindex variables in specified registers
6304 @cindex specified registers
6305 @cindex registers, global allocation
6307 GNU C allows you to put a few global variables into specified hardware
6308 registers. You can also specify the register in which an ordinary
6309 register variable should be allocated.
6313 Global register variables reserve registers throughout the program.
6314 This may be useful in programs such as programming language
6315 interpreters which have a couple of global variables that are accessed
6319 Local register variables in specific registers do not reserve the
6320 registers, except at the point where they are used as input or output
6321 operands in an @code{asm} statement and the @code{asm} statement itself is
6322 not deleted. The compiler's data flow analysis is capable of determining
6323 where the specified registers contain live values, and where they are
6324 available for other uses. Stores into local register variables may be deleted
6325 when they appear to be dead according to dataflow analysis. References
6326 to local register variables may be deleted or moved or simplified.
6328 These local variables are sometimes convenient for use with the extended
6329 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
6330 output of the assembler instruction directly into a particular register.
6331 (This will work provided the register you specify fits the constraints
6332 specified for that operand in the @code{asm}.)
6340 @node Global Reg Vars
6341 @subsection Defining Global Register Variables
6342 @cindex global register variables
6343 @cindex registers, global variables in
6345 You can define a global register variable in GNU C like this:
6348 register int *foo asm ("a5");
6352 Here @code{a5} is the name of the register which should be used. Choose a
6353 register which is normally saved and restored by function calls on your
6354 machine, so that library routines will not clobber it.
6356 Naturally the register name is cpu-dependent, so you would need to
6357 conditionalize your program according to cpu type. The register
6358 @code{a5} would be a good choice on a 68000 for a variable of pointer
6359 type. On machines with register windows, be sure to choose a ``global''
6360 register that is not affected magically by the function call mechanism.
6362 In addition, operating systems on one type of cpu may differ in how they
6363 name the registers; then you would need additional conditionals. For
6364 example, some 68000 operating systems call this register @code{%a5}.
6366 Eventually there may be a way of asking the compiler to choose a register
6367 automatically, but first we need to figure out how it should choose and
6368 how to enable you to guide the choice. No solution is evident.
6370 Defining a global register variable in a certain register reserves that
6371 register entirely for this use, at least within the current compilation.
6372 The register will not be allocated for any other purpose in the functions
6373 in the current compilation. The register will not be saved and restored by
6374 these functions. Stores into this register are never deleted even if they
6375 would appear to be dead, but references may be deleted or moved or
6378 It is not safe to access the global register variables from signal
6379 handlers, or from more than one thread of control, because the system
6380 library routines may temporarily use the register for other things (unless
6381 you recompile them specially for the task at hand).
6383 @cindex @code{qsort}, and global register variables
6384 It is not safe for one function that uses a global register variable to
6385 call another such function @code{foo} by way of a third function
6386 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
6387 different source file in which the variable wasn't declared). This is
6388 because @code{lose} might save the register and put some other value there.
6389 For example, you can't expect a global register variable to be available in
6390 the comparison-function that you pass to @code{qsort}, since @code{qsort}
6391 might have put something else in that register. (If you are prepared to
6392 recompile @code{qsort} with the same global register variable, you can
6393 solve this problem.)
6395 If you want to recompile @code{qsort} or other source files which do not
6396 actually use your global register variable, so that they will not use that
6397 register for any other purpose, then it suffices to specify the compiler
6398 option @option{-ffixed-@var{reg}}. You need not actually add a global
6399 register declaration to their source code.
6401 A function which can alter the value of a global register variable cannot
6402 safely be called from a function compiled without this variable, because it
6403 could clobber the value the caller expects to find there on return.
6404 Therefore, the function which is the entry point into the part of the
6405 program that uses the global register variable must explicitly save and
6406 restore the value which belongs to its caller.
6408 @cindex register variable after @code{longjmp}
6409 @cindex global register after @code{longjmp}
6410 @cindex value after @code{longjmp}
6413 On most machines, @code{longjmp} will restore to each global register
6414 variable the value it had at the time of the @code{setjmp}. On some
6415 machines, however, @code{longjmp} will not change the value of global
6416 register variables. To be portable, the function that called @code{setjmp}
6417 should make other arrangements to save the values of the global register
6418 variables, and to restore them in a @code{longjmp}. This way, the same
6419 thing will happen regardless of what @code{longjmp} does.
6421 All global register variable declarations must precede all function
6422 definitions. If such a declaration could appear after function
6423 definitions, the declaration would be too late to prevent the register from
6424 being used for other purposes in the preceding functions.
6426 Global register variables may not have initial values, because an
6427 executable file has no means to supply initial contents for a register.
6429 On the SPARC, there are reports that g3 @dots{} g7 are suitable
6430 registers, but certain library functions, such as @code{getwd}, as well
6431 as the subroutines for division and remainder, modify g3 and g4. g1 and
6432 g2 are local temporaries.
6434 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
6435 Of course, it will not do to use more than a few of those.
6437 @node Local Reg Vars
6438 @subsection Specifying Registers for Local Variables
6439 @cindex local variables, specifying registers
6440 @cindex specifying registers for local variables
6441 @cindex registers for local variables
6443 You can define a local register variable with a specified register
6447 register int *foo asm ("a5");
6451 Here @code{a5} is the name of the register which should be used. Note
6452 that this is the same syntax used for defining global register
6453 variables, but for a local variable it would appear within a function.
6455 Naturally the register name is cpu-dependent, but this is not a
6456 problem, since specific registers are most often useful with explicit
6457 assembler instructions (@pxref{Extended Asm}). Both of these things
6458 generally require that you conditionalize your program according to
6461 In addition, operating systems on one type of cpu may differ in how they
6462 name the registers; then you would need additional conditionals. For
6463 example, some 68000 operating systems call this register @code{%a5}.
6465 Defining such a register variable does not reserve the register; it
6466 remains available for other uses in places where flow control determines
6467 the variable's value is not live.
6469 This option does not guarantee that GCC will generate code that has
6470 this variable in the register you specify at all times. You may not
6471 code an explicit reference to this register in the @emph{assembler
6472 instruction template} part of an @code{asm} statement and assume it will
6473 always refer to this variable. However, using the variable as an
6474 @code{asm} @emph{operand} guarantees that the specified register is used
6477 Stores into local register variables may be deleted when they appear to be dead
6478 according to dataflow analysis. References to local register variables may
6479 be deleted or moved or simplified.
6481 As for global register variables, it's recommended that you choose a
6482 register which is normally saved and restored by function calls on
6483 your machine, so that library routines will not clobber it. A common
6484 pitfall is to initialize multiple call-clobbered registers with
6485 arbitrary expressions, where a function call or library call for an
6486 arithmetic operator will overwrite a register value from a previous
6487 assignment, for example @code{r0} below:
6489 register int *p1 asm ("r0") = @dots{};
6490 register int *p2 asm ("r1") = @dots{};
6492 In those cases, a solution is to use a temporary variable for
6493 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
6495 @node Alternate Keywords
6496 @section Alternate Keywords
6497 @cindex alternate keywords
6498 @cindex keywords, alternate
6500 @option{-ansi} and the various @option{-std} options disable certain
6501 keywords. This causes trouble when you want to use GNU C extensions, or
6502 a general-purpose header file that should be usable by all programs,
6503 including ISO C programs. The keywords @code{asm}, @code{typeof} and
6504 @code{inline} are not available in programs compiled with
6505 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
6506 program compiled with @option{-std=c99} or @option{-std=c11}). The
6508 @code{restrict} is only available when @option{-std=gnu99} (which will
6509 eventually be the default) or @option{-std=c99} (or the equivalent
6510 @option{-std=iso9899:1999}), or an option for a later standard
6513 The way to solve these problems is to put @samp{__} at the beginning and
6514 end of each problematical keyword. For example, use @code{__asm__}
6515 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
6517 Other C compilers won't accept these alternative keywords; if you want to
6518 compile with another compiler, you can define the alternate keywords as
6519 macros to replace them with the customary keywords. It looks like this:
6527 @findex __extension__
6529 @option{-pedantic} and other options cause warnings for many GNU C extensions.
6531 prevent such warnings within one expression by writing
6532 @code{__extension__} before the expression. @code{__extension__} has no
6533 effect aside from this.
6535 @node Incomplete Enums
6536 @section Incomplete @code{enum} Types
6538 You can define an @code{enum} tag without specifying its possible values.
6539 This results in an incomplete type, much like what you get if you write
6540 @code{struct foo} without describing the elements. A later declaration
6541 which does specify the possible values completes the type.
6543 You can't allocate variables or storage using the type while it is
6544 incomplete. However, you can work with pointers to that type.
6546 This extension may not be very useful, but it makes the handling of
6547 @code{enum} more consistent with the way @code{struct} and @code{union}
6550 This extension is not supported by GNU C++.
6552 @node Function Names
6553 @section Function Names as Strings
6554 @cindex @code{__func__} identifier
6555 @cindex @code{__FUNCTION__} identifier
6556 @cindex @code{__PRETTY_FUNCTION__} identifier
6558 GCC provides three magic variables which hold the name of the current
6559 function, as a string. The first of these is @code{__func__}, which
6560 is part of the C99 standard:
6562 The identifier @code{__func__} is implicitly declared by the translator
6563 as if, immediately following the opening brace of each function
6564 definition, the declaration
6567 static const char __func__[] = "function-name";
6571 appeared, where function-name is the name of the lexically-enclosing
6572 function. This name is the unadorned name of the function.
6574 @code{__FUNCTION__} is another name for @code{__func__}. Older
6575 versions of GCC recognize only this name. However, it is not
6576 standardized. For maximum portability, we recommend you use
6577 @code{__func__}, but provide a fallback definition with the
6581 #if __STDC_VERSION__ < 199901L
6583 # define __func__ __FUNCTION__
6585 # define __func__ "<unknown>"
6590 In C, @code{__PRETTY_FUNCTION__} is yet another name for
6591 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
6592 the type signature of the function as well as its bare name. For
6593 example, this program:
6597 extern int printf (char *, ...);
6604 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
6605 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
6623 __PRETTY_FUNCTION__ = void a::sub(int)
6626 These identifiers are not preprocessor macros. In GCC 3.3 and
6627 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
6628 were treated as string literals; they could be used to initialize
6629 @code{char} arrays, and they could be concatenated with other string
6630 literals. GCC 3.4 and later treat them as variables, like
6631 @code{__func__}. In C++, @code{__FUNCTION__} and
6632 @code{__PRETTY_FUNCTION__} have always been variables.
6634 @node Return Address
6635 @section Getting the Return or Frame Address of a Function
6637 These functions may be used to get information about the callers of a
6640 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
6641 This function returns the return address of the current function, or of
6642 one of its callers. The @var{level} argument is number of frames to
6643 scan up the call stack. A value of @code{0} yields the return address
6644 of the current function, a value of @code{1} yields the return address
6645 of the caller of the current function, and so forth. When inlining
6646 the expected behavior is that the function will return the address of
6647 the function that will be returned to. To work around this behavior use
6648 the @code{noinline} function attribute.
6650 The @var{level} argument must be a constant integer.
6652 On some machines it may be impossible to determine the return address of
6653 any function other than the current one; in such cases, or when the top
6654 of the stack has been reached, this function will return @code{0} or a
6655 random value. In addition, @code{__builtin_frame_address} may be used
6656 to determine if the top of the stack has been reached.
6658 Additional post-processing of the returned value may be needed, see
6659 @code{__builtin_extract_return_address}.
6661 This function should only be used with a nonzero argument for debugging
6665 @deftypefn {Built-in Function} {void *} __builtin_extract_return_address (void *@var{addr})
6666 The address as returned by @code{__builtin_return_address} may have to be fed
6667 through this function to get the actual encoded address. For example, on the
6668 31-bit S/390 platform the highest bit has to be masked out, or on SPARC
6669 platforms an offset has to be added for the true next instruction to be
6672 If no fixup is needed, this function simply passes through @var{addr}.
6675 @deftypefn {Built-in Function} {void *} __builtin_frob_return_address (void *@var{addr})
6676 This function does the reverse of @code{__builtin_extract_return_address}.
6679 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
6680 This function is similar to @code{__builtin_return_address}, but it
6681 returns the address of the function frame rather than the return address
6682 of the function. Calling @code{__builtin_frame_address} with a value of
6683 @code{0} yields the frame address of the current function, a value of
6684 @code{1} yields the frame address of the caller of the current function,
6687 The frame is the area on the stack which holds local variables and saved
6688 registers. The frame address is normally the address of the first word
6689 pushed on to the stack by the function. However, the exact definition
6690 depends upon the processor and the calling convention. If the processor
6691 has a dedicated frame pointer register, and the function has a frame,
6692 then @code{__builtin_frame_address} will return the value of the frame
6695 On some machines it may be impossible to determine the frame address of
6696 any function other than the current one; in such cases, or when the top
6697 of the stack has been reached, this function will return @code{0} if
6698 the first frame pointer is properly initialized by the startup code.
6700 This function should only be used with a nonzero argument for debugging
6704 @node Vector Extensions
6705 @section Using vector instructions through built-in functions
6707 On some targets, the instruction set contains SIMD vector instructions that
6708 operate on multiple values contained in one large register at the same time.
6709 For example, on the i386 the MMX, 3DNow!@: and SSE extensions can be used
6712 The first step in using these extensions is to provide the necessary data
6713 types. This should be done using an appropriate @code{typedef}:
6716 typedef int v4si __attribute__ ((vector_size (16)));
6719 The @code{int} type specifies the base type, while the attribute specifies
6720 the vector size for the variable, measured in bytes. For example, the
6721 declaration above causes the compiler to set the mode for the @code{v4si}
6722 type to be 16 bytes wide and divided into @code{int} sized units. For
6723 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
6724 corresponding mode of @code{foo} will be @acronym{V4SI}.
6726 The @code{vector_size} attribute is only applicable to integral and
6727 float scalars, although arrays, pointers, and function return values
6728 are allowed in conjunction with this construct.
6730 All the basic integer types can be used as base types, both as signed
6731 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
6732 @code{long long}. In addition, @code{float} and @code{double} can be
6733 used to build floating-point vector types.
6735 Specifying a combination that is not valid for the current architecture
6736 will cause GCC to synthesize the instructions using a narrower mode.
6737 For example, if you specify a variable of type @code{V4SI} and your
6738 architecture does not allow for this specific SIMD type, GCC will
6739 produce code that uses 4 @code{SIs}.
6741 The types defined in this manner can be used with a subset of normal C
6742 operations. Currently, GCC will allow using the following operators
6743 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
6745 The operations behave like C++ @code{valarrays}. Addition is defined as
6746 the addition of the corresponding elements of the operands. For
6747 example, in the code below, each of the 4 elements in @var{a} will be
6748 added to the corresponding 4 elements in @var{b} and the resulting
6749 vector will be stored in @var{c}.
6752 typedef int v4si __attribute__ ((vector_size (16)));
6759 Subtraction, multiplication, division, and the logical operations
6760 operate in a similar manner. Likewise, the result of using the unary
6761 minus or complement operators on a vector type is a vector whose
6762 elements are the negative or complemented values of the corresponding
6763 elements in the operand.
6765 In C it is possible to use shifting operators @code{<<}, @code{>>} on
6766 integer-type vectors. The operation is defined as following: @code{@{a0,
6767 a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
6768 @dots{}, an >> bn@}}@. Vector operands must have the same number of
6771 For the convenience in C it is allowed to use a binary vector operation
6772 where one operand is a scalar. In that case the compiler will transform
6773 the scalar operand into a vector where each element is the scalar from
6774 the operation. The transformation will happen only if the scalar could be
6775 safely converted to the vector-element type.
6776 Consider the following code.
6779 typedef int v4si __attribute__ ((vector_size (16)));
6784 a = b + 1; /* a = b + @{1,1,1,1@}; */
6785 a = 2 * b; /* a = @{2,2,2,2@} * b; */
6787 a = l + a; /* Error, cannot convert long to int. */
6790 In C vectors can be subscripted as if the vector were an array with
6791 the same number of elements and base type. Out of bound accesses
6792 invoke undefined behavior at runtime. Warnings for out of bound
6793 accesses for vector subscription can be enabled with
6794 @option{-Warray-bounds}.
6796 In GNU C vector comparison is supported within standard comparison
6797 operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
6798 vector expressions of integer-type or real-type. Comparison between
6799 integer-type vectors and real-type vectors are not supported. The
6800 result of the comparison is a vector of the same width and number of
6801 elements as the comparison operands with a signed integral element
6804 Vectors are compared element-wise producing 0 when comparison is false
6805 and -1 (constant of the appropriate type where all bits are set)
6806 otherwise. Consider the following example.
6809 typedef int v4si __attribute__ ((vector_size (16)));
6811 v4si a = @{1,2,3,4@};
6812 v4si b = @{3,2,1,4@};
6815 c = a > b; /* The result would be @{0, 0,-1, 0@} */
6816 c = a == b; /* The result would be @{0,-1, 0,-1@} */
6819 Vector shuffling is available using functions
6820 @code{__builtin_shuffle (vec, mask)} and
6821 @code{__builtin_shuffle (vec0, vec1, mask)}.
6822 Both functions construct a permutation of elements from one or two
6823 vectors and return a vector of the same type as the input vector(s).
6824 The @var{mask} is an integral vector with the same width (@var{W})
6825 and element count (@var{N}) as the output vector.
6827 The elements of the input vectors are numbered in memory ordering of
6828 @var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}. The
6829 elements of @var{mask} are considered modulo @var{N} in the single-operand
6830 case and modulo @math{2*@var{N}} in the two-operand case.
6832 Consider the following example,
6835 typedef int v4si __attribute__ ((vector_size (16)));
6837 v4si a = @{1,2,3,4@};
6838 v4si b = @{5,6,7,8@};
6839 v4si mask1 = @{0,1,1,3@};
6840 v4si mask2 = @{0,4,2,5@};
6843 res = __builtin_shuffle (a, mask1); /* res is @{1,2,2,4@} */
6844 res = __builtin_shuffle (a, b, mask2); /* res is @{1,5,3,6@} */
6847 Note that @code{__builtin_shuffle} is intentionally semantically
6848 compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.
6850 You can declare variables and use them in function calls and returns, as
6851 well as in assignments and some casts. You can specify a vector type as
6852 a return type for a function. Vector types can also be used as function
6853 arguments. It is possible to cast from one vector type to another,
6854 provided they are of the same size (in fact, you can also cast vectors
6855 to and from other datatypes of the same size).
6857 You cannot operate between vectors of different lengths or different
6858 signedness without a cast.
6862 @findex __builtin_offsetof
6864 GCC implements for both C and C++ a syntactic extension to implement
6865 the @code{offsetof} macro.
6869 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
6871 offsetof_member_designator:
6873 | offsetof_member_designator "." @code{identifier}
6874 | offsetof_member_designator "[" @code{expr} "]"
6877 This extension is sufficient such that
6880 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
6883 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
6884 may be dependent. In either case, @var{member} may consist of a single
6885 identifier, or a sequence of member accesses and array references.
6887 @node __sync Builtins
6888 @section Legacy __sync built-in functions for atomic memory access
6890 The following builtins are intended to be compatible with those described
6891 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
6892 section 7.4. As such, they depart from the normal GCC practice of using
6893 the ``__builtin_'' prefix, and further that they are overloaded such that
6894 they work on multiple types.
6896 The definition given in the Intel documentation allows only for the use of
6897 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
6898 counterparts. GCC will allow any integral scalar or pointer type that is
6899 1, 2, 4 or 8 bytes in length.
6901 Not all operations are supported by all target processors. If a particular
6902 operation cannot be implemented on the target processor, a warning will be
6903 generated and a call an external function will be generated. The external
6904 function will carry the same name as the builtin, with an additional suffix
6905 @samp{_@var{n}} where @var{n} is the size of the data type.
6907 @c ??? Should we have a mechanism to suppress this warning? This is almost
6908 @c useful for implementing the operation under the control of an external
6911 In most cases, these builtins are considered a @dfn{full barrier}. That is,
6912 no memory operand will be moved across the operation, either forward or
6913 backward. Further, instructions will be issued as necessary to prevent the
6914 processor from speculating loads across the operation and from queuing stores
6915 after the operation.
6917 All of the routines are described in the Intel documentation to take
6918 ``an optional list of variables protected by the memory barrier''. It's
6919 not clear what is meant by that; it could mean that @emph{only} the
6920 following variables are protected, or it could mean that these variables
6921 should in addition be protected. At present GCC ignores this list and
6922 protects all variables which are globally accessible. If in the future
6923 we make some use of this list, an empty list will continue to mean all
6924 globally accessible variables.
6927 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
6928 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
6929 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
6930 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
6931 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
6932 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
6933 @findex __sync_fetch_and_add
6934 @findex __sync_fetch_and_sub
6935 @findex __sync_fetch_and_or
6936 @findex __sync_fetch_and_and
6937 @findex __sync_fetch_and_xor
6938 @findex __sync_fetch_and_nand
6939 These builtins perform the operation suggested by the name, and
6940 returns the value that had previously been in memory. That is,
6943 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
6944 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
6947 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
6948 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
6950 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
6951 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
6952 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
6953 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
6954 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
6955 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
6956 @findex __sync_add_and_fetch
6957 @findex __sync_sub_and_fetch
6958 @findex __sync_or_and_fetch
6959 @findex __sync_and_and_fetch
6960 @findex __sync_xor_and_fetch
6961 @findex __sync_nand_and_fetch
6962 These builtins perform the operation suggested by the name, and
6963 return the new value. That is,
6966 @{ *ptr @var{op}= value; return *ptr; @}
6967 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
6970 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
6971 builtin as @code{*ptr = ~(*ptr & value)} instead of
6972 @code{*ptr = ~*ptr & value}.
6974 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
6975 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
6976 @findex __sync_bool_compare_and_swap
6977 @findex __sync_val_compare_and_swap
6978 These builtins perform an atomic compare and swap. That is, if the current
6979 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
6982 The ``bool'' version returns true if the comparison is successful and
6983 @var{newval} was written. The ``val'' version returns the contents
6984 of @code{*@var{ptr}} before the operation.
6986 @item __sync_synchronize (...)
6987 @findex __sync_synchronize
6988 This builtin issues a full memory barrier.
6990 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
6991 @findex __sync_lock_test_and_set
6992 This builtin, as described by Intel, is not a traditional test-and-set
6993 operation, but rather an atomic exchange operation. It writes @var{value}
6994 into @code{*@var{ptr}}, and returns the previous contents of
6997 Many targets have only minimal support for such locks, and do not support
6998 a full exchange operation. In this case, a target may support reduced
6999 functionality here by which the @emph{only} valid value to store is the
7000 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
7001 is implementation defined.
7003 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
7004 This means that references after the builtin cannot move to (or be
7005 speculated to) before the builtin, but previous memory stores may not
7006 be globally visible yet, and previous memory loads may not yet be
7009 @item void __sync_lock_release (@var{type} *ptr, ...)
7010 @findex __sync_lock_release
7011 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
7012 Normally this means writing the constant 0 to @code{*@var{ptr}}.
7014 This builtin is not a full barrier, but rather a @dfn{release barrier}.
7015 This means that all previous memory stores are globally visible, and all
7016 previous memory loads have been satisfied, but following memory reads
7017 are not prevented from being speculated to before the barrier.
7020 @node __atomic Builtins
7021 @section Built-in functions for memory model aware atomic operations
7023 The following built-in functions approximately match the requirements for
7024 C++11 memory model. Many are similar to the @samp{__sync} prefixed built-in
7025 functions, but all also have a memory model parameter. These are all
7026 identified by being prefixed with @samp{__atomic}, and most are overloaded
7027 such that they work with multiple types.
7029 GCC will allow any integral scalar or pointer type that is 1, 2, 4, or 8
7030 bytes in length. 16-byte integral types are also allowed if
7031 @samp{__int128} (@pxref{__int128}) is supported by the architecture.
7033 Target architectures are encouraged to provide their own patterns for
7034 each of these built-in functions. If no target is provided, the original
7035 non-memory model set of @samp{__sync} atomic built-in functions will be
7036 utilized, along with any required synchronization fences surrounding it in
7037 order to achieve the proper behaviour. Execution in this case is subject
7038 to the same restrictions as those built-in functions.
7040 If there is no pattern or mechanism to provide a lock free instruction
7041 sequence, a call is made to an external routine with the same parameters
7042 to be resolved at runtime.
7044 The four non-arithmetic functions (load, store, exchange, and
7045 compare_exchange) all have a generic version as well. This generic
7046 version will work on any data type. If the data type size maps to one
7047 of the integral sizes which may have lock free support, the generic
7048 version will utilize the lock free built-in function. Otherwise an
7049 external call is left to be resolved at runtime. This external call will
7050 be the same format with the addition of a @samp{size_t} parameter inserted
7051 as the first parameter indicating the size of the object being pointed to.
7052 All objects must be the same size.
7054 There are 6 different memory models which can be specified. These map
7055 to the same names in the C++11 standard. Refer there or to the
7056 @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki on
7057 atomic synchronization} for more detailed definitions. These memory
7058 models integrate both barriers to code motion as well as synchronization
7059 requirements with other threads. These are listed in approximately
7060 ascending order of strength.
7063 @item __ATOMIC_RELAXED
7064 No barriers or synchronization.
7065 @item __ATOMIC_CONSUME
7066 Data dependency only for both barrier and synchronization with another
7068 @item __ATOMIC_ACQUIRE
7069 Barrier to hoisting of code and synchronizes with release (or stronger)
7070 semantic stores from another thread.
7071 @item __ATOMIC_RELEASE
7072 Barrier to sinking of code and synchronizes with acquire (or stronger)
7073 semantic loads from another thread.
7074 @item __ATOMIC_ACQ_REL
7075 Full barrier in both directions and synchronizes with acquire loads and
7076 release stores in another thread.
7077 @item __ATOMIC_SEQ_CST
7078 Full barrier in both directions and synchronizes with acquire loads and
7079 release stores in all threads.
7082 When implementing patterns for these built-in functions , the memory model
7083 parameter can be ignored as long as the pattern implements the most
7084 restrictive @code{__ATOMIC_SEQ_CST} model. Any of the other memory models
7085 will execute correctly with this memory model but they may not execute as
7086 efficiently as they could with a more appropriate implemention of the
7087 relaxed requirements.
7089 Note that the C++11 standard allows for the memory model parameter to be
7090 determined at runtime rather than at compile time. These built-in
7091 functions will map any runtime value to @code{__ATOMIC_SEQ_CST} rather
7092 than invoke a runtime library call or inline a switch statement. This is
7093 standard compliant, safe, and the simplest approach for now.
7095 @deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memmodel)
7096 This built-in function implements an atomic load operation. It returns the
7097 contents of @code{*@var{ptr}}.
7099 The valid memory model variants are
7100 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7101 and @code{__ATOMIC_CONSUME}.
7105 @deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memmodel)
7106 This is the generic version of an atomic load. It will return the
7107 contents of @code{*@var{ptr}} in @code{*@var{ret}}.
7111 @deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memmodel)
7112 This built-in function implements an atomic store operation. It writes
7113 @code{@var{val}} into @code{*@var{ptr}}.
7115 The valid memory model variants are
7116 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.
7120 @deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memmodel)
7121 This is the generic version of an atomic store. It will store the value
7122 of @code{*@var{val}} into @code{*@var{ptr}}.
7126 @deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memmodel)
7127 This built-in function implements an atomic exchange operation. It writes
7128 @var{val} into @code{*@var{ptr}}, and returns the previous contents of
7131 The valid memory model variants are
7132 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
7133 @code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.
7137 @deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memmodel)
7138 This is the generic version of an atomic exchange. It will store the
7139 contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
7140 of @code{*@var{ptr}} will be copied into @code{*@var{ret}}.
7144 @deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memmodel, int failure_memmodel)
7145 This built-in function implements an atomic compare and exchange operation.
7146 This compares the contents of @code{*@var{ptr}} with the contents of
7147 @code{*@var{expected}} and if equal, writes @var{desired} into
7148 @code{*@var{ptr}}. If they are not equal, the current contents of
7149 @code{*@var{ptr}} is written into @code{*@var{expected}}.
7151 True is returned if @code{*@var{desired}} is written into
7152 @code{*@var{ptr}} and the execution is considered to conform to the
7153 memory model specified by @var{success_memmodel}. There are no
7154 restrictions on what memory model can be used here.
7156 False is returned otherwise, and the execution is considered to conform
7157 to @var{failure_memmodel}. This memory model cannot be
7158 @code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}. It also cannot be a
7159 stronger model than that specified by @var{success_memmodel}.
7163 @deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memmodel, int failure_memmodel)
7164 This built-in function implements the generic version of
7165 @code{__atomic_compare_exchange}. The function is virtually identical to
7166 @code{__atomic_compare_exchange_n}, except the desired value is also a
7171 @deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7172 @deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7173 @deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7174 @deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7175 @deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7176 @deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memmodel)
7177 These built-in functions perform the operation suggested by the name, and
7178 return the result of the operation. That is,
7181 @{ *ptr @var{op}= val; return *ptr; @}
7184 All memory models are valid.
7188 @deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memmodel)
7189 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memmodel)
7190 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memmodel)
7191 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memmodel)
7192 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memmodel)
7193 @deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memmodel)
7194 These built-in functions perform the operation suggested by the name, and
7195 return the value that had previously been in @code{*@var{ptr}}. That is,
7198 @{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
7201 All memory models are valid.
7205 @deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memmodel)
7207 This built-in function performs an atomic test-and-set operation on
7208 the byte at @code{*@var{ptr}}. The byte is set to some implementation
7209 defined non-zero "set" value and the return value is @code{true} if and only
7210 if the previous contents were "set".
7212 All memory models are valid.
7216 @deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memmodel)
7218 This built-in function performs an atomic clear operation on
7219 @code{*@var{ptr}}. After the operation, @code{*@var{ptr}} will contain 0.
7221 The valid memory model variants are
7222 @code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
7223 @code{__ATOMIC_RELEASE}.
7227 @deftypefn {Built-in Function} void __atomic_thread_fence (int memmodel)
7229 This built-in function acts as a synchronization fence between threads
7230 based on the specified memory model.
7232 All memory orders are valid.
7236 @deftypefn {Built-in Function} void __atomic_signal_fence (int memmodel)
7238 This built-in function acts as a synchronization fence between a thread
7239 and signal handlers based in the same thread.
7241 All memory orders are valid.
7245 @deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size)
7247 This built-in function returns true if objects of size bytes will always
7248 generate lock free atomic instructions for the target architecture.
7249 Otherwise false is returned.
7251 size must resolve to a compile time constant.
7254 if (_atomic_always_lock_free (sizeof (long long)))
7259 @deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size)
7261 This built-in function returns true if objects of size bytes will always
7262 generate lock free atomic instructions for the target architecture. If
7263 it is not known to be lock free a call is made to a runtime routine named
7264 @code{__atomic_is_lock_free}.
7268 @node Object Size Checking
7269 @section Object Size Checking Builtins
7270 @findex __builtin_object_size
7271 @findex __builtin___memcpy_chk
7272 @findex __builtin___mempcpy_chk
7273 @findex __builtin___memmove_chk
7274 @findex __builtin___memset_chk
7275 @findex __builtin___strcpy_chk
7276 @findex __builtin___stpcpy_chk
7277 @findex __builtin___strncpy_chk
7278 @findex __builtin___strcat_chk
7279 @findex __builtin___strncat_chk
7280 @findex __builtin___sprintf_chk
7281 @findex __builtin___snprintf_chk
7282 @findex __builtin___vsprintf_chk
7283 @findex __builtin___vsnprintf_chk
7284 @findex __builtin___printf_chk
7285 @findex __builtin___vprintf_chk
7286 @findex __builtin___fprintf_chk
7287 @findex __builtin___vfprintf_chk
7289 GCC implements a limited buffer overflow protection mechanism
7290 that can prevent some buffer overflow attacks.
7292 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
7293 is a built-in construct that returns a constant number of bytes from
7294 @var{ptr} to the end of the object @var{ptr} pointer points to
7295 (if known at compile time). @code{__builtin_object_size} never evaluates
7296 its arguments for side-effects. If there are any side-effects in them, it
7297 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7298 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
7299 point to and all of them are known at compile time, the returned number
7300 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
7301 0 and minimum if nonzero. If it is not possible to determine which objects
7302 @var{ptr} points to at compile time, @code{__builtin_object_size} should
7303 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
7304 for @var{type} 2 or 3.
7306 @var{type} is an integer constant from 0 to 3. If the least significant
7307 bit is clear, objects are whole variables, if it is set, a closest
7308 surrounding subobject is considered the object a pointer points to.
7309 The second bit determines if maximum or minimum of remaining bytes
7313 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
7314 char *p = &var.buf1[1], *q = &var.b;
7316 /* Here the object p points to is var. */
7317 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
7318 /* The subobject p points to is var.buf1. */
7319 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
7320 /* The object q points to is var. */
7321 assert (__builtin_object_size (q, 0)
7322 == (char *) (&var + 1) - (char *) &var.b);
7323 /* The subobject q points to is var.b. */
7324 assert (__builtin_object_size (q, 1) == sizeof (var.b));
7328 There are built-in functions added for many common string operation
7329 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
7330 built-in is provided. This built-in has an additional last argument,
7331 which is the number of bytes remaining in object the @var{dest}
7332 argument points to or @code{(size_t) -1} if the size is not known.
7334 The built-in functions are optimized into the normal string functions
7335 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
7336 it is known at compile time that the destination object will not
7337 be overflown. If the compiler can determine at compile time the
7338 object will be always overflown, it issues a warning.
7340 The intended use can be e.g.
7344 #define bos0(dest) __builtin_object_size (dest, 0)
7345 #define memcpy(dest, src, n) \
7346 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
7350 /* It is unknown what object p points to, so this is optimized
7351 into plain memcpy - no checking is possible. */
7352 memcpy (p, "abcde", n);
7353 /* Destination is known and length too. It is known at compile
7354 time there will be no overflow. */
7355 memcpy (&buf[5], "abcde", 5);
7356 /* Destination is known, but the length is not known at compile time.
7357 This will result in __memcpy_chk call that can check for overflow
7359 memcpy (&buf[5], "abcde", n);
7360 /* Destination is known and it is known at compile time there will
7361 be overflow. There will be a warning and __memcpy_chk call that
7362 will abort the program at runtime. */
7363 memcpy (&buf[6], "abcde", 5);
7366 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
7367 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
7368 @code{strcat} and @code{strncat}.
7370 There are also checking built-in functions for formatted output functions.
7372 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
7373 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7374 const char *fmt, ...);
7375 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
7377 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
7378 const char *fmt, va_list ap);
7381 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
7382 etc.@: functions and can contain implementation specific flags on what
7383 additional security measures the checking function might take, such as
7384 handling @code{%n} differently.
7386 The @var{os} argument is the object size @var{s} points to, like in the
7387 other built-in functions. There is a small difference in the behavior
7388 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
7389 optimized into the non-checking functions only if @var{flag} is 0, otherwise
7390 the checking function is called with @var{os} argument set to
7393 In addition to this, there are checking built-in functions
7394 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
7395 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
7396 These have just one additional argument, @var{flag}, right before
7397 format string @var{fmt}. If the compiler is able to optimize them to
7398 @code{fputc} etc.@: functions, it will, otherwise the checking function
7399 should be called and the @var{flag} argument passed to it.
7401 @node Other Builtins
7402 @section Other built-in functions provided by GCC
7403 @cindex built-in functions
7404 @findex __builtin_fpclassify
7405 @findex __builtin_isfinite
7406 @findex __builtin_isnormal
7407 @findex __builtin_isgreater
7408 @findex __builtin_isgreaterequal
7409 @findex __builtin_isinf_sign
7410 @findex __builtin_isless
7411 @findex __builtin_islessequal
7412 @findex __builtin_islessgreater
7413 @findex __builtin_isunordered
7414 @findex __builtin_powi
7415 @findex __builtin_powif
7416 @findex __builtin_powil
7574 @findex fprintf_unlocked
7576 @findex fputs_unlocked
7693 @findex printf_unlocked
7725 @findex significandf
7726 @findex significandl
7797 GCC provides a large number of built-in functions other than the ones
7798 mentioned above. Some of these are for internal use in the processing
7799 of exceptions or variable-length argument lists and will not be
7800 documented here because they may change from time to time; we do not
7801 recommend general use of these functions.
7803 The remaining functions are provided for optimization purposes.
7805 @opindex fno-builtin
7806 GCC includes built-in versions of many of the functions in the standard
7807 C library. The versions prefixed with @code{__builtin_} will always be
7808 treated as having the same meaning as the C library function even if you
7809 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
7810 Many of these functions are only optimized in certain cases; if they are
7811 not optimized in a particular case, a call to the library function will
7816 Outside strict ISO C mode (@option{-ansi}, @option{-std=c90},
7817 @option{-std=c99} or @option{-std=c11}), the functions
7818 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
7819 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
7820 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
7821 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
7822 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
7823 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
7824 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
7825 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
7826 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
7827 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
7828 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
7829 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
7830 @code{signbitd64}, @code{signbitd128}, @code{significandf},
7831 @code{significandl}, @code{significand}, @code{sincosf},
7832 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
7833 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
7834 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
7835 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
7837 may be handled as built-in functions.
7838 All these functions have corresponding versions
7839 prefixed with @code{__builtin_}, which may be used even in strict C90
7842 The ISO C99 functions
7843 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
7844 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
7845 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
7846 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
7847 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
7848 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
7849 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
7850 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
7851 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
7852 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
7853 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
7854 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
7855 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
7856 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
7857 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
7858 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
7859 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
7860 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
7861 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
7862 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
7863 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
7864 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
7865 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
7866 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
7867 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
7868 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
7869 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
7870 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
7871 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
7872 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
7873 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
7874 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
7875 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
7876 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
7877 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
7878 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
7879 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
7880 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
7881 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
7882 are handled as built-in functions
7883 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7885 There are also built-in versions of the ISO C99 functions
7886 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
7887 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
7888 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
7889 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
7890 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
7891 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
7892 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
7893 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
7894 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
7895 that are recognized in any mode since ISO C90 reserves these names for
7896 the purpose to which ISO C99 puts them. All these functions have
7897 corresponding versions prefixed with @code{__builtin_}.
7899 The ISO C94 functions
7900 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
7901 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
7902 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
7904 are handled as built-in functions
7905 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c90}).
7907 The ISO C90 functions
7908 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
7909 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
7910 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
7911 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
7912 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
7913 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
7914 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
7915 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
7916 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
7917 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
7918 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
7919 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
7920 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
7921 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
7922 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
7923 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
7924 are all recognized as built-in functions unless
7925 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
7926 is specified for an individual function). All of these functions have
7927 corresponding versions prefixed with @code{__builtin_}.
7929 GCC provides built-in versions of the ISO C99 floating point comparison
7930 macros that avoid raising exceptions for unordered operands. They have
7931 the same names as the standard macros ( @code{isgreater},
7932 @code{isgreaterequal}, @code{isless}, @code{islessequal},
7933 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
7934 prefixed. We intend for a library implementor to be able to simply
7935 @code{#define} each standard macro to its built-in equivalent.
7936 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
7937 @code{isinf_sign} and @code{isnormal} built-ins used with
7938 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
7939 builtins appear both with and without the @code{__builtin_} prefix.
7941 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
7943 You can use the built-in function @code{__builtin_types_compatible_p} to
7944 determine whether two types are the same.
7946 This built-in function returns 1 if the unqualified versions of the
7947 types @var{type1} and @var{type2} (which are types, not expressions) are
7948 compatible, 0 otherwise. The result of this built-in function can be
7949 used in integer constant expressions.
7951 This built-in function ignores top level qualifiers (e.g., @code{const},
7952 @code{volatile}). For example, @code{int} is equivalent to @code{const
7955 The type @code{int[]} and @code{int[5]} are compatible. On the other
7956 hand, @code{int} and @code{char *} are not compatible, even if the size
7957 of their types, on the particular architecture are the same. Also, the
7958 amount of pointer indirection is taken into account when determining
7959 similarity. Consequently, @code{short *} is not similar to
7960 @code{short **}. Furthermore, two types that are typedefed are
7961 considered compatible if their underlying types are compatible.
7963 An @code{enum} type is not considered to be compatible with another
7964 @code{enum} type even if both are compatible with the same integer
7965 type; this is what the C standard specifies.
7966 For example, @code{enum @{foo, bar@}} is not similar to
7967 @code{enum @{hot, dog@}}.
7969 You would typically use this function in code whose execution varies
7970 depending on the arguments' types. For example:
7975 typeof (x) tmp = (x); \
7976 if (__builtin_types_compatible_p (typeof (x), long double)) \
7977 tmp = foo_long_double (tmp); \
7978 else if (__builtin_types_compatible_p (typeof (x), double)) \
7979 tmp = foo_double (tmp); \
7980 else if (__builtin_types_compatible_p (typeof (x), float)) \
7981 tmp = foo_float (tmp); \
7988 @emph{Note:} This construct is only available for C@.
7992 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
7994 You can use the built-in function @code{__builtin_choose_expr} to
7995 evaluate code depending on the value of a constant expression. This
7996 built-in function returns @var{exp1} if @var{const_exp}, which is an
7997 integer constant expression, is nonzero. Otherwise it returns @var{exp2}.
7999 This built-in function is analogous to the @samp{? :} operator in C,
8000 except that the expression returned has its type unaltered by promotion
8001 rules. Also, the built-in function does not evaluate the expression
8002 that was not chosen. For example, if @var{const_exp} evaluates to true,
8003 @var{exp2} is not evaluated even if it has side-effects.
8005 This built-in function can return an lvalue if the chosen argument is an
8008 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
8009 type. Similarly, if @var{exp2} is returned, its return type is the same
8016 __builtin_choose_expr ( \
8017 __builtin_types_compatible_p (typeof (x), double), \
8019 __builtin_choose_expr ( \
8020 __builtin_types_compatible_p (typeof (x), float), \
8022 /* @r{The void expression results in a compile-time error} \
8023 @r{when assigning the result to something.} */ \
8027 @emph{Note:} This construct is only available for C@. Furthermore, the
8028 unused expression (@var{exp1} or @var{exp2} depending on the value of
8029 @var{const_exp}) may still generate syntax errors. This may change in
8034 @deftypefn {Built-in Function} @var{type} __builtin_complex (@var{real}, @var{imag})
8036 The built-in function @code{__builtin_complex} is provided for use in
8037 implementing the ISO C11 macros @code{CMPLXF}, @code{CMPLX} and
8038 @code{CMPLXL}. @var{real} and @var{imag} must have the same type, a
8039 real binary floating-point type, and the result has the corresponding
8040 complex type with real and imaginary parts @var{real} and @var{imag}.
8041 Unlike @samp{@var{real} + I * @var{imag}}, this works even when
8042 infinities, NaNs and negative zeros are involved.
8046 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
8047 You can use the built-in function @code{__builtin_constant_p} to
8048 determine if a value is known to be constant at compile-time and hence
8049 that GCC can perform constant-folding on expressions involving that
8050 value. The argument of the function is the value to test. The function
8051 returns the integer 1 if the argument is known to be a compile-time
8052 constant and 0 if it is not known to be a compile-time constant. A
8053 return of 0 does not indicate that the value is @emph{not} a constant,
8054 but merely that GCC cannot prove it is a constant with the specified
8055 value of the @option{-O} option.
8057 You would typically use this function in an embedded application where
8058 memory was a critical resource. If you have some complex calculation,
8059 you may want it to be folded if it involves constants, but need to call
8060 a function if it does not. For example:
8063 #define Scale_Value(X) \
8064 (__builtin_constant_p (X) \
8065 ? ((X) * SCALE + OFFSET) : Scale (X))
8068 You may use this built-in function in either a macro or an inline
8069 function. However, if you use it in an inlined function and pass an
8070 argument of the function as the argument to the built-in, GCC will
8071 never return 1 when you call the inline function with a string constant
8072 or compound literal (@pxref{Compound Literals}) and will not return 1
8073 when you pass a constant numeric value to the inline function unless you
8074 specify the @option{-O} option.
8076 You may also use @code{__builtin_constant_p} in initializers for static
8077 data. For instance, you can write
8080 static const int table[] = @{
8081 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
8087 This is an acceptable initializer even if @var{EXPRESSION} is not a
8088 constant expression, including the case where
8089 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
8090 folded to a constant but @var{EXPRESSION} contains operands that would
8091 not otherwise be permitted in a static initializer (for example,
8092 @code{0 && foo ()}). GCC must be more conservative about evaluating the
8093 built-in in this case, because it has no opportunity to perform
8096 Previous versions of GCC did not accept this built-in in data
8097 initializers. The earliest version where it is completely safe is
8101 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
8102 @opindex fprofile-arcs
8103 You may use @code{__builtin_expect} to provide the compiler with
8104 branch prediction information. In general, you should prefer to
8105 use actual profile feedback for this (@option{-fprofile-arcs}), as
8106 programmers are notoriously bad at predicting how their programs
8107 actually perform. However, there are applications in which this
8108 data is hard to collect.
8110 The return value is the value of @var{exp}, which should be an integral
8111 expression. The semantics of the built-in are that it is expected that
8112 @var{exp} == @var{c}. For example:
8115 if (__builtin_expect (x, 0))
8120 would indicate that we do not expect to call @code{foo}, since
8121 we expect @code{x} to be zero. Since you are limited to integral
8122 expressions for @var{exp}, you should use constructions such as
8125 if (__builtin_expect (ptr != NULL, 1))
8130 when testing pointer or floating-point values.
8133 @deftypefn {Built-in Function} void __builtin_trap (void)
8134 This function causes the program to exit abnormally. GCC implements
8135 this function by using a target-dependent mechanism (such as
8136 intentionally executing an illegal instruction) or by calling
8137 @code{abort}. The mechanism used may vary from release to release so
8138 you should not rely on any particular implementation.
8141 @deftypefn {Built-in Function} void __builtin_unreachable (void)
8142 If control flow reaches the point of the @code{__builtin_unreachable},
8143 the program is undefined. It is useful in situations where the
8144 compiler cannot deduce the unreachability of the code.
8146 One such case is immediately following an @code{asm} statement that
8147 will either never terminate, or one that transfers control elsewhere
8148 and never returns. In this example, without the
8149 @code{__builtin_unreachable}, GCC would issue a warning that control
8150 reaches the end of a non-void function. It would also generate code
8151 to return after the @code{asm}.
8154 int f (int c, int v)
8162 asm("jmp error_handler");
8163 __builtin_unreachable ();
8168 Because the @code{asm} statement unconditionally transfers control out
8169 of the function, control will never reach the end of the function
8170 body. The @code{__builtin_unreachable} is in fact unreachable and
8171 communicates this fact to the compiler.
8173 Another use for @code{__builtin_unreachable} is following a call a
8174 function that never returns but that is not declared
8175 @code{__attribute__((noreturn))}, as in this example:
8178 void function_that_never_returns (void);
8188 function_that_never_returns ();
8189 __builtin_unreachable ();
8196 @deftypefn {Built-in Function} void *__builtin_assume_aligned (const void *@var{exp}, size_t @var{align}, ...)
8197 This function returns its first argument, and allows the compiler
8198 to assume that the returned pointer is at least @var{align} bytes
8199 aligned. This built-in can have either two or three arguments,
8200 if it has three, the third argument should have integer type, and
8201 if it is non-zero means misalignment offset. For example:
8204 void *x = __builtin_assume_aligned (arg, 16);
8207 means that the compiler can assume x, set to arg, is at least
8208 16 byte aligned, while:
8211 void *x = __builtin_assume_aligned (arg, 32, 8);
8214 means that the compiler can assume for x, set to arg, that
8215 (char *) x - 8 is 32 byte aligned.
8218 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
8219 This function is used to flush the processor's instruction cache for
8220 the region of memory between @var{begin} inclusive and @var{end}
8221 exclusive. Some targets require that the instruction cache be
8222 flushed, after modifying memory containing code, in order to obtain
8223 deterministic behavior.
8225 If the target does not require instruction cache flushes,
8226 @code{__builtin___clear_cache} has no effect. Otherwise either
8227 instructions are emitted in-line to clear the instruction cache or a
8228 call to the @code{__clear_cache} function in libgcc is made.
8231 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
8232 This function is used to minimize cache-miss latency by moving data into
8233 a cache before it is accessed.
8234 You can insert calls to @code{__builtin_prefetch} into code for which
8235 you know addresses of data in memory that is likely to be accessed soon.
8236 If the target supports them, data prefetch instructions will be generated.
8237 If the prefetch is done early enough before the access then the data will
8238 be in the cache by the time it is accessed.
8240 The value of @var{addr} is the address of the memory to prefetch.
8241 There are two optional arguments, @var{rw} and @var{locality}.
8242 The value of @var{rw} is a compile-time constant one or zero; one
8243 means that the prefetch is preparing for a write to the memory address
8244 and zero, the default, means that the prefetch is preparing for a read.
8245 The value @var{locality} must be a compile-time constant integer between
8246 zero and three. A value of zero means that the data has no temporal
8247 locality, so it need not be left in the cache after the access. A value
8248 of three means that the data has a high degree of temporal locality and
8249 should be left in all levels of cache possible. Values of one and two
8250 mean, respectively, a low or moderate degree of temporal locality. The
8254 for (i = 0; i < n; i++)
8257 __builtin_prefetch (&a[i+j], 1, 1);
8258 __builtin_prefetch (&b[i+j], 0, 1);
8263 Data prefetch does not generate faults if @var{addr} is invalid, but
8264 the address expression itself must be valid. For example, a prefetch
8265 of @code{p->next} will not fault if @code{p->next} is not a valid
8266 address, but evaluation will fault if @code{p} is not a valid address.
8268 If the target does not support data prefetch, the address expression
8269 is evaluated if it includes side effects but no other code is generated
8270 and GCC does not issue a warning.
8273 @deftypefn {Built-in Function} double __builtin_huge_val (void)
8274 Returns a positive infinity, if supported by the floating-point format,
8275 else @code{DBL_MAX}. This function is suitable for implementing the
8276 ISO C macro @code{HUGE_VAL}.
8279 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
8280 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
8283 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
8284 Similar to @code{__builtin_huge_val}, except the return
8285 type is @code{long double}.
8288 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
8289 This built-in implements the C99 fpclassify functionality. The first
8290 five int arguments should be the target library's notion of the
8291 possible FP classes and are used for return values. They must be
8292 constant values and they must appear in this order: @code{FP_NAN},
8293 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
8294 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
8295 to classify. GCC treats the last argument as type-generic, which
8296 means it does not do default promotion from float to double.
8299 @deftypefn {Built-in Function} double __builtin_inf (void)
8300 Similar to @code{__builtin_huge_val}, except a warning is generated
8301 if the target floating-point format does not support infinities.
8304 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
8305 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
8308 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
8309 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
8312 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
8313 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
8316 @deftypefn {Built-in Function} float __builtin_inff (void)
8317 Similar to @code{__builtin_inf}, except the return type is @code{float}.
8318 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
8321 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
8322 Similar to @code{__builtin_inf}, except the return
8323 type is @code{long double}.
8326 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
8327 Similar to @code{isinf}, except the return value will be negative for
8328 an argument of @code{-Inf}. Note while the parameter list is an
8329 ellipsis, this function only accepts exactly one floating point
8330 argument. GCC treats this parameter as type-generic, which means it
8331 does not do default promotion from float to double.
8334 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
8335 This is an implementation of the ISO C99 function @code{nan}.
8337 Since ISO C99 defines this function in terms of @code{strtod}, which we
8338 do not implement, a description of the parsing is in order. The string
8339 is parsed as by @code{strtol}; that is, the base is recognized by
8340 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
8341 in the significand such that the least significant bit of the number
8342 is at the least significant bit of the significand. The number is
8343 truncated to fit the significand field provided. The significand is
8344 forced to be a quiet NaN@.
8346 This function, if given a string literal all of which would have been
8347 consumed by strtol, is evaluated early enough that it is considered a
8348 compile-time constant.
8351 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
8352 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
8355 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
8356 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
8359 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
8360 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
8363 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
8364 Similar to @code{__builtin_nan}, except the return type is @code{float}.
8367 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
8368 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
8371 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
8372 Similar to @code{__builtin_nan}, except the significand is forced
8373 to be a signaling NaN@. The @code{nans} function is proposed by
8374 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
8377 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
8378 Similar to @code{__builtin_nans}, except the return type is @code{float}.
8381 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
8382 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
8385 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
8386 Returns one plus the index of the least significant 1-bit of @var{x}, or
8387 if @var{x} is zero, returns zero.
8390 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
8391 Returns the number of leading 0-bits in @var{x}, starting at the most
8392 significant bit position. If @var{x} is 0, the result is undefined.
8395 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
8396 Returns the number of trailing 0-bits in @var{x}, starting at the least
8397 significant bit position. If @var{x} is 0, the result is undefined.
8400 @deftypefn {Built-in Function} int __builtin_clrsb (int x)
8401 Returns the number of leading redundant sign bits in @var{x}, i.e. the
8402 number of bits following the most significant bit which are identical
8403 to it. There are no special cases for 0 or other values.
8406 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
8407 Returns the number of 1-bits in @var{x}.
8410 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
8411 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
8415 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
8416 Similar to @code{__builtin_ffs}, except the argument type is
8417 @code{unsigned long}.
8420 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
8421 Similar to @code{__builtin_clz}, except the argument type is
8422 @code{unsigned long}.
8425 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
8426 Similar to @code{__builtin_ctz}, except the argument type is
8427 @code{unsigned long}.
8430 @deftypefn {Built-in Function} int __builtin_clrsbl (long)
8431 Similar to @code{__builtin_clrsb}, except the argument type is
8435 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
8436 Similar to @code{__builtin_popcount}, except the argument type is
8437 @code{unsigned long}.
8440 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
8441 Similar to @code{__builtin_parity}, except the argument type is
8442 @code{unsigned long}.
8445 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
8446 Similar to @code{__builtin_ffs}, except the argument type is
8447 @code{unsigned long long}.
8450 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
8451 Similar to @code{__builtin_clz}, except the argument type is
8452 @code{unsigned long long}.
8455 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
8456 Similar to @code{__builtin_ctz}, except the argument type is
8457 @code{unsigned long long}.
8460 @deftypefn {Built-in Function} int __builtin_clrsbll (long long)
8461 Similar to @code{__builtin_clrsb}, except the argument type is
8465 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
8466 Similar to @code{__builtin_popcount}, except the argument type is
8467 @code{unsigned long long}.
8470 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
8471 Similar to @code{__builtin_parity}, except the argument type is
8472 @code{unsigned long long}.
8475 @deftypefn {Built-in Function} double __builtin_powi (double, int)
8476 Returns the first argument raised to the power of the second. Unlike the
8477 @code{pow} function no guarantees about precision and rounding are made.
8480 @deftypefn {Built-in Function} float __builtin_powif (float, int)
8481 Similar to @code{__builtin_powi}, except the argument and return types
8485 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
8486 Similar to @code{__builtin_powi}, except the argument and return types
8487 are @code{long double}.
8490 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
8491 Returns @var{x} with the order of the bytes reversed; for example,
8492 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
8496 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
8497 Similar to @code{__builtin_bswap32}, except the argument and return types
8501 @node Target Builtins
8502 @section Built-in Functions Specific to Particular Target Machines
8504 On some target machines, GCC supports many built-in functions specific
8505 to those machines. Generally these generate calls to specific machine
8506 instructions, but allow the compiler to schedule those calls.
8509 * Alpha Built-in Functions::
8510 * ARM iWMMXt Built-in Functions::
8511 * ARM NEON Intrinsics::
8512 * AVR Built-in Functions::
8513 * Blackfin Built-in Functions::
8514 * FR-V Built-in Functions::
8515 * X86 Built-in Functions::
8516 * MIPS DSP Built-in Functions::
8517 * MIPS Paired-Single Support::
8518 * MIPS Loongson Built-in Functions::
8519 * Other MIPS Built-in Functions::
8520 * picoChip Built-in Functions::
8521 * PowerPC AltiVec/VSX Built-in Functions::
8522 * RX Built-in Functions::
8523 * SPARC VIS Built-in Functions::
8524 * SPU Built-in Functions::
8525 * TI C6X Built-in Functions::
8528 @node Alpha Built-in Functions
8529 @subsection Alpha Built-in Functions
8531 These built-in functions are available for the Alpha family of
8532 processors, depending on the command-line switches used.
8534 The following built-in functions are always available. They
8535 all generate the machine instruction that is part of the name.
8538 long __builtin_alpha_implver (void)
8539 long __builtin_alpha_rpcc (void)
8540 long __builtin_alpha_amask (long)
8541 long __builtin_alpha_cmpbge (long, long)
8542 long __builtin_alpha_extbl (long, long)
8543 long __builtin_alpha_extwl (long, long)
8544 long __builtin_alpha_extll (long, long)
8545 long __builtin_alpha_extql (long, long)
8546 long __builtin_alpha_extwh (long, long)
8547 long __builtin_alpha_extlh (long, long)
8548 long __builtin_alpha_extqh (long, long)
8549 long __builtin_alpha_insbl (long, long)
8550 long __builtin_alpha_inswl (long, long)
8551 long __builtin_alpha_insll (long, long)
8552 long __builtin_alpha_insql (long, long)
8553 long __builtin_alpha_inswh (long, long)
8554 long __builtin_alpha_inslh (long, long)
8555 long __builtin_alpha_insqh (long, long)
8556 long __builtin_alpha_mskbl (long, long)
8557 long __builtin_alpha_mskwl (long, long)
8558 long __builtin_alpha_mskll (long, long)
8559 long __builtin_alpha_mskql (long, long)
8560 long __builtin_alpha_mskwh (long, long)
8561 long __builtin_alpha_msklh (long, long)
8562 long __builtin_alpha_mskqh (long, long)
8563 long __builtin_alpha_umulh (long, long)
8564 long __builtin_alpha_zap (long, long)
8565 long __builtin_alpha_zapnot (long, long)
8568 The following built-in functions are always with @option{-mmax}
8569 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
8570 later. They all generate the machine instruction that is part
8574 long __builtin_alpha_pklb (long)
8575 long __builtin_alpha_pkwb (long)
8576 long __builtin_alpha_unpkbl (long)
8577 long __builtin_alpha_unpkbw (long)
8578 long __builtin_alpha_minub8 (long, long)
8579 long __builtin_alpha_minsb8 (long, long)
8580 long __builtin_alpha_minuw4 (long, long)
8581 long __builtin_alpha_minsw4 (long, long)
8582 long __builtin_alpha_maxub8 (long, long)
8583 long __builtin_alpha_maxsb8 (long, long)
8584 long __builtin_alpha_maxuw4 (long, long)
8585 long __builtin_alpha_maxsw4 (long, long)
8586 long __builtin_alpha_perr (long, long)
8589 The following built-in functions are always with @option{-mcix}
8590 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
8591 later. They all generate the machine instruction that is part
8595 long __builtin_alpha_cttz (long)
8596 long __builtin_alpha_ctlz (long)
8597 long __builtin_alpha_ctpop (long)
8600 The following builtins are available on systems that use the OSF/1
8601 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
8602 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
8603 @code{rdval} and @code{wrval}.
8606 void *__builtin_thread_pointer (void)
8607 void __builtin_set_thread_pointer (void *)
8610 @node ARM iWMMXt Built-in Functions
8611 @subsection ARM iWMMXt Built-in Functions
8613 These built-in functions are available for the ARM family of
8614 processors when the @option{-mcpu=iwmmxt} switch is used:
8617 typedef int v2si __attribute__ ((vector_size (8)));
8618 typedef short v4hi __attribute__ ((vector_size (8)));
8619 typedef char v8qi __attribute__ ((vector_size (8)));
8621 int __builtin_arm_getwcx (int)
8622 void __builtin_arm_setwcx (int, int)
8623 int __builtin_arm_textrmsb (v8qi, int)
8624 int __builtin_arm_textrmsh (v4hi, int)
8625 int __builtin_arm_textrmsw (v2si, int)
8626 int __builtin_arm_textrmub (v8qi, int)
8627 int __builtin_arm_textrmuh (v4hi, int)
8628 int __builtin_arm_textrmuw (v2si, int)
8629 v8qi __builtin_arm_tinsrb (v8qi, int)
8630 v4hi __builtin_arm_tinsrh (v4hi, int)
8631 v2si __builtin_arm_tinsrw (v2si, int)
8632 long long __builtin_arm_tmia (long long, int, int)
8633 long long __builtin_arm_tmiabb (long long, int, int)
8634 long long __builtin_arm_tmiabt (long long, int, int)
8635 long long __builtin_arm_tmiaph (long long, int, int)
8636 long long __builtin_arm_tmiatb (long long, int, int)
8637 long long __builtin_arm_tmiatt (long long, int, int)
8638 int __builtin_arm_tmovmskb (v8qi)
8639 int __builtin_arm_tmovmskh (v4hi)
8640 int __builtin_arm_tmovmskw (v2si)
8641 long long __builtin_arm_waccb (v8qi)
8642 long long __builtin_arm_wacch (v4hi)
8643 long long __builtin_arm_waccw (v2si)
8644 v8qi __builtin_arm_waddb (v8qi, v8qi)
8645 v8qi __builtin_arm_waddbss (v8qi, v8qi)
8646 v8qi __builtin_arm_waddbus (v8qi, v8qi)
8647 v4hi __builtin_arm_waddh (v4hi, v4hi)
8648 v4hi __builtin_arm_waddhss (v4hi, v4hi)
8649 v4hi __builtin_arm_waddhus (v4hi, v4hi)
8650 v2si __builtin_arm_waddw (v2si, v2si)
8651 v2si __builtin_arm_waddwss (v2si, v2si)
8652 v2si __builtin_arm_waddwus (v2si, v2si)
8653 v8qi __builtin_arm_walign (v8qi, v8qi, int)
8654 long long __builtin_arm_wand(long long, long long)
8655 long long __builtin_arm_wandn (long long, long long)
8656 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
8657 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
8658 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
8659 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
8660 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
8661 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
8662 v2si __builtin_arm_wcmpeqw (v2si, v2si)
8663 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
8664 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
8665 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
8666 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
8667 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
8668 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
8669 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
8670 long long __builtin_arm_wmacsz (v4hi, v4hi)
8671 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
8672 long long __builtin_arm_wmacuz (v4hi, v4hi)
8673 v4hi __builtin_arm_wmadds (v4hi, v4hi)
8674 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
8675 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
8676 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
8677 v2si __builtin_arm_wmaxsw (v2si, v2si)
8678 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
8679 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
8680 v2si __builtin_arm_wmaxuw (v2si, v2si)
8681 v8qi __builtin_arm_wminsb (v8qi, v8qi)
8682 v4hi __builtin_arm_wminsh (v4hi, v4hi)
8683 v2si __builtin_arm_wminsw (v2si, v2si)
8684 v8qi __builtin_arm_wminub (v8qi, v8qi)
8685 v4hi __builtin_arm_wminuh (v4hi, v4hi)
8686 v2si __builtin_arm_wminuw (v2si, v2si)
8687 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
8688 v4hi __builtin_arm_wmulul (v4hi, v4hi)
8689 v4hi __builtin_arm_wmulum (v4hi, v4hi)
8690 long long __builtin_arm_wor (long long, long long)
8691 v2si __builtin_arm_wpackdss (long long, long long)
8692 v2si __builtin_arm_wpackdus (long long, long long)
8693 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
8694 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
8695 v4hi __builtin_arm_wpackwss (v2si, v2si)
8696 v4hi __builtin_arm_wpackwus (v2si, v2si)
8697 long long __builtin_arm_wrord (long long, long long)
8698 long long __builtin_arm_wrordi (long long, int)
8699 v4hi __builtin_arm_wrorh (v4hi, long long)
8700 v4hi __builtin_arm_wrorhi (v4hi, int)
8701 v2si __builtin_arm_wrorw (v2si, long long)
8702 v2si __builtin_arm_wrorwi (v2si, int)
8703 v2si __builtin_arm_wsadb (v8qi, v8qi)
8704 v2si __builtin_arm_wsadbz (v8qi, v8qi)
8705 v2si __builtin_arm_wsadh (v4hi, v4hi)
8706 v2si __builtin_arm_wsadhz (v4hi, v4hi)
8707 v4hi __builtin_arm_wshufh (v4hi, int)
8708 long long __builtin_arm_wslld (long long, long long)
8709 long long __builtin_arm_wslldi (long long, int)
8710 v4hi __builtin_arm_wsllh (v4hi, long long)
8711 v4hi __builtin_arm_wsllhi (v4hi, int)
8712 v2si __builtin_arm_wsllw (v2si, long long)
8713 v2si __builtin_arm_wsllwi (v2si, int)
8714 long long __builtin_arm_wsrad (long long, long long)
8715 long long __builtin_arm_wsradi (long long, int)
8716 v4hi __builtin_arm_wsrah (v4hi, long long)
8717 v4hi __builtin_arm_wsrahi (v4hi, int)
8718 v2si __builtin_arm_wsraw (v2si, long long)
8719 v2si __builtin_arm_wsrawi (v2si, int)
8720 long long __builtin_arm_wsrld (long long, long long)
8721 long long __builtin_arm_wsrldi (long long, int)
8722 v4hi __builtin_arm_wsrlh (v4hi, long long)
8723 v4hi __builtin_arm_wsrlhi (v4hi, int)
8724 v2si __builtin_arm_wsrlw (v2si, long long)
8725 v2si __builtin_arm_wsrlwi (v2si, int)
8726 v8qi __builtin_arm_wsubb (v8qi, v8qi)
8727 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
8728 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
8729 v4hi __builtin_arm_wsubh (v4hi, v4hi)
8730 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
8731 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
8732 v2si __builtin_arm_wsubw (v2si, v2si)
8733 v2si __builtin_arm_wsubwss (v2si, v2si)
8734 v2si __builtin_arm_wsubwus (v2si, v2si)
8735 v4hi __builtin_arm_wunpckehsb (v8qi)
8736 v2si __builtin_arm_wunpckehsh (v4hi)
8737 long long __builtin_arm_wunpckehsw (v2si)
8738 v4hi __builtin_arm_wunpckehub (v8qi)
8739 v2si __builtin_arm_wunpckehuh (v4hi)
8740 long long __builtin_arm_wunpckehuw (v2si)
8741 v4hi __builtin_arm_wunpckelsb (v8qi)
8742 v2si __builtin_arm_wunpckelsh (v4hi)
8743 long long __builtin_arm_wunpckelsw (v2si)
8744 v4hi __builtin_arm_wunpckelub (v8qi)
8745 v2si __builtin_arm_wunpckeluh (v4hi)
8746 long long __builtin_arm_wunpckeluw (v2si)
8747 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
8748 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
8749 v2si __builtin_arm_wunpckihw (v2si, v2si)
8750 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
8751 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
8752 v2si __builtin_arm_wunpckilw (v2si, v2si)
8753 long long __builtin_arm_wxor (long long, long long)
8754 long long __builtin_arm_wzero ()
8757 @node ARM NEON Intrinsics
8758 @subsection ARM NEON Intrinsics
8760 These built-in intrinsics for the ARM Advanced SIMD extension are available
8761 when the @option{-mfpu=neon} switch is used:
8763 @include arm-neon-intrinsics.texi
8765 @node AVR Built-in Functions
8766 @subsection AVR Built-in Functions
8768 For each built-in function for AVR, there is an equally named,
8769 uppercase built-in macro defined. That way users can easily query if
8770 or if not a specific built-in is implemented or not. For example, if
8771 @code{__builtin_avr_nop} is available the macro
8772 @code{__BUILTIN_AVR_NOP} is defined to @code{1} and undefined otherwise.
8774 The following built-in functions map to the respective machine
8775 instruction, i.e. @code{nop}, @code{sei}, @code{cli}, @code{sleep},
8776 @code{wdr}, @code{swap}, @code{fmul}, @code{fmuls}
8777 resp. @code{fmulsu}. The three @code{fmul*} built-ins are implemented
8778 as library call if no hardware multiplier is available.
8781 void __builtin_avr_nop (void)
8782 void __builtin_avr_sei (void)
8783 void __builtin_avr_cli (void)
8784 void __builtin_avr_sleep (void)
8785 void __builtin_avr_wdr (void)
8786 unsigned char __builtin_avr_swap (unsigned char)
8787 unsigned int __builtin_avr_fmul (unsigned char, unsigned char)
8788 int __builtin_avr_fmuls (char, char)
8789 int __builtin_avr_fmulsu (char, unsigned char)
8792 In order to delay execution for a specific number of cycles, GCC
8795 void __builtin_avr_delay_cycles (unsigned long ticks)
8799 @code{ticks} is the number of ticks to delay execution. Note that this
8800 built-in does not take into account the effect of interrupts which
8801 might increase delay time. @code{ticks} must be a compile time
8802 integer constant; delays with a variable number of cycles are not supported.
8805 unsigned char __builtin_avr_map8 (unsigned long map, unsigned char val)
8809 Each bit of the result is copied from a specific bit of @code{val}.
8810 @code{map} is a compile time constant that represents a map composed
8811 of 8 nibbles (4-bit groups):
8812 The @var{n}-th nibble of @code{map} specifies which bit of @code{val}
8813 is to be moved to the @var{n}-th bit of the result.
8814 For example, @code{map = 0x76543210} represents identity: The MSB of
8815 the result is read from the 7-th bit of @code{val}, the LSB is
8816 read from the 0-th bit to @code{val}, etc.
8817 Two more examples: @code{0x01234567} reverses the bit order and
8818 @code{0x32107654} is equivalent to a @code{swap} instruction.
8821 One typical use case for this and the following built-in is adjusting input and
8822 output values to non-contiguous port layouts.
8825 unsigned int __builtin_avr_map16 (unsigned long long map, unsigned int val)
8829 Similar to the previous built-in except that it operates on @code{int}
8830 and thus 16 bits are involved. Again, @code{map} must be a compile
8833 @node Blackfin Built-in Functions
8834 @subsection Blackfin Built-in Functions
8836 Currently, there are two Blackfin-specific built-in functions. These are
8837 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
8838 using inline assembly; by using these built-in functions the compiler can
8839 automatically add workarounds for hardware errata involving these
8840 instructions. These functions are named as follows:
8843 void __builtin_bfin_csync (void)
8844 void __builtin_bfin_ssync (void)
8847 @node FR-V Built-in Functions
8848 @subsection FR-V Built-in Functions
8850 GCC provides many FR-V-specific built-in functions. In general,
8851 these functions are intended to be compatible with those described
8852 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
8853 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
8854 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
8855 pointer rather than by value.
8857 Most of the functions are named after specific FR-V instructions.
8858 Such functions are said to be ``directly mapped'' and are summarized
8859 here in tabular form.
8863 * Directly-mapped Integer Functions::
8864 * Directly-mapped Media Functions::
8865 * Raw read/write Functions::
8866 * Other Built-in Functions::
8869 @node Argument Types
8870 @subsubsection Argument Types
8872 The arguments to the built-in functions can be divided into three groups:
8873 register numbers, compile-time constants and run-time values. In order
8874 to make this classification clear at a glance, the arguments and return
8875 values are given the following pseudo types:
8877 @multitable @columnfractions .20 .30 .15 .35
8878 @item Pseudo type @tab Real C type @tab Constant? @tab Description
8879 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
8880 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
8881 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
8882 @item @code{uw2} @tab @code{unsigned long long} @tab No
8883 @tab an unsigned doubleword
8884 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
8885 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
8886 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
8887 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
8890 These pseudo types are not defined by GCC, they are simply a notational
8891 convenience used in this manual.
8893 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
8894 and @code{sw2} are evaluated at run time. They correspond to
8895 register operands in the underlying FR-V instructions.
8897 @code{const} arguments represent immediate operands in the underlying
8898 FR-V instructions. They must be compile-time constants.
8900 @code{acc} arguments are evaluated at compile time and specify the number
8901 of an accumulator register. For example, an @code{acc} argument of 2
8902 will select the ACC2 register.
8904 @code{iacc} arguments are similar to @code{acc} arguments but specify the
8905 number of an IACC register. See @pxref{Other Built-in Functions}
8908 @node Directly-mapped Integer Functions
8909 @subsubsection Directly-mapped Integer Functions
8911 The functions listed below map directly to FR-V I-type instructions.
8913 @multitable @columnfractions .45 .32 .23
8914 @item Function prototype @tab Example usage @tab Assembly output
8915 @item @code{sw1 __ADDSS (sw1, sw1)}
8916 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
8917 @tab @code{ADDSS @var{a},@var{b},@var{c}}
8918 @item @code{sw1 __SCAN (sw1, sw1)}
8919 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
8920 @tab @code{SCAN @var{a},@var{b},@var{c}}
8921 @item @code{sw1 __SCUTSS (sw1)}
8922 @tab @code{@var{b} = __SCUTSS (@var{a})}
8923 @tab @code{SCUTSS @var{a},@var{b}}
8924 @item @code{sw1 __SLASS (sw1, sw1)}
8925 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
8926 @tab @code{SLASS @var{a},@var{b},@var{c}}
8927 @item @code{void __SMASS (sw1, sw1)}
8928 @tab @code{__SMASS (@var{a}, @var{b})}
8929 @tab @code{SMASS @var{a},@var{b}}
8930 @item @code{void __SMSSS (sw1, sw1)}
8931 @tab @code{__SMSSS (@var{a}, @var{b})}
8932 @tab @code{SMSSS @var{a},@var{b}}
8933 @item @code{void __SMU (sw1, sw1)}
8934 @tab @code{__SMU (@var{a}, @var{b})}
8935 @tab @code{SMU @var{a},@var{b}}
8936 @item @code{sw2 __SMUL (sw1, sw1)}
8937 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
8938 @tab @code{SMUL @var{a},@var{b},@var{c}}
8939 @item @code{sw1 __SUBSS (sw1, sw1)}
8940 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
8941 @tab @code{SUBSS @var{a},@var{b},@var{c}}
8942 @item @code{uw2 __UMUL (uw1, uw1)}
8943 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
8944 @tab @code{UMUL @var{a},@var{b},@var{c}}
8947 @node Directly-mapped Media Functions
8948 @subsubsection Directly-mapped Media Functions
8950 The functions listed below map directly to FR-V M-type instructions.
8952 @multitable @columnfractions .45 .32 .23
8953 @item Function prototype @tab Example usage @tab Assembly output
8954 @item @code{uw1 __MABSHS (sw1)}
8955 @tab @code{@var{b} = __MABSHS (@var{a})}
8956 @tab @code{MABSHS @var{a},@var{b}}
8957 @item @code{void __MADDACCS (acc, acc)}
8958 @tab @code{__MADDACCS (@var{b}, @var{a})}
8959 @tab @code{MADDACCS @var{a},@var{b}}
8960 @item @code{sw1 __MADDHSS (sw1, sw1)}
8961 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
8962 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
8963 @item @code{uw1 __MADDHUS (uw1, uw1)}
8964 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
8965 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
8966 @item @code{uw1 __MAND (uw1, uw1)}
8967 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
8968 @tab @code{MAND @var{a},@var{b},@var{c}}
8969 @item @code{void __MASACCS (acc, acc)}
8970 @tab @code{__MASACCS (@var{b}, @var{a})}
8971 @tab @code{MASACCS @var{a},@var{b}}
8972 @item @code{uw1 __MAVEH (uw1, uw1)}
8973 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
8974 @tab @code{MAVEH @var{a},@var{b},@var{c}}
8975 @item @code{uw2 __MBTOH (uw1)}
8976 @tab @code{@var{b} = __MBTOH (@var{a})}
8977 @tab @code{MBTOH @var{a},@var{b}}
8978 @item @code{void __MBTOHE (uw1 *, uw1)}
8979 @tab @code{__MBTOHE (&@var{b}, @var{a})}
8980 @tab @code{MBTOHE @var{a},@var{b}}
8981 @item @code{void __MCLRACC (acc)}
8982 @tab @code{__MCLRACC (@var{a})}
8983 @tab @code{MCLRACC @var{a}}
8984 @item @code{void __MCLRACCA (void)}
8985 @tab @code{__MCLRACCA ()}
8986 @tab @code{MCLRACCA}
8987 @item @code{uw1 __Mcop1 (uw1, uw1)}
8988 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
8989 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
8990 @item @code{uw1 __Mcop2 (uw1, uw1)}
8991 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
8992 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
8993 @item @code{uw1 __MCPLHI (uw2, const)}
8994 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
8995 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
8996 @item @code{uw1 __MCPLI (uw2, const)}
8997 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
8998 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
8999 @item @code{void __MCPXIS (acc, sw1, sw1)}
9000 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
9001 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
9002 @item @code{void __MCPXIU (acc, uw1, uw1)}
9003 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
9004 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
9005 @item @code{void __MCPXRS (acc, sw1, sw1)}
9006 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
9007 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
9008 @item @code{void __MCPXRU (acc, uw1, uw1)}
9009 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
9010 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
9011 @item @code{uw1 __MCUT (acc, uw1)}
9012 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
9013 @tab @code{MCUT @var{a},@var{b},@var{c}}
9014 @item @code{uw1 __MCUTSS (acc, sw1)}
9015 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
9016 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
9017 @item @code{void __MDADDACCS (acc, acc)}
9018 @tab @code{__MDADDACCS (@var{b}, @var{a})}
9019 @tab @code{MDADDACCS @var{a},@var{b}}
9020 @item @code{void __MDASACCS (acc, acc)}
9021 @tab @code{__MDASACCS (@var{b}, @var{a})}
9022 @tab @code{MDASACCS @var{a},@var{b}}
9023 @item @code{uw2 __MDCUTSSI (acc, const)}
9024 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
9025 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
9026 @item @code{uw2 __MDPACKH (uw2, uw2)}
9027 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
9028 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
9029 @item @code{uw2 __MDROTLI (uw2, const)}
9030 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
9031 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
9032 @item @code{void __MDSUBACCS (acc, acc)}
9033 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
9034 @tab @code{MDSUBACCS @var{a},@var{b}}
9035 @item @code{void __MDUNPACKH (uw1 *, uw2)}
9036 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
9037 @tab @code{MDUNPACKH @var{a},@var{b}}
9038 @item @code{uw2 __MEXPDHD (uw1, const)}
9039 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
9040 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
9041 @item @code{uw1 __MEXPDHW (uw1, const)}
9042 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
9043 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
9044 @item @code{uw1 __MHDSETH (uw1, const)}
9045 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
9046 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
9047 @item @code{sw1 __MHDSETS (const)}
9048 @tab @code{@var{b} = __MHDSETS (@var{a})}
9049 @tab @code{MHDSETS #@var{a},@var{b}}
9050 @item @code{uw1 __MHSETHIH (uw1, const)}
9051 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
9052 @tab @code{MHSETHIH #@var{a},@var{b}}
9053 @item @code{sw1 __MHSETHIS (sw1, const)}
9054 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
9055 @tab @code{MHSETHIS #@var{a},@var{b}}
9056 @item @code{uw1 __MHSETLOH (uw1, const)}
9057 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
9058 @tab @code{MHSETLOH #@var{a},@var{b}}
9059 @item @code{sw1 __MHSETLOS (sw1, const)}
9060 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
9061 @tab @code{MHSETLOS #@var{a},@var{b}}
9062 @item @code{uw1 __MHTOB (uw2)}
9063 @tab @code{@var{b} = __MHTOB (@var{a})}
9064 @tab @code{MHTOB @var{a},@var{b}}
9065 @item @code{void __MMACHS (acc, sw1, sw1)}
9066 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
9067 @tab @code{MMACHS @var{a},@var{b},@var{c}}
9068 @item @code{void __MMACHU (acc, uw1, uw1)}
9069 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
9070 @tab @code{MMACHU @var{a},@var{b},@var{c}}
9071 @item @code{void __MMRDHS (acc, sw1, sw1)}
9072 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
9073 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
9074 @item @code{void __MMRDHU (acc, uw1, uw1)}
9075 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
9076 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
9077 @item @code{void __MMULHS (acc, sw1, sw1)}
9078 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
9079 @tab @code{MMULHS @var{a},@var{b},@var{c}}
9080 @item @code{void __MMULHU (acc, uw1, uw1)}
9081 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
9082 @tab @code{MMULHU @var{a},@var{b},@var{c}}
9083 @item @code{void __MMULXHS (acc, sw1, sw1)}
9084 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
9085 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
9086 @item @code{void __MMULXHU (acc, uw1, uw1)}
9087 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
9088 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
9089 @item @code{uw1 __MNOT (uw1)}
9090 @tab @code{@var{b} = __MNOT (@var{a})}
9091 @tab @code{MNOT @var{a},@var{b}}
9092 @item @code{uw1 __MOR (uw1, uw1)}
9093 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
9094 @tab @code{MOR @var{a},@var{b},@var{c}}
9095 @item @code{uw1 __MPACKH (uh, uh)}
9096 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
9097 @tab @code{MPACKH @var{a},@var{b},@var{c}}
9098 @item @code{sw2 __MQADDHSS (sw2, sw2)}
9099 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
9100 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
9101 @item @code{uw2 __MQADDHUS (uw2, uw2)}
9102 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
9103 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
9104 @item @code{void __MQCPXIS (acc, sw2, sw2)}
9105 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
9106 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
9107 @item @code{void __MQCPXIU (acc, uw2, uw2)}
9108 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
9109 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
9110 @item @code{void __MQCPXRS (acc, sw2, sw2)}
9111 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
9112 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
9113 @item @code{void __MQCPXRU (acc, uw2, uw2)}
9114 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
9115 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
9116 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
9117 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
9118 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
9119 @item @code{sw2 __MQLMTHS (sw2, sw2)}
9120 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
9121 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
9122 @item @code{void __MQMACHS (acc, sw2, sw2)}
9123 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
9124 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
9125 @item @code{void __MQMACHU (acc, uw2, uw2)}
9126 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
9127 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
9128 @item @code{void __MQMACXHS (acc, sw2, sw2)}
9129 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
9130 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
9131 @item @code{void __MQMULHS (acc, sw2, sw2)}
9132 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
9133 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
9134 @item @code{void __MQMULHU (acc, uw2, uw2)}
9135 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
9136 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
9137 @item @code{void __MQMULXHS (acc, sw2, sw2)}
9138 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
9139 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
9140 @item @code{void __MQMULXHU (acc, uw2, uw2)}
9141 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
9142 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
9143 @item @code{sw2 __MQSATHS (sw2, sw2)}
9144 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
9145 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
9146 @item @code{uw2 __MQSLLHI (uw2, int)}
9147 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
9148 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
9149 @item @code{sw2 __MQSRAHI (sw2, int)}
9150 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
9151 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
9152 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
9153 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
9154 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
9155 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
9156 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
9157 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
9158 @item @code{void __MQXMACHS (acc, sw2, sw2)}
9159 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
9160 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
9161 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
9162 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
9163 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
9164 @item @code{uw1 __MRDACC (acc)}
9165 @tab @code{@var{b} = __MRDACC (@var{a})}
9166 @tab @code{MRDACC @var{a},@var{b}}
9167 @item @code{uw1 __MRDACCG (acc)}
9168 @tab @code{@var{b} = __MRDACCG (@var{a})}
9169 @tab @code{MRDACCG @var{a},@var{b}}
9170 @item @code{uw1 __MROTLI (uw1, const)}
9171 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
9172 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
9173 @item @code{uw1 __MROTRI (uw1, const)}
9174 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
9175 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
9176 @item @code{sw1 __MSATHS (sw1, sw1)}
9177 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
9178 @tab @code{MSATHS @var{a},@var{b},@var{c}}
9179 @item @code{uw1 __MSATHU (uw1, uw1)}
9180 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
9181 @tab @code{MSATHU @var{a},@var{b},@var{c}}
9182 @item @code{uw1 __MSLLHI (uw1, const)}
9183 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
9184 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
9185 @item @code{sw1 __MSRAHI (sw1, const)}
9186 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
9187 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
9188 @item @code{uw1 __MSRLHI (uw1, const)}
9189 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
9190 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
9191 @item @code{void __MSUBACCS (acc, acc)}
9192 @tab @code{__MSUBACCS (@var{b}, @var{a})}
9193 @tab @code{MSUBACCS @var{a},@var{b}}
9194 @item @code{sw1 __MSUBHSS (sw1, sw1)}
9195 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
9196 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
9197 @item @code{uw1 __MSUBHUS (uw1, uw1)}
9198 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
9199 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
9200 @item @code{void __MTRAP (void)}
9201 @tab @code{__MTRAP ()}
9203 @item @code{uw2 __MUNPACKH (uw1)}
9204 @tab @code{@var{b} = __MUNPACKH (@var{a})}
9205 @tab @code{MUNPACKH @var{a},@var{b}}
9206 @item @code{uw1 __MWCUT (uw2, uw1)}
9207 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
9208 @tab @code{MWCUT @var{a},@var{b},@var{c}}
9209 @item @code{void __MWTACC (acc, uw1)}
9210 @tab @code{__MWTACC (@var{b}, @var{a})}
9211 @tab @code{MWTACC @var{a},@var{b}}
9212 @item @code{void __MWTACCG (acc, uw1)}
9213 @tab @code{__MWTACCG (@var{b}, @var{a})}
9214 @tab @code{MWTACCG @var{a},@var{b}}
9215 @item @code{uw1 __MXOR (uw1, uw1)}
9216 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
9217 @tab @code{MXOR @var{a},@var{b},@var{c}}
9220 @node Raw read/write Functions
9221 @subsubsection Raw read/write Functions
9223 This sections describes built-in functions related to read and write
9224 instructions to access memory. These functions generate
9225 @code{membar} instructions to flush the I/O load and stores where
9226 appropriate, as described in Fujitsu's manual described above.
9230 @item unsigned char __builtin_read8 (void *@var{data})
9231 @item unsigned short __builtin_read16 (void *@var{data})
9232 @item unsigned long __builtin_read32 (void *@var{data})
9233 @item unsigned long long __builtin_read64 (void *@var{data})
9235 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
9236 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
9237 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
9238 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
9241 @node Other Built-in Functions
9242 @subsubsection Other Built-in Functions
9244 This section describes built-in functions that are not named after
9245 a specific FR-V instruction.
9248 @item sw2 __IACCreadll (iacc @var{reg})
9249 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
9250 for future expansion and must be 0.
9252 @item sw1 __IACCreadl (iacc @var{reg})
9253 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
9254 Other values of @var{reg} are rejected as invalid.
9256 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
9257 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
9258 is reserved for future expansion and must be 0.
9260 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
9261 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
9262 is 1. Other values of @var{reg} are rejected as invalid.
9264 @item void __data_prefetch0 (const void *@var{x})
9265 Use the @code{dcpl} instruction to load the contents of address @var{x}
9266 into the data cache.
9268 @item void __data_prefetch (const void *@var{x})
9269 Use the @code{nldub} instruction to load the contents of address @var{x}
9270 into the data cache. The instruction will be issued in slot I1@.
9273 @node X86 Built-in Functions
9274 @subsection X86 Built-in Functions
9276 These built-in functions are available for the i386 and x86-64 family
9277 of computers, depending on the command-line switches used.
9279 Note that, if you specify command-line switches such as @option{-msse},
9280 the compiler could use the extended instruction sets even if the built-ins
9281 are not used explicitly in the program. For this reason, applications
9282 which perform runtime CPU detection must compile separate files for each
9283 supported architecture, using the appropriate flags. In particular,
9284 the file containing the CPU detection code should be compiled without
9287 The following machine modes are available for use with MMX built-in functions
9288 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
9289 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
9290 vector of eight 8-bit integers. Some of the built-in functions operate on
9291 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
9293 If 3DNow!@: extensions are enabled, @code{V2SF} is used as a mode for a vector
9294 of two 32-bit floating point values.
9296 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
9297 floating point values. Some instructions use a vector of four 32-bit
9298 integers, these use @code{V4SI}. Finally, some instructions operate on an
9299 entire vector register, interpreting it as a 128-bit integer, these use mode
9302 In 64-bit mode, the x86-64 family of processors uses additional built-in
9303 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
9304 floating point and @code{TC} 128-bit complex floating point values.
9306 The following floating point built-in functions are available in 64-bit
9307 mode. All of them implement the function that is part of the name.
9310 __float128 __builtin_fabsq (__float128)
9311 __float128 __builtin_copysignq (__float128, __float128)
9314 The following built-in function is always available.
9317 @item void __builtin_ia32_pause (void)
9318 Generates the @code{pause} machine instruction with a compiler memory
9322 The following floating point built-in functions are made available in the
9326 @item __float128 __builtin_infq (void)
9327 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
9328 @findex __builtin_infq
9330 @item __float128 __builtin_huge_valq (void)
9331 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
9332 @findex __builtin_huge_valq
9335 The following built-in functions are made available by @option{-mmmx}.
9336 All of them generate the machine instruction that is part of the name.
9339 v8qi __builtin_ia32_paddb (v8qi, v8qi)
9340 v4hi __builtin_ia32_paddw (v4hi, v4hi)
9341 v2si __builtin_ia32_paddd (v2si, v2si)
9342 v8qi __builtin_ia32_psubb (v8qi, v8qi)
9343 v4hi __builtin_ia32_psubw (v4hi, v4hi)
9344 v2si __builtin_ia32_psubd (v2si, v2si)
9345 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
9346 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
9347 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
9348 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
9349 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
9350 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
9351 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
9352 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
9353 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
9354 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
9355 di __builtin_ia32_pand (di, di)
9356 di __builtin_ia32_pandn (di,di)
9357 di __builtin_ia32_por (di, di)
9358 di __builtin_ia32_pxor (di, di)
9359 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
9360 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
9361 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
9362 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
9363 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
9364 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
9365 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
9366 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
9367 v2si __builtin_ia32_punpckhdq (v2si, v2si)
9368 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
9369 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
9370 v2si __builtin_ia32_punpckldq (v2si, v2si)
9371 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
9372 v4hi __builtin_ia32_packssdw (v2si, v2si)
9373 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
9375 v4hi __builtin_ia32_psllw (v4hi, v4hi)
9376 v2si __builtin_ia32_pslld (v2si, v2si)
9377 v1di __builtin_ia32_psllq (v1di, v1di)
9378 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
9379 v2si __builtin_ia32_psrld (v2si, v2si)
9380 v1di __builtin_ia32_psrlq (v1di, v1di)
9381 v4hi __builtin_ia32_psraw (v4hi, v4hi)
9382 v2si __builtin_ia32_psrad (v2si, v2si)
9383 v4hi __builtin_ia32_psllwi (v4hi, int)
9384 v2si __builtin_ia32_pslldi (v2si, int)
9385 v1di __builtin_ia32_psllqi (v1di, int)
9386 v4hi __builtin_ia32_psrlwi (v4hi, int)
9387 v2si __builtin_ia32_psrldi (v2si, int)
9388 v1di __builtin_ia32_psrlqi (v1di, int)
9389 v4hi __builtin_ia32_psrawi (v4hi, int)
9390 v2si __builtin_ia32_psradi (v2si, int)
9394 The following built-in functions are made available either with
9395 @option{-msse}, or with a combination of @option{-m3dnow} and
9396 @option{-march=athlon}. All of them generate the machine
9397 instruction that is part of the name.
9400 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
9401 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
9402 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
9403 v1di __builtin_ia32_psadbw (v8qi, v8qi)
9404 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
9405 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
9406 v8qi __builtin_ia32_pminub (v8qi, v8qi)
9407 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
9408 int __builtin_ia32_pextrw (v4hi, int)
9409 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
9410 int __builtin_ia32_pmovmskb (v8qi)
9411 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
9412 void __builtin_ia32_movntq (di *, di)
9413 void __builtin_ia32_sfence (void)
9416 The following built-in functions are available when @option{-msse} is used.
9417 All of them generate the machine instruction that is part of the name.
9420 int __builtin_ia32_comieq (v4sf, v4sf)
9421 int __builtin_ia32_comineq (v4sf, v4sf)
9422 int __builtin_ia32_comilt (v4sf, v4sf)
9423 int __builtin_ia32_comile (v4sf, v4sf)
9424 int __builtin_ia32_comigt (v4sf, v4sf)
9425 int __builtin_ia32_comige (v4sf, v4sf)
9426 int __builtin_ia32_ucomieq (v4sf, v4sf)
9427 int __builtin_ia32_ucomineq (v4sf, v4sf)
9428 int __builtin_ia32_ucomilt (v4sf, v4sf)
9429 int __builtin_ia32_ucomile (v4sf, v4sf)
9430 int __builtin_ia32_ucomigt (v4sf, v4sf)
9431 int __builtin_ia32_ucomige (v4sf, v4sf)
9432 v4sf __builtin_ia32_addps (v4sf, v4sf)
9433 v4sf __builtin_ia32_subps (v4sf, v4sf)
9434 v4sf __builtin_ia32_mulps (v4sf, v4sf)
9435 v4sf __builtin_ia32_divps (v4sf, v4sf)
9436 v4sf __builtin_ia32_addss (v4sf, v4sf)
9437 v4sf __builtin_ia32_subss (v4sf, v4sf)
9438 v4sf __builtin_ia32_mulss (v4sf, v4sf)
9439 v4sf __builtin_ia32_divss (v4sf, v4sf)
9440 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
9441 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
9442 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
9443 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
9444 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
9445 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
9446 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
9447 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
9448 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
9449 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
9450 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
9451 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
9452 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
9453 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
9454 v4si __builtin_ia32_cmpless (v4sf, v4sf)
9455 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
9456 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
9457 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
9458 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
9459 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
9460 v4sf __builtin_ia32_maxps (v4sf, v4sf)
9461 v4sf __builtin_ia32_maxss (v4sf, v4sf)
9462 v4sf __builtin_ia32_minps (v4sf, v4sf)
9463 v4sf __builtin_ia32_minss (v4sf, v4sf)
9464 v4sf __builtin_ia32_andps (v4sf, v4sf)
9465 v4sf __builtin_ia32_andnps (v4sf, v4sf)
9466 v4sf __builtin_ia32_orps (v4sf, v4sf)
9467 v4sf __builtin_ia32_xorps (v4sf, v4sf)
9468 v4sf __builtin_ia32_movss (v4sf, v4sf)
9469 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
9470 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
9471 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
9472 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
9473 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
9474 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
9475 v2si __builtin_ia32_cvtps2pi (v4sf)
9476 int __builtin_ia32_cvtss2si (v4sf)
9477 v2si __builtin_ia32_cvttps2pi (v4sf)
9478 int __builtin_ia32_cvttss2si (v4sf)
9479 v4sf __builtin_ia32_rcpps (v4sf)
9480 v4sf __builtin_ia32_rsqrtps (v4sf)
9481 v4sf __builtin_ia32_sqrtps (v4sf)
9482 v4sf __builtin_ia32_rcpss (v4sf)
9483 v4sf __builtin_ia32_rsqrtss (v4sf)
9484 v4sf __builtin_ia32_sqrtss (v4sf)
9485 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
9486 void __builtin_ia32_movntps (float *, v4sf)
9487 int __builtin_ia32_movmskps (v4sf)
9490 The following built-in functions are available when @option{-msse} is used.
9493 @item v4sf __builtin_ia32_loadaps (float *)
9494 Generates the @code{movaps} machine instruction as a load from memory.
9495 @item void __builtin_ia32_storeaps (float *, v4sf)
9496 Generates the @code{movaps} machine instruction as a store to memory.
9497 @item v4sf __builtin_ia32_loadups (float *)
9498 Generates the @code{movups} machine instruction as a load from memory.
9499 @item void __builtin_ia32_storeups (float *, v4sf)
9500 Generates the @code{movups} machine instruction as a store to memory.
9501 @item v4sf __builtin_ia32_loadsss (float *)
9502 Generates the @code{movss} machine instruction as a load from memory.
9503 @item void __builtin_ia32_storess (float *, v4sf)
9504 Generates the @code{movss} machine instruction as a store to memory.
9505 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
9506 Generates the @code{movhps} machine instruction as a load from memory.
9507 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
9508 Generates the @code{movlps} machine instruction as a load from memory
9509 @item void __builtin_ia32_storehps (v2sf *, v4sf)
9510 Generates the @code{movhps} machine instruction as a store to memory.
9511 @item void __builtin_ia32_storelps (v2sf *, v4sf)
9512 Generates the @code{movlps} machine instruction as a store to memory.
9515 The following built-in functions are available when @option{-msse2} is used.
9516 All of them generate the machine instruction that is part of the name.
9519 int __builtin_ia32_comisdeq (v2df, v2df)
9520 int __builtin_ia32_comisdlt (v2df, v2df)
9521 int __builtin_ia32_comisdle (v2df, v2df)
9522 int __builtin_ia32_comisdgt (v2df, v2df)
9523 int __builtin_ia32_comisdge (v2df, v2df)
9524 int __builtin_ia32_comisdneq (v2df, v2df)
9525 int __builtin_ia32_ucomisdeq (v2df, v2df)
9526 int __builtin_ia32_ucomisdlt (v2df, v2df)
9527 int __builtin_ia32_ucomisdle (v2df, v2df)
9528 int __builtin_ia32_ucomisdgt (v2df, v2df)
9529 int __builtin_ia32_ucomisdge (v2df, v2df)
9530 int __builtin_ia32_ucomisdneq (v2df, v2df)
9531 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
9532 v2df __builtin_ia32_cmpltpd (v2df, v2df)
9533 v2df __builtin_ia32_cmplepd (v2df, v2df)
9534 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
9535 v2df __builtin_ia32_cmpgepd (v2df, v2df)
9536 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
9537 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
9538 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
9539 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
9540 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
9541 v2df __builtin_ia32_cmpngepd (v2df, v2df)
9542 v2df __builtin_ia32_cmpordpd (v2df, v2df)
9543 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
9544 v2df __builtin_ia32_cmpltsd (v2df, v2df)
9545 v2df __builtin_ia32_cmplesd (v2df, v2df)
9546 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
9547 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
9548 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
9549 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
9550 v2df __builtin_ia32_cmpordsd (v2df, v2df)
9551 v2di __builtin_ia32_paddq (v2di, v2di)
9552 v2di __builtin_ia32_psubq (v2di, v2di)
9553 v2df __builtin_ia32_addpd (v2df, v2df)
9554 v2df __builtin_ia32_subpd (v2df, v2df)
9555 v2df __builtin_ia32_mulpd (v2df, v2df)
9556 v2df __builtin_ia32_divpd (v2df, v2df)
9557 v2df __builtin_ia32_addsd (v2df, v2df)
9558 v2df __builtin_ia32_subsd (v2df, v2df)
9559 v2df __builtin_ia32_mulsd (v2df, v2df)
9560 v2df __builtin_ia32_divsd (v2df, v2df)
9561 v2df __builtin_ia32_minpd (v2df, v2df)
9562 v2df __builtin_ia32_maxpd (v2df, v2df)
9563 v2df __builtin_ia32_minsd (v2df, v2df)
9564 v2df __builtin_ia32_maxsd (v2df, v2df)
9565 v2df __builtin_ia32_andpd (v2df, v2df)
9566 v2df __builtin_ia32_andnpd (v2df, v2df)
9567 v2df __builtin_ia32_orpd (v2df, v2df)
9568 v2df __builtin_ia32_xorpd (v2df, v2df)
9569 v2df __builtin_ia32_movsd (v2df, v2df)
9570 v2df __builtin_ia32_unpckhpd (v2df, v2df)
9571 v2df __builtin_ia32_unpcklpd (v2df, v2df)
9572 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
9573 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
9574 v4si __builtin_ia32_paddd128 (v4si, v4si)
9575 v2di __builtin_ia32_paddq128 (v2di, v2di)
9576 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
9577 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
9578 v4si __builtin_ia32_psubd128 (v4si, v4si)
9579 v2di __builtin_ia32_psubq128 (v2di, v2di)
9580 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
9581 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
9582 v2di __builtin_ia32_pand128 (v2di, v2di)
9583 v2di __builtin_ia32_pandn128 (v2di, v2di)
9584 v2di __builtin_ia32_por128 (v2di, v2di)
9585 v2di __builtin_ia32_pxor128 (v2di, v2di)
9586 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
9587 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
9588 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
9589 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
9590 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
9591 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
9592 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
9593 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
9594 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
9595 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
9596 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
9597 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
9598 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
9599 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
9600 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
9601 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
9602 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
9603 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
9604 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
9605 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
9606 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
9607 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
9608 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
9609 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
9610 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
9611 v2df __builtin_ia32_loadupd (double *)
9612 void __builtin_ia32_storeupd (double *, v2df)
9613 v2df __builtin_ia32_loadhpd (v2df, double const *)
9614 v2df __builtin_ia32_loadlpd (v2df, double const *)
9615 int __builtin_ia32_movmskpd (v2df)
9616 int __builtin_ia32_pmovmskb128 (v16qi)
9617 void __builtin_ia32_movnti (int *, int)
9618 void __builtin_ia32_movnti64 (long long int *, long long int)
9619 void __builtin_ia32_movntpd (double *, v2df)
9620 void __builtin_ia32_movntdq (v2df *, v2df)
9621 v4si __builtin_ia32_pshufd (v4si, int)
9622 v8hi __builtin_ia32_pshuflw (v8hi, int)
9623 v8hi __builtin_ia32_pshufhw (v8hi, int)
9624 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
9625 v2df __builtin_ia32_sqrtpd (v2df)
9626 v2df __builtin_ia32_sqrtsd (v2df)
9627 v2df __builtin_ia32_shufpd (v2df, v2df, int)
9628 v2df __builtin_ia32_cvtdq2pd (v4si)
9629 v4sf __builtin_ia32_cvtdq2ps (v4si)
9630 v4si __builtin_ia32_cvtpd2dq (v2df)
9631 v2si __builtin_ia32_cvtpd2pi (v2df)
9632 v4sf __builtin_ia32_cvtpd2ps (v2df)
9633 v4si __builtin_ia32_cvttpd2dq (v2df)
9634 v2si __builtin_ia32_cvttpd2pi (v2df)
9635 v2df __builtin_ia32_cvtpi2pd (v2si)
9636 int __builtin_ia32_cvtsd2si (v2df)
9637 int __builtin_ia32_cvttsd2si (v2df)
9638 long long __builtin_ia32_cvtsd2si64 (v2df)
9639 long long __builtin_ia32_cvttsd2si64 (v2df)
9640 v4si __builtin_ia32_cvtps2dq (v4sf)
9641 v2df __builtin_ia32_cvtps2pd (v4sf)
9642 v4si __builtin_ia32_cvttps2dq (v4sf)
9643 v2df __builtin_ia32_cvtsi2sd (v2df, int)
9644 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
9645 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
9646 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
9647 void __builtin_ia32_clflush (const void *)
9648 void __builtin_ia32_lfence (void)
9649 void __builtin_ia32_mfence (void)
9650 v16qi __builtin_ia32_loaddqu (const char *)
9651 void __builtin_ia32_storedqu (char *, v16qi)
9652 v1di __builtin_ia32_pmuludq (v2si, v2si)
9653 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
9654 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
9655 v4si __builtin_ia32_pslld128 (v4si, v4si)
9656 v2di __builtin_ia32_psllq128 (v2di, v2di)
9657 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
9658 v4si __builtin_ia32_psrld128 (v4si, v4si)
9659 v2di __builtin_ia32_psrlq128 (v2di, v2di)
9660 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
9661 v4si __builtin_ia32_psrad128 (v4si, v4si)
9662 v2di __builtin_ia32_pslldqi128 (v2di, int)
9663 v8hi __builtin_ia32_psllwi128 (v8hi, int)
9664 v4si __builtin_ia32_pslldi128 (v4si, int)
9665 v2di __builtin_ia32_psllqi128 (v2di, int)
9666 v2di __builtin_ia32_psrldqi128 (v2di, int)
9667 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
9668 v4si __builtin_ia32_psrldi128 (v4si, int)
9669 v2di __builtin_ia32_psrlqi128 (v2di, int)
9670 v8hi __builtin_ia32_psrawi128 (v8hi, int)
9671 v4si __builtin_ia32_psradi128 (v4si, int)
9672 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
9673 v2di __builtin_ia32_movq128 (v2di)
9676 The following built-in functions are available when @option{-msse3} is used.
9677 All of them generate the machine instruction that is part of the name.
9680 v2df __builtin_ia32_addsubpd (v2df, v2df)
9681 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
9682 v2df __builtin_ia32_haddpd (v2df, v2df)
9683 v4sf __builtin_ia32_haddps (v4sf, v4sf)
9684 v2df __builtin_ia32_hsubpd (v2df, v2df)
9685 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
9686 v16qi __builtin_ia32_lddqu (char const *)
9687 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
9688 v2df __builtin_ia32_movddup (v2df)
9689 v4sf __builtin_ia32_movshdup (v4sf)
9690 v4sf __builtin_ia32_movsldup (v4sf)
9691 void __builtin_ia32_mwait (unsigned int, unsigned int)
9694 The following built-in functions are available when @option{-msse3} is used.
9697 @item v2df __builtin_ia32_loadddup (double const *)
9698 Generates the @code{movddup} machine instruction as a load from memory.
9701 The following built-in functions are available when @option{-mssse3} is used.
9702 All of them generate the machine instruction that is part of the name
9706 v2si __builtin_ia32_phaddd (v2si, v2si)
9707 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
9708 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
9709 v2si __builtin_ia32_phsubd (v2si, v2si)
9710 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
9711 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
9712 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
9713 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
9714 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
9715 v8qi __builtin_ia32_psignb (v8qi, v8qi)
9716 v2si __builtin_ia32_psignd (v2si, v2si)
9717 v4hi __builtin_ia32_psignw (v4hi, v4hi)
9718 v1di __builtin_ia32_palignr (v1di, v1di, int)
9719 v8qi __builtin_ia32_pabsb (v8qi)
9720 v2si __builtin_ia32_pabsd (v2si)
9721 v4hi __builtin_ia32_pabsw (v4hi)
9724 The following built-in functions are available when @option{-mssse3} is used.
9725 All of them generate the machine instruction that is part of the name
9729 v4si __builtin_ia32_phaddd128 (v4si, v4si)
9730 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
9731 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
9732 v4si __builtin_ia32_phsubd128 (v4si, v4si)
9733 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
9734 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
9735 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
9736 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
9737 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
9738 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
9739 v4si __builtin_ia32_psignd128 (v4si, v4si)
9740 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
9741 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
9742 v16qi __builtin_ia32_pabsb128 (v16qi)
9743 v4si __builtin_ia32_pabsd128 (v4si)
9744 v8hi __builtin_ia32_pabsw128 (v8hi)
9747 The following built-in functions are available when @option{-msse4.1} is
9748 used. All of them generate the machine instruction that is part of the
9752 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
9753 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
9754 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
9755 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
9756 v2df __builtin_ia32_dppd (v2df, v2df, const int)
9757 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
9758 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
9759 v2di __builtin_ia32_movntdqa (v2di *);
9760 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
9761 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
9762 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
9763 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
9764 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
9765 v8hi __builtin_ia32_phminposuw128 (v8hi)
9766 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
9767 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
9768 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
9769 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
9770 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
9771 v4si __builtin_ia32_pminsd128 (v4si, v4si)
9772 v4si __builtin_ia32_pminud128 (v4si, v4si)
9773 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
9774 v4si __builtin_ia32_pmovsxbd128 (v16qi)
9775 v2di __builtin_ia32_pmovsxbq128 (v16qi)
9776 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
9777 v2di __builtin_ia32_pmovsxdq128 (v4si)
9778 v4si __builtin_ia32_pmovsxwd128 (v8hi)
9779 v2di __builtin_ia32_pmovsxwq128 (v8hi)
9780 v4si __builtin_ia32_pmovzxbd128 (v16qi)
9781 v2di __builtin_ia32_pmovzxbq128 (v16qi)
9782 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
9783 v2di __builtin_ia32_pmovzxdq128 (v4si)
9784 v4si __builtin_ia32_pmovzxwd128 (v8hi)
9785 v2di __builtin_ia32_pmovzxwq128 (v8hi)
9786 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
9787 v4si __builtin_ia32_pmulld128 (v4si, v4si)
9788 int __builtin_ia32_ptestc128 (v2di, v2di)
9789 int __builtin_ia32_ptestnzc128 (v2di, v2di)
9790 int __builtin_ia32_ptestz128 (v2di, v2di)
9791 v2df __builtin_ia32_roundpd (v2df, const int)
9792 v4sf __builtin_ia32_roundps (v4sf, const int)
9793 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
9794 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
9797 The following built-in functions are available when @option{-msse4.1} is
9801 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
9802 Generates the @code{insertps} machine instruction.
9803 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
9804 Generates the @code{pextrb} machine instruction.
9805 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
9806 Generates the @code{pinsrb} machine instruction.
9807 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
9808 Generates the @code{pinsrd} machine instruction.
9809 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
9810 Generates the @code{pinsrq} machine instruction in 64bit mode.
9813 The following built-in functions are changed to generate new SSE4.1
9814 instructions when @option{-msse4.1} is used.
9817 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
9818 Generates the @code{extractps} machine instruction.
9819 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
9820 Generates the @code{pextrd} machine instruction.
9821 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
9822 Generates the @code{pextrq} machine instruction in 64bit mode.
9825 The following built-in functions are available when @option{-msse4.2} is
9826 used. All of them generate the machine instruction that is part of the
9830 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
9831 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
9832 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
9833 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
9834 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
9835 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
9836 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
9837 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
9838 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
9839 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
9840 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
9841 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
9842 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
9843 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
9844 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
9847 The following built-in functions are available when @option{-msse4.2} is
9851 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
9852 Generates the @code{crc32b} machine instruction.
9853 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
9854 Generates the @code{crc32w} machine instruction.
9855 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
9856 Generates the @code{crc32l} machine instruction.
9857 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
9858 Generates the @code{crc32q} machine instruction.
9861 The following built-in functions are changed to generate new SSE4.2
9862 instructions when @option{-msse4.2} is used.
9865 @item int __builtin_popcount (unsigned int)
9866 Generates the @code{popcntl} machine instruction.
9867 @item int __builtin_popcountl (unsigned long)
9868 Generates the @code{popcntl} or @code{popcntq} machine instruction,
9869 depending on the size of @code{unsigned long}.
9870 @item int __builtin_popcountll (unsigned long long)
9871 Generates the @code{popcntq} machine instruction.
9874 The following built-in functions are available when @option{-mavx} is
9875 used. All of them generate the machine instruction that is part of the
9879 v4df __builtin_ia32_addpd256 (v4df,v4df)
9880 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
9881 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
9882 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
9883 v4df __builtin_ia32_andnpd256 (v4df,v4df)
9884 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
9885 v4df __builtin_ia32_andpd256 (v4df,v4df)
9886 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
9887 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
9888 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
9889 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
9890 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
9891 v2df __builtin_ia32_cmppd (v2df,v2df,int)
9892 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
9893 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
9894 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
9895 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
9896 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
9897 v4df __builtin_ia32_cvtdq2pd256 (v4si)
9898 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
9899 v4si __builtin_ia32_cvtpd2dq256 (v4df)
9900 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
9901 v8si __builtin_ia32_cvtps2dq256 (v8sf)
9902 v4df __builtin_ia32_cvtps2pd256 (v4sf)
9903 v4si __builtin_ia32_cvttpd2dq256 (v4df)
9904 v8si __builtin_ia32_cvttps2dq256 (v8sf)
9905 v4df __builtin_ia32_divpd256 (v4df,v4df)
9906 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
9907 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
9908 v4df __builtin_ia32_haddpd256 (v4df,v4df)
9909 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
9910 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
9911 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
9912 v32qi __builtin_ia32_lddqu256 (pcchar)
9913 v32qi __builtin_ia32_loaddqu256 (pcchar)
9914 v4df __builtin_ia32_loadupd256 (pcdouble)
9915 v8sf __builtin_ia32_loadups256 (pcfloat)
9916 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
9917 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
9918 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
9919 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
9920 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
9921 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
9922 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
9923 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
9924 v4df __builtin_ia32_maxpd256 (v4df,v4df)
9925 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
9926 v4df __builtin_ia32_minpd256 (v4df,v4df)
9927 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
9928 v4df __builtin_ia32_movddup256 (v4df)
9929 int __builtin_ia32_movmskpd256 (v4df)
9930 int __builtin_ia32_movmskps256 (v8sf)
9931 v8sf __builtin_ia32_movshdup256 (v8sf)
9932 v8sf __builtin_ia32_movsldup256 (v8sf)
9933 v4df __builtin_ia32_mulpd256 (v4df,v4df)
9934 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
9935 v4df __builtin_ia32_orpd256 (v4df,v4df)
9936 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
9937 v2df __builtin_ia32_pd_pd256 (v4df)
9938 v4df __builtin_ia32_pd256_pd (v2df)
9939 v4sf __builtin_ia32_ps_ps256 (v8sf)
9940 v8sf __builtin_ia32_ps256_ps (v4sf)
9941 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
9942 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
9943 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
9944 v8sf __builtin_ia32_rcpps256 (v8sf)
9945 v4df __builtin_ia32_roundpd256 (v4df,int)
9946 v8sf __builtin_ia32_roundps256 (v8sf,int)
9947 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
9948 v8sf __builtin_ia32_rsqrtps256 (v8sf)
9949 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
9950 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
9951 v4si __builtin_ia32_si_si256 (v8si)
9952 v8si __builtin_ia32_si256_si (v4si)
9953 v4df __builtin_ia32_sqrtpd256 (v4df)
9954 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
9955 v8sf __builtin_ia32_sqrtps256 (v8sf)
9956 void __builtin_ia32_storedqu256 (pchar,v32qi)
9957 void __builtin_ia32_storeupd256 (pdouble,v4df)
9958 void __builtin_ia32_storeups256 (pfloat,v8sf)
9959 v4df __builtin_ia32_subpd256 (v4df,v4df)
9960 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
9961 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
9962 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
9963 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
9964 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
9965 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
9966 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
9967 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
9968 v4sf __builtin_ia32_vbroadcastss (pcfloat)
9969 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
9970 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
9971 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
9972 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
9973 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
9974 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
9975 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
9976 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
9977 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
9978 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
9979 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
9980 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
9981 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
9982 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
9983 v2df __builtin_ia32_vpermilpd (v2df,int)
9984 v4df __builtin_ia32_vpermilpd256 (v4df,int)
9985 v4sf __builtin_ia32_vpermilps (v4sf,int)
9986 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
9987 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
9988 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
9989 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
9990 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
9991 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
9992 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
9993 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
9994 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
9995 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
9996 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
9997 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
9998 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
9999 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
10000 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
10001 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
10002 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
10003 void __builtin_ia32_vzeroall (void)
10004 void __builtin_ia32_vzeroupper (void)
10005 v4df __builtin_ia32_xorpd256 (v4df,v4df)
10006 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
10009 The following built-in functions are available when @option{-mavx2} is
10010 used. All of them generate the machine instruction that is part of the
10014 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,v32qi,int)
10015 v32qi __builtin_ia32_pabsb256 (v32qi)
10016 v16hi __builtin_ia32_pabsw256 (v16hi)
10017 v8si __builtin_ia32_pabsd256 (v8si)
10018 v16hi builtin_ia32_packssdw256 (v8si,v8si)
10019 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi)
10020 v16hi __builtin_ia32_packusdw256 (v8si,v8si)
10021 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi)
10022 v32qi__builtin_ia32_paddb256 (v32qi,v32qi)
10023 v16hi __builtin_ia32_paddw256 (v16hi,v16hi)
10024 v8si __builtin_ia32_paddd256 (v8si,v8si)
10025 v4di __builtin_ia32_paddq256 (v4di,v4di)
10026 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi)
10027 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi)
10028 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi)
10029 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi)
10030 v4di __builtin_ia32_palignr256 (v4di,v4di,int)
10031 v4di __builtin_ia32_andsi256 (v4di,v4di)
10032 v4di __builtin_ia32_andnotsi256 (v4di,v4di)
10033 v32qi__builtin_ia32_pavgb256 (v32qi,v32qi)
10034 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi)
10035 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi)
10036 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int)
10037 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi)
10038 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi)
10039 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si)
10040 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di)
10041 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi)
10042 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi)
10043 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si)
10044 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di)
10045 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi)
10046 v8si __builtin_ia32_phaddd256 (v8si,v8si)
10047 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi)
10048 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi)
10049 v8si __builtin_ia32_phsubd256 (v8si,v8si)
10050 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi)
10051 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi)
10052 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi)
10053 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi)
10054 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi)
10055 v8si __builtin_ia32_pmaxsd256 (v8si,v8si)
10056 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi)
10057 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi)
10058 v8si __builtin_ia32_pmaxud256 (v8si,v8si)
10059 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi)
10060 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi)
10061 v8si __builtin_ia32_pminsd256 (v8si,v8si)
10062 v32qi __builtin_ia32_pminub256 (v32qi,v32qi)
10063 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi)
10064 v8si __builtin_ia32_pminud256 (v8si,v8si)
10065 int __builtin_ia32_pmovmskb256 (v32qi)
10066 v16hi __builtin_ia32_pmovsxbw256 (v16qi)
10067 v8si __builtin_ia32_pmovsxbd256 (v16qi)
10068 v4di __builtin_ia32_pmovsxbq256 (v16qi)
10069 v8si __builtin_ia32_pmovsxwd256 (v8hi)
10070 v4di __builtin_ia32_pmovsxwq256 (v8hi)
10071 v4di __builtin_ia32_pmovsxdq256 (v4si)
10072 v16hi __builtin_ia32_pmovzxbw256 (v16qi)
10073 v8si __builtin_ia32_pmovzxbd256 (v16qi)
10074 v4di __builtin_ia32_pmovzxbq256 (v16qi)
10075 v8si __builtin_ia32_pmovzxwd256 (v8hi)
10076 v4di __builtin_ia32_pmovzxwq256 (v8hi)
10077 v4di __builtin_ia32_pmovzxdq256 (v4si)
10078 v4di __builtin_ia32_pmuldq256 (v8si,v8si)
10079 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi)
10080 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi)
10081 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi)
10082 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi)
10083 v8si __builtin_ia32_pmulld256 (v8si,v8si)
10084 v4di __builtin_ia32_pmuludq256 (v8si,v8si)
10085 v4di __builtin_ia32_por256 (v4di,v4di)
10086 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi)
10087 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi)
10088 v8si __builtin_ia32_pshufd256 (v8si,int)
10089 v16hi __builtin_ia32_pshufhw256 (v16hi,int)
10090 v16hi __builtin_ia32_pshuflw256 (v16hi,int)
10091 v32qi __builtin_ia32_psignb256 (v32qi,v32qi)
10092 v16hi __builtin_ia32_psignw256 (v16hi,v16hi)
10093 v8si __builtin_ia32_psignd256 (v8si,v8si)
10094 v4di __builtin_ia32_pslldqi256 (v4di,int)
10095 v16hi __builtin_ia32_psllwi256 (16hi,int)
10096 v16hi __builtin_ia32_psllw256(v16hi,v8hi)
10097 v8si __builtin_ia32_pslldi256 (v8si,int)
10098 v8si __builtin_ia32_pslld256(v8si,v4si)
10099 v4di __builtin_ia32_psllqi256 (v4di,int)
10100 v4di __builtin_ia32_psllq256(v4di,v2di)
10101 v16hi __builtin_ia32_psrawi256 (v16hi,int)
10102 v16hi __builtin_ia32_psraw256 (v16hi,v8hi)
10103 v8si __builtin_ia32_psradi256 (v8si,int)
10104 v8si __builtin_ia32_psrad256 (v8si,v4si)
10105 v4di __builtin_ia32_psrldqi256 (v4di, int)
10106 v16hi __builtin_ia32_psrlwi256 (v16hi,int)
10107 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi)
10108 v8si __builtin_ia32_psrldi256 (v8si,int)
10109 v8si __builtin_ia32_psrld256 (v8si,v4si)
10110 v4di __builtin_ia32_psrlqi256 (v4di,int)
10111 v4di __builtin_ia32_psrlq256(v4di,v2di)
10112 v32qi __builtin_ia32_psubb256 (v32qi,v32qi)
10113 v32hi __builtin_ia32_psubw256 (v16hi,v16hi)
10114 v8si __builtin_ia32_psubd256 (v8si,v8si)
10115 v4di __builtin_ia32_psubq256 (v4di,v4di)
10116 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi)
10117 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi)
10118 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi)
10119 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi)
10120 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi)
10121 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi)
10122 v8si __builtin_ia32_punpckhdq256 (v8si,v8si)
10123 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di)
10124 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi)
10125 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi)
10126 v8si __builtin_ia32_punpckldq256 (v8si,v8si)
10127 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di)
10128 v4di __builtin_ia32_pxor256 (v4di,v4di)
10129 v4di __builtin_ia32_movntdqa256 (pv4di)
10130 v4sf __builtin_ia32_vbroadcastss_ps (v4sf)
10131 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf)
10132 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df)
10133 v4di __builtin_ia32_vbroadcastsi256 (v2di)
10134 v4si __builtin_ia32_pblendd128 (v4si,v4si)
10135 v8si __builtin_ia32_pblendd256 (v8si,v8si)
10136 v32qi __builtin_ia32_pbroadcastb256 (v16qi)
10137 v16hi __builtin_ia32_pbroadcastw256 (v8hi)
10138 v8si __builtin_ia32_pbroadcastd256 (v4si)
10139 v4di __builtin_ia32_pbroadcastq256 (v2di)
10140 v16qi __builtin_ia32_pbroadcastb128 (v16qi)
10141 v8hi __builtin_ia32_pbroadcastw128 (v8hi)
10142 v4si __builtin_ia32_pbroadcastd128 (v4si)
10143 v2di __builtin_ia32_pbroadcastq128 (v2di)
10144 v8si __builtin_ia32_permvarsi256 (v8si,v8si)
10145 v4df __builtin_ia32_permdf256 (v4df,int)
10146 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf)
10147 v4di __builtin_ia32_permdi256 (v4di,int)
10148 v4di __builtin_ia32_permti256 (v4di,v4di,int)
10149 v4di __builtin_ia32_extract128i256 (v4di,int)
10150 v4di __builtin_ia32_insert128i256 (v4di,v2di,int)
10151 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si)
10152 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di)
10153 v4si __builtin_ia32_maskloadd (pcv4si,v4si)
10154 v2di __builtin_ia32_maskloadq (pcv2di,v2di)
10155 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si)
10156 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di)
10157 void __builtin_ia32_maskstored (pv4si,v4si,v4si)
10158 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di)
10159 v8si __builtin_ia32_psllv8si (v8si,v8si)
10160 v4si __builtin_ia32_psllv4si (v4si,v4si)
10161 v4di __builtin_ia32_psllv4di (v4di,v4di)
10162 v2di __builtin_ia32_psllv2di (v2di,v2di)
10163 v8si __builtin_ia32_psrav8si (v8si,v8si)
10164 v4si __builtin_ia32_psrav4si (v4si,v4si)
10165 v8si __builtin_ia32_psrlv8si (v8si,v8si)
10166 v4si __builtin_ia32_psrlv4si (v4si,v4si)
10167 v4di __builtin_ia32_psrlv4di (v4di,v4di)
10168 v2di __builtin_ia32_psrlv2di (v2di,v2di)
10169 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int)
10170 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int)
10171 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int)
10172 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int)
10173 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int)
10174 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int)
10175 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int)
10176 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int)
10177 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int)
10178 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int)
10179 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int)
10180 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int)
10181 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int)
10182 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int)
10183 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int)
10184 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int)
10187 The following built-in functions are available when @option{-maes} is
10188 used. All of them generate the machine instruction that is part of the
10192 v2di __builtin_ia32_aesenc128 (v2di, v2di)
10193 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
10194 v2di __builtin_ia32_aesdec128 (v2di, v2di)
10195 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
10196 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
10197 v2di __builtin_ia32_aesimc128 (v2di)
10200 The following built-in function is available when @option{-mpclmul} is
10204 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
10205 Generates the @code{pclmulqdq} machine instruction.
10208 The following built-in function is available when @option{-mfsgsbase} is
10209 used. All of them generate the machine instruction that is part of the
10213 unsigned int __builtin_ia32_rdfsbase32 (void)
10214 unsigned long long __builtin_ia32_rdfsbase64 (void)
10215 unsigned int __builtin_ia32_rdgsbase32 (void)
10216 unsigned long long __builtin_ia32_rdgsbase64 (void)
10217 void _writefsbase_u32 (unsigned int)
10218 void _writefsbase_u64 (unsigned long long)
10219 void _writegsbase_u32 (unsigned int)
10220 void _writegsbase_u64 (unsigned long long)
10223 The following built-in function is available when @option{-mrdrnd} is
10224 used. All of them generate the machine instruction that is part of the
10228 unsigned int __builtin_ia32_rdrand16_step (unsigned short *)
10229 unsigned int __builtin_ia32_rdrand32_step (unsigned int *)
10230 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *)
10233 The following built-in functions are available when @option{-msse4a} is used.
10234 All of them generate the machine instruction that is part of the name.
10237 void __builtin_ia32_movntsd (double *, v2df)
10238 void __builtin_ia32_movntss (float *, v4sf)
10239 v2di __builtin_ia32_extrq (v2di, v16qi)
10240 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
10241 v2di __builtin_ia32_insertq (v2di, v2di)
10242 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
10245 The following built-in functions are available when @option{-mxop} is used.
10247 v2df __builtin_ia32_vfrczpd (v2df)
10248 v4sf __builtin_ia32_vfrczps (v4sf)
10249 v2df __builtin_ia32_vfrczsd (v2df, v2df)
10250 v4sf __builtin_ia32_vfrczss (v4sf, v4sf)
10251 v4df __builtin_ia32_vfrczpd256 (v4df)
10252 v8sf __builtin_ia32_vfrczps256 (v8sf)
10253 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di)
10254 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di)
10255 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si)
10256 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi)
10257 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi)
10258 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df)
10259 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf)
10260 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di)
10261 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si)
10262 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi)
10263 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi)
10264 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df)
10265 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf)
10266 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi)
10267 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
10268 v4si __builtin_ia32_vpcomeqd (v4si, v4si)
10269 v2di __builtin_ia32_vpcomeqq (v2di, v2di)
10270 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi)
10271 v4si __builtin_ia32_vpcomequd (v4si, v4si)
10272 v2di __builtin_ia32_vpcomequq (v2di, v2di)
10273 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi)
10274 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi)
10275 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi)
10276 v4si __builtin_ia32_vpcomfalsed (v4si, v4si)
10277 v2di __builtin_ia32_vpcomfalseq (v2di, v2di)
10278 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi)
10279 v4si __builtin_ia32_vpcomfalseud (v4si, v4si)
10280 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di)
10281 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi)
10282 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi)
10283 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi)
10284 v4si __builtin_ia32_vpcomged (v4si, v4si)
10285 v2di __builtin_ia32_vpcomgeq (v2di, v2di)
10286 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi)
10287 v4si __builtin_ia32_vpcomgeud (v4si, v4si)
10288 v2di __builtin_ia32_vpcomgeuq (v2di, v2di)
10289 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi)
10290 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi)
10291 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi)
10292 v4si __builtin_ia32_vpcomgtd (v4si, v4si)
10293 v2di __builtin_ia32_vpcomgtq (v2di, v2di)
10294 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi)
10295 v4si __builtin_ia32_vpcomgtud (v4si, v4si)
10296 v2di __builtin_ia32_vpcomgtuq (v2di, v2di)
10297 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi)
10298 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi)
10299 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi)
10300 v4si __builtin_ia32_vpcomled (v4si, v4si)
10301 v2di __builtin_ia32_vpcomleq (v2di, v2di)
10302 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi)
10303 v4si __builtin_ia32_vpcomleud (v4si, v4si)
10304 v2di __builtin_ia32_vpcomleuq (v2di, v2di)
10305 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi)
10306 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi)
10307 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi)
10308 v4si __builtin_ia32_vpcomltd (v4si, v4si)
10309 v2di __builtin_ia32_vpcomltq (v2di, v2di)
10310 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi)
10311 v4si __builtin_ia32_vpcomltud (v4si, v4si)
10312 v2di __builtin_ia32_vpcomltuq (v2di, v2di)
10313 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi)
10314 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi)
10315 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi)
10316 v4si __builtin_ia32_vpcomned (v4si, v4si)
10317 v2di __builtin_ia32_vpcomneq (v2di, v2di)
10318 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi)
10319 v4si __builtin_ia32_vpcomneud (v4si, v4si)
10320 v2di __builtin_ia32_vpcomneuq (v2di, v2di)
10321 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi)
10322 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi)
10323 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi)
10324 v4si __builtin_ia32_vpcomtrued (v4si, v4si)
10325 v2di __builtin_ia32_vpcomtrueq (v2di, v2di)
10326 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi)
10327 v4si __builtin_ia32_vpcomtrueud (v4si, v4si)
10328 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di)
10329 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi)
10330 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi)
10331 v4si __builtin_ia32_vphaddbd (v16qi)
10332 v2di __builtin_ia32_vphaddbq (v16qi)
10333 v8hi __builtin_ia32_vphaddbw (v16qi)
10334 v2di __builtin_ia32_vphadddq (v4si)
10335 v4si __builtin_ia32_vphaddubd (v16qi)
10336 v2di __builtin_ia32_vphaddubq (v16qi)
10337 v8hi __builtin_ia32_vphaddubw (v16qi)
10338 v2di __builtin_ia32_vphaddudq (v4si)
10339 v4si __builtin_ia32_vphadduwd (v8hi)
10340 v2di __builtin_ia32_vphadduwq (v8hi)
10341 v4si __builtin_ia32_vphaddwd (v8hi)
10342 v2di __builtin_ia32_vphaddwq (v8hi)
10343 v8hi __builtin_ia32_vphsubbw (v16qi)
10344 v2di __builtin_ia32_vphsubdq (v4si)
10345 v4si __builtin_ia32_vphsubwd (v8hi)
10346 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si)
10347 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di)
10348 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di)
10349 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si)
10350 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di)
10351 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di)
10352 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si)
10353 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi)
10354 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si)
10355 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi)
10356 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si)
10357 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si)
10358 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi)
10359 v16qi __builtin_ia32_vprotb (v16qi, v16qi)
10360 v4si __builtin_ia32_vprotd (v4si, v4si)
10361 v2di __builtin_ia32_vprotq (v2di, v2di)
10362 v8hi __builtin_ia32_vprotw (v8hi, v8hi)
10363 v16qi __builtin_ia32_vpshab (v16qi, v16qi)
10364 v4si __builtin_ia32_vpshad (v4si, v4si)
10365 v2di __builtin_ia32_vpshaq (v2di, v2di)
10366 v8hi __builtin_ia32_vpshaw (v8hi, v8hi)
10367 v16qi __builtin_ia32_vpshlb (v16qi, v16qi)
10368 v4si __builtin_ia32_vpshld (v4si, v4si)
10369 v2di __builtin_ia32_vpshlq (v2di, v2di)
10370 v8hi __builtin_ia32_vpshlw (v8hi, v8hi)
10373 The following built-in functions are available when @option{-mfma4} is used.
10374 All of them generate the machine instruction that is part of the name
10375 with MMX registers.
10378 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
10379 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
10380 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
10381 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
10382 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
10383 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
10384 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
10385 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
10386 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
10387 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
10388 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
10389 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
10390 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
10391 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
10392 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
10393 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
10394 v2df __builtin_ia32_fmaddsubpd (v2df, v2df, v2df)
10395 v4sf __builtin_ia32_fmaddsubps (v4sf, v4sf, v4sf)
10396 v2df __builtin_ia32_fmsubaddpd (v2df, v2df, v2df)
10397 v4sf __builtin_ia32_fmsubaddps (v4sf, v4sf, v4sf)
10398 v4df __builtin_ia32_fmaddpd256 (v4df, v4df, v4df)
10399 v8sf __builtin_ia32_fmaddps256 (v8sf, v8sf, v8sf)
10400 v4df __builtin_ia32_fmsubpd256 (v4df, v4df, v4df)
10401 v8sf __builtin_ia32_fmsubps256 (v8sf, v8sf, v8sf)
10402 v4df __builtin_ia32_fnmaddpd256 (v4df, v4df, v4df)
10403 v8sf __builtin_ia32_fnmaddps256 (v8sf, v8sf, v8sf)
10404 v4df __builtin_ia32_fnmsubpd256 (v4df, v4df, v4df)
10405 v8sf __builtin_ia32_fnmsubps256 (v8sf, v8sf, v8sf)
10406 v4df __builtin_ia32_fmaddsubpd256 (v4df, v4df, v4df)
10407 v8sf __builtin_ia32_fmaddsubps256 (v8sf, v8sf, v8sf)
10408 v4df __builtin_ia32_fmsubaddpd256 (v4df, v4df, v4df)
10409 v8sf __builtin_ia32_fmsubaddps256 (v8sf, v8sf, v8sf)
10413 The following built-in functions are available when @option{-mlwp} is used.
10416 void __builtin_ia32_llwpcb16 (void *);
10417 void __builtin_ia32_llwpcb32 (void *);
10418 void __builtin_ia32_llwpcb64 (void *);
10419 void * __builtin_ia32_llwpcb16 (void);
10420 void * __builtin_ia32_llwpcb32 (void);
10421 void * __builtin_ia32_llwpcb64 (void);
10422 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short)
10423 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int)
10424 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int)
10425 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short)
10426 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int)
10427 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int)
10430 The following built-in functions are available when @option{-mbmi} is used.
10431 All of them generate the machine instruction that is part of the name.
10433 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int);
10434 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long);
10437 The following built-in functions are available when @option{-mbmi2} is used.
10438 All of them generate the machine instruction that is part of the name.
10440 unsigned int _bzhi_u32 (unsigned int, unsigned int)
10441 unsigned int _pdep_u32 (unsigned int, unsigned int)
10442 unsigned int _pext_u32 (unsigned int, unsigned int)
10443 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long)
10444 unsigned long long _pdep_u64 (unsigned long long, unsigned long long)
10445 unsigned long long _pext_u64 (unsigned long long, unsigned long long)
10448 The following built-in functions are available when @option{-mlzcnt} is used.
10449 All of them generate the machine instruction that is part of the name.
10451 unsigned short __builtin_ia32_lzcnt_16(unsigned short);
10452 unsigned int __builtin_ia32_lzcnt_u32(unsigned int);
10453 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long);
10456 The following built-in functions are available when @option{-mtbm} is used.
10457 Both of them generate the immediate form of the bextr machine instruction.
10459 unsigned int __builtin_ia32_bextri_u32 (unsigned int, const unsigned int);
10460 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, const unsigned long long);
10464 The following built-in functions are available when @option{-m3dnow} is used.
10465 All of them generate the machine instruction that is part of the name.
10468 void __builtin_ia32_femms (void)
10469 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
10470 v2si __builtin_ia32_pf2id (v2sf)
10471 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
10472 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
10473 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
10474 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
10475 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
10476 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
10477 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
10478 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
10479 v2sf __builtin_ia32_pfrcp (v2sf)
10480 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
10481 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
10482 v2sf __builtin_ia32_pfrsqrt (v2sf)
10483 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
10484 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
10485 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
10486 v2sf __builtin_ia32_pi2fd (v2si)
10487 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
10490 The following built-in functions are available when both @option{-m3dnow}
10491 and @option{-march=athlon} are used. All of them generate the machine
10492 instruction that is part of the name.
10495 v2si __builtin_ia32_pf2iw (v2sf)
10496 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
10497 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
10498 v2sf __builtin_ia32_pi2fw (v2si)
10499 v2sf __builtin_ia32_pswapdsf (v2sf)
10500 v2si __builtin_ia32_pswapdsi (v2si)
10503 @node MIPS DSP Built-in Functions
10504 @subsection MIPS DSP Built-in Functions
10506 The MIPS DSP Application-Specific Extension (ASE) includes new
10507 instructions that are designed to improve the performance of DSP and
10508 media applications. It provides instructions that operate on packed
10509 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
10511 GCC supports MIPS DSP operations using both the generic
10512 vector extensions (@pxref{Vector Extensions}) and a collection of
10513 MIPS-specific built-in functions. Both kinds of support are
10514 enabled by the @option{-mdsp} command-line option.
10516 Revision 2 of the ASE was introduced in the second half of 2006.
10517 This revision adds extra instructions to the original ASE, but is
10518 otherwise backwards-compatible with it. You can select revision 2
10519 using the command-line option @option{-mdspr2}; this option implies
10522 The SCOUNT and POS bits of the DSP control register are global. The
10523 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
10524 POS bits. During optimization, the compiler will not delete these
10525 instructions and it will not delete calls to functions containing
10526 these instructions.
10528 At present, GCC only provides support for operations on 32-bit
10529 vectors. The vector type associated with 8-bit integer data is
10530 usually called @code{v4i8}, the vector type associated with Q7
10531 is usually called @code{v4q7}, the vector type associated with 16-bit
10532 integer data is usually called @code{v2i16}, and the vector type
10533 associated with Q15 is usually called @code{v2q15}. They can be
10534 defined in C as follows:
10537 typedef signed char v4i8 __attribute__ ((vector_size(4)));
10538 typedef signed char v4q7 __attribute__ ((vector_size(4)));
10539 typedef short v2i16 __attribute__ ((vector_size(4)));
10540 typedef short v2q15 __attribute__ ((vector_size(4)));
10543 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
10544 initialized in the same way as aggregates. For example:
10547 v4i8 a = @{1, 2, 3, 4@};
10549 b = (v4i8) @{5, 6, 7, 8@};
10551 v2q15 c = @{0x0fcb, 0x3a75@};
10553 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
10556 @emph{Note:} The CPU's endianness determines the order in which values
10557 are packed. On little-endian targets, the first value is the least
10558 significant and the last value is the most significant. The opposite
10559 order applies to big-endian targets. For example, the code above will
10560 set the lowest byte of @code{a} to @code{1} on little-endian targets
10561 and @code{4} on big-endian targets.
10563 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
10564 representation. As shown in this example, the integer representation
10565 of a Q7 value can be obtained by multiplying the fractional value by
10566 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
10567 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
10570 The table below lists the @code{v4i8} and @code{v2q15} operations for which
10571 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
10572 and @code{c} and @code{d} are @code{v2q15} values.
10574 @multitable @columnfractions .50 .50
10575 @item C code @tab MIPS instruction
10576 @item @code{a + b} @tab @code{addu.qb}
10577 @item @code{c + d} @tab @code{addq.ph}
10578 @item @code{a - b} @tab @code{subu.qb}
10579 @item @code{c - d} @tab @code{subq.ph}
10582 The table below lists the @code{v2i16} operation for which
10583 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
10584 @code{v2i16} values.
10586 @multitable @columnfractions .50 .50
10587 @item C code @tab MIPS instruction
10588 @item @code{e * f} @tab @code{mul.ph}
10591 It is easier to describe the DSP built-in functions if we first define
10592 the following types:
10597 typedef unsigned int ui32;
10598 typedef long long a64;
10601 @code{q31} and @code{i32} are actually the same as @code{int}, but we
10602 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
10603 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
10604 @code{long long}, but we use @code{a64} to indicate values that will
10605 be placed in one of the four DSP accumulators (@code{$ac0},
10606 @code{$ac1}, @code{$ac2} or @code{$ac3}).
10608 Also, some built-in functions prefer or require immediate numbers as
10609 parameters, because the corresponding DSP instructions accept both immediate
10610 numbers and register operands, or accept immediate numbers only. The
10611 immediate parameters are listed as follows.
10619 imm0_255: 0 to 255.
10620 imm_n32_31: -32 to 31.
10621 imm_n512_511: -512 to 511.
10624 The following built-in functions map directly to a particular MIPS DSP
10625 instruction. Please refer to the architecture specification
10626 for details on what each instruction does.
10629 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
10630 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
10631 q31 __builtin_mips_addq_s_w (q31, q31)
10632 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
10633 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
10634 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
10635 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
10636 q31 __builtin_mips_subq_s_w (q31, q31)
10637 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
10638 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
10639 i32 __builtin_mips_addsc (i32, i32)
10640 i32 __builtin_mips_addwc (i32, i32)
10641 i32 __builtin_mips_modsub (i32, i32)
10642 i32 __builtin_mips_raddu_w_qb (v4i8)
10643 v2q15 __builtin_mips_absq_s_ph (v2q15)
10644 q31 __builtin_mips_absq_s_w (q31)
10645 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
10646 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
10647 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
10648 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
10649 q31 __builtin_mips_preceq_w_phl (v2q15)
10650 q31 __builtin_mips_preceq_w_phr (v2q15)
10651 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
10652 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
10653 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
10654 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
10655 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
10656 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
10657 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
10658 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
10659 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
10660 v4i8 __builtin_mips_shll_qb (v4i8, i32)
10661 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
10662 v2q15 __builtin_mips_shll_ph (v2q15, i32)
10663 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
10664 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
10665 q31 __builtin_mips_shll_s_w (q31, imm0_31)
10666 q31 __builtin_mips_shll_s_w (q31, i32)
10667 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
10668 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
10669 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
10670 v2q15 __builtin_mips_shra_ph (v2q15, i32)
10671 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
10672 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
10673 q31 __builtin_mips_shra_r_w (q31, imm0_31)
10674 q31 __builtin_mips_shra_r_w (q31, i32)
10675 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
10676 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
10677 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
10678 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
10679 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
10680 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
10681 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
10682 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
10683 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
10684 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
10685 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
10686 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
10687 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
10688 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
10689 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
10690 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
10691 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
10692 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
10693 i32 __builtin_mips_bitrev (i32)
10694 i32 __builtin_mips_insv (i32, i32)
10695 v4i8 __builtin_mips_repl_qb (imm0_255)
10696 v4i8 __builtin_mips_repl_qb (i32)
10697 v2q15 __builtin_mips_repl_ph (imm_n512_511)
10698 v2q15 __builtin_mips_repl_ph (i32)
10699 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
10700 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
10701 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
10702 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
10703 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
10704 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
10705 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
10706 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
10707 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
10708 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
10709 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
10710 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
10711 i32 __builtin_mips_extr_w (a64, imm0_31)
10712 i32 __builtin_mips_extr_w (a64, i32)
10713 i32 __builtin_mips_extr_r_w (a64, imm0_31)
10714 i32 __builtin_mips_extr_s_h (a64, i32)
10715 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
10716 i32 __builtin_mips_extr_rs_w (a64, i32)
10717 i32 __builtin_mips_extr_s_h (a64, imm0_31)
10718 i32 __builtin_mips_extr_r_w (a64, i32)
10719 i32 __builtin_mips_extp (a64, imm0_31)
10720 i32 __builtin_mips_extp (a64, i32)
10721 i32 __builtin_mips_extpdp (a64, imm0_31)
10722 i32 __builtin_mips_extpdp (a64, i32)
10723 a64 __builtin_mips_shilo (a64, imm_n32_31)
10724 a64 __builtin_mips_shilo (a64, i32)
10725 a64 __builtin_mips_mthlip (a64, i32)
10726 void __builtin_mips_wrdsp (i32, imm0_63)
10727 i32 __builtin_mips_rddsp (imm0_63)
10728 i32 __builtin_mips_lbux (void *, i32)
10729 i32 __builtin_mips_lhx (void *, i32)
10730 i32 __builtin_mips_lwx (void *, i32)
10731 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only]
10732 i32 __builtin_mips_bposge32 (void)
10733 a64 __builtin_mips_madd (a64, i32, i32);
10734 a64 __builtin_mips_maddu (a64, ui32, ui32);
10735 a64 __builtin_mips_msub (a64, i32, i32);
10736 a64 __builtin_mips_msubu (a64, ui32, ui32);
10737 a64 __builtin_mips_mult (i32, i32);
10738 a64 __builtin_mips_multu (ui32, ui32);
10741 The following built-in functions map directly to a particular MIPS DSP REV 2
10742 instruction. Please refer to the architecture specification
10743 for details on what each instruction does.
10746 v4q7 __builtin_mips_absq_s_qb (v4q7);
10747 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
10748 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
10749 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
10750 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
10751 i32 __builtin_mips_append (i32, i32, imm0_31);
10752 i32 __builtin_mips_balign (i32, i32, imm0_3);
10753 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
10754 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
10755 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
10756 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
10757 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
10758 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
10759 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
10760 q31 __builtin_mips_mulq_rs_w (q31, q31);
10761 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
10762 q31 __builtin_mips_mulq_s_w (q31, q31);
10763 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
10764 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
10765 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
10766 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
10767 i32 __builtin_mips_prepend (i32, i32, imm0_31);
10768 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
10769 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
10770 v4i8 __builtin_mips_shra_qb (v4i8, i32);
10771 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
10772 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
10773 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
10774 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
10775 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
10776 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
10777 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
10778 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
10779 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
10780 q31 __builtin_mips_addqh_w (q31, q31);
10781 q31 __builtin_mips_addqh_r_w (q31, q31);
10782 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
10783 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
10784 q31 __builtin_mips_subqh_w (q31, q31);
10785 q31 __builtin_mips_subqh_r_w (q31, q31);
10786 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
10787 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
10788 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
10789 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
10790 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
10791 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
10795 @node MIPS Paired-Single Support
10796 @subsection MIPS Paired-Single Support
10798 The MIPS64 architecture includes a number of instructions that
10799 operate on pairs of single-precision floating-point values.
10800 Each pair is packed into a 64-bit floating-point register,
10801 with one element being designated the ``upper half'' and
10802 the other being designated the ``lower half''.
10804 GCC supports paired-single operations using both the generic
10805 vector extensions (@pxref{Vector Extensions}) and a collection of
10806 MIPS-specific built-in functions. Both kinds of support are
10807 enabled by the @option{-mpaired-single} command-line option.
10809 The vector type associated with paired-single values is usually
10810 called @code{v2sf}. It can be defined in C as follows:
10813 typedef float v2sf __attribute__ ((vector_size (8)));
10816 @code{v2sf} values are initialized in the same way as aggregates.
10820 v2sf a = @{1.5, 9.1@};
10823 b = (v2sf) @{e, f@};
10826 @emph{Note:} The CPU's endianness determines which value is stored in
10827 the upper half of a register and which value is stored in the lower half.
10828 On little-endian targets, the first value is the lower one and the second
10829 value is the upper one. The opposite order applies to big-endian targets.
10830 For example, the code above will set the lower half of @code{a} to
10831 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
10833 @node MIPS Loongson Built-in Functions
10834 @subsection MIPS Loongson Built-in Functions
10836 GCC provides intrinsics to access the SIMD instructions provided by the
10837 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
10838 available after inclusion of the @code{loongson.h} header file,
10839 operate on the following 64-bit vector types:
10842 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
10843 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
10844 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
10845 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
10846 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
10847 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
10850 The intrinsics provided are listed below; each is named after the
10851 machine instruction to which it corresponds, with suffixes added as
10852 appropriate to distinguish intrinsics that expand to the same machine
10853 instruction yet have different argument types. Refer to the architecture
10854 documentation for a description of the functionality of each
10858 int16x4_t packsswh (int32x2_t s, int32x2_t t);
10859 int8x8_t packsshb (int16x4_t s, int16x4_t t);
10860 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
10861 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
10862 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
10863 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
10864 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
10865 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
10866 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
10867 uint64_t paddd_u (uint64_t s, uint64_t t);
10868 int64_t paddd_s (int64_t s, int64_t t);
10869 int16x4_t paddsh (int16x4_t s, int16x4_t t);
10870 int8x8_t paddsb (int8x8_t s, int8x8_t t);
10871 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
10872 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
10873 uint64_t pandn_ud (uint64_t s, uint64_t t);
10874 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
10875 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
10876 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
10877 int64_t pandn_sd (int64_t s, int64_t t);
10878 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
10879 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
10880 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
10881 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
10882 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
10883 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
10884 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
10885 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
10886 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
10887 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
10888 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
10889 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
10890 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
10891 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
10892 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
10893 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
10894 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
10895 uint16x4_t pextrh_u (uint16x4_t s, int field);
10896 int16x4_t pextrh_s (int16x4_t s, int field);
10897 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
10898 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
10899 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
10900 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
10901 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
10902 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
10903 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
10904 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
10905 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
10906 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
10907 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
10908 int16x4_t pminsh (int16x4_t s, int16x4_t t);
10909 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
10910 uint8x8_t pmovmskb_u (uint8x8_t s);
10911 int8x8_t pmovmskb_s (int8x8_t s);
10912 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
10913 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
10914 int16x4_t pmullh (int16x4_t s, int16x4_t t);
10915 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
10916 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
10917 uint16x4_t biadd (uint8x8_t s);
10918 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
10919 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
10920 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
10921 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
10922 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
10923 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
10924 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
10925 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
10926 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
10927 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
10928 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
10929 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
10930 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
10931 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
10932 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
10933 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
10934 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
10935 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
10936 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
10937 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
10938 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
10939 uint64_t psubd_u (uint64_t s, uint64_t t);
10940 int64_t psubd_s (int64_t s, int64_t t);
10941 int16x4_t psubsh (int16x4_t s, int16x4_t t);
10942 int8x8_t psubsb (int8x8_t s, int8x8_t t);
10943 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
10944 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
10945 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
10946 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
10947 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
10948 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
10949 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
10950 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
10951 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
10952 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
10953 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
10954 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
10955 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
10956 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
10960 * Paired-Single Arithmetic::
10961 * Paired-Single Built-in Functions::
10962 * MIPS-3D Built-in Functions::
10965 @node Paired-Single Arithmetic
10966 @subsubsection Paired-Single Arithmetic
10968 The table below lists the @code{v2sf} operations for which hardware
10969 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
10970 values and @code{x} is an integral value.
10972 @multitable @columnfractions .50 .50
10973 @item C code @tab MIPS instruction
10974 @item @code{a + b} @tab @code{add.ps}
10975 @item @code{a - b} @tab @code{sub.ps}
10976 @item @code{-a} @tab @code{neg.ps}
10977 @item @code{a * b} @tab @code{mul.ps}
10978 @item @code{a * b + c} @tab @code{madd.ps}
10979 @item @code{a * b - c} @tab @code{msub.ps}
10980 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
10981 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
10982 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
10985 Note that the multiply-accumulate instructions can be disabled
10986 using the command-line option @code{-mno-fused-madd}.
10988 @node Paired-Single Built-in Functions
10989 @subsubsection Paired-Single Built-in Functions
10991 The following paired-single functions map directly to a particular
10992 MIPS instruction. Please refer to the architecture specification
10993 for details on what each instruction does.
10996 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
10997 Pair lower lower (@code{pll.ps}).
10999 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
11000 Pair upper lower (@code{pul.ps}).
11002 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
11003 Pair lower upper (@code{plu.ps}).
11005 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
11006 Pair upper upper (@code{puu.ps}).
11008 @item v2sf __builtin_mips_cvt_ps_s (float, float)
11009 Convert pair to paired single (@code{cvt.ps.s}).
11011 @item float __builtin_mips_cvt_s_pl (v2sf)
11012 Convert pair lower to single (@code{cvt.s.pl}).
11014 @item float __builtin_mips_cvt_s_pu (v2sf)
11015 Convert pair upper to single (@code{cvt.s.pu}).
11017 @item v2sf __builtin_mips_abs_ps (v2sf)
11018 Absolute value (@code{abs.ps}).
11020 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
11021 Align variable (@code{alnv.ps}).
11023 @emph{Note:} The value of the third parameter must be 0 or 4
11024 modulo 8, otherwise the result will be unpredictable. Please read the
11025 instruction description for details.
11028 The following multi-instruction functions are also available.
11029 In each case, @var{cond} can be any of the 16 floating-point conditions:
11030 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
11031 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
11032 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
11035 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11036 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11037 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
11038 @code{movt.ps}/@code{movf.ps}).
11040 The @code{movt} functions return the value @var{x} computed by:
11043 c.@var{cond}.ps @var{cc},@var{a},@var{b}
11044 mov.ps @var{x},@var{c}
11045 movt.ps @var{x},@var{d},@var{cc}
11048 The @code{movf} functions are similar but use @code{movf.ps} instead
11051 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11052 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11053 Comparison of two paired-single values (@code{c.@var{cond}.ps},
11054 @code{bc1t}/@code{bc1f}).
11056 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
11057 and return either the upper or lower half of the result. For example:
11061 if (__builtin_mips_upper_c_eq_ps (a, b))
11062 upper_halves_are_equal ();
11064 upper_halves_are_unequal ();
11066 if (__builtin_mips_lower_c_eq_ps (a, b))
11067 lower_halves_are_equal ();
11069 lower_halves_are_unequal ();
11073 @node MIPS-3D Built-in Functions
11074 @subsubsection MIPS-3D Built-in Functions
11076 The MIPS-3D Application-Specific Extension (ASE) includes additional
11077 paired-single instructions that are designed to improve the performance
11078 of 3D graphics operations. Support for these instructions is controlled
11079 by the @option{-mips3d} command-line option.
11081 The functions listed below map directly to a particular MIPS-3D
11082 instruction. Please refer to the architecture specification for
11083 more details on what each instruction does.
11086 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
11087 Reduction add (@code{addr.ps}).
11089 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
11090 Reduction multiply (@code{mulr.ps}).
11092 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
11093 Convert paired single to paired word (@code{cvt.pw.ps}).
11095 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
11096 Convert paired word to paired single (@code{cvt.ps.pw}).
11098 @item float __builtin_mips_recip1_s (float)
11099 @itemx double __builtin_mips_recip1_d (double)
11100 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
11101 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
11103 @item float __builtin_mips_recip2_s (float, float)
11104 @itemx double __builtin_mips_recip2_d (double, double)
11105 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
11106 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
11108 @item float __builtin_mips_rsqrt1_s (float)
11109 @itemx double __builtin_mips_rsqrt1_d (double)
11110 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
11111 Reduced precision reciprocal square root (sequence step 1)
11112 (@code{rsqrt1.@var{fmt}}).
11114 @item float __builtin_mips_rsqrt2_s (float, float)
11115 @itemx double __builtin_mips_rsqrt2_d (double, double)
11116 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
11117 Reduced precision reciprocal square root (sequence step 2)
11118 (@code{rsqrt2.@var{fmt}}).
11121 The following multi-instruction functions are also available.
11122 In each case, @var{cond} can be any of the 16 floating-point conditions:
11123 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
11124 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
11125 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
11128 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
11129 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
11130 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
11131 @code{bc1t}/@code{bc1f}).
11133 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
11134 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
11139 if (__builtin_mips_cabs_eq_s (a, b))
11145 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11146 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11147 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
11148 @code{bc1t}/@code{bc1f}).
11150 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
11151 and return either the upper or lower half of the result. For example:
11155 if (__builtin_mips_upper_cabs_eq_ps (a, b))
11156 upper_halves_are_equal ();
11158 upper_halves_are_unequal ();
11160 if (__builtin_mips_lower_cabs_eq_ps (a, b))
11161 lower_halves_are_equal ();
11163 lower_halves_are_unequal ();
11166 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11167 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11168 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
11169 @code{movt.ps}/@code{movf.ps}).
11171 The @code{movt} functions return the value @var{x} computed by:
11174 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
11175 mov.ps @var{x},@var{c}
11176 movt.ps @var{x},@var{d},@var{cc}
11179 The @code{movf} functions are similar but use @code{movf.ps} instead
11182 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11183 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11184 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11185 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
11186 Comparison of two paired-single values
11187 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
11188 @code{bc1any2t}/@code{bc1any2f}).
11190 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
11191 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
11192 result is true and the @code{all} forms return true if both results are true.
11197 if (__builtin_mips_any_c_eq_ps (a, b))
11202 if (__builtin_mips_all_c_eq_ps (a, b))
11208 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11209 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11210 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11211 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
11212 Comparison of four paired-single values
11213 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
11214 @code{bc1any4t}/@code{bc1any4f}).
11216 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
11217 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
11218 The @code{any} forms return true if any of the four results are true
11219 and the @code{all} forms return true if all four results are true.
11224 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
11229 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
11236 @node picoChip Built-in Functions
11237 @subsection picoChip Built-in Functions
11239 GCC provides an interface to selected machine instructions from the
11240 picoChip instruction set.
11243 @item int __builtin_sbc (int @var{value})
11244 Sign bit count. Return the number of consecutive bits in @var{value}
11245 which have the same value as the sign-bit. The result is the number of
11246 leading sign bits minus one, giving the number of redundant sign bits in
11249 @item int __builtin_byteswap (int @var{value})
11250 Byte swap. Return the result of swapping the upper and lower bytes of
11253 @item int __builtin_brev (int @var{value})
11254 Bit reversal. Return the result of reversing the bits in
11255 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
11258 @item int __builtin_adds (int @var{x}, int @var{y})
11259 Saturating addition. Return the result of adding @var{x} and @var{y},
11260 storing the value 32767 if the result overflows.
11262 @item int __builtin_subs (int @var{x}, int @var{y})
11263 Saturating subtraction. Return the result of subtracting @var{y} from
11264 @var{x}, storing the value @minus{}32768 if the result overflows.
11266 @item void __builtin_halt (void)
11267 Halt. The processor will stop execution. This built-in is useful for
11268 implementing assertions.
11272 @node Other MIPS Built-in Functions
11273 @subsection Other MIPS Built-in Functions
11275 GCC provides other MIPS-specific built-in functions:
11278 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
11279 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
11280 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
11281 when this function is available.
11284 @node PowerPC AltiVec/VSX Built-in Functions
11285 @subsection PowerPC AltiVec Built-in Functions
11287 GCC provides an interface for the PowerPC family of processors to access
11288 the AltiVec operations described in Motorola's AltiVec Programming
11289 Interface Manual. The interface is made available by including
11290 @code{<altivec.h>} and using @option{-maltivec} and
11291 @option{-mabi=altivec}. The interface supports the following vector
11295 vector unsigned char
11299 vector unsigned short
11300 vector signed short
11304 vector unsigned int
11310 If @option{-mvsx} is used the following additional vector types are
11314 vector unsigned long
11319 The long types are only implemented for 64-bit code generation, and
11320 the long type is only used in the floating point/integer conversion
11323 GCC's implementation of the high-level language interface available from
11324 C and C++ code differs from Motorola's documentation in several ways.
11329 A vector constant is a list of constant expressions within curly braces.
11332 A vector initializer requires no cast if the vector constant is of the
11333 same type as the variable it is initializing.
11336 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11337 vector type is the default signedness of the base type. The default
11338 varies depending on the operating system, so a portable program should
11339 always specify the signedness.
11342 Compiling with @option{-maltivec} adds keywords @code{__vector},
11343 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
11344 @code{bool}. When compiling ISO C, the context-sensitive substitution
11345 of the keywords @code{vector}, @code{pixel} and @code{bool} is
11346 disabled. To use them, you must include @code{<altivec.h>} instead.
11349 GCC allows using a @code{typedef} name as the type specifier for a
11353 For C, overloaded functions are implemented with macros so the following
11357 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11360 Since @code{vec_add} is a macro, the vector constant in the example
11361 is treated as four separate arguments. Wrap the entire argument in
11362 parentheses for this to work.
11365 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
11366 Internally, GCC uses built-in functions to achieve the functionality in
11367 the aforementioned header file, but they are not supported and are
11368 subject to change without notice.
11370 The following interfaces are supported for the generic and specific
11371 AltiVec operations and the AltiVec predicates. In cases where there
11372 is a direct mapping between generic and specific operations, only the
11373 generic names are shown here, although the specific operations can also
11376 Arguments that are documented as @code{const int} require literal
11377 integral values within the range required for that operation.
11380 vector signed char vec_abs (vector signed char);
11381 vector signed short vec_abs (vector signed short);
11382 vector signed int vec_abs (vector signed int);
11383 vector float vec_abs (vector float);
11385 vector signed char vec_abss (vector signed char);
11386 vector signed short vec_abss (vector signed short);
11387 vector signed int vec_abss (vector signed int);
11389 vector signed char vec_add (vector bool char, vector signed char);
11390 vector signed char vec_add (vector signed char, vector bool char);
11391 vector signed char vec_add (vector signed char, vector signed char);
11392 vector unsigned char vec_add (vector bool char, vector unsigned char);
11393 vector unsigned char vec_add (vector unsigned char, vector bool char);
11394 vector unsigned char vec_add (vector unsigned char,
11395 vector unsigned char);
11396 vector signed short vec_add (vector bool short, vector signed short);
11397 vector signed short vec_add (vector signed short, vector bool short);
11398 vector signed short vec_add (vector signed short, vector signed short);
11399 vector unsigned short vec_add (vector bool short,
11400 vector unsigned short);
11401 vector unsigned short vec_add (vector unsigned short,
11402 vector bool short);
11403 vector unsigned short vec_add (vector unsigned short,
11404 vector unsigned short);
11405 vector signed int vec_add (vector bool int, vector signed int);
11406 vector signed int vec_add (vector signed int, vector bool int);
11407 vector signed int vec_add (vector signed int, vector signed int);
11408 vector unsigned int vec_add (vector bool int, vector unsigned int);
11409 vector unsigned int vec_add (vector unsigned int, vector bool int);
11410 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
11411 vector float vec_add (vector float, vector float);
11413 vector float vec_vaddfp (vector float, vector float);
11415 vector signed int vec_vadduwm (vector bool int, vector signed int);
11416 vector signed int vec_vadduwm (vector signed int, vector bool int);
11417 vector signed int vec_vadduwm (vector signed int, vector signed int);
11418 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
11419 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
11420 vector unsigned int vec_vadduwm (vector unsigned int,
11421 vector unsigned int);
11423 vector signed short vec_vadduhm (vector bool short,
11424 vector signed short);
11425 vector signed short vec_vadduhm (vector signed short,
11426 vector bool short);
11427 vector signed short vec_vadduhm (vector signed short,
11428 vector signed short);
11429 vector unsigned short vec_vadduhm (vector bool short,
11430 vector unsigned short);
11431 vector unsigned short vec_vadduhm (vector unsigned short,
11432 vector bool short);
11433 vector unsigned short vec_vadduhm (vector unsigned short,
11434 vector unsigned short);
11436 vector signed char vec_vaddubm (vector bool char, vector signed char);
11437 vector signed char vec_vaddubm (vector signed char, vector bool char);
11438 vector signed char vec_vaddubm (vector signed char, vector signed char);
11439 vector unsigned char vec_vaddubm (vector bool char,
11440 vector unsigned char);
11441 vector unsigned char vec_vaddubm (vector unsigned char,
11443 vector unsigned char vec_vaddubm (vector unsigned char,
11444 vector unsigned char);
11446 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
11448 vector unsigned char vec_adds (vector bool char, vector unsigned char);
11449 vector unsigned char vec_adds (vector unsigned char, vector bool char);
11450 vector unsigned char vec_adds (vector unsigned char,
11451 vector unsigned char);
11452 vector signed char vec_adds (vector bool char, vector signed char);
11453 vector signed char vec_adds (vector signed char, vector bool char);
11454 vector signed char vec_adds (vector signed char, vector signed char);
11455 vector unsigned short vec_adds (vector bool short,
11456 vector unsigned short);
11457 vector unsigned short vec_adds (vector unsigned short,
11458 vector bool short);
11459 vector unsigned short vec_adds (vector unsigned short,
11460 vector unsigned short);
11461 vector signed short vec_adds (vector bool short, vector signed short);
11462 vector signed short vec_adds (vector signed short, vector bool short);
11463 vector signed short vec_adds (vector signed short, vector signed short);
11464 vector unsigned int vec_adds (vector bool int, vector unsigned int);
11465 vector unsigned int vec_adds (vector unsigned int, vector bool int);
11466 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
11467 vector signed int vec_adds (vector bool int, vector signed int);
11468 vector signed int vec_adds (vector signed int, vector bool int);
11469 vector signed int vec_adds (vector signed int, vector signed int);
11471 vector signed int vec_vaddsws (vector bool int, vector signed int);
11472 vector signed int vec_vaddsws (vector signed int, vector bool int);
11473 vector signed int vec_vaddsws (vector signed int, vector signed int);
11475 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
11476 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
11477 vector unsigned int vec_vadduws (vector unsigned int,
11478 vector unsigned int);
11480 vector signed short vec_vaddshs (vector bool short,
11481 vector signed short);
11482 vector signed short vec_vaddshs (vector signed short,
11483 vector bool short);
11484 vector signed short vec_vaddshs (vector signed short,
11485 vector signed short);
11487 vector unsigned short vec_vadduhs (vector bool short,
11488 vector unsigned short);
11489 vector unsigned short vec_vadduhs (vector unsigned short,
11490 vector bool short);
11491 vector unsigned short vec_vadduhs (vector unsigned short,
11492 vector unsigned short);
11494 vector signed char vec_vaddsbs (vector bool char, vector signed char);
11495 vector signed char vec_vaddsbs (vector signed char, vector bool char);
11496 vector signed char vec_vaddsbs (vector signed char, vector signed char);
11498 vector unsigned char vec_vaddubs (vector bool char,
11499 vector unsigned char);
11500 vector unsigned char vec_vaddubs (vector unsigned char,
11502 vector unsigned char vec_vaddubs (vector unsigned char,
11503 vector unsigned char);
11505 vector float vec_and (vector float, vector float);
11506 vector float vec_and (vector float, vector bool int);
11507 vector float vec_and (vector bool int, vector float);
11508 vector bool int vec_and (vector bool int, vector bool int);
11509 vector signed int vec_and (vector bool int, vector signed int);
11510 vector signed int vec_and (vector signed int, vector bool int);
11511 vector signed int vec_and (vector signed int, vector signed int);
11512 vector unsigned int vec_and (vector bool int, vector unsigned int);
11513 vector unsigned int vec_and (vector unsigned int, vector bool int);
11514 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
11515 vector bool short vec_and (vector bool short, vector bool short);
11516 vector signed short vec_and (vector bool short, vector signed short);
11517 vector signed short vec_and (vector signed short, vector bool short);
11518 vector signed short vec_and (vector signed short, vector signed short);
11519 vector unsigned short vec_and (vector bool short,
11520 vector unsigned short);
11521 vector unsigned short vec_and (vector unsigned short,
11522 vector bool short);
11523 vector unsigned short vec_and (vector unsigned short,
11524 vector unsigned short);
11525 vector signed char vec_and (vector bool char, vector signed char);
11526 vector bool char vec_and (vector bool char, vector bool char);
11527 vector signed char vec_and (vector signed char, vector bool char);
11528 vector signed char vec_and (vector signed char, vector signed char);
11529 vector unsigned char vec_and (vector bool char, vector unsigned char);
11530 vector unsigned char vec_and (vector unsigned char, vector bool char);
11531 vector unsigned char vec_and (vector unsigned char,
11532 vector unsigned char);
11534 vector float vec_andc (vector float, vector float);
11535 vector float vec_andc (vector float, vector bool int);
11536 vector float vec_andc (vector bool int, vector float);
11537 vector bool int vec_andc (vector bool int, vector bool int);
11538 vector signed int vec_andc (vector bool int, vector signed int);
11539 vector signed int vec_andc (vector signed int, vector bool int);
11540 vector signed int vec_andc (vector signed int, vector signed int);
11541 vector unsigned int vec_andc (vector bool int, vector unsigned int);
11542 vector unsigned int vec_andc (vector unsigned int, vector bool int);
11543 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
11544 vector bool short vec_andc (vector bool short, vector bool short);
11545 vector signed short vec_andc (vector bool short, vector signed short);
11546 vector signed short vec_andc (vector signed short, vector bool short);
11547 vector signed short vec_andc (vector signed short, vector signed short);
11548 vector unsigned short vec_andc (vector bool short,
11549 vector unsigned short);
11550 vector unsigned short vec_andc (vector unsigned short,
11551 vector bool short);
11552 vector unsigned short vec_andc (vector unsigned short,
11553 vector unsigned short);
11554 vector signed char vec_andc (vector bool char, vector signed char);
11555 vector bool char vec_andc (vector bool char, vector bool char);
11556 vector signed char vec_andc (vector signed char, vector bool char);
11557 vector signed char vec_andc (vector signed char, vector signed char);
11558 vector unsigned char vec_andc (vector bool char, vector unsigned char);
11559 vector unsigned char vec_andc (vector unsigned char, vector bool char);
11560 vector unsigned char vec_andc (vector unsigned char,
11561 vector unsigned char);
11563 vector unsigned char vec_avg (vector unsigned char,
11564 vector unsigned char);
11565 vector signed char vec_avg (vector signed char, vector signed char);
11566 vector unsigned short vec_avg (vector unsigned short,
11567 vector unsigned short);
11568 vector signed short vec_avg (vector signed short, vector signed short);
11569 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
11570 vector signed int vec_avg (vector signed int, vector signed int);
11572 vector signed int vec_vavgsw (vector signed int, vector signed int);
11574 vector unsigned int vec_vavguw (vector unsigned int,
11575 vector unsigned int);
11577 vector signed short vec_vavgsh (vector signed short,
11578 vector signed short);
11580 vector unsigned short vec_vavguh (vector unsigned short,
11581 vector unsigned short);
11583 vector signed char vec_vavgsb (vector signed char, vector signed char);
11585 vector unsigned char vec_vavgub (vector unsigned char,
11586 vector unsigned char);
11588 vector float vec_copysign (vector float);
11590 vector float vec_ceil (vector float);
11592 vector signed int vec_cmpb (vector float, vector float);
11594 vector bool char vec_cmpeq (vector signed char, vector signed char);
11595 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
11596 vector bool short vec_cmpeq (vector signed short, vector signed short);
11597 vector bool short vec_cmpeq (vector unsigned short,
11598 vector unsigned short);
11599 vector bool int vec_cmpeq (vector signed int, vector signed int);
11600 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
11601 vector bool int vec_cmpeq (vector float, vector float);
11603 vector bool int vec_vcmpeqfp (vector float, vector float);
11605 vector bool int vec_vcmpequw (vector signed int, vector signed int);
11606 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
11608 vector bool short vec_vcmpequh (vector signed short,
11609 vector signed short);
11610 vector bool short vec_vcmpequh (vector unsigned short,
11611 vector unsigned short);
11613 vector bool char vec_vcmpequb (vector signed char, vector signed char);
11614 vector bool char vec_vcmpequb (vector unsigned char,
11615 vector unsigned char);
11617 vector bool int vec_cmpge (vector float, vector float);
11619 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
11620 vector bool char vec_cmpgt (vector signed char, vector signed char);
11621 vector bool short vec_cmpgt (vector unsigned short,
11622 vector unsigned short);
11623 vector bool short vec_cmpgt (vector signed short, vector signed short);
11624 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
11625 vector bool int vec_cmpgt (vector signed int, vector signed int);
11626 vector bool int vec_cmpgt (vector float, vector float);
11628 vector bool int vec_vcmpgtfp (vector float, vector float);
11630 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
11632 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
11634 vector bool short vec_vcmpgtsh (vector signed short,
11635 vector signed short);
11637 vector bool short vec_vcmpgtuh (vector unsigned short,
11638 vector unsigned short);
11640 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
11642 vector bool char vec_vcmpgtub (vector unsigned char,
11643 vector unsigned char);
11645 vector bool int vec_cmple (vector float, vector float);
11647 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
11648 vector bool char vec_cmplt (vector signed char, vector signed char);
11649 vector bool short vec_cmplt (vector unsigned short,
11650 vector unsigned short);
11651 vector bool short vec_cmplt (vector signed short, vector signed short);
11652 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
11653 vector bool int vec_cmplt (vector signed int, vector signed int);
11654 vector bool int vec_cmplt (vector float, vector float);
11656 vector float vec_ctf (vector unsigned int, const int);
11657 vector float vec_ctf (vector signed int, const int);
11659 vector float vec_vcfsx (vector signed int, const int);
11661 vector float vec_vcfux (vector unsigned int, const int);
11663 vector signed int vec_cts (vector float, const int);
11665 vector unsigned int vec_ctu (vector float, const int);
11667 void vec_dss (const int);
11669 void vec_dssall (void);
11671 void vec_dst (const vector unsigned char *, int, const int);
11672 void vec_dst (const vector signed char *, int, const int);
11673 void vec_dst (const vector bool char *, int, const int);
11674 void vec_dst (const vector unsigned short *, int, const int);
11675 void vec_dst (const vector signed short *, int, const int);
11676 void vec_dst (const vector bool short *, int, const int);
11677 void vec_dst (const vector pixel *, int, const int);
11678 void vec_dst (const vector unsigned int *, int, const int);
11679 void vec_dst (const vector signed int *, int, const int);
11680 void vec_dst (const vector bool int *, int, const int);
11681 void vec_dst (const vector float *, int, const int);
11682 void vec_dst (const unsigned char *, int, const int);
11683 void vec_dst (const signed char *, int, const int);
11684 void vec_dst (const unsigned short *, int, const int);
11685 void vec_dst (const short *, int, const int);
11686 void vec_dst (const unsigned int *, int, const int);
11687 void vec_dst (const int *, int, const int);
11688 void vec_dst (const unsigned long *, int, const int);
11689 void vec_dst (const long *, int, const int);
11690 void vec_dst (const float *, int, const int);
11692 void vec_dstst (const vector unsigned char *, int, const int);
11693 void vec_dstst (const vector signed char *, int, const int);
11694 void vec_dstst (const vector bool char *, int, const int);
11695 void vec_dstst (const vector unsigned short *, int, const int);
11696 void vec_dstst (const vector signed short *, int, const int);
11697 void vec_dstst (const vector bool short *, int, const int);
11698 void vec_dstst (const vector pixel *, int, const int);
11699 void vec_dstst (const vector unsigned int *, int, const int);
11700 void vec_dstst (const vector signed int *, int, const int);
11701 void vec_dstst (const vector bool int *, int, const int);
11702 void vec_dstst (const vector float *, int, const int);
11703 void vec_dstst (const unsigned char *, int, const int);
11704 void vec_dstst (const signed char *, int, const int);
11705 void vec_dstst (const unsigned short *, int, const int);
11706 void vec_dstst (const short *, int, const int);
11707 void vec_dstst (const unsigned int *, int, const int);
11708 void vec_dstst (const int *, int, const int);
11709 void vec_dstst (const unsigned long *, int, const int);
11710 void vec_dstst (const long *, int, const int);
11711 void vec_dstst (const float *, int, const int);
11713 void vec_dststt (const vector unsigned char *, int, const int);
11714 void vec_dststt (const vector signed char *, int, const int);
11715 void vec_dststt (const vector bool char *, int, const int);
11716 void vec_dststt (const vector unsigned short *, int, const int);
11717 void vec_dststt (const vector signed short *, int, const int);
11718 void vec_dststt (const vector bool short *, int, const int);
11719 void vec_dststt (const vector pixel *, int, const int);
11720 void vec_dststt (const vector unsigned int *, int, const int);
11721 void vec_dststt (const vector signed int *, int, const int);
11722 void vec_dststt (const vector bool int *, int, const int);
11723 void vec_dststt (const vector float *, int, const int);
11724 void vec_dststt (const unsigned char *, int, const int);
11725 void vec_dststt (const signed char *, int, const int);
11726 void vec_dststt (const unsigned short *, int, const int);
11727 void vec_dststt (const short *, int, const int);
11728 void vec_dststt (const unsigned int *, int, const int);
11729 void vec_dststt (const int *, int, const int);
11730 void vec_dststt (const unsigned long *, int, const int);
11731 void vec_dststt (const long *, int, const int);
11732 void vec_dststt (const float *, int, const int);
11734 void vec_dstt (const vector unsigned char *, int, const int);
11735 void vec_dstt (const vector signed char *, int, const int);
11736 void vec_dstt (const vector bool char *, int, const int);
11737 void vec_dstt (const vector unsigned short *, int, const int);
11738 void vec_dstt (const vector signed short *, int, const int);
11739 void vec_dstt (const vector bool short *, int, const int);
11740 void vec_dstt (const vector pixel *, int, const int);
11741 void vec_dstt (const vector unsigned int *, int, const int);
11742 void vec_dstt (const vector signed int *, int, const int);
11743 void vec_dstt (const vector bool int *, int, const int);
11744 void vec_dstt (const vector float *, int, const int);
11745 void vec_dstt (const unsigned char *, int, const int);
11746 void vec_dstt (const signed char *, int, const int);
11747 void vec_dstt (const unsigned short *, int, const int);
11748 void vec_dstt (const short *, int, const int);
11749 void vec_dstt (const unsigned int *, int, const int);
11750 void vec_dstt (const int *, int, const int);
11751 void vec_dstt (const unsigned long *, int, const int);
11752 void vec_dstt (const long *, int, const int);
11753 void vec_dstt (const float *, int, const int);
11755 vector float vec_expte (vector float);
11757 vector float vec_floor (vector float);
11759 vector float vec_ld (int, const vector float *);
11760 vector float vec_ld (int, const float *);
11761 vector bool int vec_ld (int, const vector bool int *);
11762 vector signed int vec_ld (int, const vector signed int *);
11763 vector signed int vec_ld (int, const int *);
11764 vector signed int vec_ld (int, const long *);
11765 vector unsigned int vec_ld (int, const vector unsigned int *);
11766 vector unsigned int vec_ld (int, const unsigned int *);
11767 vector unsigned int vec_ld (int, const unsigned long *);
11768 vector bool short vec_ld (int, const vector bool short *);
11769 vector pixel vec_ld (int, const vector pixel *);
11770 vector signed short vec_ld (int, const vector signed short *);
11771 vector signed short vec_ld (int, const short *);
11772 vector unsigned short vec_ld (int, const vector unsigned short *);
11773 vector unsigned short vec_ld (int, const unsigned short *);
11774 vector bool char vec_ld (int, const vector bool char *);
11775 vector signed char vec_ld (int, const vector signed char *);
11776 vector signed char vec_ld (int, const signed char *);
11777 vector unsigned char vec_ld (int, const vector unsigned char *);
11778 vector unsigned char vec_ld (int, const unsigned char *);
11780 vector signed char vec_lde (int, const signed char *);
11781 vector unsigned char vec_lde (int, const unsigned char *);
11782 vector signed short vec_lde (int, const short *);
11783 vector unsigned short vec_lde (int, const unsigned short *);
11784 vector float vec_lde (int, const float *);
11785 vector signed int vec_lde (int, const int *);
11786 vector unsigned int vec_lde (int, const unsigned int *);
11787 vector signed int vec_lde (int, const long *);
11788 vector unsigned int vec_lde (int, const unsigned long *);
11790 vector float vec_lvewx (int, float *);
11791 vector signed int vec_lvewx (int, int *);
11792 vector unsigned int vec_lvewx (int, unsigned int *);
11793 vector signed int vec_lvewx (int, long *);
11794 vector unsigned int vec_lvewx (int, unsigned long *);
11796 vector signed short vec_lvehx (int, short *);
11797 vector unsigned short vec_lvehx (int, unsigned short *);
11799 vector signed char vec_lvebx (int, char *);
11800 vector unsigned char vec_lvebx (int, unsigned char *);
11802 vector float vec_ldl (int, const vector float *);
11803 vector float vec_ldl (int, const float *);
11804 vector bool int vec_ldl (int, const vector bool int *);
11805 vector signed int vec_ldl (int, const vector signed int *);
11806 vector signed int vec_ldl (int, const int *);
11807 vector signed int vec_ldl (int, const long *);
11808 vector unsigned int vec_ldl (int, const vector unsigned int *);
11809 vector unsigned int vec_ldl (int, const unsigned int *);
11810 vector unsigned int vec_ldl (int, const unsigned long *);
11811 vector bool short vec_ldl (int, const vector bool short *);
11812 vector pixel vec_ldl (int, const vector pixel *);
11813 vector signed short vec_ldl (int, const vector signed short *);
11814 vector signed short vec_ldl (int, const short *);
11815 vector unsigned short vec_ldl (int, const vector unsigned short *);
11816 vector unsigned short vec_ldl (int, const unsigned short *);
11817 vector bool char vec_ldl (int, const vector bool char *);
11818 vector signed char vec_ldl (int, const vector signed char *);
11819 vector signed char vec_ldl (int, const signed char *);
11820 vector unsigned char vec_ldl (int, const vector unsigned char *);
11821 vector unsigned char vec_ldl (int, const unsigned char *);
11823 vector float vec_loge (vector float);
11825 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
11826 vector unsigned char vec_lvsl (int, const volatile signed char *);
11827 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
11828 vector unsigned char vec_lvsl (int, const volatile short *);
11829 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
11830 vector unsigned char vec_lvsl (int, const volatile int *);
11831 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
11832 vector unsigned char vec_lvsl (int, const volatile long *);
11833 vector unsigned char vec_lvsl (int, const volatile float *);
11835 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
11836 vector unsigned char vec_lvsr (int, const volatile signed char *);
11837 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
11838 vector unsigned char vec_lvsr (int, const volatile short *);
11839 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
11840 vector unsigned char vec_lvsr (int, const volatile int *);
11841 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
11842 vector unsigned char vec_lvsr (int, const volatile long *);
11843 vector unsigned char vec_lvsr (int, const volatile float *);
11845 vector float vec_madd (vector float, vector float, vector float);
11847 vector signed short vec_madds (vector signed short,
11848 vector signed short,
11849 vector signed short);
11851 vector unsigned char vec_max (vector bool char, vector unsigned char);
11852 vector unsigned char vec_max (vector unsigned char, vector bool char);
11853 vector unsigned char vec_max (vector unsigned char,
11854 vector unsigned char);
11855 vector signed char vec_max (vector bool char, vector signed char);
11856 vector signed char vec_max (vector signed char, vector bool char);
11857 vector signed char vec_max (vector signed char, vector signed char);
11858 vector unsigned short vec_max (vector bool short,
11859 vector unsigned short);
11860 vector unsigned short vec_max (vector unsigned short,
11861 vector bool short);
11862 vector unsigned short vec_max (vector unsigned short,
11863 vector unsigned short);
11864 vector signed short vec_max (vector bool short, vector signed short);
11865 vector signed short vec_max (vector signed short, vector bool short);
11866 vector signed short vec_max (vector signed short, vector signed short);
11867 vector unsigned int vec_max (vector bool int, vector unsigned int);
11868 vector unsigned int vec_max (vector unsigned int, vector bool int);
11869 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
11870 vector signed int vec_max (vector bool int, vector signed int);
11871 vector signed int vec_max (vector signed int, vector bool int);
11872 vector signed int vec_max (vector signed int, vector signed int);
11873 vector float vec_max (vector float, vector float);
11875 vector float vec_vmaxfp (vector float, vector float);
11877 vector signed int vec_vmaxsw (vector bool int, vector signed int);
11878 vector signed int vec_vmaxsw (vector signed int, vector bool int);
11879 vector signed int vec_vmaxsw (vector signed int, vector signed int);
11881 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
11882 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
11883 vector unsigned int vec_vmaxuw (vector unsigned int,
11884 vector unsigned int);
11886 vector signed short vec_vmaxsh (vector bool short, vector signed short);
11887 vector signed short vec_vmaxsh (vector signed short, vector bool short);
11888 vector signed short vec_vmaxsh (vector signed short,
11889 vector signed short);
11891 vector unsigned short vec_vmaxuh (vector bool short,
11892 vector unsigned short);
11893 vector unsigned short vec_vmaxuh (vector unsigned short,
11894 vector bool short);
11895 vector unsigned short vec_vmaxuh (vector unsigned short,
11896 vector unsigned short);
11898 vector signed char vec_vmaxsb (vector bool char, vector signed char);
11899 vector signed char vec_vmaxsb (vector signed char, vector bool char);
11900 vector signed char vec_vmaxsb (vector signed char, vector signed char);
11902 vector unsigned char vec_vmaxub (vector bool char,
11903 vector unsigned char);
11904 vector unsigned char vec_vmaxub (vector unsigned char,
11906 vector unsigned char vec_vmaxub (vector unsigned char,
11907 vector unsigned char);
11909 vector bool char vec_mergeh (vector bool char, vector bool char);
11910 vector signed char vec_mergeh (vector signed char, vector signed char);
11911 vector unsigned char vec_mergeh (vector unsigned char,
11912 vector unsigned char);
11913 vector bool short vec_mergeh (vector bool short, vector bool short);
11914 vector pixel vec_mergeh (vector pixel, vector pixel);
11915 vector signed short vec_mergeh (vector signed short,
11916 vector signed short);
11917 vector unsigned short vec_mergeh (vector unsigned short,
11918 vector unsigned short);
11919 vector float vec_mergeh (vector float, vector float);
11920 vector bool int vec_mergeh (vector bool int, vector bool int);
11921 vector signed int vec_mergeh (vector signed int, vector signed int);
11922 vector unsigned int vec_mergeh (vector unsigned int,
11923 vector unsigned int);
11925 vector float vec_vmrghw (vector float, vector float);
11926 vector bool int vec_vmrghw (vector bool int, vector bool int);
11927 vector signed int vec_vmrghw (vector signed int, vector signed int);
11928 vector unsigned int vec_vmrghw (vector unsigned int,
11929 vector unsigned int);
11931 vector bool short vec_vmrghh (vector bool short, vector bool short);
11932 vector signed short vec_vmrghh (vector signed short,
11933 vector signed short);
11934 vector unsigned short vec_vmrghh (vector unsigned short,
11935 vector unsigned short);
11936 vector pixel vec_vmrghh (vector pixel, vector pixel);
11938 vector bool char vec_vmrghb (vector bool char, vector bool char);
11939 vector signed char vec_vmrghb (vector signed char, vector signed char);
11940 vector unsigned char vec_vmrghb (vector unsigned char,
11941 vector unsigned char);
11943 vector bool char vec_mergel (vector bool char, vector bool char);
11944 vector signed char vec_mergel (vector signed char, vector signed char);
11945 vector unsigned char vec_mergel (vector unsigned char,
11946 vector unsigned char);
11947 vector bool short vec_mergel (vector bool short, vector bool short);
11948 vector pixel vec_mergel (vector pixel, vector pixel);
11949 vector signed short vec_mergel (vector signed short,
11950 vector signed short);
11951 vector unsigned short vec_mergel (vector unsigned short,
11952 vector unsigned short);
11953 vector float vec_mergel (vector float, vector float);
11954 vector bool int vec_mergel (vector bool int, vector bool int);
11955 vector signed int vec_mergel (vector signed int, vector signed int);
11956 vector unsigned int vec_mergel (vector unsigned int,
11957 vector unsigned int);
11959 vector float vec_vmrglw (vector float, vector float);
11960 vector signed int vec_vmrglw (vector signed int, vector signed int);
11961 vector unsigned int vec_vmrglw (vector unsigned int,
11962 vector unsigned int);
11963 vector bool int vec_vmrglw (vector bool int, vector bool int);
11965 vector bool short vec_vmrglh (vector bool short, vector bool short);
11966 vector signed short vec_vmrglh (vector signed short,
11967 vector signed short);
11968 vector unsigned short vec_vmrglh (vector unsigned short,
11969 vector unsigned short);
11970 vector pixel vec_vmrglh (vector pixel, vector pixel);
11972 vector bool char vec_vmrglb (vector bool char, vector bool char);
11973 vector signed char vec_vmrglb (vector signed char, vector signed char);
11974 vector unsigned char vec_vmrglb (vector unsigned char,
11975 vector unsigned char);
11977 vector unsigned short vec_mfvscr (void);
11979 vector unsigned char vec_min (vector bool char, vector unsigned char);
11980 vector unsigned char vec_min (vector unsigned char, vector bool char);
11981 vector unsigned char vec_min (vector unsigned char,
11982 vector unsigned char);
11983 vector signed char vec_min (vector bool char, vector signed char);
11984 vector signed char vec_min (vector signed char, vector bool char);
11985 vector signed char vec_min (vector signed char, vector signed char);
11986 vector unsigned short vec_min (vector bool short,
11987 vector unsigned short);
11988 vector unsigned short vec_min (vector unsigned short,
11989 vector bool short);
11990 vector unsigned short vec_min (vector unsigned short,
11991 vector unsigned short);
11992 vector signed short vec_min (vector bool short, vector signed short);
11993 vector signed short vec_min (vector signed short, vector bool short);
11994 vector signed short vec_min (vector signed short, vector signed short);
11995 vector unsigned int vec_min (vector bool int, vector unsigned int);
11996 vector unsigned int vec_min (vector unsigned int, vector bool int);
11997 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
11998 vector signed int vec_min (vector bool int, vector signed int);
11999 vector signed int vec_min (vector signed int, vector bool int);
12000 vector signed int vec_min (vector signed int, vector signed int);
12001 vector float vec_min (vector float, vector float);
12003 vector float vec_vminfp (vector float, vector float);
12005 vector signed int vec_vminsw (vector bool int, vector signed int);
12006 vector signed int vec_vminsw (vector signed int, vector bool int);
12007 vector signed int vec_vminsw (vector signed int, vector signed int);
12009 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
12010 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
12011 vector unsigned int vec_vminuw (vector unsigned int,
12012 vector unsigned int);
12014 vector signed short vec_vminsh (vector bool short, vector signed short);
12015 vector signed short vec_vminsh (vector signed short, vector bool short);
12016 vector signed short vec_vminsh (vector signed short,
12017 vector signed short);
12019 vector unsigned short vec_vminuh (vector bool short,
12020 vector unsigned short);
12021 vector unsigned short vec_vminuh (vector unsigned short,
12022 vector bool short);
12023 vector unsigned short vec_vminuh (vector unsigned short,
12024 vector unsigned short);
12026 vector signed char vec_vminsb (vector bool char, vector signed char);
12027 vector signed char vec_vminsb (vector signed char, vector bool char);
12028 vector signed char vec_vminsb (vector signed char, vector signed char);
12030 vector unsigned char vec_vminub (vector bool char,
12031 vector unsigned char);
12032 vector unsigned char vec_vminub (vector unsigned char,
12034 vector unsigned char vec_vminub (vector unsigned char,
12035 vector unsigned char);
12037 vector signed short vec_mladd (vector signed short,
12038 vector signed short,
12039 vector signed short);
12040 vector signed short vec_mladd (vector signed short,
12041 vector unsigned short,
12042 vector unsigned short);
12043 vector signed short vec_mladd (vector unsigned short,
12044 vector signed short,
12045 vector signed short);
12046 vector unsigned short vec_mladd (vector unsigned short,
12047 vector unsigned short,
12048 vector unsigned short);
12050 vector signed short vec_mradds (vector signed short,
12051 vector signed short,
12052 vector signed short);
12054 vector unsigned int vec_msum (vector unsigned char,
12055 vector unsigned char,
12056 vector unsigned int);
12057 vector signed int vec_msum (vector signed char,
12058 vector unsigned char,
12059 vector signed int);
12060 vector unsigned int vec_msum (vector unsigned short,
12061 vector unsigned short,
12062 vector unsigned int);
12063 vector signed int vec_msum (vector signed short,
12064 vector signed short,
12065 vector signed int);
12067 vector signed int vec_vmsumshm (vector signed short,
12068 vector signed short,
12069 vector signed int);
12071 vector unsigned int vec_vmsumuhm (vector unsigned short,
12072 vector unsigned short,
12073 vector unsigned int);
12075 vector signed int vec_vmsummbm (vector signed char,
12076 vector unsigned char,
12077 vector signed int);
12079 vector unsigned int vec_vmsumubm (vector unsigned char,
12080 vector unsigned char,
12081 vector unsigned int);
12083 vector unsigned int vec_msums (vector unsigned short,
12084 vector unsigned short,
12085 vector unsigned int);
12086 vector signed int vec_msums (vector signed short,
12087 vector signed short,
12088 vector signed int);
12090 vector signed int vec_vmsumshs (vector signed short,
12091 vector signed short,
12092 vector signed int);
12094 vector unsigned int vec_vmsumuhs (vector unsigned short,
12095 vector unsigned short,
12096 vector unsigned int);
12098 void vec_mtvscr (vector signed int);
12099 void vec_mtvscr (vector unsigned int);
12100 void vec_mtvscr (vector bool int);
12101 void vec_mtvscr (vector signed short);
12102 void vec_mtvscr (vector unsigned short);
12103 void vec_mtvscr (vector bool short);
12104 void vec_mtvscr (vector pixel);
12105 void vec_mtvscr (vector signed char);
12106 void vec_mtvscr (vector unsigned char);
12107 void vec_mtvscr (vector bool char);
12109 vector unsigned short vec_mule (vector unsigned char,
12110 vector unsigned char);
12111 vector signed short vec_mule (vector signed char,
12112 vector signed char);
12113 vector unsigned int vec_mule (vector unsigned short,
12114 vector unsigned short);
12115 vector signed int vec_mule (vector signed short, vector signed short);
12117 vector signed int vec_vmulesh (vector signed short,
12118 vector signed short);
12120 vector unsigned int vec_vmuleuh (vector unsigned short,
12121 vector unsigned short);
12123 vector signed short vec_vmulesb (vector signed char,
12124 vector signed char);
12126 vector unsigned short vec_vmuleub (vector unsigned char,
12127 vector unsigned char);
12129 vector unsigned short vec_mulo (vector unsigned char,
12130 vector unsigned char);
12131 vector signed short vec_mulo (vector signed char, vector signed char);
12132 vector unsigned int vec_mulo (vector unsigned short,
12133 vector unsigned short);
12134 vector signed int vec_mulo (vector signed short, vector signed short);
12136 vector signed int vec_vmulosh (vector signed short,
12137 vector signed short);
12139 vector unsigned int vec_vmulouh (vector unsigned short,
12140 vector unsigned short);
12142 vector signed short vec_vmulosb (vector signed char,
12143 vector signed char);
12145 vector unsigned short vec_vmuloub (vector unsigned char,
12146 vector unsigned char);
12148 vector float vec_nmsub (vector float, vector float, vector float);
12150 vector float vec_nor (vector float, vector float);
12151 vector signed int vec_nor (vector signed int, vector signed int);
12152 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
12153 vector bool int vec_nor (vector bool int, vector bool int);
12154 vector signed short vec_nor (vector signed short, vector signed short);
12155 vector unsigned short vec_nor (vector unsigned short,
12156 vector unsigned short);
12157 vector bool short vec_nor (vector bool short, vector bool short);
12158 vector signed char vec_nor (vector signed char, vector signed char);
12159 vector unsigned char vec_nor (vector unsigned char,
12160 vector unsigned char);
12161 vector bool char vec_nor (vector bool char, vector bool char);
12163 vector float vec_or (vector float, vector float);
12164 vector float vec_or (vector float, vector bool int);
12165 vector float vec_or (vector bool int, vector float);
12166 vector bool int vec_or (vector bool int, vector bool int);
12167 vector signed int vec_or (vector bool int, vector signed int);
12168 vector signed int vec_or (vector signed int, vector bool int);
12169 vector signed int vec_or (vector signed int, vector signed int);
12170 vector unsigned int vec_or (vector bool int, vector unsigned int);
12171 vector unsigned int vec_or (vector unsigned int, vector bool int);
12172 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
12173 vector bool short vec_or (vector bool short, vector bool short);
12174 vector signed short vec_or (vector bool short, vector signed short);
12175 vector signed short vec_or (vector signed short, vector bool short);
12176 vector signed short vec_or (vector signed short, vector signed short);
12177 vector unsigned short vec_or (vector bool short, vector unsigned short);
12178 vector unsigned short vec_or (vector unsigned short, vector bool short);
12179 vector unsigned short vec_or (vector unsigned short,
12180 vector unsigned short);
12181 vector signed char vec_or (vector bool char, vector signed char);
12182 vector bool char vec_or (vector bool char, vector bool char);
12183 vector signed char vec_or (vector signed char, vector bool char);
12184 vector signed char vec_or (vector signed char, vector signed char);
12185 vector unsigned char vec_or (vector bool char, vector unsigned char);
12186 vector unsigned char vec_or (vector unsigned char, vector bool char);
12187 vector unsigned char vec_or (vector unsigned char,
12188 vector unsigned char);
12190 vector signed char vec_pack (vector signed short, vector signed short);
12191 vector unsigned char vec_pack (vector unsigned short,
12192 vector unsigned short);
12193 vector bool char vec_pack (vector bool short, vector bool short);
12194 vector signed short vec_pack (vector signed int, vector signed int);
12195 vector unsigned short vec_pack (vector unsigned int,
12196 vector unsigned int);
12197 vector bool short vec_pack (vector bool int, vector bool int);
12199 vector bool short vec_vpkuwum (vector bool int, vector bool int);
12200 vector signed short vec_vpkuwum (vector signed int, vector signed int);
12201 vector unsigned short vec_vpkuwum (vector unsigned int,
12202 vector unsigned int);
12204 vector bool char vec_vpkuhum (vector bool short, vector bool short);
12205 vector signed char vec_vpkuhum (vector signed short,
12206 vector signed short);
12207 vector unsigned char vec_vpkuhum (vector unsigned short,
12208 vector unsigned short);
12210 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
12212 vector unsigned char vec_packs (vector unsigned short,
12213 vector unsigned short);
12214 vector signed char vec_packs (vector signed short, vector signed short);
12215 vector unsigned short vec_packs (vector unsigned int,
12216 vector unsigned int);
12217 vector signed short vec_packs (vector signed int, vector signed int);
12219 vector signed short vec_vpkswss (vector signed int, vector signed int);
12221 vector unsigned short vec_vpkuwus (vector unsigned int,
12222 vector unsigned int);
12224 vector signed char vec_vpkshss (vector signed short,
12225 vector signed short);
12227 vector unsigned char vec_vpkuhus (vector unsigned short,
12228 vector unsigned short);
12230 vector unsigned char vec_packsu (vector unsigned short,
12231 vector unsigned short);
12232 vector unsigned char vec_packsu (vector signed short,
12233 vector signed short);
12234 vector unsigned short vec_packsu (vector unsigned int,
12235 vector unsigned int);
12236 vector unsigned short vec_packsu (vector signed int, vector signed int);
12238 vector unsigned short vec_vpkswus (vector signed int,
12239 vector signed int);
12241 vector unsigned char vec_vpkshus (vector signed short,
12242 vector signed short);
12244 vector float vec_perm (vector float,
12246 vector unsigned char);
12247 vector signed int vec_perm (vector signed int,
12249 vector unsigned char);
12250 vector unsigned int vec_perm (vector unsigned int,
12251 vector unsigned int,
12252 vector unsigned char);
12253 vector bool int vec_perm (vector bool int,
12255 vector unsigned char);
12256 vector signed short vec_perm (vector signed short,
12257 vector signed short,
12258 vector unsigned char);
12259 vector unsigned short vec_perm (vector unsigned short,
12260 vector unsigned short,
12261 vector unsigned char);
12262 vector bool short vec_perm (vector bool short,
12264 vector unsigned char);
12265 vector pixel vec_perm (vector pixel,
12267 vector unsigned char);
12268 vector signed char vec_perm (vector signed char,
12269 vector signed char,
12270 vector unsigned char);
12271 vector unsigned char vec_perm (vector unsigned char,
12272 vector unsigned char,
12273 vector unsigned char);
12274 vector bool char vec_perm (vector bool char,
12276 vector unsigned char);
12278 vector float vec_re (vector float);
12280 vector signed char vec_rl (vector signed char,
12281 vector unsigned char);
12282 vector unsigned char vec_rl (vector unsigned char,
12283 vector unsigned char);
12284 vector signed short vec_rl (vector signed short, vector unsigned short);
12285 vector unsigned short vec_rl (vector unsigned short,
12286 vector unsigned short);
12287 vector signed int vec_rl (vector signed int, vector unsigned int);
12288 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
12290 vector signed int vec_vrlw (vector signed int, vector unsigned int);
12291 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
12293 vector signed short vec_vrlh (vector signed short,
12294 vector unsigned short);
12295 vector unsigned short vec_vrlh (vector unsigned short,
12296 vector unsigned short);
12298 vector signed char vec_vrlb (vector signed char, vector unsigned char);
12299 vector unsigned char vec_vrlb (vector unsigned char,
12300 vector unsigned char);
12302 vector float vec_round (vector float);
12304 vector float vec_recip (vector float, vector float);
12306 vector float vec_rsqrt (vector float);
12308 vector float vec_rsqrte (vector float);
12310 vector float vec_sel (vector float, vector float, vector bool int);
12311 vector float vec_sel (vector float, vector float, vector unsigned int);
12312 vector signed int vec_sel (vector signed int,
12315 vector signed int vec_sel (vector signed int,
12317 vector unsigned int);
12318 vector unsigned int vec_sel (vector unsigned int,
12319 vector unsigned int,
12321 vector unsigned int vec_sel (vector unsigned int,
12322 vector unsigned int,
12323 vector unsigned int);
12324 vector bool int vec_sel (vector bool int,
12327 vector bool int vec_sel (vector bool int,
12329 vector unsigned int);
12330 vector signed short vec_sel (vector signed short,
12331 vector signed short,
12332 vector bool short);
12333 vector signed short vec_sel (vector signed short,
12334 vector signed short,
12335 vector unsigned short);
12336 vector unsigned short vec_sel (vector unsigned short,
12337 vector unsigned short,
12338 vector bool short);
12339 vector unsigned short vec_sel (vector unsigned short,
12340 vector unsigned short,
12341 vector unsigned short);
12342 vector bool short vec_sel (vector bool short,
12344 vector bool short);
12345 vector bool short vec_sel (vector bool short,
12347 vector unsigned short);
12348 vector signed char vec_sel (vector signed char,
12349 vector signed char,
12351 vector signed char vec_sel (vector signed char,
12352 vector signed char,
12353 vector unsigned char);
12354 vector unsigned char vec_sel (vector unsigned char,
12355 vector unsigned char,
12357 vector unsigned char vec_sel (vector unsigned char,
12358 vector unsigned char,
12359 vector unsigned char);
12360 vector bool char vec_sel (vector bool char,
12363 vector bool char vec_sel (vector bool char,
12365 vector unsigned char);
12367 vector signed char vec_sl (vector signed char,
12368 vector unsigned char);
12369 vector unsigned char vec_sl (vector unsigned char,
12370 vector unsigned char);
12371 vector signed short vec_sl (vector signed short, vector unsigned short);
12372 vector unsigned short vec_sl (vector unsigned short,
12373 vector unsigned short);
12374 vector signed int vec_sl (vector signed int, vector unsigned int);
12375 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
12377 vector signed int vec_vslw (vector signed int, vector unsigned int);
12378 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
12380 vector signed short vec_vslh (vector signed short,
12381 vector unsigned short);
12382 vector unsigned short vec_vslh (vector unsigned short,
12383 vector unsigned short);
12385 vector signed char vec_vslb (vector signed char, vector unsigned char);
12386 vector unsigned char vec_vslb (vector unsigned char,
12387 vector unsigned char);
12389 vector float vec_sld (vector float, vector float, const int);
12390 vector signed int vec_sld (vector signed int,
12393 vector unsigned int vec_sld (vector unsigned int,
12394 vector unsigned int,
12396 vector bool int vec_sld (vector bool int,
12399 vector signed short vec_sld (vector signed short,
12400 vector signed short,
12402 vector unsigned short vec_sld (vector unsigned short,
12403 vector unsigned short,
12405 vector bool short vec_sld (vector bool short,
12408 vector pixel vec_sld (vector pixel,
12411 vector signed char vec_sld (vector signed char,
12412 vector signed char,
12414 vector unsigned char vec_sld (vector unsigned char,
12415 vector unsigned char,
12417 vector bool char vec_sld (vector bool char,
12421 vector signed int vec_sll (vector signed int,
12422 vector unsigned int);
12423 vector signed int vec_sll (vector signed int,
12424 vector unsigned short);
12425 vector signed int vec_sll (vector signed int,
12426 vector unsigned char);
12427 vector unsigned int vec_sll (vector unsigned int,
12428 vector unsigned int);
12429 vector unsigned int vec_sll (vector unsigned int,
12430 vector unsigned short);
12431 vector unsigned int vec_sll (vector unsigned int,
12432 vector unsigned char);
12433 vector bool int vec_sll (vector bool int,
12434 vector unsigned int);
12435 vector bool int vec_sll (vector bool int,
12436 vector unsigned short);
12437 vector bool int vec_sll (vector bool int,
12438 vector unsigned char);
12439 vector signed short vec_sll (vector signed short,
12440 vector unsigned int);
12441 vector signed short vec_sll (vector signed short,
12442 vector unsigned short);
12443 vector signed short vec_sll (vector signed short,
12444 vector unsigned char);
12445 vector unsigned short vec_sll (vector unsigned short,
12446 vector unsigned int);
12447 vector unsigned short vec_sll (vector unsigned short,
12448 vector unsigned short);
12449 vector unsigned short vec_sll (vector unsigned short,
12450 vector unsigned char);
12451 vector bool short vec_sll (vector bool short, vector unsigned int);
12452 vector bool short vec_sll (vector bool short, vector unsigned short);
12453 vector bool short vec_sll (vector bool short, vector unsigned char);
12454 vector pixel vec_sll (vector pixel, vector unsigned int);
12455 vector pixel vec_sll (vector pixel, vector unsigned short);
12456 vector pixel vec_sll (vector pixel, vector unsigned char);
12457 vector signed char vec_sll (vector signed char, vector unsigned int);
12458 vector signed char vec_sll (vector signed char, vector unsigned short);
12459 vector signed char vec_sll (vector signed char, vector unsigned char);
12460 vector unsigned char vec_sll (vector unsigned char,
12461 vector unsigned int);
12462 vector unsigned char vec_sll (vector unsigned char,
12463 vector unsigned short);
12464 vector unsigned char vec_sll (vector unsigned char,
12465 vector unsigned char);
12466 vector bool char vec_sll (vector bool char, vector unsigned int);
12467 vector bool char vec_sll (vector bool char, vector unsigned short);
12468 vector bool char vec_sll (vector bool char, vector unsigned char);
12470 vector float vec_slo (vector float, vector signed char);
12471 vector float vec_slo (vector float, vector unsigned char);
12472 vector signed int vec_slo (vector signed int, vector signed char);
12473 vector signed int vec_slo (vector signed int, vector unsigned char);
12474 vector unsigned int vec_slo (vector unsigned int, vector signed char);
12475 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
12476 vector signed short vec_slo (vector signed short, vector signed char);
12477 vector signed short vec_slo (vector signed short, vector unsigned char);
12478 vector unsigned short vec_slo (vector unsigned short,
12479 vector signed char);
12480 vector unsigned short vec_slo (vector unsigned short,
12481 vector unsigned char);
12482 vector pixel vec_slo (vector pixel, vector signed char);
12483 vector pixel vec_slo (vector pixel, vector unsigned char);
12484 vector signed char vec_slo (vector signed char, vector signed char);
12485 vector signed char vec_slo (vector signed char, vector unsigned char);
12486 vector unsigned char vec_slo (vector unsigned char, vector signed char);
12487 vector unsigned char vec_slo (vector unsigned char,
12488 vector unsigned char);
12490 vector signed char vec_splat (vector signed char, const int);
12491 vector unsigned char vec_splat (vector unsigned char, const int);
12492 vector bool char vec_splat (vector bool char, const int);
12493 vector signed short vec_splat (vector signed short, const int);
12494 vector unsigned short vec_splat (vector unsigned short, const int);
12495 vector bool short vec_splat (vector bool short, const int);
12496 vector pixel vec_splat (vector pixel, const int);
12497 vector float vec_splat (vector float, const int);
12498 vector signed int vec_splat (vector signed int, const int);
12499 vector unsigned int vec_splat (vector unsigned int, const int);
12500 vector bool int vec_splat (vector bool int, const int);
12502 vector float vec_vspltw (vector float, const int);
12503 vector signed int vec_vspltw (vector signed int, const int);
12504 vector unsigned int vec_vspltw (vector unsigned int, const int);
12505 vector bool int vec_vspltw (vector bool int, const int);
12507 vector bool short vec_vsplth (vector bool short, const int);
12508 vector signed short vec_vsplth (vector signed short, const int);
12509 vector unsigned short vec_vsplth (vector unsigned short, const int);
12510 vector pixel vec_vsplth (vector pixel, const int);
12512 vector signed char vec_vspltb (vector signed char, const int);
12513 vector unsigned char vec_vspltb (vector unsigned char, const int);
12514 vector bool char vec_vspltb (vector bool char, const int);
12516 vector signed char vec_splat_s8 (const int);
12518 vector signed short vec_splat_s16 (const int);
12520 vector signed int vec_splat_s32 (const int);
12522 vector unsigned char vec_splat_u8 (const int);
12524 vector unsigned short vec_splat_u16 (const int);
12526 vector unsigned int vec_splat_u32 (const int);
12528 vector signed char vec_sr (vector signed char, vector unsigned char);
12529 vector unsigned char vec_sr (vector unsigned char,
12530 vector unsigned char);
12531 vector signed short vec_sr (vector signed short,
12532 vector unsigned short);
12533 vector unsigned short vec_sr (vector unsigned short,
12534 vector unsigned short);
12535 vector signed int vec_sr (vector signed int, vector unsigned int);
12536 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
12538 vector signed int vec_vsrw (vector signed int, vector unsigned int);
12539 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
12541 vector signed short vec_vsrh (vector signed short,
12542 vector unsigned short);
12543 vector unsigned short vec_vsrh (vector unsigned short,
12544 vector unsigned short);
12546 vector signed char vec_vsrb (vector signed char, vector unsigned char);
12547 vector unsigned char vec_vsrb (vector unsigned char,
12548 vector unsigned char);
12550 vector signed char vec_sra (vector signed char, vector unsigned char);
12551 vector unsigned char vec_sra (vector unsigned char,
12552 vector unsigned char);
12553 vector signed short vec_sra (vector signed short,
12554 vector unsigned short);
12555 vector unsigned short vec_sra (vector unsigned short,
12556 vector unsigned short);
12557 vector signed int vec_sra (vector signed int, vector unsigned int);
12558 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
12560 vector signed int vec_vsraw (vector signed int, vector unsigned int);
12561 vector unsigned int vec_vsraw (vector unsigned int,
12562 vector unsigned int);
12564 vector signed short vec_vsrah (vector signed short,
12565 vector unsigned short);
12566 vector unsigned short vec_vsrah (vector unsigned short,
12567 vector unsigned short);
12569 vector signed char vec_vsrab (vector signed char, vector unsigned char);
12570 vector unsigned char vec_vsrab (vector unsigned char,
12571 vector unsigned char);
12573 vector signed int vec_srl (vector signed int, vector unsigned int);
12574 vector signed int vec_srl (vector signed int, vector unsigned short);
12575 vector signed int vec_srl (vector signed int, vector unsigned char);
12576 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
12577 vector unsigned int vec_srl (vector unsigned int,
12578 vector unsigned short);
12579 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
12580 vector bool int vec_srl (vector bool int, vector unsigned int);
12581 vector bool int vec_srl (vector bool int, vector unsigned short);
12582 vector bool int vec_srl (vector bool int, vector unsigned char);
12583 vector signed short vec_srl (vector signed short, vector unsigned int);
12584 vector signed short vec_srl (vector signed short,
12585 vector unsigned short);
12586 vector signed short vec_srl (vector signed short, vector unsigned char);
12587 vector unsigned short vec_srl (vector unsigned short,
12588 vector unsigned int);
12589 vector unsigned short vec_srl (vector unsigned short,
12590 vector unsigned short);
12591 vector unsigned short vec_srl (vector unsigned short,
12592 vector unsigned char);
12593 vector bool short vec_srl (vector bool short, vector unsigned int);
12594 vector bool short vec_srl (vector bool short, vector unsigned short);
12595 vector bool short vec_srl (vector bool short, vector unsigned char);
12596 vector pixel vec_srl (vector pixel, vector unsigned int);
12597 vector pixel vec_srl (vector pixel, vector unsigned short);
12598 vector pixel vec_srl (vector pixel, vector unsigned char);
12599 vector signed char vec_srl (vector signed char, vector unsigned int);
12600 vector signed char vec_srl (vector signed char, vector unsigned short);
12601 vector signed char vec_srl (vector signed char, vector unsigned char);
12602 vector unsigned char vec_srl (vector unsigned char,
12603 vector unsigned int);
12604 vector unsigned char vec_srl (vector unsigned char,
12605 vector unsigned short);
12606 vector unsigned char vec_srl (vector unsigned char,
12607 vector unsigned char);
12608 vector bool char vec_srl (vector bool char, vector unsigned int);
12609 vector bool char vec_srl (vector bool char, vector unsigned short);
12610 vector bool char vec_srl (vector bool char, vector unsigned char);
12612 vector float vec_sro (vector float, vector signed char);
12613 vector float vec_sro (vector float, vector unsigned char);
12614 vector signed int vec_sro (vector signed int, vector signed char);
12615 vector signed int vec_sro (vector signed int, vector unsigned char);
12616 vector unsigned int vec_sro (vector unsigned int, vector signed char);
12617 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
12618 vector signed short vec_sro (vector signed short, vector signed char);
12619 vector signed short vec_sro (vector signed short, vector unsigned char);
12620 vector unsigned short vec_sro (vector unsigned short,
12621 vector signed char);
12622 vector unsigned short vec_sro (vector unsigned short,
12623 vector unsigned char);
12624 vector pixel vec_sro (vector pixel, vector signed char);
12625 vector pixel vec_sro (vector pixel, vector unsigned char);
12626 vector signed char vec_sro (vector signed char, vector signed char);
12627 vector signed char vec_sro (vector signed char, vector unsigned char);
12628 vector unsigned char vec_sro (vector unsigned char, vector signed char);
12629 vector unsigned char vec_sro (vector unsigned char,
12630 vector unsigned char);
12632 void vec_st (vector float, int, vector float *);
12633 void vec_st (vector float, int, float *);
12634 void vec_st (vector signed int, int, vector signed int *);
12635 void vec_st (vector signed int, int, int *);
12636 void vec_st (vector unsigned int, int, vector unsigned int *);
12637 void vec_st (vector unsigned int, int, unsigned int *);
12638 void vec_st (vector bool int, int, vector bool int *);
12639 void vec_st (vector bool int, int, unsigned int *);
12640 void vec_st (vector bool int, int, int *);
12641 void vec_st (vector signed short, int, vector signed short *);
12642 void vec_st (vector signed short, int, short *);
12643 void vec_st (vector unsigned short, int, vector unsigned short *);
12644 void vec_st (vector unsigned short, int, unsigned short *);
12645 void vec_st (vector bool short, int, vector bool short *);
12646 void vec_st (vector bool short, int, unsigned short *);
12647 void vec_st (vector pixel, int, vector pixel *);
12648 void vec_st (vector pixel, int, unsigned short *);
12649 void vec_st (vector pixel, int, short *);
12650 void vec_st (vector bool short, int, short *);
12651 void vec_st (vector signed char, int, vector signed char *);
12652 void vec_st (vector signed char, int, signed char *);
12653 void vec_st (vector unsigned char, int, vector unsigned char *);
12654 void vec_st (vector unsigned char, int, unsigned char *);
12655 void vec_st (vector bool char, int, vector bool char *);
12656 void vec_st (vector bool char, int, unsigned char *);
12657 void vec_st (vector bool char, int, signed char *);
12659 void vec_ste (vector signed char, int, signed char *);
12660 void vec_ste (vector unsigned char, int, unsigned char *);
12661 void vec_ste (vector bool char, int, signed char *);
12662 void vec_ste (vector bool char, int, unsigned char *);
12663 void vec_ste (vector signed short, int, short *);
12664 void vec_ste (vector unsigned short, int, unsigned short *);
12665 void vec_ste (vector bool short, int, short *);
12666 void vec_ste (vector bool short, int, unsigned short *);
12667 void vec_ste (vector pixel, int, short *);
12668 void vec_ste (vector pixel, int, unsigned short *);
12669 void vec_ste (vector float, int, float *);
12670 void vec_ste (vector signed int, int, int *);
12671 void vec_ste (vector unsigned int, int, unsigned int *);
12672 void vec_ste (vector bool int, int, int *);
12673 void vec_ste (vector bool int, int, unsigned int *);
12675 void vec_stvewx (vector float, int, float *);
12676 void vec_stvewx (vector signed int, int, int *);
12677 void vec_stvewx (vector unsigned int, int, unsigned int *);
12678 void vec_stvewx (vector bool int, int, int *);
12679 void vec_stvewx (vector bool int, int, unsigned int *);
12681 void vec_stvehx (vector signed short, int, short *);
12682 void vec_stvehx (vector unsigned short, int, unsigned short *);
12683 void vec_stvehx (vector bool short, int, short *);
12684 void vec_stvehx (vector bool short, int, unsigned short *);
12685 void vec_stvehx (vector pixel, int, short *);
12686 void vec_stvehx (vector pixel, int, unsigned short *);
12688 void vec_stvebx (vector signed char, int, signed char *);
12689 void vec_stvebx (vector unsigned char, int, unsigned char *);
12690 void vec_stvebx (vector bool char, int, signed char *);
12691 void vec_stvebx (vector bool char, int, unsigned char *);
12693 void vec_stl (vector float, int, vector float *);
12694 void vec_stl (vector float, int, float *);
12695 void vec_stl (vector signed int, int, vector signed int *);
12696 void vec_stl (vector signed int, int, int *);
12697 void vec_stl (vector unsigned int, int, vector unsigned int *);
12698 void vec_stl (vector unsigned int, int, unsigned int *);
12699 void vec_stl (vector bool int, int, vector bool int *);
12700 void vec_stl (vector bool int, int, unsigned int *);
12701 void vec_stl (vector bool int, int, int *);
12702 void vec_stl (vector signed short, int, vector signed short *);
12703 void vec_stl (vector signed short, int, short *);
12704 void vec_stl (vector unsigned short, int, vector unsigned short *);
12705 void vec_stl (vector unsigned short, int, unsigned short *);
12706 void vec_stl (vector bool short, int, vector bool short *);
12707 void vec_stl (vector bool short, int, unsigned short *);
12708 void vec_stl (vector bool short, int, short *);
12709 void vec_stl (vector pixel, int, vector pixel *);
12710 void vec_stl (vector pixel, int, unsigned short *);
12711 void vec_stl (vector pixel, int, short *);
12712 void vec_stl (vector signed char, int, vector signed char *);
12713 void vec_stl (vector signed char, int, signed char *);
12714 void vec_stl (vector unsigned char, int, vector unsigned char *);
12715 void vec_stl (vector unsigned char, int, unsigned char *);
12716 void vec_stl (vector bool char, int, vector bool char *);
12717 void vec_stl (vector bool char, int, unsigned char *);
12718 void vec_stl (vector bool char, int, signed char *);
12720 vector signed char vec_sub (vector bool char, vector signed char);
12721 vector signed char vec_sub (vector signed char, vector bool char);
12722 vector signed char vec_sub (vector signed char, vector signed char);
12723 vector unsigned char vec_sub (vector bool char, vector unsigned char);
12724 vector unsigned char vec_sub (vector unsigned char, vector bool char);
12725 vector unsigned char vec_sub (vector unsigned char,
12726 vector unsigned char);
12727 vector signed short vec_sub (vector bool short, vector signed short);
12728 vector signed short vec_sub (vector signed short, vector bool short);
12729 vector signed short vec_sub (vector signed short, vector signed short);
12730 vector unsigned short vec_sub (vector bool short,
12731 vector unsigned short);
12732 vector unsigned short vec_sub (vector unsigned short,
12733 vector bool short);
12734 vector unsigned short vec_sub (vector unsigned short,
12735 vector unsigned short);
12736 vector signed int vec_sub (vector bool int, vector signed int);
12737 vector signed int vec_sub (vector signed int, vector bool int);
12738 vector signed int vec_sub (vector signed int, vector signed int);
12739 vector unsigned int vec_sub (vector bool int, vector unsigned int);
12740 vector unsigned int vec_sub (vector unsigned int, vector bool int);
12741 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
12742 vector float vec_sub (vector float, vector float);
12744 vector float vec_vsubfp (vector float, vector float);
12746 vector signed int vec_vsubuwm (vector bool int, vector signed int);
12747 vector signed int vec_vsubuwm (vector signed int, vector bool int);
12748 vector signed int vec_vsubuwm (vector signed int, vector signed int);
12749 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
12750 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
12751 vector unsigned int vec_vsubuwm (vector unsigned int,
12752 vector unsigned int);
12754 vector signed short vec_vsubuhm (vector bool short,
12755 vector signed short);
12756 vector signed short vec_vsubuhm (vector signed short,
12757 vector bool short);
12758 vector signed short vec_vsubuhm (vector signed short,
12759 vector signed short);
12760 vector unsigned short vec_vsubuhm (vector bool short,
12761 vector unsigned short);
12762 vector unsigned short vec_vsubuhm (vector unsigned short,
12763 vector bool short);
12764 vector unsigned short vec_vsubuhm (vector unsigned short,
12765 vector unsigned short);
12767 vector signed char vec_vsububm (vector bool char, vector signed char);
12768 vector signed char vec_vsububm (vector signed char, vector bool char);
12769 vector signed char vec_vsububm (vector signed char, vector signed char);
12770 vector unsigned char vec_vsububm (vector bool char,
12771 vector unsigned char);
12772 vector unsigned char vec_vsububm (vector unsigned char,
12774 vector unsigned char vec_vsububm (vector unsigned char,
12775 vector unsigned char);
12777 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
12779 vector unsigned char vec_subs (vector bool char, vector unsigned char);
12780 vector unsigned char vec_subs (vector unsigned char, vector bool char);
12781 vector unsigned char vec_subs (vector unsigned char,
12782 vector unsigned char);
12783 vector signed char vec_subs (vector bool char, vector signed char);
12784 vector signed char vec_subs (vector signed char, vector bool char);
12785 vector signed char vec_subs (vector signed char, vector signed char);
12786 vector unsigned short vec_subs (vector bool short,
12787 vector unsigned short);
12788 vector unsigned short vec_subs (vector unsigned short,
12789 vector bool short);
12790 vector unsigned short vec_subs (vector unsigned short,
12791 vector unsigned short);
12792 vector signed short vec_subs (vector bool short, vector signed short);
12793 vector signed short vec_subs (vector signed short, vector bool short);
12794 vector signed short vec_subs (vector signed short, vector signed short);
12795 vector unsigned int vec_subs (vector bool int, vector unsigned int);
12796 vector unsigned int vec_subs (vector unsigned int, vector bool int);
12797 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
12798 vector signed int vec_subs (vector bool int, vector signed int);
12799 vector signed int vec_subs (vector signed int, vector bool int);
12800 vector signed int vec_subs (vector signed int, vector signed int);
12802 vector signed int vec_vsubsws (vector bool int, vector signed int);
12803 vector signed int vec_vsubsws (vector signed int, vector bool int);
12804 vector signed int vec_vsubsws (vector signed int, vector signed int);
12806 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
12807 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
12808 vector unsigned int vec_vsubuws (vector unsigned int,
12809 vector unsigned int);
12811 vector signed short vec_vsubshs (vector bool short,
12812 vector signed short);
12813 vector signed short vec_vsubshs (vector signed short,
12814 vector bool short);
12815 vector signed short vec_vsubshs (vector signed short,
12816 vector signed short);
12818 vector unsigned short vec_vsubuhs (vector bool short,
12819 vector unsigned short);
12820 vector unsigned short vec_vsubuhs (vector unsigned short,
12821 vector bool short);
12822 vector unsigned short vec_vsubuhs (vector unsigned short,
12823 vector unsigned short);
12825 vector signed char vec_vsubsbs (vector bool char, vector signed char);
12826 vector signed char vec_vsubsbs (vector signed char, vector bool char);
12827 vector signed char vec_vsubsbs (vector signed char, vector signed char);
12829 vector unsigned char vec_vsububs (vector bool char,
12830 vector unsigned char);
12831 vector unsigned char vec_vsububs (vector unsigned char,
12833 vector unsigned char vec_vsububs (vector unsigned char,
12834 vector unsigned char);
12836 vector unsigned int vec_sum4s (vector unsigned char,
12837 vector unsigned int);
12838 vector signed int vec_sum4s (vector signed char, vector signed int);
12839 vector signed int vec_sum4s (vector signed short, vector signed int);
12841 vector signed int vec_vsum4shs (vector signed short, vector signed int);
12843 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
12845 vector unsigned int vec_vsum4ubs (vector unsigned char,
12846 vector unsigned int);
12848 vector signed int vec_sum2s (vector signed int, vector signed int);
12850 vector signed int vec_sums (vector signed int, vector signed int);
12852 vector float vec_trunc (vector float);
12854 vector signed short vec_unpackh (vector signed char);
12855 vector bool short vec_unpackh (vector bool char);
12856 vector signed int vec_unpackh (vector signed short);
12857 vector bool int vec_unpackh (vector bool short);
12858 vector unsigned int vec_unpackh (vector pixel);
12860 vector bool int vec_vupkhsh (vector bool short);
12861 vector signed int vec_vupkhsh (vector signed short);
12863 vector unsigned int vec_vupkhpx (vector pixel);
12865 vector bool short vec_vupkhsb (vector bool char);
12866 vector signed short vec_vupkhsb (vector signed char);
12868 vector signed short vec_unpackl (vector signed char);
12869 vector bool short vec_unpackl (vector bool char);
12870 vector unsigned int vec_unpackl (vector pixel);
12871 vector signed int vec_unpackl (vector signed short);
12872 vector bool int vec_unpackl (vector bool short);
12874 vector unsigned int vec_vupklpx (vector pixel);
12876 vector bool int vec_vupklsh (vector bool short);
12877 vector signed int vec_vupklsh (vector signed short);
12879 vector bool short vec_vupklsb (vector bool char);
12880 vector signed short vec_vupklsb (vector signed char);
12882 vector float vec_xor (vector float, vector float);
12883 vector float vec_xor (vector float, vector bool int);
12884 vector float vec_xor (vector bool int, vector float);
12885 vector bool int vec_xor (vector bool int, vector bool int);
12886 vector signed int vec_xor (vector bool int, vector signed int);
12887 vector signed int vec_xor (vector signed int, vector bool int);
12888 vector signed int vec_xor (vector signed int, vector signed int);
12889 vector unsigned int vec_xor (vector bool int, vector unsigned int);
12890 vector unsigned int vec_xor (vector unsigned int, vector bool int);
12891 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
12892 vector bool short vec_xor (vector bool short, vector bool short);
12893 vector signed short vec_xor (vector bool short, vector signed short);
12894 vector signed short vec_xor (vector signed short, vector bool short);
12895 vector signed short vec_xor (vector signed short, vector signed short);
12896 vector unsigned short vec_xor (vector bool short,
12897 vector unsigned short);
12898 vector unsigned short vec_xor (vector unsigned short,
12899 vector bool short);
12900 vector unsigned short vec_xor (vector unsigned short,
12901 vector unsigned short);
12902 vector signed char vec_xor (vector bool char, vector signed char);
12903 vector bool char vec_xor (vector bool char, vector bool char);
12904 vector signed char vec_xor (vector signed char, vector bool char);
12905 vector signed char vec_xor (vector signed char, vector signed char);
12906 vector unsigned char vec_xor (vector bool char, vector unsigned char);
12907 vector unsigned char vec_xor (vector unsigned char, vector bool char);
12908 vector unsigned char vec_xor (vector unsigned char,
12909 vector unsigned char);
12911 int vec_all_eq (vector signed char, vector bool char);
12912 int vec_all_eq (vector signed char, vector signed char);
12913 int vec_all_eq (vector unsigned char, vector bool char);
12914 int vec_all_eq (vector unsigned char, vector unsigned char);
12915 int vec_all_eq (vector bool char, vector bool char);
12916 int vec_all_eq (vector bool char, vector unsigned char);
12917 int vec_all_eq (vector bool char, vector signed char);
12918 int vec_all_eq (vector signed short, vector bool short);
12919 int vec_all_eq (vector signed short, vector signed short);
12920 int vec_all_eq (vector unsigned short, vector bool short);
12921 int vec_all_eq (vector unsigned short, vector unsigned short);
12922 int vec_all_eq (vector bool short, vector bool short);
12923 int vec_all_eq (vector bool short, vector unsigned short);
12924 int vec_all_eq (vector bool short, vector signed short);
12925 int vec_all_eq (vector pixel, vector pixel);
12926 int vec_all_eq (vector signed int, vector bool int);
12927 int vec_all_eq (vector signed int, vector signed int);
12928 int vec_all_eq (vector unsigned int, vector bool int);
12929 int vec_all_eq (vector unsigned int, vector unsigned int);
12930 int vec_all_eq (vector bool int, vector bool int);
12931 int vec_all_eq (vector bool int, vector unsigned int);
12932 int vec_all_eq (vector bool int, vector signed int);
12933 int vec_all_eq (vector float, vector float);
12935 int vec_all_ge (vector bool char, vector unsigned char);
12936 int vec_all_ge (vector unsigned char, vector bool char);
12937 int vec_all_ge (vector unsigned char, vector unsigned char);
12938 int vec_all_ge (vector bool char, vector signed char);
12939 int vec_all_ge (vector signed char, vector bool char);
12940 int vec_all_ge (vector signed char, vector signed char);
12941 int vec_all_ge (vector bool short, vector unsigned short);
12942 int vec_all_ge (vector unsigned short, vector bool short);
12943 int vec_all_ge (vector unsigned short, vector unsigned short);
12944 int vec_all_ge (vector signed short, vector signed short);
12945 int vec_all_ge (vector bool short, vector signed short);
12946 int vec_all_ge (vector signed short, vector bool short);
12947 int vec_all_ge (vector bool int, vector unsigned int);
12948 int vec_all_ge (vector unsigned int, vector bool int);
12949 int vec_all_ge (vector unsigned int, vector unsigned int);
12950 int vec_all_ge (vector bool int, vector signed int);
12951 int vec_all_ge (vector signed int, vector bool int);
12952 int vec_all_ge (vector signed int, vector signed int);
12953 int vec_all_ge (vector float, vector float);
12955 int vec_all_gt (vector bool char, vector unsigned char);
12956 int vec_all_gt (vector unsigned char, vector bool char);
12957 int vec_all_gt (vector unsigned char, vector unsigned char);
12958 int vec_all_gt (vector bool char, vector signed char);
12959 int vec_all_gt (vector signed char, vector bool char);
12960 int vec_all_gt (vector signed char, vector signed char);
12961 int vec_all_gt (vector bool short, vector unsigned short);
12962 int vec_all_gt (vector unsigned short, vector bool short);
12963 int vec_all_gt (vector unsigned short, vector unsigned short);
12964 int vec_all_gt (vector bool short, vector signed short);
12965 int vec_all_gt (vector signed short, vector bool short);
12966 int vec_all_gt (vector signed short, vector signed short);
12967 int vec_all_gt (vector bool int, vector unsigned int);
12968 int vec_all_gt (vector unsigned int, vector bool int);
12969 int vec_all_gt (vector unsigned int, vector unsigned int);
12970 int vec_all_gt (vector bool int, vector signed int);
12971 int vec_all_gt (vector signed int, vector bool int);
12972 int vec_all_gt (vector signed int, vector signed int);
12973 int vec_all_gt (vector float, vector float);
12975 int vec_all_in (vector float, vector float);
12977 int vec_all_le (vector bool char, vector unsigned char);
12978 int vec_all_le (vector unsigned char, vector bool char);
12979 int vec_all_le (vector unsigned char, vector unsigned char);
12980 int vec_all_le (vector bool char, vector signed char);
12981 int vec_all_le (vector signed char, vector bool char);
12982 int vec_all_le (vector signed char, vector signed char);
12983 int vec_all_le (vector bool short, vector unsigned short);
12984 int vec_all_le (vector unsigned short, vector bool short);
12985 int vec_all_le (vector unsigned short, vector unsigned short);
12986 int vec_all_le (vector bool short, vector signed short);
12987 int vec_all_le (vector signed short, vector bool short);
12988 int vec_all_le (vector signed short, vector signed short);
12989 int vec_all_le (vector bool int, vector unsigned int);
12990 int vec_all_le (vector unsigned int, vector bool int);
12991 int vec_all_le (vector unsigned int, vector unsigned int);
12992 int vec_all_le (vector bool int, vector signed int);
12993 int vec_all_le (vector signed int, vector bool int);
12994 int vec_all_le (vector signed int, vector signed int);
12995 int vec_all_le (vector float, vector float);
12997 int vec_all_lt (vector bool char, vector unsigned char);
12998 int vec_all_lt (vector unsigned char, vector bool char);
12999 int vec_all_lt (vector unsigned char, vector unsigned char);
13000 int vec_all_lt (vector bool char, vector signed char);
13001 int vec_all_lt (vector signed char, vector bool char);
13002 int vec_all_lt (vector signed char, vector signed char);
13003 int vec_all_lt (vector bool short, vector unsigned short);
13004 int vec_all_lt (vector unsigned short, vector bool short);
13005 int vec_all_lt (vector unsigned short, vector unsigned short);
13006 int vec_all_lt (vector bool short, vector signed short);
13007 int vec_all_lt (vector signed short, vector bool short);
13008 int vec_all_lt (vector signed short, vector signed short);
13009 int vec_all_lt (vector bool int, vector unsigned int);
13010 int vec_all_lt (vector unsigned int, vector bool int);
13011 int vec_all_lt (vector unsigned int, vector unsigned int);
13012 int vec_all_lt (vector bool int, vector signed int);
13013 int vec_all_lt (vector signed int, vector bool int);
13014 int vec_all_lt (vector signed int, vector signed int);
13015 int vec_all_lt (vector float, vector float);
13017 int vec_all_nan (vector float);
13019 int vec_all_ne (vector signed char, vector bool char);
13020 int vec_all_ne (vector signed char, vector signed char);
13021 int vec_all_ne (vector unsigned char, vector bool char);
13022 int vec_all_ne (vector unsigned char, vector unsigned char);
13023 int vec_all_ne (vector bool char, vector bool char);
13024 int vec_all_ne (vector bool char, vector unsigned char);
13025 int vec_all_ne (vector bool char, vector signed char);
13026 int vec_all_ne (vector signed short, vector bool short);
13027 int vec_all_ne (vector signed short, vector signed short);
13028 int vec_all_ne (vector unsigned short, vector bool short);
13029 int vec_all_ne (vector unsigned short, vector unsigned short);
13030 int vec_all_ne (vector bool short, vector bool short);
13031 int vec_all_ne (vector bool short, vector unsigned short);
13032 int vec_all_ne (vector bool short, vector signed short);
13033 int vec_all_ne (vector pixel, vector pixel);
13034 int vec_all_ne (vector signed int, vector bool int);
13035 int vec_all_ne (vector signed int, vector signed int);
13036 int vec_all_ne (vector unsigned int, vector bool int);
13037 int vec_all_ne (vector unsigned int, vector unsigned int);
13038 int vec_all_ne (vector bool int, vector bool int);
13039 int vec_all_ne (vector bool int, vector unsigned int);
13040 int vec_all_ne (vector bool int, vector signed int);
13041 int vec_all_ne (vector float, vector float);
13043 int vec_all_nge (vector float, vector float);
13045 int vec_all_ngt (vector float, vector float);
13047 int vec_all_nle (vector float, vector float);
13049 int vec_all_nlt (vector float, vector float);
13051 int vec_all_numeric (vector float);
13053 int vec_any_eq (vector signed char, vector bool char);
13054 int vec_any_eq (vector signed char, vector signed char);
13055 int vec_any_eq (vector unsigned char, vector bool char);
13056 int vec_any_eq (vector unsigned char, vector unsigned char);
13057 int vec_any_eq (vector bool char, vector bool char);
13058 int vec_any_eq (vector bool char, vector unsigned char);
13059 int vec_any_eq (vector bool char, vector signed char);
13060 int vec_any_eq (vector signed short, vector bool short);
13061 int vec_any_eq (vector signed short, vector signed short);
13062 int vec_any_eq (vector unsigned short, vector bool short);
13063 int vec_any_eq (vector unsigned short, vector unsigned short);
13064 int vec_any_eq (vector bool short, vector bool short);
13065 int vec_any_eq (vector bool short, vector unsigned short);
13066 int vec_any_eq (vector bool short, vector signed short);
13067 int vec_any_eq (vector pixel, vector pixel);
13068 int vec_any_eq (vector signed int, vector bool int);
13069 int vec_any_eq (vector signed int, vector signed int);
13070 int vec_any_eq (vector unsigned int, vector bool int);
13071 int vec_any_eq (vector unsigned int, vector unsigned int);
13072 int vec_any_eq (vector bool int, vector bool int);
13073 int vec_any_eq (vector bool int, vector unsigned int);
13074 int vec_any_eq (vector bool int, vector signed int);
13075 int vec_any_eq (vector float, vector float);
13077 int vec_any_ge (vector signed char, vector bool char);
13078 int vec_any_ge (vector unsigned char, vector bool char);
13079 int vec_any_ge (vector unsigned char, vector unsigned char);
13080 int vec_any_ge (vector signed char, vector signed char);
13081 int vec_any_ge (vector bool char, vector unsigned char);
13082 int vec_any_ge (vector bool char, vector signed char);
13083 int vec_any_ge (vector unsigned short, vector bool short);
13084 int vec_any_ge (vector unsigned short, vector unsigned short);
13085 int vec_any_ge (vector signed short, vector signed short);
13086 int vec_any_ge (vector signed short, vector bool short);
13087 int vec_any_ge (vector bool short, vector unsigned short);
13088 int vec_any_ge (vector bool short, vector signed short);
13089 int vec_any_ge (vector signed int, vector bool int);
13090 int vec_any_ge (vector unsigned int, vector bool int);
13091 int vec_any_ge (vector unsigned int, vector unsigned int);
13092 int vec_any_ge (vector signed int, vector signed int);
13093 int vec_any_ge (vector bool int, vector unsigned int);
13094 int vec_any_ge (vector bool int, vector signed int);
13095 int vec_any_ge (vector float, vector float);
13097 int vec_any_gt (vector bool char, vector unsigned char);
13098 int vec_any_gt (vector unsigned char, vector bool char);
13099 int vec_any_gt (vector unsigned char, vector unsigned char);
13100 int vec_any_gt (vector bool char, vector signed char);
13101 int vec_any_gt (vector signed char, vector bool char);
13102 int vec_any_gt (vector signed char, vector signed char);
13103 int vec_any_gt (vector bool short, vector unsigned short);
13104 int vec_any_gt (vector unsigned short, vector bool short);
13105 int vec_any_gt (vector unsigned short, vector unsigned short);
13106 int vec_any_gt (vector bool short, vector signed short);
13107 int vec_any_gt (vector signed short, vector bool short);
13108 int vec_any_gt (vector signed short, vector signed short);
13109 int vec_any_gt (vector bool int, vector unsigned int);
13110 int vec_any_gt (vector unsigned int, vector bool int);
13111 int vec_any_gt (vector unsigned int, vector unsigned int);
13112 int vec_any_gt (vector bool int, vector signed int);
13113 int vec_any_gt (vector signed int, vector bool int);
13114 int vec_any_gt (vector signed int, vector signed int);
13115 int vec_any_gt (vector float, vector float);
13117 int vec_any_le (vector bool char, vector unsigned char);
13118 int vec_any_le (vector unsigned char, vector bool char);
13119 int vec_any_le (vector unsigned char, vector unsigned char);
13120 int vec_any_le (vector bool char, vector signed char);
13121 int vec_any_le (vector signed char, vector bool char);
13122 int vec_any_le (vector signed char, vector signed char);
13123 int vec_any_le (vector bool short, vector unsigned short);
13124 int vec_any_le (vector unsigned short, vector bool short);
13125 int vec_any_le (vector unsigned short, vector unsigned short);
13126 int vec_any_le (vector bool short, vector signed short);
13127 int vec_any_le (vector signed short, vector bool short);
13128 int vec_any_le (vector signed short, vector signed short);
13129 int vec_any_le (vector bool int, vector unsigned int);
13130 int vec_any_le (vector unsigned int, vector bool int);
13131 int vec_any_le (vector unsigned int, vector unsigned int);
13132 int vec_any_le (vector bool int, vector signed int);
13133 int vec_any_le (vector signed int, vector bool int);
13134 int vec_any_le (vector signed int, vector signed int);
13135 int vec_any_le (vector float, vector float);
13137 int vec_any_lt (vector bool char, vector unsigned char);
13138 int vec_any_lt (vector unsigned char, vector bool char);
13139 int vec_any_lt (vector unsigned char, vector unsigned char);
13140 int vec_any_lt (vector bool char, vector signed char);
13141 int vec_any_lt (vector signed char, vector bool char);
13142 int vec_any_lt (vector signed char, vector signed char);
13143 int vec_any_lt (vector bool short, vector unsigned short);
13144 int vec_any_lt (vector unsigned short, vector bool short);
13145 int vec_any_lt (vector unsigned short, vector unsigned short);
13146 int vec_any_lt (vector bool short, vector signed short);
13147 int vec_any_lt (vector signed short, vector bool short);
13148 int vec_any_lt (vector signed short, vector signed short);
13149 int vec_any_lt (vector bool int, vector unsigned int);
13150 int vec_any_lt (vector unsigned int, vector bool int);
13151 int vec_any_lt (vector unsigned int, vector unsigned int);
13152 int vec_any_lt (vector bool int, vector signed int);
13153 int vec_any_lt (vector signed int, vector bool int);
13154 int vec_any_lt (vector signed int, vector signed int);
13155 int vec_any_lt (vector float, vector float);
13157 int vec_any_nan (vector float);
13159 int vec_any_ne (vector signed char, vector bool char);
13160 int vec_any_ne (vector signed char, vector signed char);
13161 int vec_any_ne (vector unsigned char, vector bool char);
13162 int vec_any_ne (vector unsigned char, vector unsigned char);
13163 int vec_any_ne (vector bool char, vector bool char);
13164 int vec_any_ne (vector bool char, vector unsigned char);
13165 int vec_any_ne (vector bool char, vector signed char);
13166 int vec_any_ne (vector signed short, vector bool short);
13167 int vec_any_ne (vector signed short, vector signed short);
13168 int vec_any_ne (vector unsigned short, vector bool short);
13169 int vec_any_ne (vector unsigned short, vector unsigned short);
13170 int vec_any_ne (vector bool short, vector bool short);
13171 int vec_any_ne (vector bool short, vector unsigned short);
13172 int vec_any_ne (vector bool short, vector signed short);
13173 int vec_any_ne (vector pixel, vector pixel);
13174 int vec_any_ne (vector signed int, vector bool int);
13175 int vec_any_ne (vector signed int, vector signed int);
13176 int vec_any_ne (vector unsigned int, vector bool int);
13177 int vec_any_ne (vector unsigned int, vector unsigned int);
13178 int vec_any_ne (vector bool int, vector bool int);
13179 int vec_any_ne (vector bool int, vector unsigned int);
13180 int vec_any_ne (vector bool int, vector signed int);
13181 int vec_any_ne (vector float, vector float);
13183 int vec_any_nge (vector float, vector float);
13185 int vec_any_ngt (vector float, vector float);
13187 int vec_any_nle (vector float, vector float);
13189 int vec_any_nlt (vector float, vector float);
13191 int vec_any_numeric (vector float);
13193 int vec_any_out (vector float, vector float);
13196 If the vector/scalar (VSX) instruction set is available, the following
13197 additional functions are available:
13200 vector double vec_abs (vector double);
13201 vector double vec_add (vector double, vector double);
13202 vector double vec_and (vector double, vector double);
13203 vector double vec_and (vector double, vector bool long);
13204 vector double vec_and (vector bool long, vector double);
13205 vector double vec_andc (vector double, vector double);
13206 vector double vec_andc (vector double, vector bool long);
13207 vector double vec_andc (vector bool long, vector double);
13208 vector double vec_ceil (vector double);
13209 vector bool long vec_cmpeq (vector double, vector double);
13210 vector bool long vec_cmpge (vector double, vector double);
13211 vector bool long vec_cmpgt (vector double, vector double);
13212 vector bool long vec_cmple (vector double, vector double);
13213 vector bool long vec_cmplt (vector double, vector double);
13214 vector float vec_div (vector float, vector float);
13215 vector double vec_div (vector double, vector double);
13216 vector double vec_floor (vector double);
13217 vector double vec_ld (int, const vector double *);
13218 vector double vec_ld (int, const double *);
13219 vector double vec_ldl (int, const vector double *);
13220 vector double vec_ldl (int, const double *);
13221 vector unsigned char vec_lvsl (int, const volatile double *);
13222 vector unsigned char vec_lvsr (int, const volatile double *);
13223 vector double vec_madd (vector double, vector double, vector double);
13224 vector double vec_max (vector double, vector double);
13225 vector double vec_min (vector double, vector double);
13226 vector float vec_msub (vector float, vector float, vector float);
13227 vector double vec_msub (vector double, vector double, vector double);
13228 vector float vec_mul (vector float, vector float);
13229 vector double vec_mul (vector double, vector double);
13230 vector float vec_nearbyint (vector float);
13231 vector double vec_nearbyint (vector double);
13232 vector float vec_nmadd (vector float, vector float, vector float);
13233 vector double vec_nmadd (vector double, vector double, vector double);
13234 vector double vec_nmsub (vector double, vector double, vector double);
13235 vector double vec_nor (vector double, vector double);
13236 vector double vec_or (vector double, vector double);
13237 vector double vec_or (vector double, vector bool long);
13238 vector double vec_or (vector bool long, vector double);
13239 vector double vec_perm (vector double,
13241 vector unsigned char);
13242 vector double vec_rint (vector double);
13243 vector double vec_recip (vector double, vector double);
13244 vector double vec_rsqrt (vector double);
13245 vector double vec_rsqrte (vector double);
13246 vector double vec_sel (vector double, vector double, vector bool long);
13247 vector double vec_sel (vector double, vector double, vector unsigned long);
13248 vector double vec_sub (vector double, vector double);
13249 vector float vec_sqrt (vector float);
13250 vector double vec_sqrt (vector double);
13251 void vec_st (vector double, int, vector double *);
13252 void vec_st (vector double, int, double *);
13253 vector double vec_trunc (vector double);
13254 vector double vec_xor (vector double, vector double);
13255 vector double vec_xor (vector double, vector bool long);
13256 vector double vec_xor (vector bool long, vector double);
13257 int vec_all_eq (vector double, vector double);
13258 int vec_all_ge (vector double, vector double);
13259 int vec_all_gt (vector double, vector double);
13260 int vec_all_le (vector double, vector double);
13261 int vec_all_lt (vector double, vector double);
13262 int vec_all_nan (vector double);
13263 int vec_all_ne (vector double, vector double);
13264 int vec_all_nge (vector double, vector double);
13265 int vec_all_ngt (vector double, vector double);
13266 int vec_all_nle (vector double, vector double);
13267 int vec_all_nlt (vector double, vector double);
13268 int vec_all_numeric (vector double);
13269 int vec_any_eq (vector double, vector double);
13270 int vec_any_ge (vector double, vector double);
13271 int vec_any_gt (vector double, vector double);
13272 int vec_any_le (vector double, vector double);
13273 int vec_any_lt (vector double, vector double);
13274 int vec_any_nan (vector double);
13275 int vec_any_ne (vector double, vector double);
13276 int vec_any_nge (vector double, vector double);
13277 int vec_any_ngt (vector double, vector double);
13278 int vec_any_nle (vector double, vector double);
13279 int vec_any_nlt (vector double, vector double);
13280 int vec_any_numeric (vector double);
13282 vector double vec_vsx_ld (int, const vector double *);
13283 vector double vec_vsx_ld (int, const double *);
13284 vector float vec_vsx_ld (int, const vector float *);
13285 vector float vec_vsx_ld (int, const float *);
13286 vector bool int vec_vsx_ld (int, const vector bool int *);
13287 vector signed int vec_vsx_ld (int, const vector signed int *);
13288 vector signed int vec_vsx_ld (int, const int *);
13289 vector signed int vec_vsx_ld (int, const long *);
13290 vector unsigned int vec_vsx_ld (int, const vector unsigned int *);
13291 vector unsigned int vec_vsx_ld (int, const unsigned int *);
13292 vector unsigned int vec_vsx_ld (int, const unsigned long *);
13293 vector bool short vec_vsx_ld (int, const vector bool short *);
13294 vector pixel vec_vsx_ld (int, const vector pixel *);
13295 vector signed short vec_vsx_ld (int, const vector signed short *);
13296 vector signed short vec_vsx_ld (int, const short *);
13297 vector unsigned short vec_vsx_ld (int, const vector unsigned short *);
13298 vector unsigned short vec_vsx_ld (int, const unsigned short *);
13299 vector bool char vec_vsx_ld (int, const vector bool char *);
13300 vector signed char vec_vsx_ld (int, const vector signed char *);
13301 vector signed char vec_vsx_ld (int, const signed char *);
13302 vector unsigned char vec_vsx_ld (int, const vector unsigned char *);
13303 vector unsigned char vec_vsx_ld (int, const unsigned char *);
13305 void vec_vsx_st (vector double, int, vector double *);
13306 void vec_vsx_st (vector double, int, double *);
13307 void vec_vsx_st (vector float, int, vector float *);
13308 void vec_vsx_st (vector float, int, float *);
13309 void vec_vsx_st (vector signed int, int, vector signed int *);
13310 void vec_vsx_st (vector signed int, int, int *);
13311 void vec_vsx_st (vector unsigned int, int, vector unsigned int *);
13312 void vec_vsx_st (vector unsigned int, int, unsigned int *);
13313 void vec_vsx_st (vector bool int, int, vector bool int *);
13314 void vec_vsx_st (vector bool int, int, unsigned int *);
13315 void vec_vsx_st (vector bool int, int, int *);
13316 void vec_vsx_st (vector signed short, int, vector signed short *);
13317 void vec_vsx_st (vector signed short, int, short *);
13318 void vec_vsx_st (vector unsigned short, int, vector unsigned short *);
13319 void vec_vsx_st (vector unsigned short, int, unsigned short *);
13320 void vec_vsx_st (vector bool short, int, vector bool short *);
13321 void vec_vsx_st (vector bool short, int, unsigned short *);
13322 void vec_vsx_st (vector pixel, int, vector pixel *);
13323 void vec_vsx_st (vector pixel, int, unsigned short *);
13324 void vec_vsx_st (vector pixel, int, short *);
13325 void vec_vsx_st (vector bool short, int, short *);
13326 void vec_vsx_st (vector signed char, int, vector signed char *);
13327 void vec_vsx_st (vector signed char, int, signed char *);
13328 void vec_vsx_st (vector unsigned char, int, vector unsigned char *);
13329 void vec_vsx_st (vector unsigned char, int, unsigned char *);
13330 void vec_vsx_st (vector bool char, int, vector bool char *);
13331 void vec_vsx_st (vector bool char, int, unsigned char *);
13332 void vec_vsx_st (vector bool char, int, signed char *);
13335 Note that the @samp{vec_ld} and @samp{vec_st} builtins will always
13336 generate the Altivec @samp{LVX} and @samp{STVX} instructions even
13337 if the VSX instruction set is available. The @samp{vec_vsx_ld} and
13338 @samp{vec_vsx_st} builtins will always generate the VSX @samp{LXVD2X},
13339 @samp{LXVW4X}, @samp{STXVD2X}, and @samp{STXVW4X} instructions.
13341 GCC provides a few other builtins on Powerpc to access certain instructions:
13343 float __builtin_recipdivf (float, float);
13344 float __builtin_rsqrtf (float);
13345 double __builtin_recipdiv (double, double);
13346 double __builtin_rsqrt (double);
13347 long __builtin_bpermd (long, long);
13348 int __builtin_bswap16 (int);
13351 The @code{vec_rsqrt}, @code{__builtin_rsqrt}, and
13352 @code{__builtin_rsqrtf} functions generate multiple instructions to
13353 implement the reciprocal sqrt functionality using reciprocal sqrt
13354 estimate instructions.
13356 The @code{__builtin_recipdiv}, and @code{__builtin_recipdivf}
13357 functions generate multiple instructions to implement division using
13358 the reciprocal estimate instructions.
13360 @node RX Built-in Functions
13361 @subsection RX Built-in Functions
13362 GCC supports some of the RX instructions which cannot be expressed in
13363 the C programming language via the use of built-in functions. The
13364 following functions are supported:
13366 @deftypefn {Built-in Function} void __builtin_rx_brk (void)
13367 Generates the @code{brk} machine instruction.
13370 @deftypefn {Built-in Function} void __builtin_rx_clrpsw (int)
13371 Generates the @code{clrpsw} machine instruction to clear the specified
13372 bit in the processor status word.
13375 @deftypefn {Built-in Function} void __builtin_rx_int (int)
13376 Generates the @code{int} machine instruction to generate an interrupt
13377 with the specified value.
13380 @deftypefn {Built-in Function} void __builtin_rx_machi (int, int)
13381 Generates the @code{machi} machine instruction to add the result of
13382 multiplying the top 16-bits of the two arguments into the
13386 @deftypefn {Built-in Function} void __builtin_rx_maclo (int, int)
13387 Generates the @code{maclo} machine instruction to add the result of
13388 multiplying the bottom 16-bits of the two arguments into the
13392 @deftypefn {Built-in Function} void __builtin_rx_mulhi (int, int)
13393 Generates the @code{mulhi} machine instruction to place the result of
13394 multiplying the top 16-bits of the two arguments into the
13398 @deftypefn {Built-in Function} void __builtin_rx_mullo (int, int)
13399 Generates the @code{mullo} machine instruction to place the result of
13400 multiplying the bottom 16-bits of the two arguments into the
13404 @deftypefn {Built-in Function} int __builtin_rx_mvfachi (void)
13405 Generates the @code{mvfachi} machine instruction to read the top
13406 32-bits of the accumulator.
13409 @deftypefn {Built-in Function} int __builtin_rx_mvfacmi (void)
13410 Generates the @code{mvfacmi} machine instruction to read the middle
13411 32-bits of the accumulator.
13414 @deftypefn {Built-in Function} int __builtin_rx_mvfc (int)
13415 Generates the @code{mvfc} machine instruction which reads the control
13416 register specified in its argument and returns its value.
13419 @deftypefn {Built-in Function} void __builtin_rx_mvtachi (int)
13420 Generates the @code{mvtachi} machine instruction to set the top
13421 32-bits of the accumulator.
13424 @deftypefn {Built-in Function} void __builtin_rx_mvtaclo (int)
13425 Generates the @code{mvtaclo} machine instruction to set the bottom
13426 32-bits of the accumulator.
13429 @deftypefn {Built-in Function} void __builtin_rx_mvtc (int reg, int val)
13430 Generates the @code{mvtc} machine instruction which sets control
13431 register number @code{reg} to @code{val}.
13434 @deftypefn {Built-in Function} void __builtin_rx_mvtipl (int)
13435 Generates the @code{mvtipl} machine instruction set the interrupt
13439 @deftypefn {Built-in Function} void __builtin_rx_racw (int)
13440 Generates the @code{racw} machine instruction to round the accumulator
13441 according to the specified mode.
13444 @deftypefn {Built-in Function} int __builtin_rx_revw (int)
13445 Generates the @code{revw} machine instruction which swaps the bytes in
13446 the argument so that bits 0--7 now occupy bits 8--15 and vice versa,
13447 and also bits 16--23 occupy bits 24--31 and vice versa.
13450 @deftypefn {Built-in Function} void __builtin_rx_rmpa (void)
13451 Generates the @code{rmpa} machine instruction which initiates a
13452 repeated multiply and accumulate sequence.
13455 @deftypefn {Built-in Function} void __builtin_rx_round (float)
13456 Generates the @code{round} machine instruction which returns the
13457 floating point argument rounded according to the current rounding mode
13458 set in the floating point status word register.
13461 @deftypefn {Built-in Function} int __builtin_rx_sat (int)
13462 Generates the @code{sat} machine instruction which returns the
13463 saturated value of the argument.
13466 @deftypefn {Built-in Function} void __builtin_rx_setpsw (int)
13467 Generates the @code{setpsw} machine instruction to set the specified
13468 bit in the processor status word.
13471 @deftypefn {Built-in Function} void __builtin_rx_wait (void)
13472 Generates the @code{wait} machine instruction.
13475 @node SPARC VIS Built-in Functions
13476 @subsection SPARC VIS Built-in Functions
13478 GCC supports SIMD operations on the SPARC using both the generic vector
13479 extensions (@pxref{Vector Extensions}) as well as built-in functions for
13480 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
13481 switch, the VIS extension is exposed as the following built-in functions:
13484 typedef int v1si __attribute__ ((vector_size (4)));
13485 typedef int v2si __attribute__ ((vector_size (8)));
13486 typedef short v4hi __attribute__ ((vector_size (8)));
13487 typedef short v2hi __attribute__ ((vector_size (4)));
13488 typedef unsigned char v8qi __attribute__ ((vector_size (8)));
13489 typedef unsigned char v4qi __attribute__ ((vector_size (4)));
13491 void __builtin_vis_write_gsr (int64_t);
13492 int64_t __builtin_vis_read_gsr (void);
13494 void * __builtin_vis_alignaddr (void *, long);
13495 void * __builtin_vis_alignaddrl (void *, long);
13496 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
13497 v2si __builtin_vis_faligndatav2si (v2si, v2si);
13498 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
13499 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
13501 v4hi __builtin_vis_fexpand (v4qi);
13503 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
13504 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi);
13505 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi);
13506 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
13507 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
13508 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
13509 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
13511 v4qi __builtin_vis_fpack16 (v4hi);
13512 v8qi __builtin_vis_fpack32 (v2si, v8qi);
13513 v2hi __builtin_vis_fpackfix (v2si);
13514 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
13516 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
13518 long __builtin_vis_edge8 (void *, void *);
13519 long __builtin_vis_edge8l (void *, void *);
13520 long __builtin_vis_edge16 (void *, void *);
13521 long __builtin_vis_edge16l (void *, void *);
13522 long __builtin_vis_edge32 (void *, void *);
13523 long __builtin_vis_edge32l (void *, void *);
13525 long __builtin_vis_fcmple16 (v4hi, v4hi);
13526 long __builtin_vis_fcmple32 (v2si, v2si);
13527 long __builtin_vis_fcmpne16 (v4hi, v4hi);
13528 long __builtin_vis_fcmpne32 (v2si, v2si);
13529 long __builtin_vis_fcmpgt16 (v4hi, v4hi);
13530 long __builtin_vis_fcmpgt32 (v2si, v2si);
13531 long __builtin_vis_fcmpeq16 (v4hi, v4hi);
13532 long __builtin_vis_fcmpeq32 (v2si, v2si);
13534 v4hi __builtin_vis_fpadd16 (v4hi, v4hi);
13535 v2hi __builtin_vis_fpadd16s (v2hi, v2hi);
13536 v2si __builtin_vis_fpadd32 (v2si, v2si);
13537 v1si __builtin_vis_fpadd32s (v1si, v1si);
13538 v4hi __builtin_vis_fpsub16 (v4hi, v4hi);
13539 v2hi __builtin_vis_fpsub16s (v2hi, v2hi);
13540 v2si __builtin_vis_fpsub32 (v2si, v2si);
13541 v1si __builtin_vis_fpsub32s (v1si, v1si);
13543 long __builtin_vis_array8 (long, long);
13544 long __builtin_vis_array16 (long, long);
13545 long __builtin_vis_array32 (long, long);
13548 When you use the @option{-mvis2} switch, the VIS version 2.0 built-in
13549 functions also become available:
13552 long __builtin_vis_bmask (long, long);
13553 int64_t __builtin_vis_bshuffledi (int64_t, int64_t);
13554 v2si __builtin_vis_bshufflev2si (v2si, v2si);
13555 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi);
13556 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi);
13558 long __builtin_vis_edge8n (void *, void *);
13559 long __builtin_vis_edge8ln (void *, void *);
13560 long __builtin_vis_edge16n (void *, void *);
13561 long __builtin_vis_edge16ln (void *, void *);
13562 long __builtin_vis_edge32n (void *, void *);
13563 long __builtin_vis_edge32ln (void *, void *);
13566 When you use the @option{-mvis3} switch, the VIS version 3.0 built-in
13567 functions also become available:
13570 void __builtin_vis_cmask8 (long);
13571 void __builtin_vis_cmask16 (long);
13572 void __builtin_vis_cmask32 (long);
13574 v4hi __builtin_vis_fchksm16 (v4hi, v4hi);
13576 v4hi __builtin_vis_fsll16 (v4hi, v4hi);
13577 v4hi __builtin_vis_fslas16 (v4hi, v4hi);
13578 v4hi __builtin_vis_fsrl16 (v4hi, v4hi);
13579 v4hi __builtin_vis_fsra16 (v4hi, v4hi);
13580 v2si __builtin_vis_fsll16 (v2si, v2si);
13581 v2si __builtin_vis_fslas16 (v2si, v2si);
13582 v2si __builtin_vis_fsrl16 (v2si, v2si);
13583 v2si __builtin_vis_fsra16 (v2si, v2si);
13585 long __builtin_vis_pdistn (v8qi, v8qi);
13587 v4hi __builtin_vis_fmean16 (v4hi, v4hi);
13589 int64_t __builtin_vis_fpadd64 (int64_t, int64_t);
13590 int64_t __builtin_vis_fpsub64 (int64_t, int64_t);
13592 v4hi __builtin_vis_fpadds16 (v4hi, v4hi);
13593 v2hi __builtin_vis_fpadds16s (v2hi, v2hi);
13594 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi);
13595 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi);
13596 v2si __builtin_vis_fpadds32 (v2si, v2si);
13597 v1si __builtin_vis_fpadds32s (v1si, v1si);
13598 v2si __builtin_vis_fpsubs32 (v2si, v2si);
13599 v1si __builtin_vis_fpsubs32s (v1si, v1si);
13601 long __builtin_vis_fucmple8 (v8qi, v8qi);
13602 long __builtin_vis_fucmpne8 (v8qi, v8qi);
13603 long __builtin_vis_fucmpgt8 (v8qi, v8qi);
13604 long __builtin_vis_fucmpeq8 (v8qi, v8qi);
13606 float __builtin_vis_fhadds (float, float);
13607 double __builtin_vis_fhaddd (double, double);
13608 float __builtin_vis_fhsubs (float, float);
13609 double __builtin_vis_fhsubd (double, double);
13610 float __builtin_vis_fnhadds (float, float);
13611 double __builtin_vis_fnhaddd (double, double);
13613 int64_t __builtin_vis_umulxhi (int64_t, int64_t);
13614 int64_t __builtin_vis_xmulx (int64_t, int64_t);
13615 int64_t __builtin_vis_xmulxhi (int64_t, int64_t);
13618 @node SPU Built-in Functions
13619 @subsection SPU Built-in Functions
13621 GCC provides extensions for the SPU processor as described in the
13622 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
13623 found at @uref{http://cell.scei.co.jp/} or
13624 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
13625 implementation differs in several ways.
13630 The optional extension of specifying vector constants in parentheses is
13634 A vector initializer requires no cast if the vector constant is of the
13635 same type as the variable it is initializing.
13638 If @code{signed} or @code{unsigned} is omitted, the signedness of the
13639 vector type is the default signedness of the base type. The default
13640 varies depending on the operating system, so a portable program should
13641 always specify the signedness.
13644 By default, the keyword @code{__vector} is added. The macro
13645 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
13649 GCC allows using a @code{typedef} name as the type specifier for a
13653 For C, overloaded functions are implemented with macros so the following
13657 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
13660 Since @code{spu_add} is a macro, the vector constant in the example
13661 is treated as four separate arguments. Wrap the entire argument in
13662 parentheses for this to work.
13665 The extended version of @code{__builtin_expect} is not supported.
13669 @emph{Note:} Only the interface described in the aforementioned
13670 specification is supported. Internally, GCC uses built-in functions to
13671 implement the required functionality, but these are not supported and
13672 are subject to change without notice.
13674 @node TI C6X Built-in Functions
13675 @subsection TI C6X Built-in Functions
13677 GCC provides intrinsics to access certain instructions of the TI C6X
13678 processors. These intrinsics, listed below, are available after
13679 inclusion of the @code{c6x_intrinsics.h} header file. They map directly
13680 to C6X instructions.
13684 int _sadd (int, int)
13685 int _ssub (int, int)
13686 int _sadd2 (int, int)
13687 int _ssub2 (int, int)
13688 long long _mpy2 (int, int)
13689 long long _smpy2 (int, int)
13690 int _add4 (int, int)
13691 int _sub4 (int, int)
13692 int _saddu4 (int, int)
13694 int _smpy (int, int)
13695 int _smpyh (int, int)
13696 int _smpyhl (int, int)
13697 int _smpylh (int, int)
13699 int _sshl (int, int)
13700 int _subc (int, int)
13702 int _avg2 (int, int)
13703 int _avgu4 (int, int)
13705 int _clrr (int, int)
13706 int _extr (int, int)
13707 int _extru (int, int)
13713 @node Target Format Checks
13714 @section Format Checks Specific to Particular Target Machines
13716 For some target machines, GCC supports additional options to the
13718 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
13721 * Solaris Format Checks::
13722 * Darwin Format Checks::
13725 @node Solaris Format Checks
13726 @subsection Solaris Format Checks
13728 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
13729 check. @code{cmn_err} accepts a subset of the standard @code{printf}
13730 conversions, and the two-argument @code{%b} conversion for displaying
13731 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
13733 @node Darwin Format Checks
13734 @subsection Darwin Format Checks
13736 Darwin targets support the @code{CFString} (or @code{__CFString__}) in the format
13737 attribute context. Declarations made with such attribution will be parsed for correct syntax
13738 and format argument types. However, parsing of the format string itself is currently undefined
13739 and will not be carried out by this version of the compiler.
13741 Additionally, @code{CFStringRefs} (defined by the @code{CoreFoundation} headers) may
13742 also be used as format arguments. Note that the relevant headers are only likely to be
13743 available on Darwin (OSX) installations. On such installations, the XCode and system
13744 documentation provide descriptions of @code{CFString}, @code{CFStringRefs} and
13745 associated functions.
13748 @section Pragmas Accepted by GCC
13750 @cindex @code{#pragma}
13752 GCC supports several types of pragmas, primarily in order to compile
13753 code originally written for other compilers. Note that in general
13754 we do not recommend the use of pragmas; @xref{Function Attributes},
13755 for further explanation.
13761 * RS/6000 and PowerPC Pragmas::
13763 * Solaris Pragmas::
13764 * Symbol-Renaming Pragmas::
13765 * Structure-Packing Pragmas::
13767 * Diagnostic Pragmas::
13768 * Visibility Pragmas::
13769 * Push/Pop Macro Pragmas::
13770 * Function Specific Option Pragmas::
13774 @subsection ARM Pragmas
13776 The ARM target defines pragmas for controlling the default addition of
13777 @code{long_call} and @code{short_call} attributes to functions.
13778 @xref{Function Attributes}, for information about the effects of these
13783 @cindex pragma, long_calls
13784 Set all subsequent functions to have the @code{long_call} attribute.
13786 @item no_long_calls
13787 @cindex pragma, no_long_calls
13788 Set all subsequent functions to have the @code{short_call} attribute.
13790 @item long_calls_off
13791 @cindex pragma, long_calls_off
13792 Do not affect the @code{long_call} or @code{short_call} attributes of
13793 subsequent functions.
13797 @subsection M32C Pragmas
13800 @item GCC memregs @var{number}
13801 @cindex pragma, memregs
13802 Overrides the command-line option @code{-memregs=} for the current
13803 file. Use with care! This pragma must be before any function in the
13804 file, and mixing different memregs values in different objects may
13805 make them incompatible. This pragma is useful when a
13806 performance-critical function uses a memreg for temporary values,
13807 as it may allow you to reduce the number of memregs used.
13809 @item ADDRESS @var{name} @var{address}
13810 @cindex pragma, address
13811 For any declared symbols matching @var{name}, this does three things
13812 to that symbol: it forces the symbol to be located at the given
13813 address (a number), it forces the symbol to be volatile, and it
13814 changes the symbol's scope to be static. This pragma exists for
13815 compatibility with other compilers, but note that the common
13816 @code{1234H} numeric syntax is not supported (use @code{0x1234}
13820 #pragma ADDRESS port3 0x103
13827 @subsection MeP Pragmas
13831 @item custom io_volatile (on|off)
13832 @cindex pragma, custom io_volatile
13833 Overrides the command line option @code{-mio-volatile} for the current
13834 file. Note that for compatibility with future GCC releases, this
13835 option should only be used once before any @code{io} variables in each
13838 @item GCC coprocessor available @var{registers}
13839 @cindex pragma, coprocessor available
13840 Specifies which coprocessor registers are available to the register
13841 allocator. @var{registers} may be a single register, register range
13842 separated by ellipses, or comma-separated list of those. Example:
13845 #pragma GCC coprocessor available $c0...$c10, $c28
13848 @item GCC coprocessor call_saved @var{registers}
13849 @cindex pragma, coprocessor call_saved
13850 Specifies which coprocessor registers are to be saved and restored by
13851 any function using them. @var{registers} may be a single register,
13852 register range separated by ellipses, or comma-separated list of
13856 #pragma GCC coprocessor call_saved $c4...$c6, $c31
13859 @item GCC coprocessor subclass '(A|B|C|D)' = @var{registers}
13860 @cindex pragma, coprocessor subclass
13861 Creates and defines a register class. These register classes can be
13862 used by inline @code{asm} constructs. @var{registers} may be a single
13863 register, register range separated by ellipses, or comma-separated
13864 list of those. Example:
13867 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6
13869 asm ("cpfoo %0" : "=B" (x));
13872 @item GCC disinterrupt @var{name} , @var{name} @dots{}
13873 @cindex pragma, disinterrupt
13874 For the named functions, the compiler adds code to disable interrupts
13875 for the duration of those functions. Any functions so named, which
13876 are not encountered in the source, cause a warning that the pragma was
13877 not used. Examples:
13880 #pragma disinterrupt foo
13881 #pragma disinterrupt bar, grill
13882 int foo () @{ @dots{} @}
13885 @item GCC call @var{name} , @var{name} @dots{}
13886 @cindex pragma, call
13887 For the named functions, the compiler always uses a register-indirect
13888 call model when calling the named functions. Examples:
13897 @node RS/6000 and PowerPC Pragmas
13898 @subsection RS/6000 and PowerPC Pragmas
13900 The RS/6000 and PowerPC targets define one pragma for controlling
13901 whether or not the @code{longcall} attribute is added to function
13902 declarations by default. This pragma overrides the @option{-mlongcall}
13903 option, but not the @code{longcall} and @code{shortcall} attributes.
13904 @xref{RS/6000 and PowerPC Options}, for more information about when long
13905 calls are and are not necessary.
13909 @cindex pragma, longcall
13910 Apply the @code{longcall} attribute to all subsequent function
13914 Do not apply the @code{longcall} attribute to subsequent function
13918 @c Describe h8300 pragmas here.
13919 @c Describe sh pragmas here.
13920 @c Describe v850 pragmas here.
13922 @node Darwin Pragmas
13923 @subsection Darwin Pragmas
13925 The following pragmas are available for all architectures running the
13926 Darwin operating system. These are useful for compatibility with other
13930 @item mark @var{tokens}@dots{}
13931 @cindex pragma, mark
13932 This pragma is accepted, but has no effect.
13934 @item options align=@var{alignment}
13935 @cindex pragma, options align
13936 This pragma sets the alignment of fields in structures. The values of
13937 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
13938 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
13939 properly; to restore the previous setting, use @code{reset} for the
13942 @item segment @var{tokens}@dots{}
13943 @cindex pragma, segment
13944 This pragma is accepted, but has no effect.
13946 @item unused (@var{var} [, @var{var}]@dots{})
13947 @cindex pragma, unused
13948 This pragma declares variables to be possibly unused. GCC will not
13949 produce warnings for the listed variables. The effect is similar to
13950 that of the @code{unused} attribute, except that this pragma may appear
13951 anywhere within the variables' scopes.
13954 @node Solaris Pragmas
13955 @subsection Solaris Pragmas
13957 The Solaris target supports @code{#pragma redefine_extname}
13958 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
13959 @code{#pragma} directives for compatibility with the system compiler.
13962 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
13963 @cindex pragma, align
13965 Increase the minimum alignment of each @var{variable} to @var{alignment}.
13966 This is the same as GCC's @code{aligned} attribute @pxref{Variable
13967 Attributes}). Macro expansion occurs on the arguments to this pragma
13968 when compiling C and Objective-C@. It does not currently occur when
13969 compiling C++, but this is a bug which may be fixed in a future
13972 @item fini (@var{function} [, @var{function}]...)
13973 @cindex pragma, fini
13975 This pragma causes each listed @var{function} to be called after
13976 main, or during shared module unloading, by adding a call to the
13977 @code{.fini} section.
13979 @item init (@var{function} [, @var{function}]...)
13980 @cindex pragma, init
13982 This pragma causes each listed @var{function} to be called during
13983 initialization (before @code{main}) or during shared module loading, by
13984 adding a call to the @code{.init} section.
13988 @node Symbol-Renaming Pragmas
13989 @subsection Symbol-Renaming Pragmas
13991 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
13992 supports two @code{#pragma} directives which change the name used in
13993 assembly for a given declaration. @code{#pragma extern_prefix} is only
13994 available on platforms whose system headers need it. To get this effect
13995 on all platforms supported by GCC, use the asm labels extension (@pxref{Asm
13999 @item redefine_extname @var{oldname} @var{newname}
14000 @cindex pragma, redefine_extname
14002 This pragma gives the C function @var{oldname} the assembly symbol
14003 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
14004 will be defined if this pragma is available (currently on all platforms).
14006 @item extern_prefix @var{string}
14007 @cindex pragma, extern_prefix
14009 This pragma causes all subsequent external function and variable
14010 declarations to have @var{string} prepended to their assembly symbols.
14011 This effect may be terminated with another @code{extern_prefix} pragma
14012 whose argument is an empty string. The preprocessor macro
14013 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
14014 available (currently only on Tru64 UNIX)@.
14017 These pragmas and the asm labels extension interact in a complicated
14018 manner. Here are some corner cases you may want to be aware of.
14021 @item Both pragmas silently apply only to declarations with external
14022 linkage. Asm labels do not have this restriction.
14024 @item In C++, both pragmas silently apply only to declarations with
14025 ``C'' linkage. Again, asm labels do not have this restriction.
14027 @item If any of the three ways of changing the assembly name of a
14028 declaration is applied to a declaration whose assembly name has
14029 already been determined (either by a previous use of one of these
14030 features, or because the compiler needed the assembly name in order to
14031 generate code), and the new name is different, a warning issues and
14032 the name does not change.
14034 @item The @var{oldname} used by @code{#pragma redefine_extname} is
14035 always the C-language name.
14037 @item If @code{#pragma extern_prefix} is in effect, and a declaration
14038 occurs with an asm label attached, the prefix is silently ignored for
14041 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
14042 apply to the same declaration, whichever triggered first wins, and a
14043 warning issues if they contradict each other. (We would like to have
14044 @code{#pragma redefine_extname} always win, for consistency with asm
14045 labels, but if @code{#pragma extern_prefix} triggers first we have no
14046 way of knowing that that happened.)
14049 @node Structure-Packing Pragmas
14050 @subsection Structure-Packing Pragmas
14052 For compatibility with Microsoft Windows compilers, GCC supports a
14053 set of @code{#pragma} directives which change the maximum alignment of
14054 members of structures (other than zero-width bitfields), unions, and
14055 classes subsequently defined. The @var{n} value below always is required
14056 to be a small power of two and specifies the new alignment in bytes.
14059 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
14060 @item @code{#pragma pack()} sets the alignment to the one that was in
14061 effect when compilation started (see also command-line option
14062 @option{-fpack-struct[=@var{n}]} @pxref{Code Gen Options}).
14063 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
14064 setting on an internal stack and then optionally sets the new alignment.
14065 @item @code{#pragma pack(pop)} restores the alignment setting to the one
14066 saved at the top of the internal stack (and removes that stack entry).
14067 Note that @code{#pragma pack([@var{n}])} does not influence this internal
14068 stack; thus it is possible to have @code{#pragma pack(push)} followed by
14069 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
14070 @code{#pragma pack(pop)}.
14073 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
14074 @code{#pragma} which lays out a structure as the documented
14075 @code{__attribute__ ((ms_struct))}.
14077 @item @code{#pragma ms_struct on} turns on the layout for structures
14079 @item @code{#pragma ms_struct off} turns off the layout for structures
14081 @item @code{#pragma ms_struct reset} goes back to the default layout.
14085 @subsection Weak Pragmas
14087 For compatibility with SVR4, GCC supports a set of @code{#pragma}
14088 directives for declaring symbols to be weak, and defining weak
14092 @item #pragma weak @var{symbol}
14093 @cindex pragma, weak
14094 This pragma declares @var{symbol} to be weak, as if the declaration
14095 had the attribute of the same name. The pragma may appear before
14096 or after the declaration of @var{symbol}. It is not an error for
14097 @var{symbol} to never be defined at all.
14099 @item #pragma weak @var{symbol1} = @var{symbol2}
14100 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
14101 It is an error if @var{symbol2} is not defined in the current
14105 @node Diagnostic Pragmas
14106 @subsection Diagnostic Pragmas
14108 GCC allows the user to selectively enable or disable certain types of
14109 diagnostics, and change the kind of the diagnostic. For example, a
14110 project's policy might require that all sources compile with
14111 @option{-Werror} but certain files might have exceptions allowing
14112 specific types of warnings. Or, a project might selectively enable
14113 diagnostics and treat them as errors depending on which preprocessor
14114 macros are defined.
14117 @item #pragma GCC diagnostic @var{kind} @var{option}
14118 @cindex pragma, diagnostic
14120 Modifies the disposition of a diagnostic. Note that not all
14121 diagnostics are modifiable; at the moment only warnings (normally
14122 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
14123 Use @option{-fdiagnostics-show-option} to determine which diagnostics
14124 are controllable and which option controls them.
14126 @var{kind} is @samp{error} to treat this diagnostic as an error,
14127 @samp{warning} to treat it like a warning (even if @option{-Werror} is
14128 in effect), or @samp{ignored} if the diagnostic is to be ignored.
14129 @var{option} is a double quoted string which matches the command-line
14133 #pragma GCC diagnostic warning "-Wformat"
14134 #pragma GCC diagnostic error "-Wformat"
14135 #pragma GCC diagnostic ignored "-Wformat"
14138 Note that these pragmas override any command-line options. GCC keeps
14139 track of the location of each pragma, and issues diagnostics according
14140 to the state as of that point in the source file. Thus, pragmas occurring
14141 after a line do not affect diagnostics caused by that line.
14143 @item #pragma GCC diagnostic push
14144 @itemx #pragma GCC diagnostic pop
14146 Causes GCC to remember the state of the diagnostics as of each
14147 @code{push}, and restore to that point at each @code{pop}. If a
14148 @code{pop} has no matching @code{push}, the command line options are
14152 #pragma GCC diagnostic error "-Wuninitialized"
14153 foo(a); /* error is given for this one */
14154 #pragma GCC diagnostic push
14155 #pragma GCC diagnostic ignored "-Wuninitialized"
14156 foo(b); /* no diagnostic for this one */
14157 #pragma GCC diagnostic pop
14158 foo(c); /* error is given for this one */
14159 #pragma GCC diagnostic pop
14160 foo(d); /* depends on command line options */
14165 GCC also offers a simple mechanism for printing messages during
14169 @item #pragma message @var{string}
14170 @cindex pragma, diagnostic
14172 Prints @var{string} as a compiler message on compilation. The message
14173 is informational only, and is neither a compilation warning nor an error.
14176 #pragma message "Compiling " __FILE__ "..."
14179 @var{string} may be parenthesized, and is printed with location
14180 information. For example,
14183 #define DO_PRAGMA(x) _Pragma (#x)
14184 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
14186 TODO(Remember to fix this)
14189 prints @samp{/tmp/file.c:4: note: #pragma message:
14190 TODO - Remember to fix this}.
14194 @node Visibility Pragmas
14195 @subsection Visibility Pragmas
14198 @item #pragma GCC visibility push(@var{visibility})
14199 @itemx #pragma GCC visibility pop
14200 @cindex pragma, visibility
14202 This pragma allows the user to set the visibility for multiple
14203 declarations without having to give each a visibility attribute
14204 @xref{Function Attributes}, for more information about visibility and
14205 the attribute syntax.
14207 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
14208 declarations. Class members and template specializations are not
14209 affected; if you want to override the visibility for a particular
14210 member or instantiation, you must use an attribute.
14215 @node Push/Pop Macro Pragmas
14216 @subsection Push/Pop Macro Pragmas
14218 For compatibility with Microsoft Windows compilers, GCC supports
14219 @samp{#pragma push_macro(@var{"macro_name"})}
14220 and @samp{#pragma pop_macro(@var{"macro_name"})}.
14223 @item #pragma push_macro(@var{"macro_name"})
14224 @cindex pragma, push_macro
14225 This pragma saves the value of the macro named as @var{macro_name} to
14226 the top of the stack for this macro.
14228 @item #pragma pop_macro(@var{"macro_name"})
14229 @cindex pragma, pop_macro
14230 This pragma sets the value of the macro named as @var{macro_name} to
14231 the value on top of the stack for this macro. If the stack for
14232 @var{macro_name} is empty, the value of the macro remains unchanged.
14239 #pragma push_macro("X")
14242 #pragma pop_macro("X")
14246 In this example, the definition of X as 1 is saved by @code{#pragma
14247 push_macro} and restored by @code{#pragma pop_macro}.
14249 @node Function Specific Option Pragmas
14250 @subsection Function Specific Option Pragmas
14253 @item #pragma GCC target (@var{"string"}...)
14254 @cindex pragma GCC target
14256 This pragma allows you to set target specific options for functions
14257 defined later in the source file. One or more strings can be
14258 specified. Each function that is defined after this point will be as
14259 if @code{attribute((target("STRING")))} was specified for that
14260 function. The parenthesis around the options is optional.
14261 @xref{Function Attributes}, for more information about the
14262 @code{target} attribute and the attribute syntax.
14264 The @code{#pragma GCC target} attribute is not implemented in GCC versions earlier
14265 than 4.4 for the i386/x86_64 and 4.6 for the PowerPC backends. At
14266 present, it is not implemented for other backends.
14270 @item #pragma GCC optimize (@var{"string"}...)
14271 @cindex pragma GCC optimize
14273 This pragma allows you to set global optimization options for functions
14274 defined later in the source file. One or more strings can be
14275 specified. Each function that is defined after this point will be as
14276 if @code{attribute((optimize("STRING")))} was specified for that
14277 function. The parenthesis around the options is optional.
14278 @xref{Function Attributes}, for more information about the
14279 @code{optimize} attribute and the attribute syntax.
14281 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
14282 versions earlier than 4.4.
14286 @item #pragma GCC push_options
14287 @itemx #pragma GCC pop_options
14288 @cindex pragma GCC push_options
14289 @cindex pragma GCC pop_options
14291 These pragmas maintain a stack of the current target and optimization
14292 options. It is intended for include files where you temporarily want
14293 to switch to using a different @samp{#pragma GCC target} or
14294 @samp{#pragma GCC optimize} and then to pop back to the previous
14297 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
14298 pragmas are not implemented in GCC versions earlier than 4.4.
14302 @item #pragma GCC reset_options
14303 @cindex pragma GCC reset_options
14305 This pragma clears the current @code{#pragma GCC target} and
14306 @code{#pragma GCC optimize} to use the default switches as specified
14307 on the command line.
14309 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
14310 versions earlier than 4.4.
14313 @node Unnamed Fields
14314 @section Unnamed struct/union fields within structs/unions
14315 @cindex @code{struct}
14316 @cindex @code{union}
14318 As permitted by ISO C11 and for compatibility with other compilers,
14319 GCC allows you to define
14320 a structure or union that contains, as fields, structures and unions
14321 without names. For example:
14334 In this example, the user would be able to access members of the unnamed
14335 union with code like @samp{foo.b}. Note that only unnamed structs and
14336 unions are allowed, you may not have, for example, an unnamed
14339 You must never create such structures that cause ambiguous field definitions.
14340 For example, this structure:
14351 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
14352 The compiler gives errors for such constructs.
14354 @opindex fms-extensions
14355 Unless @option{-fms-extensions} is used, the unnamed field must be a
14356 structure or union definition without a tag (for example, @samp{struct
14357 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
14358 also be a definition with a tag such as @samp{struct foo @{ int a;
14359 @};}, a reference to a previously defined structure or union such as
14360 @samp{struct foo;}, or a reference to a @code{typedef} name for a
14361 previously defined structure or union type.
14363 @opindex fplan9-extensions
14364 The option @option{-fplan9-extensions} enables
14365 @option{-fms-extensions} as well as two other extensions. First, a
14366 pointer to a structure is automatically converted to a pointer to an
14367 anonymous field for assignments and function calls. For example:
14370 struct s1 @{ int a; @};
14371 struct s2 @{ struct s1; @};
14372 extern void f1 (struct s1 *);
14373 void f2 (struct s2 *p) @{ f1 (p); @}
14376 In the call to @code{f1} inside @code{f2}, the pointer @code{p} is
14377 converted into a pointer to the anonymous field.
14379 Second, when the type of an anonymous field is a @code{typedef} for a
14380 @code{struct} or @code{union}, code may refer to the field using the
14381 name of the @code{typedef}.
14384 typedef struct @{ int a; @} s1;
14385 struct s2 @{ s1; @};
14386 s1 f1 (struct s2 *p) @{ return p->s1; @}
14389 These usages are only permitted when they are not ambiguous.
14392 @section Thread-Local Storage
14393 @cindex Thread-Local Storage
14394 @cindex @acronym{TLS}
14395 @cindex @code{__thread}
14397 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
14398 are allocated such that there is one instance of the variable per extant
14399 thread. The run-time model GCC uses to implement this originates
14400 in the IA-64 processor-specific ABI, but has since been migrated
14401 to other processors as well. It requires significant support from
14402 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
14403 system libraries (@file{libc.so} and @file{libpthread.so}), so it
14404 is not available everywhere.
14406 At the user level, the extension is visible with a new storage
14407 class keyword: @code{__thread}. For example:
14411 extern __thread struct state s;
14412 static __thread char *p;
14415 The @code{__thread} specifier may be used alone, with the @code{extern}
14416 or @code{static} specifiers, but with no other storage class specifier.
14417 When used with @code{extern} or @code{static}, @code{__thread} must appear
14418 immediately after the other storage class specifier.
14420 The @code{__thread} specifier may be applied to any global, file-scoped
14421 static, function-scoped static, or static data member of a class. It may
14422 not be applied to block-scoped automatic or non-static data member.
14424 When the address-of operator is applied to a thread-local variable, it is
14425 evaluated at run-time and returns the address of the current thread's
14426 instance of that variable. An address so obtained may be used by any
14427 thread. When a thread terminates, any pointers to thread-local variables
14428 in that thread become invalid.
14430 No static initialization may refer to the address of a thread-local variable.
14432 In C++, if an initializer is present for a thread-local variable, it must
14433 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
14436 See @uref{http://www.akkadia.org/drepper/tls.pdf,
14437 ELF Handling For Thread-Local Storage} for a detailed explanation of
14438 the four thread-local storage addressing models, and how the run-time
14439 is expected to function.
14442 * C99 Thread-Local Edits::
14443 * C++98 Thread-Local Edits::
14446 @node C99 Thread-Local Edits
14447 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
14449 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
14450 that document the exact semantics of the language extension.
14454 @cite{5.1.2 Execution environments}
14456 Add new text after paragraph 1
14459 Within either execution environment, a @dfn{thread} is a flow of
14460 control within a program. It is implementation defined whether
14461 or not there may be more than one thread associated with a program.
14462 It is implementation defined how threads beyond the first are
14463 created, the name and type of the function called at thread
14464 startup, and how threads may be terminated. However, objects
14465 with thread storage duration shall be initialized before thread
14470 @cite{6.2.4 Storage durations of objects}
14472 Add new text before paragraph 3
14475 An object whose identifier is declared with the storage-class
14476 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
14477 Its lifetime is the entire execution of the thread, and its
14478 stored value is initialized only once, prior to thread startup.
14482 @cite{6.4.1 Keywords}
14484 Add @code{__thread}.
14487 @cite{6.7.1 Storage-class specifiers}
14489 Add @code{__thread} to the list of storage class specifiers in
14492 Change paragraph 2 to
14495 With the exception of @code{__thread}, at most one storage-class
14496 specifier may be given [@dots{}]. The @code{__thread} specifier may
14497 be used alone, or immediately following @code{extern} or
14501 Add new text after paragraph 6
14504 The declaration of an identifier for a variable that has
14505 block scope that specifies @code{__thread} shall also
14506 specify either @code{extern} or @code{static}.
14508 The @code{__thread} specifier shall be used only with
14513 @node C++98 Thread-Local Edits
14514 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
14516 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
14517 that document the exact semantics of the language extension.
14521 @b{[intro.execution]}
14523 New text after paragraph 4
14526 A @dfn{thread} is a flow of control within the abstract machine.
14527 It is implementation defined whether or not there may be more than
14531 New text after paragraph 7
14534 It is unspecified whether additional action must be taken to
14535 ensure when and whether side effects are visible to other threads.
14541 Add @code{__thread}.
14544 @b{[basic.start.main]}
14546 Add after paragraph 5
14549 The thread that begins execution at the @code{main} function is called
14550 the @dfn{main thread}. It is implementation defined how functions
14551 beginning threads other than the main thread are designated or typed.
14552 A function so designated, as well as the @code{main} function, is called
14553 a @dfn{thread startup function}. It is implementation defined what
14554 happens if a thread startup function returns. It is implementation
14555 defined what happens to other threads when any thread calls @code{exit}.
14559 @b{[basic.start.init]}
14561 Add after paragraph 4
14564 The storage for an object of thread storage duration shall be
14565 statically initialized before the first statement of the thread startup
14566 function. An object of thread storage duration shall not require
14567 dynamic initialization.
14571 @b{[basic.start.term]}
14573 Add after paragraph 3
14576 The type of an object with thread storage duration shall not have a
14577 non-trivial destructor, nor shall it be an array type whose elements
14578 (directly or indirectly) have non-trivial destructors.
14584 Add ``thread storage duration'' to the list in paragraph 1.
14589 Thread, static, and automatic storage durations are associated with
14590 objects introduced by declarations [@dots{}].
14593 Add @code{__thread} to the list of specifiers in paragraph 3.
14596 @b{[basic.stc.thread]}
14598 New section before @b{[basic.stc.static]}
14601 The keyword @code{__thread} applied to a non-local object gives the
14602 object thread storage duration.
14604 A local variable or class data member declared both @code{static}
14605 and @code{__thread} gives the variable or member thread storage
14610 @b{[basic.stc.static]}
14615 All objects which have neither thread storage duration, dynamic
14616 storage duration nor are local [@dots{}].
14622 Add @code{__thread} to the list in paragraph 1.
14627 With the exception of @code{__thread}, at most one
14628 @var{storage-class-specifier} shall appear in a given
14629 @var{decl-specifier-seq}. The @code{__thread} specifier may
14630 be used alone, or immediately following the @code{extern} or
14631 @code{static} specifiers. [@dots{}]
14634 Add after paragraph 5
14637 The @code{__thread} specifier can be applied only to the names of objects
14638 and to anonymous unions.
14644 Add after paragraph 6
14647 Non-@code{static} members shall not be @code{__thread}.
14651 @node Binary constants
14652 @section Binary constants using the @samp{0b} prefix
14653 @cindex Binary constants using the @samp{0b} prefix
14655 Integer constants can be written as binary constants, consisting of a
14656 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
14657 @samp{0B}. This is particularly useful in environments that operate a
14658 lot on the bit-level (like microcontrollers).
14660 The following statements are identical:
14669 The type of these constants follows the same rules as for octal or
14670 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
14673 @node C++ Extensions
14674 @chapter Extensions to the C++ Language
14675 @cindex extensions, C++ language
14676 @cindex C++ language extensions
14678 The GNU compiler provides these extensions to the C++ language (and you
14679 can also use most of the C language extensions in your C++ programs). If you
14680 want to write code that checks whether these features are available, you can
14681 test for the GNU compiler the same way as for C programs: check for a
14682 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
14683 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
14684 Predefined Macros,cpp,The GNU C Preprocessor}).
14687 * C++ Volatiles:: What constitutes an access to a volatile object.
14688 * Restricted Pointers:: C99 restricted pointers and references.
14689 * Vague Linkage:: Where G++ puts inlines, vtables and such.
14690 * C++ Interface:: You can use a single C++ header file for both
14691 declarations and definitions.
14692 * Template Instantiation:: Methods for ensuring that exactly one copy of
14693 each needed template instantiation is emitted.
14694 * Bound member functions:: You can extract a function pointer to the
14695 method denoted by a @samp{->*} or @samp{.*} expression.
14696 * C++ Attributes:: Variable, function, and type attributes for C++ only.
14697 * Namespace Association:: Strong using-directives for namespace association.
14698 * Type Traits:: Compiler support for type traits
14699 * Java Exceptions:: Tweaking exception handling to work with Java.
14700 * Deprecated Features:: Things will disappear from g++.
14701 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
14704 @node C++ Volatiles
14705 @section When is a Volatile C++ Object Accessed?
14706 @cindex accessing volatiles
14707 @cindex volatile read
14708 @cindex volatile write
14709 @cindex volatile access
14711 The C++ standard differs from the C standard in its treatment of
14712 volatile objects. It fails to specify what constitutes a volatile
14713 access, except to say that C++ should behave in a similar manner to C
14714 with respect to volatiles, where possible. However, the different
14715 lvalueness of expressions between C and C++ complicate the behavior.
14716 G++ behaves the same as GCC for volatile access, @xref{C
14717 Extensions,,Volatiles}, for a description of GCC's behavior.
14719 The C and C++ language specifications differ when an object is
14720 accessed in a void context:
14723 volatile int *src = @var{somevalue};
14727 The C++ standard specifies that such expressions do not undergo lvalue
14728 to rvalue conversion, and that the type of the dereferenced object may
14729 be incomplete. The C++ standard does not specify explicitly that it
14730 is lvalue to rvalue conversion which is responsible for causing an
14731 access. There is reason to believe that it is, because otherwise
14732 certain simple expressions become undefined. However, because it
14733 would surprise most programmers, G++ treats dereferencing a pointer to
14734 volatile object of complete type as GCC would do for an equivalent
14735 type in C@. When the object has incomplete type, G++ issues a
14736 warning; if you wish to force an error, you must force a conversion to
14737 rvalue with, for instance, a static cast.
14739 When using a reference to volatile, G++ does not treat equivalent
14740 expressions as accesses to volatiles, but instead issues a warning that
14741 no volatile is accessed. The rationale for this is that otherwise it
14742 becomes difficult to determine where volatile access occur, and not
14743 possible to ignore the return value from functions returning volatile
14744 references. Again, if you wish to force a read, cast the reference to
14747 G++ implements the same behavior as GCC does when assigning to a
14748 volatile object -- there is no reread of the assigned-to object, the
14749 assigned rvalue is reused. Note that in C++ assignment expressions
14750 are lvalues, and if used as an lvalue, the volatile object will be
14751 referred to. For instance, @var{vref} will refer to @var{vobj}, as
14752 expected, in the following example:
14756 volatile int &vref = vobj = @var{something};
14759 @node Restricted Pointers
14760 @section Restricting Pointer Aliasing
14761 @cindex restricted pointers
14762 @cindex restricted references
14763 @cindex restricted this pointer
14765 As with the C front end, G++ understands the C99 feature of restricted pointers,
14766 specified with the @code{__restrict__}, or @code{__restrict} type
14767 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
14768 language flag, @code{restrict} is not a keyword in C++.
14770 In addition to allowing restricted pointers, you can specify restricted
14771 references, which indicate that the reference is not aliased in the local
14775 void fn (int *__restrict__ rptr, int &__restrict__ rref)
14782 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
14783 @var{rref} refers to a (different) unaliased integer.
14785 You may also specify whether a member function's @var{this} pointer is
14786 unaliased by using @code{__restrict__} as a member function qualifier.
14789 void T::fn () __restrict__
14796 Within the body of @code{T::fn}, @var{this} will have the effective
14797 definition @code{T *__restrict__ const this}. Notice that the
14798 interpretation of a @code{__restrict__} member function qualifier is
14799 different to that of @code{const} or @code{volatile} qualifier, in that it
14800 is applied to the pointer rather than the object. This is consistent with
14801 other compilers which implement restricted pointers.
14803 As with all outermost parameter qualifiers, @code{__restrict__} is
14804 ignored in function definition matching. This means you only need to
14805 specify @code{__restrict__} in a function definition, rather than
14806 in a function prototype as well.
14808 @node Vague Linkage
14809 @section Vague Linkage
14810 @cindex vague linkage
14812 There are several constructs in C++ which require space in the object
14813 file but are not clearly tied to a single translation unit. We say that
14814 these constructs have ``vague linkage''. Typically such constructs are
14815 emitted wherever they are needed, though sometimes we can be more
14819 @item Inline Functions
14820 Inline functions are typically defined in a header file which can be
14821 included in many different compilations. Hopefully they can usually be
14822 inlined, but sometimes an out-of-line copy is necessary, if the address
14823 of the function is taken or if inlining fails. In general, we emit an
14824 out-of-line copy in all translation units where one is needed. As an
14825 exception, we only emit inline virtual functions with the vtable, since
14826 it will always require a copy.
14828 Local static variables and string constants used in an inline function
14829 are also considered to have vague linkage, since they must be shared
14830 between all inlined and out-of-line instances of the function.
14834 C++ virtual functions are implemented in most compilers using a lookup
14835 table, known as a vtable. The vtable contains pointers to the virtual
14836 functions provided by a class, and each object of the class contains a
14837 pointer to its vtable (or vtables, in some multiple-inheritance
14838 situations). If the class declares any non-inline, non-pure virtual
14839 functions, the first one is chosen as the ``key method'' for the class,
14840 and the vtable is only emitted in the translation unit where the key
14843 @emph{Note:} If the chosen key method is later defined as inline, the
14844 vtable will still be emitted in every translation unit which defines it.
14845 Make sure that any inline virtuals are declared inline in the class
14846 body, even if they are not defined there.
14848 @item @code{type_info} objects
14849 @cindex @code{type_info}
14851 C++ requires information about types to be written out in order to
14852 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
14853 For polymorphic classes (classes with virtual functions), the @samp{type_info}
14854 object is written out along with the vtable so that @samp{dynamic_cast}
14855 can determine the dynamic type of a class object at runtime. For all
14856 other types, we write out the @samp{type_info} object when it is used: when
14857 applying @samp{typeid} to an expression, throwing an object, or
14858 referring to a type in a catch clause or exception specification.
14860 @item Template Instantiations
14861 Most everything in this section also applies to template instantiations,
14862 but there are other options as well.
14863 @xref{Template Instantiation,,Where's the Template?}.
14867 When used with GNU ld version 2.8 or later on an ELF system such as
14868 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
14869 these constructs will be discarded at link time. This is known as
14872 On targets that don't support COMDAT, but do support weak symbols, GCC
14873 will use them. This way one copy will override all the others, but
14874 the unused copies will still take up space in the executable.
14876 For targets which do not support either COMDAT or weak symbols,
14877 most entities with vague linkage will be emitted as local symbols to
14878 avoid duplicate definition errors from the linker. This will not happen
14879 for local statics in inlines, however, as having multiple copies will
14880 almost certainly break things.
14882 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
14883 another way to control placement of these constructs.
14885 @node C++ Interface
14886 @section #pragma interface and implementation
14888 @cindex interface and implementation headers, C++
14889 @cindex C++ interface and implementation headers
14890 @cindex pragmas, interface and implementation
14892 @code{#pragma interface} and @code{#pragma implementation} provide the
14893 user with a way of explicitly directing the compiler to emit entities
14894 with vague linkage (and debugging information) in a particular
14897 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
14898 most cases, because of COMDAT support and the ``key method'' heuristic
14899 mentioned in @ref{Vague Linkage}. Using them can actually cause your
14900 program to grow due to unnecessary out-of-line copies of inline
14901 functions. Currently (3.4) the only benefit of these
14902 @code{#pragma}s is reduced duplication of debugging information, and
14903 that should be addressed soon on DWARF 2 targets with the use of
14907 @item #pragma interface
14908 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
14909 @kindex #pragma interface
14910 Use this directive in @emph{header files} that define object classes, to save
14911 space in most of the object files that use those classes. Normally,
14912 local copies of certain information (backup copies of inline member
14913 functions, debugging information, and the internal tables that implement
14914 virtual functions) must be kept in each object file that includes class
14915 definitions. You can use this pragma to avoid such duplication. When a
14916 header file containing @samp{#pragma interface} is included in a
14917 compilation, this auxiliary information will not be generated (unless
14918 the main input source file itself uses @samp{#pragma implementation}).
14919 Instead, the object files will contain references to be resolved at link
14922 The second form of this directive is useful for the case where you have
14923 multiple headers with the same name in different directories. If you
14924 use this form, you must specify the same string to @samp{#pragma
14927 @item #pragma implementation
14928 @itemx #pragma implementation "@var{objects}.h"
14929 @kindex #pragma implementation
14930 Use this pragma in a @emph{main input file}, when you want full output from
14931 included header files to be generated (and made globally visible). The
14932 included header file, in turn, should use @samp{#pragma interface}.
14933 Backup copies of inline member functions, debugging information, and the
14934 internal tables used to implement virtual functions are all generated in
14935 implementation files.
14937 @cindex implied @code{#pragma implementation}
14938 @cindex @code{#pragma implementation}, implied
14939 @cindex naming convention, implementation headers
14940 If you use @samp{#pragma implementation} with no argument, it applies to
14941 an include file with the same basename@footnote{A file's @dfn{basename}
14942 was the name stripped of all leading path information and of trailing
14943 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
14944 file. For example, in @file{allclass.cc}, giving just
14945 @samp{#pragma implementation}
14946 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
14948 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
14949 an implementation file whenever you would include it from
14950 @file{allclass.cc} even if you never specified @samp{#pragma
14951 implementation}. This was deemed to be more trouble than it was worth,
14952 however, and disabled.
14954 Use the string argument if you want a single implementation file to
14955 include code from multiple header files. (You must also use
14956 @samp{#include} to include the header file; @samp{#pragma
14957 implementation} only specifies how to use the file---it doesn't actually
14960 There is no way to split up the contents of a single header file into
14961 multiple implementation files.
14964 @cindex inlining and C++ pragmas
14965 @cindex C++ pragmas, effect on inlining
14966 @cindex pragmas in C++, effect on inlining
14967 @samp{#pragma implementation} and @samp{#pragma interface} also have an
14968 effect on function inlining.
14970 If you define a class in a header file marked with @samp{#pragma
14971 interface}, the effect on an inline function defined in that class is
14972 similar to an explicit @code{extern} declaration---the compiler emits
14973 no code at all to define an independent version of the function. Its
14974 definition is used only for inlining with its callers.
14976 @opindex fno-implement-inlines
14977 Conversely, when you include the same header file in a main source file
14978 that declares it as @samp{#pragma implementation}, the compiler emits
14979 code for the function itself; this defines a version of the function
14980 that can be found via pointers (or by callers compiled without
14981 inlining). If all calls to the function can be inlined, you can avoid
14982 emitting the function by compiling with @option{-fno-implement-inlines}.
14983 If any calls were not inlined, you will get linker errors.
14985 @node Template Instantiation
14986 @section Where's the Template?
14987 @cindex template instantiation
14989 C++ templates are the first language feature to require more
14990 intelligence from the environment than one usually finds on a UNIX
14991 system. Somehow the compiler and linker have to make sure that each
14992 template instance occurs exactly once in the executable if it is needed,
14993 and not at all otherwise. There are two basic approaches to this
14994 problem, which are referred to as the Borland model and the Cfront model.
14997 @item Borland model
14998 Borland C++ solved the template instantiation problem by adding the code
14999 equivalent of common blocks to their linker; the compiler emits template
15000 instances in each translation unit that uses them, and the linker
15001 collapses them together. The advantage of this model is that the linker
15002 only has to consider the object files themselves; there is no external
15003 complexity to worry about. This disadvantage is that compilation time
15004 is increased because the template code is being compiled repeatedly.
15005 Code written for this model tends to include definitions of all
15006 templates in the header file, since they must be seen to be
15010 The AT&T C++ translator, Cfront, solved the template instantiation
15011 problem by creating the notion of a template repository, an
15012 automatically maintained place where template instances are stored. A
15013 more modern version of the repository works as follows: As individual
15014 object files are built, the compiler places any template definitions and
15015 instantiations encountered in the repository. At link time, the link
15016 wrapper adds in the objects in the repository and compiles any needed
15017 instances that were not previously emitted. The advantages of this
15018 model are more optimal compilation speed and the ability to use the
15019 system linker; to implement the Borland model a compiler vendor also
15020 needs to replace the linker. The disadvantages are vastly increased
15021 complexity, and thus potential for error; for some code this can be
15022 just as transparent, but in practice it can been very difficult to build
15023 multiple programs in one directory and one program in multiple
15024 directories. Code written for this model tends to separate definitions
15025 of non-inline member templates into a separate file, which should be
15026 compiled separately.
15029 When used with GNU ld version 2.8 or later on an ELF system such as
15030 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
15031 Borland model. On other systems, G++ implements neither automatic
15034 A future version of G++ will support a hybrid model whereby the compiler
15035 will emit any instantiations for which the template definition is
15036 included in the compile, and store template definitions and
15037 instantiation context information into the object file for the rest.
15038 The link wrapper will extract that information as necessary and invoke
15039 the compiler to produce the remaining instantiations. The linker will
15040 then combine duplicate instantiations.
15042 In the mean time, you have the following options for dealing with
15043 template instantiations:
15048 Compile your template-using code with @option{-frepo}. The compiler will
15049 generate files with the extension @samp{.rpo} listing all of the
15050 template instantiations used in the corresponding object files which
15051 could be instantiated there; the link wrapper, @samp{collect2}, will
15052 then update the @samp{.rpo} files to tell the compiler where to place
15053 those instantiations and rebuild any affected object files. The
15054 link-time overhead is negligible after the first pass, as the compiler
15055 will continue to place the instantiations in the same files.
15057 This is your best option for application code written for the Borland
15058 model, as it will just work. Code written for the Cfront model will
15059 need to be modified so that the template definitions are available at
15060 one or more points of instantiation; usually this is as simple as adding
15061 @code{#include <tmethods.cc>} to the end of each template header.
15063 For library code, if you want the library to provide all of the template
15064 instantiations it needs, just try to link all of its object files
15065 together; the link will fail, but cause the instantiations to be
15066 generated as a side effect. Be warned, however, that this may cause
15067 conflicts if multiple libraries try to provide the same instantiations.
15068 For greater control, use explicit instantiation as described in the next
15072 @opindex fno-implicit-templates
15073 Compile your code with @option{-fno-implicit-templates} to disable the
15074 implicit generation of template instances, and explicitly instantiate
15075 all the ones you use. This approach requires more knowledge of exactly
15076 which instances you need than do the others, but it's less
15077 mysterious and allows greater control. You can scatter the explicit
15078 instantiations throughout your program, perhaps putting them in the
15079 translation units where the instances are used or the translation units
15080 that define the templates themselves; you can put all of the explicit
15081 instantiations you need into one big file; or you can create small files
15088 template class Foo<int>;
15089 template ostream& operator <<
15090 (ostream&, const Foo<int>&);
15093 for each of the instances you need, and create a template instantiation
15094 library from those.
15096 If you are using Cfront-model code, you can probably get away with not
15097 using @option{-fno-implicit-templates} when compiling files that don't
15098 @samp{#include} the member template definitions.
15100 If you use one big file to do the instantiations, you may want to
15101 compile it without @option{-fno-implicit-templates} so you get all of the
15102 instances required by your explicit instantiations (but not by any
15103 other files) without having to specify them as well.
15105 G++ has extended the template instantiation syntax given in the ISO
15106 standard to allow forward declaration of explicit instantiations
15107 (with @code{extern}), instantiation of the compiler support data for a
15108 template class (i.e.@: the vtable) without instantiating any of its
15109 members (with @code{inline}), and instantiation of only the static data
15110 members of a template class, without the support data or member
15111 functions (with (@code{static}):
15114 extern template int max (int, int);
15115 inline template class Foo<int>;
15116 static template class Foo<int>;
15120 Do nothing. Pretend G++ does implement automatic instantiation
15121 management. Code written for the Borland model will work fine, but
15122 each translation unit will contain instances of each of the templates it
15123 uses. In a large program, this can lead to an unacceptable amount of code
15127 @node Bound member functions
15128 @section Extracting the function pointer from a bound pointer to member function
15130 @cindex pointer to member function
15131 @cindex bound pointer to member function
15133 In C++, pointer to member functions (PMFs) are implemented using a wide
15134 pointer of sorts to handle all the possible call mechanisms; the PMF
15135 needs to store information about how to adjust the @samp{this} pointer,
15136 and if the function pointed to is virtual, where to find the vtable, and
15137 where in the vtable to look for the member function. If you are using
15138 PMFs in an inner loop, you should really reconsider that decision. If
15139 that is not an option, you can extract the pointer to the function that
15140 would be called for a given object/PMF pair and call it directly inside
15141 the inner loop, to save a bit of time.
15143 Note that you will still be paying the penalty for the call through a
15144 function pointer; on most modern architectures, such a call defeats the
15145 branch prediction features of the CPU@. This is also true of normal
15146 virtual function calls.
15148 The syntax for this extension is
15152 extern int (A::*fp)();
15153 typedef int (*fptr)(A *);
15155 fptr p = (fptr)(a.*fp);
15158 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
15159 no object is needed to obtain the address of the function. They can be
15160 converted to function pointers directly:
15163 fptr p1 = (fptr)(&A::foo);
15166 @opindex Wno-pmf-conversions
15167 You must specify @option{-Wno-pmf-conversions} to use this extension.
15169 @node C++ Attributes
15170 @section C++-Specific Variable, Function, and Type Attributes
15172 Some attributes only make sense for C++ programs.
15175 @item init_priority (@var{priority})
15176 @cindex @code{init_priority} attribute
15179 In Standard C++, objects defined at namespace scope are guaranteed to be
15180 initialized in an order in strict accordance with that of their definitions
15181 @emph{in a given translation unit}. No guarantee is made for initializations
15182 across translation units. However, GNU C++ allows users to control the
15183 order of initialization of objects defined at namespace scope with the
15184 @code{init_priority} attribute by specifying a relative @var{priority},
15185 a constant integral expression currently bounded between 101 and 65535
15186 inclusive. Lower numbers indicate a higher priority.
15188 In the following example, @code{A} would normally be created before
15189 @code{B}, but the @code{init_priority} attribute has reversed that order:
15192 Some_Class A __attribute__ ((init_priority (2000)));
15193 Some_Class B __attribute__ ((init_priority (543)));
15197 Note that the particular values of @var{priority} do not matter; only their
15200 @item java_interface
15201 @cindex @code{java_interface} attribute
15203 This type attribute informs C++ that the class is a Java interface. It may
15204 only be applied to classes declared within an @code{extern "Java"} block.
15205 Calls to methods declared in this interface will be dispatched using GCJ's
15206 interface table mechanism, instead of regular virtual table dispatch.
15210 See also @ref{Namespace Association}.
15212 @node Namespace Association
15213 @section Namespace Association
15215 @strong{Caution:} The semantics of this extension are not fully
15216 defined. Users should refrain from using this extension as its
15217 semantics may change subtly over time. It is possible that this
15218 extension will be removed in future versions of G++.
15220 A using-directive with @code{__attribute ((strong))} is stronger
15221 than a normal using-directive in two ways:
15225 Templates from the used namespace can be specialized and explicitly
15226 instantiated as though they were members of the using namespace.
15229 The using namespace is considered an associated namespace of all
15230 templates in the used namespace for purposes of argument-dependent
15234 The used namespace must be nested within the using namespace so that
15235 normal unqualified lookup works properly.
15237 This is useful for composing a namespace transparently from
15238 implementation namespaces. For example:
15243 template <class T> struct A @{ @};
15245 using namespace debug __attribute ((__strong__));
15246 template <> struct A<int> @{ @}; // @r{ok to specialize}
15248 template <class T> void f (A<T>);
15253 f (std::A<float>()); // @r{lookup finds} std::f
15259 @section Type Traits
15261 The C++ front-end implements syntactic extensions that allow to
15262 determine at compile time various characteristics of a type (or of a
15266 @item __has_nothrow_assign (type)
15267 If @code{type} is const qualified or is a reference type then the trait is
15268 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
15269 is true, else if @code{type} is a cv class or union type with copy assignment
15270 operators that are known not to throw an exception then the trait is true,
15271 else it is false. Requires: @code{type} shall be a complete type,
15272 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15274 @item __has_nothrow_copy (type)
15275 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
15276 @code{type} is a cv class or union type with copy constructors that
15277 are known not to throw an exception then the trait is true, else it is false.
15278 Requires: @code{type} shall be a complete type, (possibly cv-qualified)
15279 @code{void}, or an array of unknown bound.
15281 @item __has_nothrow_constructor (type)
15282 If @code{__has_trivial_constructor (type)} is true then the trait is
15283 true, else if @code{type} is a cv class or union type (or array
15284 thereof) with a default constructor that is known not to throw an
15285 exception then the trait is true, else it is false. Requires:
15286 @code{type} shall be a complete type, (possibly cv-qualified)
15287 @code{void}, or an array of unknown bound.
15289 @item __has_trivial_assign (type)
15290 If @code{type} is const qualified or is a reference type then the trait is
15291 false. Otherwise if @code{__is_pod (type)} is true then the trait is
15292 true, else if @code{type} is a cv class or union type with a trivial
15293 copy assignment ([class.copy]) then the trait is true, else it is
15294 false. Requires: @code{type} shall be a complete type, (possibly
15295 cv-qualified) @code{void}, or an array of unknown bound.
15297 @item __has_trivial_copy (type)
15298 If @code{__is_pod (type)} is true or @code{type} is a reference type
15299 then the trait is true, else if @code{type} is a cv class or union type
15300 with a trivial copy constructor ([class.copy]) then the trait
15301 is true, else it is false. Requires: @code{type} shall be a complete
15302 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15304 @item __has_trivial_constructor (type)
15305 If @code{__is_pod (type)} is true then the trait is true, else if
15306 @code{type} is a cv class or union type (or array thereof) with a
15307 trivial default constructor ([class.ctor]) then the trait is true,
15308 else it is false. Requires: @code{type} shall be a complete
15309 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15311 @item __has_trivial_destructor (type)
15312 If @code{__is_pod (type)} is true or @code{type} is a reference type then
15313 the trait is true, else if @code{type} is a cv class or union type (or
15314 array thereof) with a trivial destructor ([class.dtor]) then the trait
15315 is true, else it is false. Requires: @code{type} shall be a complete
15316 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15318 @item __has_virtual_destructor (type)
15319 If @code{type} is a class type with a virtual destructor
15320 ([class.dtor]) then the trait is true, else it is false. Requires:
15321 @code{type} shall be a complete type, (possibly cv-qualified)
15322 @code{void}, or an array of unknown bound.
15324 @item __is_abstract (type)
15325 If @code{type} is an abstract class ([class.abstract]) then the trait
15326 is true, else it is false. Requires: @code{type} shall be a complete
15327 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15329 @item __is_base_of (base_type, derived_type)
15330 If @code{base_type} is a base class of @code{derived_type}
15331 ([class.derived]) then the trait is true, otherwise it is false.
15332 Top-level cv qualifications of @code{base_type} and
15333 @code{derived_type} are ignored. For the purposes of this trait, a
15334 class type is considered is own base. Requires: if @code{__is_class
15335 (base_type)} and @code{__is_class (derived_type)} are true and
15336 @code{base_type} and @code{derived_type} are not the same type
15337 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
15338 type. Diagnostic is produced if this requirement is not met.
15340 @item __is_class (type)
15341 If @code{type} is a cv class type, and not a union type
15342 ([basic.compound]) the trait is true, else it is false.
15344 @item __is_empty (type)
15345 If @code{__is_class (type)} is false then the trait is false.
15346 Otherwise @code{type} is considered empty if and only if: @code{type}
15347 has no non-static data members, or all non-static data members, if
15348 any, are bit-fields of length 0, and @code{type} has no virtual
15349 members, and @code{type} has no virtual base classes, and @code{type}
15350 has no base classes @code{base_type} for which
15351 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
15352 be a complete type, (possibly cv-qualified) @code{void}, or an array
15355 @item __is_enum (type)
15356 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
15357 true, else it is false.
15359 @item __is_literal_type (type)
15360 If @code{type} is a literal type ([basic.types]) the trait is
15361 true, else it is false. Requires: @code{type} shall be a complete type,
15362 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15364 @item __is_pod (type)
15365 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
15366 else it is false. Requires: @code{type} shall be a complete type,
15367 (possibly cv-qualified) @code{void}, or an array of unknown bound.
15369 @item __is_polymorphic (type)
15370 If @code{type} is a polymorphic class ([class.virtual]) then the trait
15371 is true, else it is false. Requires: @code{type} shall be a complete
15372 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15374 @item __is_standard_layout (type)
15375 If @code{type} is a standard-layout type ([basic.types]) the trait is
15376 true, else it is false. Requires: @code{type} shall be a complete
15377 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15379 @item __is_trivial (type)
15380 If @code{type} is a trivial type ([basic.types]) the trait is
15381 true, else it is false. Requires: @code{type} shall be a complete
15382 type, (possibly cv-qualified) @code{void}, or an array of unknown bound.
15384 @item __is_union (type)
15385 If @code{type} is a cv union type ([basic.compound]) the trait is
15386 true, else it is false.
15388 @item __underlying_type (type)
15389 The underlying type of @code{type}. Requires: @code{type} shall be
15390 an enumeration type ([dcl.enum]).
15394 @node Java Exceptions
15395 @section Java Exceptions
15397 The Java language uses a slightly different exception handling model
15398 from C++. Normally, GNU C++ will automatically detect when you are
15399 writing C++ code that uses Java exceptions, and handle them
15400 appropriately. However, if C++ code only needs to execute destructors
15401 when Java exceptions are thrown through it, GCC will guess incorrectly.
15402 Sample problematic code is:
15405 struct S @{ ~S(); @};
15406 extern void bar(); // @r{is written in Java, and may throw exceptions}
15415 The usual effect of an incorrect guess is a link failure, complaining of
15416 a missing routine called @samp{__gxx_personality_v0}.
15418 You can inform the compiler that Java exceptions are to be used in a
15419 translation unit, irrespective of what it might think, by writing
15420 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
15421 @samp{#pragma} must appear before any functions that throw or catch
15422 exceptions, or run destructors when exceptions are thrown through them.
15424 You cannot mix Java and C++ exceptions in the same translation unit. It
15425 is believed to be safe to throw a C++ exception from one file through
15426 another file compiled for the Java exception model, or vice versa, but
15427 there may be bugs in this area.
15429 @node Deprecated Features
15430 @section Deprecated Features
15432 In the past, the GNU C++ compiler was extended to experiment with new
15433 features, at a time when the C++ language was still evolving. Now that
15434 the C++ standard is complete, some of those features are superseded by
15435 superior alternatives. Using the old features might cause a warning in
15436 some cases that the feature will be dropped in the future. In other
15437 cases, the feature might be gone already.
15439 While the list below is not exhaustive, it documents some of the options
15440 that are now deprecated:
15443 @item -fexternal-templates
15444 @itemx -falt-external-templates
15445 These are two of the many ways for G++ to implement template
15446 instantiation. @xref{Template Instantiation}. The C++ standard clearly
15447 defines how template definitions have to be organized across
15448 implementation units. G++ has an implicit instantiation mechanism that
15449 should work just fine for standard-conforming code.
15451 @item -fstrict-prototype
15452 @itemx -fno-strict-prototype
15453 Previously it was possible to use an empty prototype parameter list to
15454 indicate an unspecified number of parameters (like C), rather than no
15455 parameters, as C++ demands. This feature has been removed, except where
15456 it is required for backwards compatibility. @xref{Backwards Compatibility}.
15459 G++ allows a virtual function returning @samp{void *} to be overridden
15460 by one returning a different pointer type. This extension to the
15461 covariant return type rules is now deprecated and will be removed from a
15464 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
15465 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
15466 and are now removed from G++. Code using these operators should be
15467 modified to use @code{std::min} and @code{std::max} instead.
15469 The named return value extension has been deprecated, and is now
15472 The use of initializer lists with new expressions has been deprecated,
15473 and is now removed from G++.
15475 Floating and complex non-type template parameters have been deprecated,
15476 and are now removed from G++.
15478 The implicit typename extension has been deprecated and is now
15481 The use of default arguments in function pointers, function typedefs
15482 and other places where they are not permitted by the standard is
15483 deprecated and will be removed from a future version of G++.
15485 G++ allows floating-point literals to appear in integral constant expressions,
15486 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
15487 This extension is deprecated and will be removed from a future version.
15489 G++ allows static data members of const floating-point type to be declared
15490 with an initializer in a class definition. The standard only allows
15491 initializers for static members of const integral types and const
15492 enumeration types so this extension has been deprecated and will be removed
15493 from a future version.
15495 @node Backwards Compatibility
15496 @section Backwards Compatibility
15497 @cindex Backwards Compatibility
15498 @cindex ARM [Annotated C++ Reference Manual]
15500 Now that there is a definitive ISO standard C++, G++ has a specification
15501 to adhere to. The C++ language evolved over time, and features that
15502 used to be acceptable in previous drafts of the standard, such as the ARM
15503 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
15504 compilation of C++ written to such drafts, G++ contains some backwards
15505 compatibilities. @emph{All such backwards compatibility features are
15506 liable to disappear in future versions of G++.} They should be considered
15507 deprecated. @xref{Deprecated Features}.
15511 If a variable is declared at for scope, it used to remain in scope until
15512 the end of the scope which contained the for statement (rather than just
15513 within the for scope). G++ retains this, but issues a warning, if such a
15514 variable is accessed outside the for scope.
15516 @item Implicit C language
15517 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
15518 scope to set the language. On such systems, all header files are
15519 implicitly scoped inside a C language scope. Also, an empty prototype
15520 @code{()} will be treated as an unspecified number of arguments, rather
15521 than no arguments, as C++ demands.