1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
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 C89 or C++ are also, as
25 extensions, accepted by GCC in C89 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 * Complex:: Data types for complex numbers.
37 * Floating Types:: Additional Floating Types.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Zero Length:: Zero-length arrays.
42 * Variable Length:: Arrays whose length is computed at run time.
43 * Empty Structures:: Structures with no members.
44 * Variadic Macros:: Macros with a variable number of arguments.
45 * Escaped Newlines:: Slightly looser rules for escaped newlines.
46 * Subscripting:: Any array can be subscripted, even if not an lvalue.
47 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
48 * Initializers:: Non-constant initializers.
49 * Compound Literals:: Compound literals give structures, unions
51 * Designated Inits:: Labeling elements of initializers.
52 * Cast to Union:: Casting to union type from any member of the union.
53 * Case Ranges:: `case 1 ... 9' and such.
54 * Mixed Declarations:: Mixing declarations and code.
55 * Function Attributes:: Declaring that functions have no side effects,
56 or that they can never return.
57 * Attribute Syntax:: Formal syntax for attributes.
58 * Function Prototypes:: Prototype declarations and old-style definitions.
59 * C++ Comments:: C++ comments are recognized.
60 * Dollar Signs:: Dollar sign is allowed in identifiers.
61 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
62 * Variable Attributes:: Specifying attributes of variables.
63 * Type Attributes:: Specifying attributes of types.
64 * Alignment:: Inquiring about the alignment of a type or variable.
65 * Inline:: Defining inline functions (as fast as macros).
66 * Extended Asm:: Assembler instructions with C expressions as operands.
67 (With them you can define ``built-in'' functions.)
68 * Constraints:: Constraints for asm operands
69 * Asm Labels:: Specifying the assembler name to use for a C symbol.
70 * Explicit Reg Vars:: Defining variables residing in specified registers.
71 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
72 * Incomplete Enums:: @code{enum foo;}, with details to follow.
73 * Function Names:: Printable strings which are the name of the current
75 * Return Address:: Getting the return or frame address of a function.
76 * Vector Extensions:: Using vector instructions through built-in functions.
77 * Offsetof:: Special syntax for implementing @code{offsetof}.
78 * Atomic Builtins:: Built-in functions for atomic memory access.
79 * Object Size Checking:: Built-in functions for limited buffer overflow
81 * Other Builtins:: Other built-in functions.
82 * Target Builtins:: Built-in functions specific to particular targets.
83 * Target Format Checks:: Format checks specific to particular targets.
84 * Pragmas:: Pragmas accepted by GCC.
85 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
86 * Thread-Local:: Per-thread variables.
87 * Binary constants:: Binary constants using the @samp{0b} prefix.
91 @section Statements and Declarations in Expressions
92 @cindex statements inside expressions
93 @cindex declarations inside expressions
94 @cindex expressions containing statements
95 @cindex macros, statements in expressions
97 @c the above section title wrapped and causes an underfull hbox.. i
98 @c changed it from "within" to "in". --mew 4feb93
99 A compound statement enclosed in parentheses may appear as an expression
100 in GNU C@. This allows you to use loops, switches, and local variables
101 within an expression.
103 Recall that a compound statement is a sequence of statements surrounded
104 by braces; in this construct, parentheses go around the braces. For
108 (@{ int y = foo (); int z;
115 is a valid (though slightly more complex than necessary) expression
116 for the absolute value of @code{foo ()}.
118 The last thing in the compound statement should be an expression
119 followed by a semicolon; the value of this subexpression serves as the
120 value of the entire construct. (If you use some other kind of statement
121 last within the braces, the construct has type @code{void}, and thus
122 effectively no value.)
124 This feature is especially useful in making macro definitions ``safe'' (so
125 that they evaluate each operand exactly once). For example, the
126 ``maximum'' function is commonly defined as a macro in standard C as
130 #define max(a,b) ((a) > (b) ? (a) : (b))
134 @cindex side effects, macro argument
135 But this definition computes either @var{a} or @var{b} twice, with bad
136 results if the operand has side effects. In GNU C, if you know the
137 type of the operands (here taken as @code{int}), you can define
138 the macro safely as follows:
141 #define maxint(a,b) \
142 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
145 Embedded statements are not allowed in constant expressions, such as
146 the value of an enumeration constant, the width of a bit-field, or
147 the initial value of a static variable.
149 If you don't know the type of the operand, you can still do this, but you
150 must use @code{typeof} (@pxref{Typeof}).
152 In G++, the result value of a statement expression undergoes array and
153 function pointer decay, and is returned by value to the enclosing
154 expression. For instance, if @code{A} is a class, then
163 will construct a temporary @code{A} object to hold the result of the
164 statement expression, and that will be used to invoke @code{Foo}.
165 Therefore the @code{this} pointer observed by @code{Foo} will not be the
168 Any temporaries created within a statement within a statement expression
169 will be destroyed at the statement's end. This makes statement
170 expressions inside macros slightly different from function calls. In
171 the latter case temporaries introduced during argument evaluation will
172 be destroyed at the end of the statement that includes the function
173 call. In the statement expression case they will be destroyed during
174 the statement expression. For instance,
177 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
178 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
188 will have different places where temporaries are destroyed. For the
189 @code{macro} case, the temporary @code{X} will be destroyed just after
190 the initialization of @code{b}. In the @code{function} case that
191 temporary will be destroyed when the function returns.
193 These considerations mean that it is probably a bad idea to use
194 statement-expressions of this form in header files that are designed to
195 work with C++. (Note that some versions of the GNU C Library contained
196 header files using statement-expression that lead to precisely this
199 Jumping into a statement expression with @code{goto} or using a
200 @code{switch} statement outside the statement expression with a
201 @code{case} or @code{default} label inside the statement expression is
202 not permitted. Jumping into a statement expression with a computed
203 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
204 Jumping out of a statement expression is permitted, but if the
205 statement expression is part of a larger expression then it is
206 unspecified which other subexpressions of that expression have been
207 evaluated except where the language definition requires certain
208 subexpressions to be evaluated before or after the statement
209 expression. In any case, as with a function call the evaluation of a
210 statement expression is not interleaved with the evaluation of other
211 parts of the containing expression. For example,
214 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
218 will call @code{foo} and @code{bar1} and will not call @code{baz} but
219 may or may not call @code{bar2}. If @code{bar2} is called, it will be
220 called after @code{foo} and before @code{bar1}
223 @section Locally Declared Labels
225 @cindex macros, local labels
227 GCC allows you to declare @dfn{local labels} in any nested block
228 scope. A local label is just like an ordinary label, but you can
229 only reference it (with a @code{goto} statement, or by taking its
230 address) within the block in which it was declared.
232 A local label declaration looks like this:
235 __label__ @var{label};
242 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
245 Local label declarations must come at the beginning of the block,
246 before any ordinary declarations or statements.
248 The label declaration defines the label @emph{name}, but does not define
249 the label itself. You must do this in the usual way, with
250 @code{@var{label}:}, within the statements of the statement expression.
252 The local label feature is useful for complex macros. If a macro
253 contains nested loops, a @code{goto} can be useful for breaking out of
254 them. However, an ordinary label whose scope is the whole function
255 cannot be used: if the macro can be expanded several times in one
256 function, the label will be multiply defined in that function. A
257 local label avoids this problem. For example:
260 #define SEARCH(value, array, target) \
263 typeof (target) _SEARCH_target = (target); \
264 typeof (*(array)) *_SEARCH_array = (array); \
267 for (i = 0; i < max; i++) \
268 for (j = 0; j < max; j++) \
269 if (_SEARCH_array[i][j] == _SEARCH_target) \
270 @{ (value) = i; goto found; @} \
276 This could also be written using a statement-expression:
279 #define SEARCH(array, target) \
282 typeof (target) _SEARCH_target = (target); \
283 typeof (*(array)) *_SEARCH_array = (array); \
286 for (i = 0; i < max; i++) \
287 for (j = 0; j < max; j++) \
288 if (_SEARCH_array[i][j] == _SEARCH_target) \
289 @{ value = i; goto found; @} \
296 Local label declarations also make the labels they declare visible to
297 nested functions, if there are any. @xref{Nested Functions}, for details.
299 @node Labels as Values
300 @section Labels as Values
301 @cindex labels as values
302 @cindex computed gotos
303 @cindex goto with computed label
304 @cindex address of a label
306 You can get the address of a label defined in the current function
307 (or a containing function) with the unary operator @samp{&&}. The
308 value has type @code{void *}. This value is a constant and can be used
309 wherever a constant of that type is valid. For example:
317 To use these values, you need to be able to jump to one. This is done
318 with the computed goto statement@footnote{The analogous feature in
319 Fortran is called an assigned goto, but that name seems inappropriate in
320 C, where one can do more than simply store label addresses in label
321 variables.}, @code{goto *@var{exp};}. For example,
328 Any expression of type @code{void *} is allowed.
330 One way of using these constants is in initializing a static array that
331 will serve as a jump table:
334 static void *array[] = @{ &&foo, &&bar, &&hack @};
337 Then you can select a label with indexing, like this:
344 Note that this does not check whether the subscript is in bounds---array
345 indexing in C never does that.
347 Such an array of label values serves a purpose much like that of the
348 @code{switch} statement. The @code{switch} statement is cleaner, so
349 use that rather than an array unless the problem does not fit a
350 @code{switch} statement very well.
352 Another use of label values is in an interpreter for threaded code.
353 The labels within the interpreter function can be stored in the
354 threaded code for super-fast dispatching.
356 You may not use this mechanism to jump to code in a different function.
357 If you do that, totally unpredictable things will happen. The best way to
358 avoid this is to store the label address only in automatic variables and
359 never pass it as an argument.
361 An alternate way to write the above example is
364 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
366 goto *(&&foo + array[i]);
370 This is more friendly to code living in shared libraries, as it reduces
371 the number of dynamic relocations that are needed, and by consequence,
372 allows the data to be read-only.
374 The @code{&&foo} expressions for the same label might have different values
375 if the containing function is inlined or cloned. If a program relies on
376 them being always the same, @code{__attribute__((__noinline__))} should
377 be used to prevent inlining. If @code{&&foo} is used
378 in a static variable initializer, inlining is forbidden.
380 @node Nested Functions
381 @section Nested Functions
382 @cindex nested functions
383 @cindex downward funargs
386 A @dfn{nested function} is a function defined inside another function.
387 (Nested functions are not supported for GNU C++.) The nested function's
388 name is local to the block where it is defined. For example, here we
389 define a nested function named @code{square}, and call it twice:
393 foo (double a, double b)
395 double square (double z) @{ return z * z; @}
397 return square (a) + square (b);
402 The nested function can access all the variables of the containing
403 function that are visible at the point of its definition. This is
404 called @dfn{lexical scoping}. For example, here we show a nested
405 function which uses an inherited variable named @code{offset}:
409 bar (int *array, int offset, int size)
411 int access (int *array, int index)
412 @{ return array[index + offset]; @}
415 for (i = 0; i < size; i++)
416 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
421 Nested function definitions are permitted within functions in the places
422 where variable definitions are allowed; that is, in any block, mixed
423 with the other declarations and statements in the block.
425 It is possible to call the nested function from outside the scope of its
426 name by storing its address or passing the address to another function:
429 hack (int *array, int size)
431 void store (int index, int value)
432 @{ array[index] = value; @}
434 intermediate (store, size);
438 Here, the function @code{intermediate} receives the address of
439 @code{store} as an argument. If @code{intermediate} calls @code{store},
440 the arguments given to @code{store} are used to store into @code{array}.
441 But this technique works only so long as the containing function
442 (@code{hack}, in this example) does not exit.
444 If you try to call the nested function through its address after the
445 containing function has exited, all hell will break loose. If you try
446 to call it after a containing scope level has exited, and if it refers
447 to some of the variables that are no longer in scope, you may be lucky,
448 but it's not wise to take the risk. If, however, the nested function
449 does not refer to anything that has gone out of scope, you should be
452 GCC implements taking the address of a nested function using a technique
453 called @dfn{trampolines}. A paper describing them is available as
456 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
458 A nested function can jump to a label inherited from a containing
459 function, provided the label was explicitly declared in the containing
460 function (@pxref{Local Labels}). Such a jump returns instantly to the
461 containing function, exiting the nested function which did the
462 @code{goto} and any intermediate functions as well. Here is an example:
466 bar (int *array, int offset, int size)
469 int access (int *array, int index)
473 return array[index + offset];
477 for (i = 0; i < size; i++)
478 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
482 /* @r{Control comes here from @code{access}
483 if it detects an error.} */
490 A nested function always has no linkage. Declaring one with
491 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
492 before its definition, use @code{auto} (which is otherwise meaningless
493 for function declarations).
496 bar (int *array, int offset, int size)
499 auto int access (int *, int);
501 int access (int *array, int index)
505 return array[index + offset];
511 @node Constructing Calls
512 @section Constructing Function Calls
513 @cindex constructing calls
514 @cindex forwarding calls
516 Using the built-in functions described below, you can record
517 the arguments a function received, and call another function
518 with the same arguments, without knowing the number or types
521 You can also record the return value of that function call,
522 and later return that value, without knowing what data type
523 the function tried to return (as long as your caller expects
526 However, these built-in functions may interact badly with some
527 sophisticated features or other extensions of the language. It
528 is, therefore, not recommended to use them outside very simple
529 functions acting as mere forwarders for their arguments.
531 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
532 This built-in function returns a pointer to data
533 describing how to perform a call with the same arguments as were passed
534 to the current function.
536 The function saves the arg pointer register, structure value address,
537 and all registers that might be used to pass arguments to a function
538 into a block of memory allocated on the stack. Then it returns the
539 address of that block.
542 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
543 This built-in function invokes @var{function}
544 with a copy of the parameters described by @var{arguments}
547 The value of @var{arguments} should be the value returned by
548 @code{__builtin_apply_args}. The argument @var{size} specifies the size
549 of the stack argument data, in bytes.
551 This function returns a pointer to data describing
552 how to return whatever value was returned by @var{function}. The data
553 is saved in a block of memory allocated on the stack.
555 It is not always simple to compute the proper value for @var{size}. The
556 value is used by @code{__builtin_apply} to compute the amount of data
557 that should be pushed on the stack and copied from the incoming argument
561 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
562 This built-in function returns the value described by @var{result} from
563 the containing function. You should specify, for @var{result}, a value
564 returned by @code{__builtin_apply}.
567 @deftypefn {Built-in Function} __builtin_va_arg_pack ()
568 This built-in function represents all anonymous arguments of an inline
569 function. It can be used only in inline functions which will be always
570 inlined, never compiled as a separate function, such as those using
571 @code{__attribute__ ((__always_inline__))} or
572 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
573 It must be only passed as last argument to some other function
574 with variable arguments. This is useful for writing small wrapper
575 inlines for variable argument functions, when using preprocessor
576 macros is undesirable. For example:
578 extern int myprintf (FILE *f, const char *format, ...);
579 extern inline __attribute__ ((__gnu_inline__)) int
580 myprintf (FILE *f, const char *format, ...)
582 int r = fprintf (f, "myprintf: ");
585 int s = fprintf (f, format, __builtin_va_arg_pack ());
593 @deftypefn {Built-in Function} __builtin_va_arg_pack_len ()
594 This built-in function returns the number of anonymous arguments of
595 an inline function. It can be used only in inline functions which
596 will be always inlined, never compiled as a separate function, such
597 as those using @code{__attribute__ ((__always_inline__))} or
598 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
599 For example following will do link or runtime checking of open
600 arguments for optimized code:
603 extern inline __attribute__((__gnu_inline__)) int
604 myopen (const char *path, int oflag, ...)
606 if (__builtin_va_arg_pack_len () > 1)
607 warn_open_too_many_arguments ();
609 if (__builtin_constant_p (oflag))
611 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
613 warn_open_missing_mode ();
614 return __open_2 (path, oflag);
616 return open (path, oflag, __builtin_va_arg_pack ());
619 if (__builtin_va_arg_pack_len () < 1)
620 return __open_2 (path, oflag);
622 return open (path, oflag, __builtin_va_arg_pack ());
629 @section Referring to a Type with @code{typeof}
632 @cindex macros, types of arguments
634 Another way to refer to the type of an expression is with @code{typeof}.
635 The syntax of using of this keyword looks like @code{sizeof}, but the
636 construct acts semantically like a type name defined with @code{typedef}.
638 There are two ways of writing the argument to @code{typeof}: with an
639 expression or with a type. Here is an example with an expression:
646 This assumes that @code{x} is an array of pointers to functions;
647 the type described is that of the values of the functions.
649 Here is an example with a typename as the argument:
656 Here the type described is that of pointers to @code{int}.
658 If you are writing a header file that must work when included in ISO C
659 programs, write @code{__typeof__} instead of @code{typeof}.
660 @xref{Alternate Keywords}.
662 A @code{typeof}-construct can be used anywhere a typedef name could be
663 used. For example, you can use it in a declaration, in a cast, or inside
664 of @code{sizeof} or @code{typeof}.
666 The operand of @code{typeof} is evaluated for its side effects if and
667 only if it is an expression of variably modified type or the name of
670 @code{typeof} is often useful in conjunction with the
671 statements-within-expressions feature. Here is how the two together can
672 be used to define a safe ``maximum'' macro that operates on any
673 arithmetic type and evaluates each of its arguments exactly once:
677 (@{ typeof (a) _a = (a); \
678 typeof (b) _b = (b); \
679 _a > _b ? _a : _b; @})
682 @cindex underscores in variables in macros
683 @cindex @samp{_} in variables in macros
684 @cindex local variables in macros
685 @cindex variables, local, in macros
686 @cindex macros, local variables in
688 The reason for using names that start with underscores for the local
689 variables is to avoid conflicts with variable names that occur within the
690 expressions that are substituted for @code{a} and @code{b}. Eventually we
691 hope to design a new form of declaration syntax that allows you to declare
692 variables whose scopes start only after their initializers; this will be a
693 more reliable way to prevent such conflicts.
696 Some more examples of the use of @code{typeof}:
700 This declares @code{y} with the type of what @code{x} points to.
707 This declares @code{y} as an array of such values.
714 This declares @code{y} as an array of pointers to characters:
717 typeof (typeof (char *)[4]) y;
721 It is equivalent to the following traditional C declaration:
727 To see the meaning of the declaration using @code{typeof}, and why it
728 might be a useful way to write, rewrite it with these macros:
731 #define pointer(T) typeof(T *)
732 #define array(T, N) typeof(T [N])
736 Now the declaration can be rewritten this way:
739 array (pointer (char), 4) y;
743 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
744 pointers to @code{char}.
747 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
748 a more limited extension which permitted one to write
751 typedef @var{T} = @var{expr};
755 with the effect of declaring @var{T} to have the type of the expression
756 @var{expr}. This extension does not work with GCC 3 (versions between
757 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
758 relies on it should be rewritten to use @code{typeof}:
761 typedef typeof(@var{expr}) @var{T};
765 This will work with all versions of GCC@.
768 @section Conditionals with Omitted Operands
769 @cindex conditional expressions, extensions
770 @cindex omitted middle-operands
771 @cindex middle-operands, omitted
772 @cindex extensions, @code{?:}
773 @cindex @code{?:} extensions
775 The middle operand in a conditional expression may be omitted. Then
776 if the first operand is nonzero, its value is the value of the conditional
779 Therefore, the expression
786 has the value of @code{x} if that is nonzero; otherwise, the value of
789 This example is perfectly equivalent to
795 @cindex side effect in ?:
796 @cindex ?: side effect
798 In this simple case, the ability to omit the middle operand is not
799 especially useful. When it becomes useful is when the first operand does,
800 or may (if it is a macro argument), contain a side effect. Then repeating
801 the operand in the middle would perform the side effect twice. Omitting
802 the middle operand uses the value already computed without the undesirable
803 effects of recomputing it.
806 @section Double-Word Integers
807 @cindex @code{long long} data types
808 @cindex double-word arithmetic
809 @cindex multiprecision arithmetic
810 @cindex @code{LL} integer suffix
811 @cindex @code{ULL} integer suffix
813 ISO C99 supports data types for integers that are at least 64 bits wide,
814 and as an extension GCC supports them in C89 mode and in C++.
815 Simply write @code{long long int} for a signed integer, or
816 @code{unsigned long long int} for an unsigned integer. To make an
817 integer constant of type @code{long long int}, add the suffix @samp{LL}
818 to the integer. To make an integer constant of type @code{unsigned long
819 long int}, add the suffix @samp{ULL} to the integer.
821 You can use these types in arithmetic like any other integer types.
822 Addition, subtraction, and bitwise boolean operations on these types
823 are open-coded on all types of machines. Multiplication is open-coded
824 if the machine supports fullword-to-doubleword a widening multiply
825 instruction. Division and shifts are open-coded only on machines that
826 provide special support. The operations that are not open-coded use
827 special library routines that come with GCC@.
829 There may be pitfalls when you use @code{long long} types for function
830 arguments, unless you declare function prototypes. If a function
831 expects type @code{int} for its argument, and you pass a value of type
832 @code{long long int}, confusion will result because the caller and the
833 subroutine will disagree about the number of bytes for the argument.
834 Likewise, if the function expects @code{long long int} and you pass
835 @code{int}. The best way to avoid such problems is to use prototypes.
838 @section Complex Numbers
839 @cindex complex numbers
840 @cindex @code{_Complex} keyword
841 @cindex @code{__complex__} keyword
843 ISO C99 supports complex floating data types, and as an extension GCC
844 supports them in C89 mode and in C++, and supports complex integer data
845 types which are not part of ISO C99. You can declare complex types
846 using the keyword @code{_Complex}. As an extension, the older GNU
847 keyword @code{__complex__} is also supported.
849 For example, @samp{_Complex double x;} declares @code{x} as a
850 variable whose real part and imaginary part are both of type
851 @code{double}. @samp{_Complex short int y;} declares @code{y} to
852 have real and imaginary parts of type @code{short int}; this is not
853 likely to be useful, but it shows that the set of complex types is
856 To write a constant with a complex data type, use the suffix @samp{i} or
857 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
858 has type @code{_Complex float} and @code{3i} has type
859 @code{_Complex int}. Such a constant always has a pure imaginary
860 value, but you can form any complex value you like by adding one to a
861 real constant. This is a GNU extension; if you have an ISO C99
862 conforming C library (such as GNU libc), and want to construct complex
863 constants of floating type, you should include @code{<complex.h>} and
864 use the macros @code{I} or @code{_Complex_I} instead.
866 @cindex @code{__real__} keyword
867 @cindex @code{__imag__} keyword
868 To extract the real part of a complex-valued expression @var{exp}, write
869 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
870 extract the imaginary part. This is a GNU extension; for values of
871 floating type, you should use the ISO C99 functions @code{crealf},
872 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
873 @code{cimagl}, declared in @code{<complex.h>} and also provided as
874 built-in functions by GCC@.
876 @cindex complex conjugation
877 The operator @samp{~} performs complex conjugation when used on a value
878 with a complex type. This is a GNU extension; for values of
879 floating type, you should use the ISO C99 functions @code{conjf},
880 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
881 provided as built-in functions by GCC@.
883 GCC can allocate complex automatic variables in a noncontiguous
884 fashion; it's even possible for the real part to be in a register while
885 the imaginary part is on the stack (or vice-versa). Only the DWARF2
886 debug info format can represent this, so use of DWARF2 is recommended.
887 If you are using the stabs debug info format, GCC describes a noncontiguous
888 complex variable as if it were two separate variables of noncomplex type.
889 If the variable's actual name is @code{foo}, the two fictitious
890 variables are named @code{foo$real} and @code{foo$imag}. You can
891 examine and set these two fictitious variables with your debugger.
894 @section Additional Floating Types
895 @cindex additional floating types
896 @cindex @code{__float80} data type
897 @cindex @code{__float128} data type
898 @cindex @code{w} floating point suffix
899 @cindex @code{q} floating point suffix
900 @cindex @code{W} floating point suffix
901 @cindex @code{Q} floating point suffix
903 As an extension, the GNU C compiler supports additional floating
904 types, @code{__float80} and @code{__float128} to support 80bit
905 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
906 Support for additional types includes the arithmetic operators:
907 add, subtract, multiply, divide; unary arithmetic operators;
908 relational operators; equality operators; and conversions to and from
909 integer and other floating types. Use a suffix @samp{w} or @samp{W}
910 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
911 for @code{_float128}. You can declare complex types using the
912 corresponding internal complex type, @code{XCmode} for @code{__float80}
913 type and @code{TCmode} for @code{__float128} type:
916 typedef _Complex float __attribute__((mode(TC))) _Complex128;
917 typedef _Complex float __attribute__((mode(XC))) _Complex80;
920 Not all targets support additional floating point types. @code{__float80}
921 is supported on i386, x86_64 and ia64 targets and target @code{__float128}
922 is supported on x86_64 and ia64 targets.
925 @section Decimal Floating Types
926 @cindex decimal floating types
927 @cindex @code{_Decimal32} data type
928 @cindex @code{_Decimal64} data type
929 @cindex @code{_Decimal128} data type
930 @cindex @code{df} integer suffix
931 @cindex @code{dd} integer suffix
932 @cindex @code{dl} integer suffix
933 @cindex @code{DF} integer suffix
934 @cindex @code{DD} integer suffix
935 @cindex @code{DL} integer suffix
937 As an extension, the GNU C compiler supports decimal floating types as
938 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
939 floating types in GCC will evolve as the draft technical report changes.
940 Calling conventions for any target might also change. Not all targets
941 support decimal floating types.
943 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
944 @code{_Decimal128}. They use a radix of ten, unlike the floating types
945 @code{float}, @code{double}, and @code{long double} whose radix is not
946 specified by the C standard but is usually two.
948 Support for decimal floating types includes the arithmetic operators
949 add, subtract, multiply, divide; unary arithmetic operators;
950 relational operators; equality operators; and conversions to and from
951 integer and other floating types. Use a suffix @samp{df} or
952 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
953 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
956 GCC support of decimal float as specified by the draft technical report
961 Pragma @code{FLOAT_CONST_DECIMAL64} is not supported, nor is the @samp{d}
962 suffix for literal constants of type @code{double}.
965 When the value of a decimal floating type cannot be represented in the
966 integer type to which it is being converted, the result is undefined
967 rather than the result value specified by the draft technical report.
970 GCC does not provide the C library functionality associated with
971 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
972 @file{wchar.h}, which must come from a separate C library implementation.
973 Because of this the GNU C compiler does not define macro
974 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
975 the technical report.
978 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
979 are supported by the DWARF2 debug information format.
985 ISO C99 supports floating-point numbers written not only in the usual
986 decimal notation, such as @code{1.55e1}, but also numbers such as
987 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
988 supports this in C89 mode (except in some cases when strictly
989 conforming) and in C++. In that format the
990 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
991 mandatory. The exponent is a decimal number that indicates the power of
992 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
999 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1000 is the same as @code{1.55e1}.
1002 Unlike for floating-point numbers in the decimal notation the exponent
1003 is always required in the hexadecimal notation. Otherwise the compiler
1004 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1005 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1006 extension for floating-point constants of type @code{float}.
1009 @section Fixed-Point Types
1010 @cindex fixed-point types
1011 @cindex @code{_Fract} data type
1012 @cindex @code{_Accum} data type
1013 @cindex @code{_Sat} data type
1014 @cindex @code{hr} fixed-suffix
1015 @cindex @code{r} fixed-suffix
1016 @cindex @code{lr} fixed-suffix
1017 @cindex @code{llr} fixed-suffix
1018 @cindex @code{uhr} fixed-suffix
1019 @cindex @code{ur} fixed-suffix
1020 @cindex @code{ulr} fixed-suffix
1021 @cindex @code{ullr} fixed-suffix
1022 @cindex @code{hk} fixed-suffix
1023 @cindex @code{k} fixed-suffix
1024 @cindex @code{lk} fixed-suffix
1025 @cindex @code{llk} fixed-suffix
1026 @cindex @code{uhk} fixed-suffix
1027 @cindex @code{uk} fixed-suffix
1028 @cindex @code{ulk} fixed-suffix
1029 @cindex @code{ullk} fixed-suffix
1030 @cindex @code{HR} fixed-suffix
1031 @cindex @code{R} fixed-suffix
1032 @cindex @code{LR} fixed-suffix
1033 @cindex @code{LLR} fixed-suffix
1034 @cindex @code{UHR} fixed-suffix
1035 @cindex @code{UR} fixed-suffix
1036 @cindex @code{ULR} fixed-suffix
1037 @cindex @code{ULLR} fixed-suffix
1038 @cindex @code{HK} fixed-suffix
1039 @cindex @code{K} fixed-suffix
1040 @cindex @code{LK} fixed-suffix
1041 @cindex @code{LLK} fixed-suffix
1042 @cindex @code{UHK} fixed-suffix
1043 @cindex @code{UK} fixed-suffix
1044 @cindex @code{ULK} fixed-suffix
1045 @cindex @code{ULLK} fixed-suffix
1047 As an extension, the GNU C compiler supports fixed-point types as
1048 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1049 types in GCC will evolve as the draft technical report changes.
1050 Calling conventions for any target might also change. Not all targets
1051 support fixed-point types.
1053 The fixed-point types are
1054 @code{short _Fract},
1057 @code{long long _Fract},
1058 @code{unsigned short _Fract},
1059 @code{unsigned _Fract},
1060 @code{unsigned long _Fract},
1061 @code{unsigned long long _Fract},
1062 @code{_Sat short _Fract},
1064 @code{_Sat long _Fract},
1065 @code{_Sat long long _Fract},
1066 @code{_Sat unsigned short _Fract},
1067 @code{_Sat unsigned _Fract},
1068 @code{_Sat unsigned long _Fract},
1069 @code{_Sat unsigned long long _Fract},
1070 @code{short _Accum},
1073 @code{long long _Accum},
1074 @code{unsigned short _Accum},
1075 @code{unsigned _Accum},
1076 @code{unsigned long _Accum},
1077 @code{unsigned long long _Accum},
1078 @code{_Sat short _Accum},
1080 @code{_Sat long _Accum},
1081 @code{_Sat long long _Accum},
1082 @code{_Sat unsigned short _Accum},
1083 @code{_Sat unsigned _Accum},
1084 @code{_Sat unsigned long _Accum},
1085 @code{_Sat unsigned long long _Accum}.
1087 Fixed-point data values contain fractional and optional integral parts.
1088 The format of fixed-point data varies and depends on the target machine.
1090 Support for fixed-point types includes:
1093 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1095 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1097 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1099 binary shift operators (@code{<<}, @code{>>})
1101 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1103 equality operators (@code{==}, @code{!=})
1105 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1106 @code{<<=}, @code{>>=})
1108 conversions to and from integer, floating-point, or fixed-point types
1111 Use a suffix in a fixed-point literal constant:
1113 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1114 @code{_Sat short _Fract}
1115 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1116 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1117 @code{_Sat long _Fract}
1118 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1119 @code{_Sat long long _Fract}
1120 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1121 @code{_Sat unsigned short _Fract}
1122 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1123 @code{_Sat unsigned _Fract}
1124 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1125 @code{_Sat unsigned long _Fract}
1126 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1127 and @code{_Sat unsigned long long _Fract}
1128 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1129 @code{_Sat short _Accum}
1130 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1131 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1132 @code{_Sat long _Accum}
1133 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1134 @code{_Sat long long _Accum}
1135 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1136 @code{_Sat unsigned short _Accum}
1137 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1138 @code{_Sat unsigned _Accum}
1139 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1140 @code{_Sat unsigned long _Accum}
1141 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1142 and @code{_Sat unsigned long long _Accum}
1145 GCC support of fixed-point types as specified by the draft technical report
1150 Pragmas to control overflow and rounding behaviors are not implemented.
1153 Fixed-point types are supported by the DWARF2 debug information format.
1156 @section Arrays of Length Zero
1157 @cindex arrays of length zero
1158 @cindex zero-length arrays
1159 @cindex length-zero arrays
1160 @cindex flexible array members
1162 Zero-length arrays are allowed in GNU C@. They are very useful as the
1163 last element of a structure which is really a header for a variable-length
1172 struct line *thisline = (struct line *)
1173 malloc (sizeof (struct line) + this_length);
1174 thisline->length = this_length;
1177 In ISO C90, you would have to give @code{contents} a length of 1, which
1178 means either you waste space or complicate the argument to @code{malloc}.
1180 In ISO C99, you would use a @dfn{flexible array member}, which is
1181 slightly different in syntax and semantics:
1185 Flexible array members are written as @code{contents[]} without
1189 Flexible array members have incomplete type, and so the @code{sizeof}
1190 operator may not be applied. As a quirk of the original implementation
1191 of zero-length arrays, @code{sizeof} evaluates to zero.
1194 Flexible array members may only appear as the last member of a
1195 @code{struct} that is otherwise non-empty.
1198 A structure containing a flexible array member, or a union containing
1199 such a structure (possibly recursively), may not be a member of a
1200 structure or an element of an array. (However, these uses are
1201 permitted by GCC as extensions.)
1204 GCC versions before 3.0 allowed zero-length arrays to be statically
1205 initialized, as if they were flexible arrays. In addition to those
1206 cases that were useful, it also allowed initializations in situations
1207 that would corrupt later data. Non-empty initialization of zero-length
1208 arrays is now treated like any case where there are more initializer
1209 elements than the array holds, in that a suitable warning about "excess
1210 elements in array" is given, and the excess elements (all of them, in
1211 this case) are ignored.
1213 Instead GCC allows static initialization of flexible array members.
1214 This is equivalent to defining a new structure containing the original
1215 structure followed by an array of sufficient size to contain the data.
1216 I.e.@: in the following, @code{f1} is constructed as if it were declared
1222 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1225 struct f1 f1; int data[3];
1226 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1230 The convenience of this extension is that @code{f1} has the desired
1231 type, eliminating the need to consistently refer to @code{f2.f1}.
1233 This has symmetry with normal static arrays, in that an array of
1234 unknown size is also written with @code{[]}.
1236 Of course, this extension only makes sense if the extra data comes at
1237 the end of a top-level object, as otherwise we would be overwriting
1238 data at subsequent offsets. To avoid undue complication and confusion
1239 with initialization of deeply nested arrays, we simply disallow any
1240 non-empty initialization except when the structure is the top-level
1241 object. For example:
1244 struct foo @{ int x; int y[]; @};
1245 struct bar @{ struct foo z; @};
1247 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1248 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1249 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1250 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1253 @node Empty Structures
1254 @section Structures With No Members
1255 @cindex empty structures
1256 @cindex zero-size structures
1258 GCC permits a C structure to have no members:
1265 The structure will have size zero. In C++, empty structures are part
1266 of the language. G++ treats empty structures as if they had a single
1267 member of type @code{char}.
1269 @node Variable Length
1270 @section Arrays of Variable Length
1271 @cindex variable-length arrays
1272 @cindex arrays of variable length
1275 Variable-length automatic arrays are allowed in ISO C99, and as an
1276 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1277 implementation of variable-length arrays does not yet conform in detail
1278 to the ISO C99 standard.) These arrays are
1279 declared like any other automatic arrays, but with a length that is not
1280 a constant expression. The storage is allocated at the point of
1281 declaration and deallocated when the brace-level is exited. For
1286 concat_fopen (char *s1, char *s2, char *mode)
1288 char str[strlen (s1) + strlen (s2) + 1];
1291 return fopen (str, mode);
1295 @cindex scope of a variable length array
1296 @cindex variable-length array scope
1297 @cindex deallocating variable length arrays
1298 Jumping or breaking out of the scope of the array name deallocates the
1299 storage. Jumping into the scope is not allowed; you get an error
1302 @cindex @code{alloca} vs variable-length arrays
1303 You can use the function @code{alloca} to get an effect much like
1304 variable-length arrays. The function @code{alloca} is available in
1305 many other C implementations (but not in all). On the other hand,
1306 variable-length arrays are more elegant.
1308 There are other differences between these two methods. Space allocated
1309 with @code{alloca} exists until the containing @emph{function} returns.
1310 The space for a variable-length array is deallocated as soon as the array
1311 name's scope ends. (If you use both variable-length arrays and
1312 @code{alloca} in the same function, deallocation of a variable-length array
1313 will also deallocate anything more recently allocated with @code{alloca}.)
1315 You can also use variable-length arrays as arguments to functions:
1319 tester (int len, char data[len][len])
1325 The length of an array is computed once when the storage is allocated
1326 and is remembered for the scope of the array in case you access it with
1329 If you want to pass the array first and the length afterward, you can
1330 use a forward declaration in the parameter list---another GNU extension.
1334 tester (int len; char data[len][len], int len)
1340 @cindex parameter forward declaration
1341 The @samp{int len} before the semicolon is a @dfn{parameter forward
1342 declaration}, and it serves the purpose of making the name @code{len}
1343 known when the declaration of @code{data} is parsed.
1345 You can write any number of such parameter forward declarations in the
1346 parameter list. They can be separated by commas or semicolons, but the
1347 last one must end with a semicolon, which is followed by the ``real''
1348 parameter declarations. Each forward declaration must match a ``real''
1349 declaration in parameter name and data type. ISO C99 does not support
1350 parameter forward declarations.
1352 @node Variadic Macros
1353 @section Macros with a Variable Number of Arguments.
1354 @cindex variable number of arguments
1355 @cindex macro with variable arguments
1356 @cindex rest argument (in macro)
1357 @cindex variadic macros
1359 In the ISO C standard of 1999, a macro can be declared to accept a
1360 variable number of arguments much as a function can. The syntax for
1361 defining the macro is similar to that of a function. Here is an
1365 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1368 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1369 such a macro, it represents the zero or more tokens until the closing
1370 parenthesis that ends the invocation, including any commas. This set of
1371 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1372 wherever it appears. See the CPP manual for more information.
1374 GCC has long supported variadic macros, and used a different syntax that
1375 allowed you to give a name to the variable arguments just like any other
1376 argument. Here is an example:
1379 #define debug(format, args...) fprintf (stderr, format, args)
1382 This is in all ways equivalent to the ISO C example above, but arguably
1383 more readable and descriptive.
1385 GNU CPP has two further variadic macro extensions, and permits them to
1386 be used with either of the above forms of macro definition.
1388 In standard C, you are not allowed to leave the variable argument out
1389 entirely; but you are allowed to pass an empty argument. For example,
1390 this invocation is invalid in ISO C, because there is no comma after
1397 GNU CPP permits you to completely omit the variable arguments in this
1398 way. In the above examples, the compiler would complain, though since
1399 the expansion of the macro still has the extra comma after the format
1402 To help solve this problem, CPP behaves specially for variable arguments
1403 used with the token paste operator, @samp{##}. If instead you write
1406 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1409 and if the variable arguments are omitted or empty, the @samp{##}
1410 operator causes the preprocessor to remove the comma before it. If you
1411 do provide some variable arguments in your macro invocation, GNU CPP
1412 does not complain about the paste operation and instead places the
1413 variable arguments after the comma. Just like any other pasted macro
1414 argument, these arguments are not macro expanded.
1416 @node Escaped Newlines
1417 @section Slightly Looser Rules for Escaped Newlines
1418 @cindex escaped newlines
1419 @cindex newlines (escaped)
1421 Recently, the preprocessor has relaxed its treatment of escaped
1422 newlines. Previously, the newline had to immediately follow a
1423 backslash. The current implementation allows whitespace in the form
1424 of spaces, horizontal and vertical tabs, and form feeds between the
1425 backslash and the subsequent newline. The preprocessor issues a
1426 warning, but treats it as a valid escaped newline and combines the two
1427 lines to form a single logical line. This works within comments and
1428 tokens, as well as between tokens. Comments are @emph{not} treated as
1429 whitespace for the purposes of this relaxation, since they have not
1430 yet been replaced with spaces.
1433 @section Non-Lvalue Arrays May Have Subscripts
1434 @cindex subscripting
1435 @cindex arrays, non-lvalue
1437 @cindex subscripting and function values
1438 In ISO C99, arrays that are not lvalues still decay to pointers, and
1439 may be subscripted, although they may not be modified or used after
1440 the next sequence point and the unary @samp{&} operator may not be
1441 applied to them. As an extension, GCC allows such arrays to be
1442 subscripted in C89 mode, though otherwise they do not decay to
1443 pointers outside C99 mode. For example,
1444 this is valid in GNU C though not valid in C89:
1448 struct foo @{int a[4];@};
1454 return f().a[index];
1460 @section Arithmetic on @code{void}- and Function-Pointers
1461 @cindex void pointers, arithmetic
1462 @cindex void, size of pointer to
1463 @cindex function pointers, arithmetic
1464 @cindex function, size of pointer to
1466 In GNU C, addition and subtraction operations are supported on pointers to
1467 @code{void} and on pointers to functions. This is done by treating the
1468 size of a @code{void} or of a function as 1.
1470 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1471 and on function types, and returns 1.
1473 @opindex Wpointer-arith
1474 The option @option{-Wpointer-arith} requests a warning if these extensions
1478 @section Non-Constant Initializers
1479 @cindex initializers, non-constant
1480 @cindex non-constant initializers
1482 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1483 automatic variable are not required to be constant expressions in GNU C@.
1484 Here is an example of an initializer with run-time varying elements:
1487 foo (float f, float g)
1489 float beat_freqs[2] = @{ f-g, f+g @};
1494 @node Compound Literals
1495 @section Compound Literals
1496 @cindex constructor expressions
1497 @cindex initializations in expressions
1498 @cindex structures, constructor expression
1499 @cindex expressions, constructor
1500 @cindex compound literals
1501 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1503 ISO C99 supports compound literals. A compound literal looks like
1504 a cast containing an initializer. Its value is an object of the
1505 type specified in the cast, containing the elements specified in
1506 the initializer; it is an lvalue. As an extension, GCC supports
1507 compound literals in C89 mode and in C++.
1509 Usually, the specified type is a structure. Assume that
1510 @code{struct foo} and @code{structure} are declared as shown:
1513 struct foo @{int a; char b[2];@} structure;
1517 Here is an example of constructing a @code{struct foo} with a compound literal:
1520 structure = ((struct foo) @{x + y, 'a', 0@});
1524 This is equivalent to writing the following:
1528 struct foo temp = @{x + y, 'a', 0@};
1533 You can also construct an array. If all the elements of the compound literal
1534 are (made up of) simple constant expressions, suitable for use in
1535 initializers of objects of static storage duration, then the compound
1536 literal can be coerced to a pointer to its first element and used in
1537 such an initializer, as shown here:
1540 char **foo = (char *[]) @{ "x", "y", "z" @};
1543 Compound literals for scalar types and union types are is
1544 also allowed, but then the compound literal is equivalent
1547 As a GNU extension, GCC allows initialization of objects with static storage
1548 duration by compound literals (which is not possible in ISO C99, because
1549 the initializer is not a constant).
1550 It is handled as if the object was initialized only with the bracket
1551 enclosed list if the types of the compound literal and the object match.
1552 The initializer list of the compound literal must be constant.
1553 If the object being initialized has array type of unknown size, the size is
1554 determined by compound literal size.
1557 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1558 static int y[] = (int []) @{1, 2, 3@};
1559 static int z[] = (int [3]) @{1@};
1563 The above lines are equivalent to the following:
1565 static struct foo x = @{1, 'a', 'b'@};
1566 static int y[] = @{1, 2, 3@};
1567 static int z[] = @{1, 0, 0@};
1570 @node Designated Inits
1571 @section Designated Initializers
1572 @cindex initializers with labeled elements
1573 @cindex labeled elements in initializers
1574 @cindex case labels in initializers
1575 @cindex designated initializers
1577 Standard C89 requires the elements of an initializer to appear in a fixed
1578 order, the same as the order of the elements in the array or structure
1581 In ISO C99 you can give the elements in any order, specifying the array
1582 indices or structure field names they apply to, and GNU C allows this as
1583 an extension in C89 mode as well. This extension is not
1584 implemented in GNU C++.
1586 To specify an array index, write
1587 @samp{[@var{index}] =} before the element value. For example,
1590 int a[6] = @{ [4] = 29, [2] = 15 @};
1597 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1601 The index values must be constant expressions, even if the array being
1602 initialized is automatic.
1604 An alternative syntax for this which has been obsolete since GCC 2.5 but
1605 GCC still accepts is to write @samp{[@var{index}]} before the element
1606 value, with no @samp{=}.
1608 To initialize a range of elements to the same value, write
1609 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1610 extension. For example,
1613 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1617 If the value in it has side-effects, the side-effects will happen only once,
1618 not for each initialized field by the range initializer.
1621 Note that the length of the array is the highest value specified
1624 In a structure initializer, specify the name of a field to initialize
1625 with @samp{.@var{fieldname} =} before the element value. For example,
1626 given the following structure,
1629 struct point @{ int x, y; @};
1633 the following initialization
1636 struct point p = @{ .y = yvalue, .x = xvalue @};
1643 struct point p = @{ xvalue, yvalue @};
1646 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1647 @samp{@var{fieldname}:}, as shown here:
1650 struct point p = @{ y: yvalue, x: xvalue @};
1654 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1655 @dfn{designator}. You can also use a designator (or the obsolete colon
1656 syntax) when initializing a union, to specify which element of the union
1657 should be used. For example,
1660 union foo @{ int i; double d; @};
1662 union foo f = @{ .d = 4 @};
1666 will convert 4 to a @code{double} to store it in the union using
1667 the second element. By contrast, casting 4 to type @code{union foo}
1668 would store it into the union as the integer @code{i}, since it is
1669 an integer. (@xref{Cast to Union}.)
1671 You can combine this technique of naming elements with ordinary C
1672 initialization of successive elements. Each initializer element that
1673 does not have a designator applies to the next consecutive element of the
1674 array or structure. For example,
1677 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1684 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1687 Labeling the elements of an array initializer is especially useful
1688 when the indices are characters or belong to an @code{enum} type.
1693 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1694 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1697 @cindex designator lists
1698 You can also write a series of @samp{.@var{fieldname}} and
1699 @samp{[@var{index}]} designators before an @samp{=} to specify a
1700 nested subobject to initialize; the list is taken relative to the
1701 subobject corresponding to the closest surrounding brace pair. For
1702 example, with the @samp{struct point} declaration above:
1705 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1709 If the same field is initialized multiple times, it will have value from
1710 the last initialization. If any such overridden initialization has
1711 side-effect, it is unspecified whether the side-effect happens or not.
1712 Currently, GCC will discard them and issue a warning.
1715 @section Case Ranges
1717 @cindex ranges in case statements
1719 You can specify a range of consecutive values in a single @code{case} label,
1723 case @var{low} ... @var{high}:
1727 This has the same effect as the proper number of individual @code{case}
1728 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1730 This feature is especially useful for ranges of ASCII character codes:
1736 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1737 it may be parsed wrong when you use it with integer values. For example,
1752 @section Cast to a Union Type
1753 @cindex cast to a union
1754 @cindex union, casting to a
1756 A cast to union type is similar to other casts, except that the type
1757 specified is a union type. You can specify the type either with
1758 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1759 a constructor though, not a cast, and hence does not yield an lvalue like
1760 normal casts. (@xref{Compound Literals}.)
1762 The types that may be cast to the union type are those of the members
1763 of the union. Thus, given the following union and variables:
1766 union foo @{ int i; double d; @};
1772 both @code{x} and @code{y} can be cast to type @code{union foo}.
1774 Using the cast as the right-hand side of an assignment to a variable of
1775 union type is equivalent to storing in a member of the union:
1780 u = (union foo) x @equiv{} u.i = x
1781 u = (union foo) y @equiv{} u.d = y
1784 You can also use the union cast as a function argument:
1787 void hack (union foo);
1789 hack ((union foo) x);
1792 @node Mixed Declarations
1793 @section Mixed Declarations and Code
1794 @cindex mixed declarations and code
1795 @cindex declarations, mixed with code
1796 @cindex code, mixed with declarations
1798 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1799 within compound statements. As an extension, GCC also allows this in
1800 C89 mode. For example, you could do:
1809 Each identifier is visible from where it is declared until the end of
1810 the enclosing block.
1812 @node Function Attributes
1813 @section Declaring Attributes of Functions
1814 @cindex function attributes
1815 @cindex declaring attributes of functions
1816 @cindex functions that never return
1817 @cindex functions that return more than once
1818 @cindex functions that have no side effects
1819 @cindex functions in arbitrary sections
1820 @cindex functions that behave like malloc
1821 @cindex @code{volatile} applied to function
1822 @cindex @code{const} applied to function
1823 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1824 @cindex functions with non-null pointer arguments
1825 @cindex functions that are passed arguments in registers on the 386
1826 @cindex functions that pop the argument stack on the 386
1827 @cindex functions that do not pop the argument stack on the 386
1828 @cindex functions that have different compilation options on the 386
1829 @cindex functions that have different optimization options
1831 In GNU C, you declare certain things about functions called in your program
1832 which help the compiler optimize function calls and check your code more
1835 The keyword @code{__attribute__} allows you to specify special
1836 attributes when making a declaration. This keyword is followed by an
1837 attribute specification inside double parentheses. The following
1838 attributes are currently defined for functions on all targets:
1839 @code{aligned}, @code{alloc_size}, @code{noreturn},
1840 @code{returns_twice}, @code{noinline}, @code{always_inline},
1841 @code{flatten}, @code{pure}, @code{const}, @code{nothrow},
1842 @code{sentinel}, @code{format}, @code{format_arg},
1843 @code{no_instrument_function}, @code{section}, @code{constructor},
1844 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1845 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1846 @code{nonnull}, @code{gnu_inline}, @code{externally_visible},
1847 @code{hot}, @code{cold}, @code{artificial}, @code{error}
1849 Several other attributes are defined for functions on particular
1850 target systems. Other attributes, including @code{section} are
1851 supported for variables declarations (@pxref{Variable Attributes}) and
1852 for types (@pxref{Type Attributes}).
1854 You may also specify attributes with @samp{__} preceding and following
1855 each keyword. This allows you to use them in header files without
1856 being concerned about a possible macro of the same name. For example,
1857 you may use @code{__noreturn__} instead of @code{noreturn}.
1859 @xref{Attribute Syntax}, for details of the exact syntax for using
1863 @c Keep this table alphabetized by attribute name. Treat _ as space.
1865 @item alias ("@var{target}")
1866 @cindex @code{alias} attribute
1867 The @code{alias} attribute causes the declaration to be emitted as an
1868 alias for another symbol, which must be specified. For instance,
1871 void __f () @{ /* @r{Do something.} */; @}
1872 void f () __attribute__ ((weak, alias ("__f")));
1875 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1876 mangled name for the target must be used. It is an error if @samp{__f}
1877 is not defined in the same translation unit.
1879 Not all target machines support this attribute.
1881 @item aligned (@var{alignment})
1882 @cindex @code{aligned} attribute
1883 This attribute specifies a minimum alignment for the function,
1886 You cannot use this attribute to decrease the alignment of a function,
1887 only to increase it. However, when you explicitly specify a function
1888 alignment this will override the effect of the
1889 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1892 Note that the effectiveness of @code{aligned} attributes may be
1893 limited by inherent limitations in your linker. On many systems, the
1894 linker is only able to arrange for functions to be aligned up to a
1895 certain maximum alignment. (For some linkers, the maximum supported
1896 alignment may be very very small.) See your linker documentation for
1897 further information.
1899 The @code{aligned} attribute can also be used for variables and fields
1900 (@pxref{Variable Attributes}.)
1903 @cindex @code{alloc_size} attribute
1904 The @code{alloc_size} attribute is used to tell the compiler that the
1905 function return value points to memory, where the size is given by
1906 one or two of the functions parameters. GCC uses this
1907 information to improve the correctness of @code{__builtin_object_size}.
1909 The function parameter(s) denoting the allocated size are specified by
1910 one or two integer arguments supplied to the attribute. The allocated size
1911 is either the value of the single function argument specified or the product
1912 of the two function arguments specified. Argument numbering starts at
1918 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1919 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
1922 declares that my_calloc will return memory of the size given by
1923 the product of parameter 1 and 2 and that my_realloc will return memory
1924 of the size given by parameter 2.
1927 @cindex @code{always_inline} function attribute
1928 Generally, functions are not inlined unless optimization is specified.
1929 For functions declared inline, this attribute inlines the function even
1930 if no optimization level was specified.
1933 @cindex @code{gnu_inline} function attribute
1934 This attribute should be used with a function which is also declared
1935 with the @code{inline} keyword. It directs GCC to treat the function
1936 as if it were defined in gnu89 mode even when compiling in C99 or
1939 If the function is declared @code{extern}, then this definition of the
1940 function is used only for inlining. In no case is the function
1941 compiled as a standalone function, not even if you take its address
1942 explicitly. Such an address becomes an external reference, as if you
1943 had only declared the function, and had not defined it. This has
1944 almost the effect of a macro. The way to use this is to put a
1945 function definition in a header file with this attribute, and put
1946 another copy of the function, without @code{extern}, in a library
1947 file. The definition in the header file will cause most calls to the
1948 function to be inlined. If any uses of the function remain, they will
1949 refer to the single copy in the library. Note that the two
1950 definitions of the functions need not be precisely the same, although
1951 if they do not have the same effect your program may behave oddly.
1953 In C, if the function is neither @code{extern} nor @code{static}, then
1954 the function is compiled as a standalone function, as well as being
1955 inlined where possible.
1957 This is how GCC traditionally handled functions declared
1958 @code{inline}. Since ISO C99 specifies a different semantics for
1959 @code{inline}, this function attribute is provided as a transition
1960 measure and as a useful feature in its own right. This attribute is
1961 available in GCC 4.1.3 and later. It is available if either of the
1962 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1963 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1964 Function is As Fast As a Macro}.
1966 In C++, this attribute does not depend on @code{extern} in any way,
1967 but it still requires the @code{inline} keyword to enable its special
1971 @cindex @code{artificial} function attribute
1972 This attribute is useful for small inline wrappers which if possible
1973 should appear during debugging as a unit, depending on the debug
1974 info format it will either mean marking the function as artificial
1975 or using the caller location for all instructions within the inlined
1979 @cindex @code{flatten} function attribute
1980 Generally, inlining into a function is limited. For a function marked with
1981 this attribute, every call inside this function will be inlined, if possible.
1982 Whether the function itself is considered for inlining depends on its size and
1983 the current inlining parameters.
1985 @item error ("@var{message}")
1986 @cindex @code{error} function attribute
1987 If this attribute is used on a function declaration and a call to such a function
1988 is not eliminated through dead code elimination or other optimizations, an error
1989 which will include @var{message} will be diagnosed. This is useful
1990 for compile time checking, especially together with @code{__builtin_constant_p}
1991 and inline functions where checking the inline function arguments is not
1992 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
1993 While it is possible to leave the function undefined and thus invoke
1994 a link failure, when using this attribute the problem will be diagnosed
1995 earlier and with exact location of the call even in presence of inline
1996 functions or when not emitting debugging information.
1998 @item warning ("@var{message}")
1999 @cindex @code{warning} function attribute
2000 If this attribute is used on a function declaration and a call to such a function
2001 is not eliminated through dead code elimination or other optimizations, a warning
2002 which will include @var{message} will be diagnosed. This is useful
2003 for compile time checking, especially together with @code{__builtin_constant_p}
2004 and inline functions. While it is possible to define the function with
2005 a message in @code{.gnu.warning*} section, when using this attribute the problem
2006 will be diagnosed earlier and with exact location of the call even in presence
2007 of inline functions or when not emitting debugging information.
2010 @cindex functions that do pop the argument stack on the 386
2012 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2013 assume that the calling function will pop off the stack space used to
2014 pass arguments. This is
2015 useful to override the effects of the @option{-mrtd} switch.
2018 @cindex @code{const} function attribute
2019 Many functions do not examine any values except their arguments, and
2020 have no effects except the return value. Basically this is just slightly
2021 more strict class than the @code{pure} attribute below, since function is not
2022 allowed to read global memory.
2024 @cindex pointer arguments
2025 Note that a function that has pointer arguments and examines the data
2026 pointed to must @emph{not} be declared @code{const}. Likewise, a
2027 function that calls a non-@code{const} function usually must not be
2028 @code{const}. It does not make sense for a @code{const} function to
2031 The attribute @code{const} is not implemented in GCC versions earlier
2032 than 2.5. An alternative way to declare that a function has no side
2033 effects, which works in the current version and in some older versions,
2037 typedef int intfn ();
2039 extern const intfn square;
2042 This approach does not work in GNU C++ from 2.6.0 on, since the language
2043 specifies that the @samp{const} must be attached to the return value.
2047 @itemx constructor (@var{priority})
2048 @itemx destructor (@var{priority})
2049 @cindex @code{constructor} function attribute
2050 @cindex @code{destructor} function attribute
2051 The @code{constructor} attribute causes the function to be called
2052 automatically before execution enters @code{main ()}. Similarly, the
2053 @code{destructor} attribute causes the function to be called
2054 automatically after @code{main ()} has completed or @code{exit ()} has
2055 been called. Functions with these attributes are useful for
2056 initializing data that will be used implicitly during the execution of
2059 You may provide an optional integer priority to control the order in
2060 which constructor and destructor functions are run. A constructor
2061 with a smaller priority number runs before a constructor with a larger
2062 priority number; the opposite relationship holds for destructors. So,
2063 if you have a constructor that allocates a resource and a destructor
2064 that deallocates the same resource, both functions typically have the
2065 same priority. The priorities for constructor and destructor
2066 functions are the same as those specified for namespace-scope C++
2067 objects (@pxref{C++ Attributes}).
2069 These attributes are not currently implemented for Objective-C@.
2072 @cindex @code{deprecated} attribute.
2073 The @code{deprecated} attribute results in a warning if the function
2074 is used anywhere in the source file. This is useful when identifying
2075 functions that are expected to be removed in a future version of a
2076 program. The warning also includes the location of the declaration
2077 of the deprecated function, to enable users to easily find further
2078 information about why the function is deprecated, or what they should
2079 do instead. Note that the warnings only occurs for uses:
2082 int old_fn () __attribute__ ((deprecated));
2084 int (*fn_ptr)() = old_fn;
2087 results in a warning on line 3 but not line 2.
2089 The @code{deprecated} attribute can also be used for variables and
2090 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2093 @cindex @code{__declspec(dllexport)}
2094 On Microsoft Windows targets and Symbian OS targets the
2095 @code{dllexport} attribute causes the compiler to provide a global
2096 pointer to a pointer in a DLL, so that it can be referenced with the
2097 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2098 name is formed by combining @code{_imp__} and the function or variable
2101 You can use @code{__declspec(dllexport)} as a synonym for
2102 @code{__attribute__ ((dllexport))} for compatibility with other
2105 On systems that support the @code{visibility} attribute, this
2106 attribute also implies ``default'' visibility. It is an error to
2107 explicitly specify any other visibility.
2109 Currently, the @code{dllexport} attribute is ignored for inlined
2110 functions, unless the @option{-fkeep-inline-functions} flag has been
2111 used. The attribute is also ignored for undefined symbols.
2113 When applied to C++ classes, the attribute marks defined non-inlined
2114 member functions and static data members as exports. Static consts
2115 initialized in-class are not marked unless they are also defined
2118 For Microsoft Windows targets there are alternative methods for
2119 including the symbol in the DLL's export table such as using a
2120 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2121 the @option{--export-all} linker flag.
2124 @cindex @code{__declspec(dllimport)}
2125 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2126 attribute causes the compiler to reference a function or variable via
2127 a global pointer to a pointer that is set up by the DLL exporting the
2128 symbol. The attribute implies @code{extern}. On Microsoft Windows
2129 targets, the pointer name is formed by combining @code{_imp__} and the
2130 function or variable name.
2132 You can use @code{__declspec(dllimport)} as a synonym for
2133 @code{__attribute__ ((dllimport))} for compatibility with other
2136 On systems that support the @code{visibility} attribute, this
2137 attribute also implies ``default'' visibility. It is an error to
2138 explicitly specify any other visibility.
2140 Currently, the attribute is ignored for inlined functions. If the
2141 attribute is applied to a symbol @emph{definition}, an error is reported.
2142 If a symbol previously declared @code{dllimport} is later defined, the
2143 attribute is ignored in subsequent references, and a warning is emitted.
2144 The attribute is also overridden by a subsequent declaration as
2147 When applied to C++ classes, the attribute marks non-inlined
2148 member functions and static data members as imports. However, the
2149 attribute is ignored for virtual methods to allow creation of vtables
2152 On the SH Symbian OS target the @code{dllimport} attribute also has
2153 another affect---it can cause the vtable and run-time type information
2154 for a class to be exported. This happens when the class has a
2155 dllimport'ed constructor or a non-inline, non-pure virtual function
2156 and, for either of those two conditions, the class also has a inline
2157 constructor or destructor and has a key function that is defined in
2158 the current translation unit.
2160 For Microsoft Windows based targets the use of the @code{dllimport}
2161 attribute on functions is not necessary, but provides a small
2162 performance benefit by eliminating a thunk in the DLL@. The use of the
2163 @code{dllimport} attribute on imported variables was required on older
2164 versions of the GNU linker, but can now be avoided by passing the
2165 @option{--enable-auto-import} switch to the GNU linker. As with
2166 functions, using the attribute for a variable eliminates a thunk in
2169 One drawback to using this attribute is that a pointer to a
2170 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2171 address. However, a pointer to a @emph{function} with the
2172 @code{dllimport} attribute can be used as a constant initializer; in
2173 this case, the address of a stub function in the import lib is
2174 referenced. On Microsoft Windows targets, the attribute can be disabled
2175 for functions by setting the @option{-mnop-fun-dllimport} flag.
2178 @cindex eight bit data on the H8/300, H8/300H, and H8S
2179 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2180 variable should be placed into the eight bit data section.
2181 The compiler will generate more efficient code for certain operations
2182 on data in the eight bit data area. Note the eight bit data area is limited to
2185 You must use GAS and GLD from GNU binutils version 2.7 or later for
2186 this attribute to work correctly.
2188 @item exception_handler
2189 @cindex exception handler functions on the Blackfin processor
2190 Use this attribute on the Blackfin to indicate that the specified function
2191 is an exception handler. The compiler will generate function entry and
2192 exit sequences suitable for use in an exception handler when this
2193 attribute is present.
2195 @item externally_visible
2196 @cindex @code{externally_visible} attribute.
2197 This attribute, attached to a global variable or function, nullifies
2198 the effect of the @option{-fwhole-program} command-line option, so the
2199 object remains visible outside the current compilation unit.
2202 @cindex functions which handle memory bank switching
2203 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2204 use a calling convention that takes care of switching memory banks when
2205 entering and leaving a function. This calling convention is also the
2206 default when using the @option{-mlong-calls} option.
2208 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2209 to call and return from a function.
2211 On 68HC11 the compiler will generate a sequence of instructions
2212 to invoke a board-specific routine to switch the memory bank and call the
2213 real function. The board-specific routine simulates a @code{call}.
2214 At the end of a function, it will jump to a board-specific routine
2215 instead of using @code{rts}. The board-specific return routine simulates
2219 @cindex functions that pop the argument stack on the 386
2220 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2221 pass the first argument (if of integral type) in the register ECX and
2222 the second argument (if of integral type) in the register EDX@. Subsequent
2223 and other typed arguments are passed on the stack. The called function will
2224 pop the arguments off the stack. If the number of arguments is variable all
2225 arguments are pushed on the stack.
2227 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2228 @cindex @code{format} function attribute
2230 The @code{format} attribute specifies that a function takes @code{printf},
2231 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2232 should be type-checked against a format string. For example, the
2237 my_printf (void *my_object, const char *my_format, ...)
2238 __attribute__ ((format (printf, 2, 3)));
2242 causes the compiler to check the arguments in calls to @code{my_printf}
2243 for consistency with the @code{printf} style format string argument
2246 The parameter @var{archetype} determines how the format string is
2247 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2248 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2249 @code{strfmon}. (You can also use @code{__printf__},
2250 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2251 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2252 @code{ms_strftime} are also present.
2253 @var{archtype} values such as @code{printf} refer to the formats accepted
2254 by the system's C run-time library, while @code{gnu_} values always refer
2255 to the formats accepted by the GNU C Library. On Microsoft Windows
2256 targets, @code{ms_} values refer to the formats accepted by the
2257 @file{msvcrt.dll} library.
2258 The parameter @var{string-index}
2259 specifies which argument is the format string argument (starting
2260 from 1), while @var{first-to-check} is the number of the first
2261 argument to check against the format string. For functions
2262 where the arguments are not available to be checked (such as
2263 @code{vprintf}), specify the third parameter as zero. In this case the
2264 compiler only checks the format string for consistency. For
2265 @code{strftime} formats, the third parameter is required to be zero.
2266 Since non-static C++ methods have an implicit @code{this} argument, the
2267 arguments of such methods should be counted from two, not one, when
2268 giving values for @var{string-index} and @var{first-to-check}.
2270 In the example above, the format string (@code{my_format}) is the second
2271 argument of the function @code{my_print}, and the arguments to check
2272 start with the third argument, so the correct parameters for the format
2273 attribute are 2 and 3.
2275 @opindex ffreestanding
2276 @opindex fno-builtin
2277 The @code{format} attribute allows you to identify your own functions
2278 which take format strings as arguments, so that GCC can check the
2279 calls to these functions for errors. The compiler always (unless
2280 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2281 for the standard library functions @code{printf}, @code{fprintf},
2282 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2283 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2284 warnings are requested (using @option{-Wformat}), so there is no need to
2285 modify the header file @file{stdio.h}. In C99 mode, the functions
2286 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2287 @code{vsscanf} are also checked. Except in strictly conforming C
2288 standard modes, the X/Open function @code{strfmon} is also checked as
2289 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2290 @xref{C Dialect Options,,Options Controlling C Dialect}.
2292 The target may provide additional types of format checks.
2293 @xref{Target Format Checks,,Format Checks Specific to Particular
2296 @item format_arg (@var{string-index})
2297 @cindex @code{format_arg} function attribute
2298 @opindex Wformat-nonliteral
2299 The @code{format_arg} attribute specifies that a function takes a format
2300 string for a @code{printf}, @code{scanf}, @code{strftime} or
2301 @code{strfmon} style function and modifies it (for example, to translate
2302 it into another language), so the result can be passed to a
2303 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2304 function (with the remaining arguments to the format function the same
2305 as they would have been for the unmodified string). For example, the
2310 my_dgettext (char *my_domain, const char *my_format)
2311 __attribute__ ((format_arg (2)));
2315 causes the compiler to check the arguments in calls to a @code{printf},
2316 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2317 format string argument is a call to the @code{my_dgettext} function, for
2318 consistency with the format string argument @code{my_format}. If the
2319 @code{format_arg} attribute had not been specified, all the compiler
2320 could tell in such calls to format functions would be that the format
2321 string argument is not constant; this would generate a warning when
2322 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2323 without the attribute.
2325 The parameter @var{string-index} specifies which argument is the format
2326 string argument (starting from one). Since non-static C++ methods have
2327 an implicit @code{this} argument, the arguments of such methods should
2328 be counted from two.
2330 The @code{format-arg} attribute allows you to identify your own
2331 functions which modify format strings, so that GCC can check the
2332 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2333 type function whose operands are a call to one of your own function.
2334 The compiler always treats @code{gettext}, @code{dgettext}, and
2335 @code{dcgettext} in this manner except when strict ISO C support is
2336 requested by @option{-ansi} or an appropriate @option{-std} option, or
2337 @option{-ffreestanding} or @option{-fno-builtin}
2338 is used. @xref{C Dialect Options,,Options
2339 Controlling C Dialect}.
2341 @item function_vector
2342 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2343 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2344 function should be called through the function vector. Calling a
2345 function through the function vector will reduce code size, however;
2346 the function vector has a limited size (maximum 128 entries on the H8/300
2347 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2349 In SH2A target, this attribute declares a function to be called using the
2350 TBR relative addressing mode. The argument to this attribute is the entry
2351 number of the same function in a vector table containing all the TBR
2352 relative addressable functions. For the successful jump, register TBR
2353 should contain the start address of this TBR relative vector table.
2354 In the startup routine of the user application, user needs to care of this
2355 TBR register initialization. The TBR relative vector table can have at
2356 max 256 function entries. The jumps to these functions will be generated
2357 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2358 You must use GAS and GLD from GNU binutils version 2.7 or later for
2359 this attribute to work correctly.
2361 Please refer the example of M16C target, to see the use of this
2362 attribute while declaring a function,
2364 In an application, for a function being called once, this attribute will
2365 save at least 8 bytes of code; and if other successive calls are being
2366 made to the same function, it will save 2 bytes of code per each of these
2369 On M16C/M32C targets, the @code{function_vector} attribute declares a
2370 special page subroutine call function. Use of this attribute reduces
2371 the code size by 2 bytes for each call generated to the
2372 subroutine. The argument to the attribute is the vector number entry
2373 from the special page vector table which contains the 16 low-order
2374 bits of the subroutine's entry address. Each vector table has special
2375 page number (18 to 255) which are used in @code{jsrs} instruction.
2376 Jump addresses of the routines are generated by adding 0x0F0000 (in
2377 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2378 byte addresses set in the vector table. Therefore you need to ensure
2379 that all the special page vector routines should get mapped within the
2380 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2383 In the following example 2 bytes will be saved for each call to
2384 function @code{foo}.
2387 void foo (void) __attribute__((function_vector(0x18)));
2398 If functions are defined in one file and are called in another file,
2399 then be sure to write this declaration in both files.
2401 This attribute is ignored for R8C target.
2404 @cindex interrupt handler functions
2405 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k,
2406 and Xstormy16 ports to indicate that the specified function is an
2407 interrupt handler. The compiler will generate function entry and exit
2408 sequences suitable for use in an interrupt handler when this attribute
2411 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2412 SH processors can be specified via the @code{interrupt_handler} attribute.
2414 Note, on the AVR, interrupts will be enabled inside the function.
2416 Note, for the ARM, you can specify the kind of interrupt to be handled by
2417 adding an optional parameter to the interrupt attribute like this:
2420 void f () __attribute__ ((interrupt ("IRQ")));
2423 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2425 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2426 may be called with a word aligned stack pointer.
2428 @item interrupt_handler
2429 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2430 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2431 indicate that the specified function is an interrupt handler. The compiler
2432 will generate function entry and exit sequences suitable for use in an
2433 interrupt handler when this attribute is present.
2435 @item interrupt_thread
2436 @cindex interrupt thread functions on fido
2437 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2438 that the specified function is an interrupt handler that is designed
2439 to run as a thread. The compiler omits generate prologue/epilogue
2440 sequences and replaces the return instruction with a @code{sleep}
2441 instruction. This attribute is available only on fido.
2444 @cindex interrupt service routines on ARM
2445 Use this attribute on ARM to write Interrupt Service Routines. This is an
2446 alias to the @code{interrupt} attribute above.
2449 @cindex User stack pointer in interrupts on the Blackfin
2450 When used together with @code{interrupt_handler}, @code{exception_handler}
2451 or @code{nmi_handler}, code will be generated to load the stack pointer
2452 from the USP register in the function prologue.
2455 @cindex @code{l1_text} function attribute
2456 This attribute specifies a function to be placed into L1 Instruction
2457 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2458 With @option{-mfdpic}, function calls with a such function as the callee
2459 or caller will use inlined PLT.
2461 @item long_call/short_call
2462 @cindex indirect calls on ARM
2463 This attribute specifies how a particular function is called on
2464 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2465 command line switch and @code{#pragma long_calls} settings. The
2466 @code{long_call} attribute indicates that the function might be far
2467 away from the call site and require a different (more expensive)
2468 calling sequence. The @code{short_call} attribute always places
2469 the offset to the function from the call site into the @samp{BL}
2470 instruction directly.
2472 @item longcall/shortcall
2473 @cindex functions called via pointer on the RS/6000 and PowerPC
2474 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2475 indicates that the function might be far away from the call site and
2476 require a different (more expensive) calling sequence. The
2477 @code{shortcall} attribute indicates that the function is always close
2478 enough for the shorter calling sequence to be used. These attributes
2479 override both the @option{-mlongcall} switch and, on the RS/6000 and
2480 PowerPC, the @code{#pragma longcall} setting.
2482 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2483 calls are necessary.
2485 @item long_call/near/far
2486 @cindex indirect calls on MIPS
2487 These attributes specify how a particular function is called on MIPS@.
2488 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2489 command-line switch. The @code{long_call} and @code{far} attributes are
2490 synonyms, and cause the compiler to always call
2491 the function by first loading its address into a register, and then using
2492 the contents of that register. The @code{near} attribute has the opposite
2493 effect; it specifies that non-PIC calls should be made using the more
2494 efficient @code{jal} instruction.
2497 @cindex @code{malloc} attribute
2498 The @code{malloc} attribute is used to tell the compiler that a function
2499 may be treated as if any non-@code{NULL} pointer it returns cannot
2500 alias any other pointer valid when the function returns.
2501 This will often improve optimization.
2502 Standard functions with this property include @code{malloc} and
2503 @code{calloc}. @code{realloc}-like functions have this property as
2504 long as the old pointer is never referred to (including comparing it
2505 to the new pointer) after the function returns a non-@code{NULL}
2508 @item mips16/nomips16
2509 @cindex @code{mips16} attribute
2510 @cindex @code{nomips16} attribute
2512 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2513 function attributes to locally select or turn off MIPS16 code generation.
2514 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2515 while MIPS16 code generation is disabled for functions with the
2516 @code{nomips16} attribute. These attributes override the
2517 @option{-mips16} and @option{-mno-mips16} options on the command line
2518 (@pxref{MIPS Options}).
2520 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2521 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2522 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2523 may interact badly with some GCC extensions such as @code{__builtin_apply}
2524 (@pxref{Constructing Calls}).
2526 @item model (@var{model-name})
2527 @cindex function addressability on the M32R/D
2528 @cindex variable addressability on the IA-64
2530 On the M32R/D, use this attribute to set the addressability of an
2531 object, and of the code generated for a function. The identifier
2532 @var{model-name} is one of @code{small}, @code{medium}, or
2533 @code{large}, representing each of the code models.
2535 Small model objects live in the lower 16MB of memory (so that their
2536 addresses can be loaded with the @code{ld24} instruction), and are
2537 callable with the @code{bl} instruction.
2539 Medium model objects may live anywhere in the 32-bit address space (the
2540 compiler will generate @code{seth/add3} instructions to load their addresses),
2541 and are callable with the @code{bl} instruction.
2543 Large model objects may live anywhere in the 32-bit address space (the
2544 compiler will generate @code{seth/add3} instructions to load their addresses),
2545 and may not be reachable with the @code{bl} instruction (the compiler will
2546 generate the much slower @code{seth/add3/jl} instruction sequence).
2548 On IA-64, use this attribute to set the addressability of an object.
2549 At present, the only supported identifier for @var{model-name} is
2550 @code{small}, indicating addressability via ``small'' (22-bit)
2551 addresses (so that their addresses can be loaded with the @code{addl}
2552 instruction). Caveat: such addressing is by definition not position
2553 independent and hence this attribute must not be used for objects
2554 defined by shared libraries.
2556 @item ms_abi/sysv_abi
2557 @cindex @code{ms_abi} attribute
2558 @cindex @code{sysv_abi} attribute
2560 On 64-bit x86_64-*-* targets, you can use an ABI attribute to indicate
2561 which calling convention should be used for a function. The @code{ms_abi}
2562 attribute tells the compiler to use the Microsoft ABI, while the
2563 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2564 GNU/Linux and other systems. The default is to use the Microsoft ABI
2565 when targeting Windows. On all other systems, the default is the AMD ABI.
2567 Note, This feature is currently sorried out for Windows targets trying to
2570 @cindex function without a prologue/epilogue code
2571 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate that
2572 the specified function does not need prologue/epilogue sequences generated by
2573 the compiler. It is up to the programmer to provide these sequences. The
2574 only statements that can be safely included in naked functions are
2575 @code{asm} statements that do not have operands. All other statements,
2576 including declarations of local variables, @code{if} statements, and so
2577 forth, should be avoided. Naked functions should be used to implement the
2578 body of an assembly function, while allowing the compiler to construct
2579 the requisite function declaration for the assembler.
2582 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2583 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2584 use the normal calling convention based on @code{jsr} and @code{rts}.
2585 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2589 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2590 Use this attribute together with @code{interrupt_handler},
2591 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2592 entry code should enable nested interrupts or exceptions.
2595 @cindex NMI handler functions on the Blackfin processor
2596 Use this attribute on the Blackfin to indicate that the specified function
2597 is an NMI handler. The compiler will generate function entry and
2598 exit sequences suitable for use in an NMI handler when this
2599 attribute is present.
2601 @item no_instrument_function
2602 @cindex @code{no_instrument_function} function attribute
2603 @opindex finstrument-functions
2604 If @option{-finstrument-functions} is given, profiling function calls will
2605 be generated at entry and exit of most user-compiled functions.
2606 Functions with this attribute will not be so instrumented.
2609 @cindex @code{noinline} function attribute
2610 This function attribute prevents a function from being considered for
2612 @c Don't enumerate the optimizations by name here; we try to be
2613 @c future-compatible with this mechanism.
2614 If the function does not have side-effects, there are optimizations
2615 other than inlining that causes function calls to be optimized away,
2616 although the function call is live. To keep such calls from being
2621 (@pxref{Extended Asm}) in the called function, to serve as a special
2624 @item nonnull (@var{arg-index}, @dots{})
2625 @cindex @code{nonnull} function attribute
2626 The @code{nonnull} attribute specifies that some function parameters should
2627 be non-null pointers. For instance, the declaration:
2631 my_memcpy (void *dest, const void *src, size_t len)
2632 __attribute__((nonnull (1, 2)));
2636 causes the compiler to check that, in calls to @code{my_memcpy},
2637 arguments @var{dest} and @var{src} are non-null. If the compiler
2638 determines that a null pointer is passed in an argument slot marked
2639 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2640 is issued. The compiler may also choose to make optimizations based
2641 on the knowledge that certain function arguments will not be null.
2643 If no argument index list is given to the @code{nonnull} attribute,
2644 all pointer arguments are marked as non-null. To illustrate, the
2645 following declaration is equivalent to the previous example:
2649 my_memcpy (void *dest, const void *src, size_t len)
2650 __attribute__((nonnull));
2654 @cindex @code{noreturn} function attribute
2655 A few standard library functions, such as @code{abort} and @code{exit},
2656 cannot return. GCC knows this automatically. Some programs define
2657 their own functions that never return. You can declare them
2658 @code{noreturn} to tell the compiler this fact. For example,
2662 void fatal () __attribute__ ((noreturn));
2665 fatal (/* @r{@dots{}} */)
2667 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2673 The @code{noreturn} keyword tells the compiler to assume that
2674 @code{fatal} cannot return. It can then optimize without regard to what
2675 would happen if @code{fatal} ever did return. This makes slightly
2676 better code. More importantly, it helps avoid spurious warnings of
2677 uninitialized variables.
2679 The @code{noreturn} keyword does not affect the exceptional path when that
2680 applies: a @code{noreturn}-marked function may still return to the caller
2681 by throwing an exception or calling @code{longjmp}.
2683 Do not assume that registers saved by the calling function are
2684 restored before calling the @code{noreturn} function.
2686 It does not make sense for a @code{noreturn} function to have a return
2687 type other than @code{void}.
2689 The attribute @code{noreturn} is not implemented in GCC versions
2690 earlier than 2.5. An alternative way to declare that a function does
2691 not return, which works in the current version and in some older
2692 versions, is as follows:
2695 typedef void voidfn ();
2697 volatile voidfn fatal;
2700 This approach does not work in GNU C++.
2703 @cindex @code{nothrow} function attribute
2704 The @code{nothrow} attribute is used to inform the compiler that a
2705 function cannot throw an exception. For example, most functions in
2706 the standard C library can be guaranteed not to throw an exception
2707 with the notable exceptions of @code{qsort} and @code{bsearch} that
2708 take function pointer arguments. The @code{nothrow} attribute is not
2709 implemented in GCC versions earlier than 3.3.
2712 @cindex @code{optimize} function attribute
2713 The @code{optimize} attribute is used to specify that a function is to
2714 be compiled with different optimization options than specified on the
2715 command line. Arguments can either be numbers or strings. Numbers
2716 are assumed to be an optimization level. Strings that begin with
2717 @code{O} are assumed to be an optimization option, while other options
2718 are assumed to be used with a @code{-f} prefix. You can also use the
2719 @samp{#pragma GCC optimize} pragma to set the optimization options
2720 that affect more than one function.
2721 @xref{Function Specific Option Pragmas}, for details about the
2722 @samp{#pragma GCC optimize} pragma.
2724 This can be used for instance to have frequently executed functions
2725 compiled with more aggressive optimization options that produce faster
2726 and larger code, while other functions can be called with less
2730 @cindex @code{pure} function attribute
2731 Many functions have no effects except the return value and their
2732 return value depends only on the parameters and/or global variables.
2733 Such a function can be subject
2734 to common subexpression elimination and loop optimization just as an
2735 arithmetic operator would be. These functions should be declared
2736 with the attribute @code{pure}. For example,
2739 int square (int) __attribute__ ((pure));
2743 says that the hypothetical function @code{square} is safe to call
2744 fewer times than the program says.
2746 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2747 Interesting non-pure functions are functions with infinite loops or those
2748 depending on volatile memory or other system resource, that may change between
2749 two consecutive calls (such as @code{feof} in a multithreading environment).
2751 The attribute @code{pure} is not implemented in GCC versions earlier
2755 @cindex @code{hot} function attribute
2756 The @code{hot} attribute is used to inform the compiler that a function is a
2757 hot spot of the compiled program. The function is optimized more aggressively
2758 and on many target it is placed into special subsection of the text section so
2759 all hot functions appears close together improving locality.
2761 When profile feedback is available, via @option{-fprofile-use}, hot functions
2762 are automatically detected and this attribute is ignored.
2764 The @code{hot} attribute is not implemented in GCC versions earlier
2768 @cindex @code{cold} function attribute
2769 The @code{cold} attribute is used to inform the compiler that a function is
2770 unlikely executed. The function is optimized for size rather than speed and on
2771 many targets it is placed into special subsection of the text section so all
2772 cold functions appears close together improving code locality of non-cold parts
2773 of program. The paths leading to call of cold functions within code are marked
2774 as unlikely by the branch prediction mechanism. It is thus useful to mark
2775 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2776 improve optimization of hot functions that do call marked functions in rare
2779 When profile feedback is available, via @option{-fprofile-use}, hot functions
2780 are automatically detected and this attribute is ignored.
2782 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
2784 @item regparm (@var{number})
2785 @cindex @code{regparm} attribute
2786 @cindex functions that are passed arguments in registers on the 386
2787 On the Intel 386, the @code{regparm} attribute causes the compiler to
2788 pass arguments number one to @var{number} if they are of integral type
2789 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2790 take a variable number of arguments will continue to be passed all of their
2791 arguments on the stack.
2793 Beware that on some ELF systems this attribute is unsuitable for
2794 global functions in shared libraries with lazy binding (which is the
2795 default). Lazy binding will send the first call via resolving code in
2796 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2797 per the standard calling conventions. Solaris 8 is affected by this.
2798 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2799 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
2800 disabled with the linker or the loader if desired, to avoid the
2804 @cindex @code{sseregparm} attribute
2805 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2806 causes the compiler to pass up to 3 floating point arguments in
2807 SSE registers instead of on the stack. Functions that take a
2808 variable number of arguments will continue to pass all of their
2809 floating point arguments on the stack.
2811 @item force_align_arg_pointer
2812 @cindex @code{force_align_arg_pointer} attribute
2813 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2814 applied to individual function definitions, generating an alternate
2815 prologue and epilogue that realigns the runtime stack if necessary.
2816 This supports mixing legacy codes that run with a 4-byte aligned stack
2817 with modern codes that keep a 16-byte stack for SSE compatibility.
2820 @cindex @code{resbank} attribute
2821 On the SH2A target, this attribute enables the high-speed register
2822 saving and restoration using a register bank for @code{interrupt_handler}
2823 routines. Saving to the bank is performed automatically after the CPU
2824 accepts an interrupt that uses a register bank.
2826 The nineteen 32-bit registers comprising general register R0 to R14,
2827 control register GBR, and system registers MACH, MACL, and PR and the
2828 vector table address offset are saved into a register bank. Register
2829 banks are stacked in first-in last-out (FILO) sequence. Restoration
2830 from the bank is executed by issuing a RESBANK instruction.
2833 @cindex @code{returns_twice} attribute
2834 The @code{returns_twice} attribute tells the compiler that a function may
2835 return more than one time. The compiler will ensure that all registers
2836 are dead before calling such a function and will emit a warning about
2837 the variables that may be clobbered after the second return from the
2838 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2839 The @code{longjmp}-like counterpart of such function, if any, might need
2840 to be marked with the @code{noreturn} attribute.
2843 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2844 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2845 all registers except the stack pointer should be saved in the prologue
2846 regardless of whether they are used or not.
2848 @item section ("@var{section-name}")
2849 @cindex @code{section} function attribute
2850 Normally, the compiler places the code it generates in the @code{text} section.
2851 Sometimes, however, you need additional sections, or you need certain
2852 particular functions to appear in special sections. The @code{section}
2853 attribute specifies that a function lives in a particular section.
2854 For example, the declaration:
2857 extern void foobar (void) __attribute__ ((section ("bar")));
2861 puts the function @code{foobar} in the @code{bar} section.
2863 Some file formats do not support arbitrary sections so the @code{section}
2864 attribute is not available on all platforms.
2865 If you need to map the entire contents of a module to a particular
2866 section, consider using the facilities of the linker instead.
2869 @cindex @code{sentinel} function attribute
2870 This function attribute ensures that a parameter in a function call is
2871 an explicit @code{NULL}. The attribute is only valid on variadic
2872 functions. By default, the sentinel is located at position zero, the
2873 last parameter of the function call. If an optional integer position
2874 argument P is supplied to the attribute, the sentinel must be located at
2875 position P counting backwards from the end of the argument list.
2878 __attribute__ ((sentinel))
2880 __attribute__ ((sentinel(0)))
2883 The attribute is automatically set with a position of 0 for the built-in
2884 functions @code{execl} and @code{execlp}. The built-in function
2885 @code{execle} has the attribute set with a position of 1.
2887 A valid @code{NULL} in this context is defined as zero with any pointer
2888 type. If your system defines the @code{NULL} macro with an integer type
2889 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2890 with a copy that redefines NULL appropriately.
2892 The warnings for missing or incorrect sentinels are enabled with
2896 See long_call/short_call.
2899 See longcall/shortcall.
2902 @cindex signal handler functions on the AVR processors
2903 Use this attribute on the AVR to indicate that the specified
2904 function is a signal handler. The compiler will generate function
2905 entry and exit sequences suitable for use in a signal handler when this
2906 attribute is present. Interrupts will be disabled inside the function.
2909 Use this attribute on the SH to indicate an @code{interrupt_handler}
2910 function should switch to an alternate stack. It expects a string
2911 argument that names a global variable holding the address of the
2916 void f () __attribute__ ((interrupt_handler,
2917 sp_switch ("alt_stack")));
2921 @cindex functions that pop the argument stack on the 386
2922 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2923 assume that the called function will pop off the stack space used to
2924 pass arguments, unless it takes a variable number of arguments.
2926 @item syscall_linkage
2927 @cindex @code{syscall_linkage} attribute
2928 This attribute is used to modify the IA64 calling convention by marking
2929 all input registers as live at all function exits. This makes it possible
2930 to restart a system call after an interrupt without having to save/restore
2931 the input registers. This also prevents kernel data from leaking into
2935 @cindex @code{target} function attribute
2936 The @code{target} attribute is used to specify that a function is to
2937 be compiled with different target options than specified on the
2938 command line. This can be used for instance to have functions
2939 compiled with a different ISA (instruction set architecture) than the
2940 default. You can also use the @samp{#pragma GCC target} pragma to set
2941 more than one function to be compiled with specific target options.
2942 @xref{Function Specific Option Pragmas}, for details about the
2943 @samp{#pragma GCC target} pragma.
2945 For instance on a 386, you could compile one function with
2946 @code{target("sse4.1,arch=core2")} and another with
2947 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
2948 compiling the first function with @option{-msse4.1} and
2949 @option{-march=core2} options, and the second function with
2950 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
2951 user to make sure that a function is only invoked on a machine that
2952 supports the particular ISA it was compiled for (for example by using
2953 @code{cpuid} on 386 to determine what feature bits and architecture
2957 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
2958 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
2961 On the 386, the following options are allowed:
2966 @cindex @code{target("abm")} attribute
2967 Enable/disable the generation of the advanced bit instructions.
2971 @cindex @code{target("aes")} attribute
2972 Enable/disable the generation of the AES instructions.
2976 @cindex @code{target("mmx")} attribute
2977 Enable/disable the generation of the MMX instructions.
2981 @cindex @code{target("pclmul")} attribute
2982 Enable/disable the generation of the PCLMUL instructions.
2986 @cindex @code{target("popcnt")} attribute
2987 Enable/disable the generation of the POPCNT instruction.
2991 @cindex @code{target("sse")} attribute
2992 Enable/disable the generation of the SSE instructions.
2996 @cindex @code{target("sse2")} attribute
2997 Enable/disable the generation of the SSE2 instructions.
3001 @cindex @code{target("sse3")} attribute
3002 Enable/disable the generation of the SSE3 instructions.
3006 @cindex @code{target("sse4")} attribute
3007 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3012 @cindex @code{target("sse4.1")} attribute
3013 Enable/disable the generation of the sse4.1 instructions.
3017 @cindex @code{target("sse4.2")} attribute
3018 Enable/disable the generation of the sse4.2 instructions.
3022 @cindex @code{target("sse4a")} attribute
3023 Enable/disable the generation of the SSE4A instructions.
3027 @cindex @code{target("sse5")} attribute
3028 Enable/disable the generation of the SSE5 instructions.
3032 @cindex @code{target("ssse3")} attribute
3033 Enable/disable the generation of the SSSE3 instructions.
3037 @cindex @code{target("cld")} attribute
3038 Enable/disable the generation of the CLD before string moves.
3040 @item fancy-math-387
3041 @itemx no-fancy-math-387
3042 @cindex @code{target("fancy-math-387")} attribute
3043 Enable/disable the generation of the @code{sin}, @code{cos}, and
3044 @code{sqrt} instructions on the 387 floating point unit.
3047 @itemx no-fused-madd
3048 @cindex @code{target("fused-madd")} attribute
3049 Enable/disable the generation of the fused multiply/add instructions.
3053 @cindex @code{target("ieee-fp")} attribute
3054 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3056 @item inline-all-stringops
3057 @itemx no-inline-all-stringops
3058 @cindex @code{target("inline-all-stringops")} attribute
3059 Enable/disable inlining of string operations.
3061 @item inline-stringops-dynamically
3062 @itemx no-inline-stringops-dynamically
3063 @cindex @code{target("inline-stringops-dynamically")} attribute
3064 Enable/disable the generation of the inline code to do small string
3065 operations and calling the library routines for large operations.
3067 @item align-stringops
3068 @itemx no-align-stringops
3069 @cindex @code{target("align-stringops")} attribute
3070 Do/do not align destination of inlined string operations.
3074 @cindex @code{target("recip")} attribute
3075 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3076 instructions followed an additional Newton-Raphson step instead of
3077 doing a floating point division.
3079 @item arch=@var{ARCH}
3080 @cindex @code{target("arch=@var{ARCH}")} attribute
3081 Specify the architecture to generate code for in compiling the function.
3083 @item tune=@var{TUNE}
3084 @cindex @code{target("tune=@var{TUNE}")} attribute
3085 Specify the architecture to tune for in compiling the function.
3087 @item fpmath=@var{FPMATH}
3088 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3089 Specify which floating point unit to use. The
3090 @code{target("fpmath=sse,387")} option must be specified as
3091 @code{target("fpmath=sse+387")} because the comma would separate
3095 On the 386, you can use either multiple strings to specify multiple
3096 options, or you can separate the option with a comma (@code{,}).
3098 On the 386, the inliner will not inline a function that has different
3099 target options than the caller, unless the callee has a subset of the
3100 target options of the caller. For example a function declared with
3101 @code{target("sse5")} can inline a function with
3102 @code{target("sse2")}, since @code{-msse5} implies @code{-msse2}.
3104 The @code{target} attribute is not implemented in GCC versions earlier
3105 than 4.4, and at present only the 386 uses it.
3108 @cindex tiny data section on the H8/300H and H8S
3109 Use this attribute on the H8/300H and H8S to indicate that the specified
3110 variable should be placed into the tiny data section.
3111 The compiler will generate more efficient code for loads and stores
3112 on data in the tiny data section. Note the tiny data area is limited to
3113 slightly under 32kbytes of data.
3116 Use this attribute on the SH for an @code{interrupt_handler} to return using
3117 @code{trapa} instead of @code{rte}. This attribute expects an integer
3118 argument specifying the trap number to be used.
3121 @cindex @code{unused} attribute.
3122 This attribute, attached to a function, means that the function is meant
3123 to be possibly unused. GCC will not produce a warning for this
3127 @cindex @code{used} attribute.
3128 This attribute, attached to a function, means that code must be emitted
3129 for the function even if it appears that the function is not referenced.
3130 This is useful, for example, when the function is referenced only in
3134 @cindex @code{version_id} attribute
3135 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3136 symbol to contain a version string, thus allowing for function level
3137 versioning. HP-UX system header files may use version level functioning
3138 for some system calls.
3141 extern int foo () __attribute__((version_id ("20040821")));
3144 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3146 @item visibility ("@var{visibility_type}")
3147 @cindex @code{visibility} attribute
3148 This attribute affects the linkage of the declaration to which it is attached.
3149 There are four supported @var{visibility_type} values: default,
3150 hidden, protected or internal visibility.
3153 void __attribute__ ((visibility ("protected")))
3154 f () @{ /* @r{Do something.} */; @}
3155 int i __attribute__ ((visibility ("hidden")));
3158 The possible values of @var{visibility_type} correspond to the
3159 visibility settings in the ELF gABI.
3162 @c keep this list of visibilities in alphabetical order.
3165 Default visibility is the normal case for the object file format.
3166 This value is available for the visibility attribute to override other
3167 options that may change the assumed visibility of entities.
3169 On ELF, default visibility means that the declaration is visible to other
3170 modules and, in shared libraries, means that the declared entity may be
3173 On Darwin, default visibility means that the declaration is visible to
3176 Default visibility corresponds to ``external linkage'' in the language.
3179 Hidden visibility indicates that the entity declared will have a new
3180 form of linkage, which we'll call ``hidden linkage''. Two
3181 declarations of an object with hidden linkage refer to the same object
3182 if they are in the same shared object.
3185 Internal visibility is like hidden visibility, but with additional
3186 processor specific semantics. Unless otherwise specified by the
3187 psABI, GCC defines internal visibility to mean that a function is
3188 @emph{never} called from another module. Compare this with hidden
3189 functions which, while they cannot be referenced directly by other
3190 modules, can be referenced indirectly via function pointers. By
3191 indicating that a function cannot be called from outside the module,
3192 GCC may for instance omit the load of a PIC register since it is known
3193 that the calling function loaded the correct value.
3196 Protected visibility is like default visibility except that it
3197 indicates that references within the defining module will bind to the
3198 definition in that module. That is, the declared entity cannot be
3199 overridden by another module.
3203 All visibilities are supported on many, but not all, ELF targets
3204 (supported when the assembler supports the @samp{.visibility}
3205 pseudo-op). Default visibility is supported everywhere. Hidden
3206 visibility is supported on Darwin targets.
3208 The visibility attribute should be applied only to declarations which
3209 would otherwise have external linkage. The attribute should be applied
3210 consistently, so that the same entity should not be declared with
3211 different settings of the attribute.
3213 In C++, the visibility attribute applies to types as well as functions
3214 and objects, because in C++ types have linkage. A class must not have
3215 greater visibility than its non-static data member types and bases,
3216 and class members default to the visibility of their class. Also, a
3217 declaration without explicit visibility is limited to the visibility
3220 In C++, you can mark member functions and static member variables of a
3221 class with the visibility attribute. This is useful if you know a
3222 particular method or static member variable should only be used from
3223 one shared object; then you can mark it hidden while the rest of the
3224 class has default visibility. Care must be taken to avoid breaking
3225 the One Definition Rule; for example, it is usually not useful to mark
3226 an inline method as hidden without marking the whole class as hidden.
3228 A C++ namespace declaration can also have the visibility attribute.
3229 This attribute applies only to the particular namespace body, not to
3230 other definitions of the same namespace; it is equivalent to using
3231 @samp{#pragma GCC visibility} before and after the namespace
3232 definition (@pxref{Visibility Pragmas}).
3234 In C++, if a template argument has limited visibility, this
3235 restriction is implicitly propagated to the template instantiation.
3236 Otherwise, template instantiations and specializations default to the
3237 visibility of their template.
3239 If both the template and enclosing class have explicit visibility, the
3240 visibility from the template is used.
3242 @item warn_unused_result
3243 @cindex @code{warn_unused_result} attribute
3244 The @code{warn_unused_result} attribute causes a warning to be emitted
3245 if a caller of the function with this attribute does not use its
3246 return value. This is useful for functions where not checking
3247 the result is either a security problem or always a bug, such as
3251 int fn () __attribute__ ((warn_unused_result));
3254 if (fn () < 0) return -1;
3260 results in warning on line 5.
3263 @cindex @code{weak} attribute
3264 The @code{weak} attribute causes the declaration to be emitted as a weak
3265 symbol rather than a global. This is primarily useful in defining
3266 library functions which can be overridden in user code, though it can
3267 also be used with non-function declarations. Weak symbols are supported
3268 for ELF targets, and also for a.out targets when using the GNU assembler
3272 @itemx weakref ("@var{target}")
3273 @cindex @code{weakref} attribute
3274 The @code{weakref} attribute marks a declaration as a weak reference.
3275 Without arguments, it should be accompanied by an @code{alias} attribute
3276 naming the target symbol. Optionally, the @var{target} may be given as
3277 an argument to @code{weakref} itself. In either case, @code{weakref}
3278 implicitly marks the declaration as @code{weak}. Without a
3279 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3280 @code{weakref} is equivalent to @code{weak}.
3283 static int x() __attribute__ ((weakref ("y")));
3284 /* is equivalent to... */
3285 static int x() __attribute__ ((weak, weakref, alias ("y")));
3287 static int x() __attribute__ ((weakref));
3288 static int x() __attribute__ ((alias ("y")));
3291 A weak reference is an alias that does not by itself require a
3292 definition to be given for the target symbol. If the target symbol is
3293 only referenced through weak references, then the becomes a @code{weak}
3294 undefined symbol. If it is directly referenced, however, then such
3295 strong references prevail, and a definition will be required for the
3296 symbol, not necessarily in the same translation unit.
3298 The effect is equivalent to moving all references to the alias to a
3299 separate translation unit, renaming the alias to the aliased symbol,
3300 declaring it as weak, compiling the two separate translation units and
3301 performing a reloadable link on them.
3303 At present, a declaration to which @code{weakref} is attached can
3304 only be @code{static}.
3308 You can specify multiple attributes in a declaration by separating them
3309 by commas within the double parentheses or by immediately following an
3310 attribute declaration with another attribute declaration.
3312 @cindex @code{#pragma}, reason for not using
3313 @cindex pragma, reason for not using
3314 Some people object to the @code{__attribute__} feature, suggesting that
3315 ISO C's @code{#pragma} should be used instead. At the time
3316 @code{__attribute__} was designed, there were two reasons for not doing
3321 It is impossible to generate @code{#pragma} commands from a macro.
3324 There is no telling what the same @code{#pragma} might mean in another
3328 These two reasons applied to almost any application that might have been
3329 proposed for @code{#pragma}. It was basically a mistake to use
3330 @code{#pragma} for @emph{anything}.
3332 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3333 to be generated from macros. In addition, a @code{#pragma GCC}
3334 namespace is now in use for GCC-specific pragmas. However, it has been
3335 found convenient to use @code{__attribute__} to achieve a natural
3336 attachment of attributes to their corresponding declarations, whereas
3337 @code{#pragma GCC} is of use for constructs that do not naturally form
3338 part of the grammar. @xref{Other Directives,,Miscellaneous
3339 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3341 @node Attribute Syntax
3342 @section Attribute Syntax
3343 @cindex attribute syntax
3345 This section describes the syntax with which @code{__attribute__} may be
3346 used, and the constructs to which attribute specifiers bind, for the C
3347 language. Some details may vary for C++ and Objective-C@. Because of
3348 infelicities in the grammar for attributes, some forms described here
3349 may not be successfully parsed in all cases.
3351 There are some problems with the semantics of attributes in C++. For
3352 example, there are no manglings for attributes, although they may affect
3353 code generation, so problems may arise when attributed types are used in
3354 conjunction with templates or overloading. Similarly, @code{typeid}
3355 does not distinguish between types with different attributes. Support
3356 for attributes in C++ may be restricted in future to attributes on
3357 declarations only, but not on nested declarators.
3359 @xref{Function Attributes}, for details of the semantics of attributes
3360 applying to functions. @xref{Variable Attributes}, for details of the
3361 semantics of attributes applying to variables. @xref{Type Attributes},
3362 for details of the semantics of attributes applying to structure, union
3363 and enumerated types.
3365 An @dfn{attribute specifier} is of the form
3366 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3367 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3368 each attribute is one of the following:
3372 Empty. Empty attributes are ignored.
3375 A word (which may be an identifier such as @code{unused}, or a reserved
3376 word such as @code{const}).
3379 A word, followed by, in parentheses, parameters for the attribute.
3380 These parameters take one of the following forms:
3384 An identifier. For example, @code{mode} attributes use this form.
3387 An identifier followed by a comma and a non-empty comma-separated list
3388 of expressions. For example, @code{format} attributes use this form.
3391 A possibly empty comma-separated list of expressions. For example,
3392 @code{format_arg} attributes use this form with the list being a single
3393 integer constant expression, and @code{alias} attributes use this form
3394 with the list being a single string constant.
3398 An @dfn{attribute specifier list} is a sequence of one or more attribute
3399 specifiers, not separated by any other tokens.
3401 In GNU C, an attribute specifier list may appear after the colon following a
3402 label, other than a @code{case} or @code{default} label. The only
3403 attribute it makes sense to use after a label is @code{unused}. This
3404 feature is intended for code generated by programs which contains labels
3405 that may be unused but which is compiled with @option{-Wall}. It would
3406 not normally be appropriate to use in it human-written code, though it
3407 could be useful in cases where the code that jumps to the label is
3408 contained within an @code{#ifdef} conditional. GNU C++ does not permit
3409 such placement of attribute lists, as it is permissible for a
3410 declaration, which could begin with an attribute list, to be labelled in
3411 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
3412 does not arise there.
3414 An attribute specifier list may appear as part of a @code{struct},
3415 @code{union} or @code{enum} specifier. It may go either immediately
3416 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3417 the closing brace. The former syntax is preferred.
3418 Where attribute specifiers follow the closing brace, they are considered
3419 to relate to the structure, union or enumerated type defined, not to any
3420 enclosing declaration the type specifier appears in, and the type
3421 defined is not complete until after the attribute specifiers.
3422 @c Otherwise, there would be the following problems: a shift/reduce
3423 @c conflict between attributes binding the struct/union/enum and
3424 @c binding to the list of specifiers/qualifiers; and "aligned"
3425 @c attributes could use sizeof for the structure, but the size could be
3426 @c changed later by "packed" attributes.
3428 Otherwise, an attribute specifier appears as part of a declaration,
3429 counting declarations of unnamed parameters and type names, and relates
3430 to that declaration (which may be nested in another declaration, for
3431 example in the case of a parameter declaration), or to a particular declarator
3432 within a declaration. Where an
3433 attribute specifier is applied to a parameter declared as a function or
3434 an array, it should apply to the function or array rather than the
3435 pointer to which the parameter is implicitly converted, but this is not
3436 yet correctly implemented.
3438 Any list of specifiers and qualifiers at the start of a declaration may
3439 contain attribute specifiers, whether or not such a list may in that
3440 context contain storage class specifiers. (Some attributes, however,
3441 are essentially in the nature of storage class specifiers, and only make
3442 sense where storage class specifiers may be used; for example,
3443 @code{section}.) There is one necessary limitation to this syntax: the
3444 first old-style parameter declaration in a function definition cannot
3445 begin with an attribute specifier, because such an attribute applies to
3446 the function instead by syntax described below (which, however, is not
3447 yet implemented in this case). In some other cases, attribute
3448 specifiers are permitted by this grammar but not yet supported by the
3449 compiler. All attribute specifiers in this place relate to the
3450 declaration as a whole. In the obsolescent usage where a type of
3451 @code{int} is implied by the absence of type specifiers, such a list of
3452 specifiers and qualifiers may be an attribute specifier list with no
3453 other specifiers or qualifiers.
3455 At present, the first parameter in a function prototype must have some
3456 type specifier which is not an attribute specifier; this resolves an
3457 ambiguity in the interpretation of @code{void f(int
3458 (__attribute__((foo)) x))}, but is subject to change. At present, if
3459 the parentheses of a function declarator contain only attributes then
3460 those attributes are ignored, rather than yielding an error or warning
3461 or implying a single parameter of type int, but this is subject to
3464 An attribute specifier list may appear immediately before a declarator
3465 (other than the first) in a comma-separated list of declarators in a
3466 declaration of more than one identifier using a single list of
3467 specifiers and qualifiers. Such attribute specifiers apply
3468 only to the identifier before whose declarator they appear. For
3472 __attribute__((noreturn)) void d0 (void),
3473 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3478 the @code{noreturn} attribute applies to all the functions
3479 declared; the @code{format} attribute only applies to @code{d1}.
3481 An attribute specifier list may appear immediately before the comma,
3482 @code{=} or semicolon terminating the declaration of an identifier other
3483 than a function definition. Such attribute specifiers apply
3484 to the declared object or function. Where an
3485 assembler name for an object or function is specified (@pxref{Asm
3486 Labels}), the attribute must follow the @code{asm}
3489 An attribute specifier list may, in future, be permitted to appear after
3490 the declarator in a function definition (before any old-style parameter
3491 declarations or the function body).
3493 Attribute specifiers may be mixed with type qualifiers appearing inside
3494 the @code{[]} of a parameter array declarator, in the C99 construct by
3495 which such qualifiers are applied to the pointer to which the array is
3496 implicitly converted. Such attribute specifiers apply to the pointer,
3497 not to the array, but at present this is not implemented and they are
3500 An attribute specifier list may appear at the start of a nested
3501 declarator. At present, there are some limitations in this usage: the
3502 attributes correctly apply to the declarator, but for most individual
3503 attributes the semantics this implies are not implemented.
3504 When attribute specifiers follow the @code{*} of a pointer
3505 declarator, they may be mixed with any type qualifiers present.
3506 The following describes the formal semantics of this syntax. It will make the
3507 most sense if you are familiar with the formal specification of
3508 declarators in the ISO C standard.
3510 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3511 D1}, where @code{T} contains declaration specifiers that specify a type
3512 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3513 contains an identifier @var{ident}. The type specified for @var{ident}
3514 for derived declarators whose type does not include an attribute
3515 specifier is as in the ISO C standard.
3517 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3518 and the declaration @code{T D} specifies the type
3519 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3520 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3521 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3523 If @code{D1} has the form @code{*
3524 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3525 declaration @code{T D} specifies the type
3526 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3527 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3528 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3534 void (__attribute__((noreturn)) ****f) (void);
3538 specifies the type ``pointer to pointer to pointer to pointer to
3539 non-returning function returning @code{void}''. As another example,
3542 char *__attribute__((aligned(8))) *f;
3546 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3547 Note again that this does not work with most attributes; for example,
3548 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3549 is not yet supported.
3551 For compatibility with existing code written for compiler versions that
3552 did not implement attributes on nested declarators, some laxity is
3553 allowed in the placing of attributes. If an attribute that only applies
3554 to types is applied to a declaration, it will be treated as applying to
3555 the type of that declaration. If an attribute that only applies to
3556 declarations is applied to the type of a declaration, it will be treated
3557 as applying to that declaration; and, for compatibility with code
3558 placing the attributes immediately before the identifier declared, such
3559 an attribute applied to a function return type will be treated as
3560 applying to the function type, and such an attribute applied to an array
3561 element type will be treated as applying to the array type. If an
3562 attribute that only applies to function types is applied to a
3563 pointer-to-function type, it will be treated as applying to the pointer
3564 target type; if such an attribute is applied to a function return type
3565 that is not a pointer-to-function type, it will be treated as applying
3566 to the function type.
3568 @node Function Prototypes
3569 @section Prototypes and Old-Style Function Definitions
3570 @cindex function prototype declarations
3571 @cindex old-style function definitions
3572 @cindex promotion of formal parameters
3574 GNU C extends ISO C to allow a function prototype to override a later
3575 old-style non-prototype definition. Consider the following example:
3578 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3585 /* @r{Prototype function declaration.} */
3586 int isroot P((uid_t));
3588 /* @r{Old-style function definition.} */
3590 isroot (x) /* @r{??? lossage here ???} */
3597 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3598 not allow this example, because subword arguments in old-style
3599 non-prototype definitions are promoted. Therefore in this example the
3600 function definition's argument is really an @code{int}, which does not
3601 match the prototype argument type of @code{short}.
3603 This restriction of ISO C makes it hard to write code that is portable
3604 to traditional C compilers, because the programmer does not know
3605 whether the @code{uid_t} type is @code{short}, @code{int}, or
3606 @code{long}. Therefore, in cases like these GNU C allows a prototype
3607 to override a later old-style definition. More precisely, in GNU C, a
3608 function prototype argument type overrides the argument type specified
3609 by a later old-style definition if the former type is the same as the
3610 latter type before promotion. Thus in GNU C the above example is
3611 equivalent to the following:
3624 GNU C++ does not support old-style function definitions, so this
3625 extension is irrelevant.
3628 @section C++ Style Comments
3630 @cindex C++ comments
3631 @cindex comments, C++ style
3633 In GNU C, you may use C++ style comments, which start with @samp{//} and
3634 continue until the end of the line. Many other C implementations allow
3635 such comments, and they are included in the 1999 C standard. However,
3636 C++ style comments are not recognized if you specify an @option{-std}
3637 option specifying a version of ISO C before C99, or @option{-ansi}
3638 (equivalent to @option{-std=c89}).
3641 @section Dollar Signs in Identifier Names
3643 @cindex dollar signs in identifier names
3644 @cindex identifier names, dollar signs in
3646 In GNU C, you may normally use dollar signs in identifier names.
3647 This is because many traditional C implementations allow such identifiers.
3648 However, dollar signs in identifiers are not supported on a few target
3649 machines, typically because the target assembler does not allow them.
3651 @node Character Escapes
3652 @section The Character @key{ESC} in Constants
3654 You can use the sequence @samp{\e} in a string or character constant to
3655 stand for the ASCII character @key{ESC}.
3658 @section Inquiring on Alignment of Types or Variables
3660 @cindex type alignment
3661 @cindex variable alignment
3663 The keyword @code{__alignof__} allows you to inquire about how an object
3664 is aligned, or the minimum alignment usually required by a type. Its
3665 syntax is just like @code{sizeof}.
3667 For example, if the target machine requires a @code{double} value to be
3668 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3669 This is true on many RISC machines. On more traditional machine
3670 designs, @code{__alignof__ (double)} is 4 or even 2.
3672 Some machines never actually require alignment; they allow reference to any
3673 data type even at an odd address. For these machines, @code{__alignof__}
3674 reports the smallest alignment that GCC will give the data type, usually as
3675 mandated by the target ABI.
3677 If the operand of @code{__alignof__} is an lvalue rather than a type,
3678 its value is the required alignment for its type, taking into account
3679 any minimum alignment specified with GCC's @code{__attribute__}
3680 extension (@pxref{Variable Attributes}). For example, after this
3684 struct foo @{ int x; char y; @} foo1;
3688 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3689 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3691 It is an error to ask for the alignment of an incomplete type.
3693 @node Variable Attributes
3694 @section Specifying Attributes of Variables
3695 @cindex attribute of variables
3696 @cindex variable attributes
3698 The keyword @code{__attribute__} allows you to specify special
3699 attributes of variables or structure fields. This keyword is followed
3700 by an attribute specification inside double parentheses. Some
3701 attributes are currently defined generically for variables.
3702 Other attributes are defined for variables on particular target
3703 systems. Other attributes are available for functions
3704 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3705 Other front ends might define more attributes
3706 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3708 You may also specify attributes with @samp{__} preceding and following
3709 each keyword. This allows you to use them in header files without
3710 being concerned about a possible macro of the same name. For example,
3711 you may use @code{__aligned__} instead of @code{aligned}.
3713 @xref{Attribute Syntax}, for details of the exact syntax for using
3717 @cindex @code{aligned} attribute
3718 @item aligned (@var{alignment})
3719 This attribute specifies a minimum alignment for the variable or
3720 structure field, measured in bytes. For example, the declaration:
3723 int x __attribute__ ((aligned (16))) = 0;
3727 causes the compiler to allocate the global variable @code{x} on a
3728 16-byte boundary. On a 68040, this could be used in conjunction with
3729 an @code{asm} expression to access the @code{move16} instruction which
3730 requires 16-byte aligned operands.
3732 You can also specify the alignment of structure fields. For example, to
3733 create a double-word aligned @code{int} pair, you could write:
3736 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3740 This is an alternative to creating a union with a @code{double} member
3741 that forces the union to be double-word aligned.
3743 As in the preceding examples, you can explicitly specify the alignment
3744 (in bytes) that you wish the compiler to use for a given variable or
3745 structure field. Alternatively, you can leave out the alignment factor
3746 and just ask the compiler to align a variable or field to the
3747 default alignment for the target architecture you are compiling for.
3748 The default alignment is sufficient for all scalar types, but may not be
3749 enough for all vector types on a target which supports vector operations.
3750 The default alignment is fixed for a particular target ABI.
3752 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
3753 which is the largest alignment ever used for any data type on the
3754 target machine you are compiling for. For example, you could write:
3757 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
3760 The compiler automatically sets the alignment for the declared
3761 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
3762 often make copy operations more efficient, because the compiler can
3763 use whatever instructions copy the biggest chunks of memory when
3764 performing copies to or from the variables or fields that you have
3765 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
3766 may change depending on command line options.
3768 When used on a struct, or struct member, the @code{aligned} attribute can
3769 only increase the alignment; in order to decrease it, the @code{packed}
3770 attribute must be specified as well. When used as part of a typedef, the
3771 @code{aligned} attribute can both increase and decrease alignment, and
3772 specifying the @code{packed} attribute will generate a warning.
3774 Note that the effectiveness of @code{aligned} attributes may be limited
3775 by inherent limitations in your linker. On many systems, the linker is
3776 only able to arrange for variables to be aligned up to a certain maximum
3777 alignment. (For some linkers, the maximum supported alignment may
3778 be very very small.) If your linker is only able to align variables
3779 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3780 in an @code{__attribute__} will still only provide you with 8 byte
3781 alignment. See your linker documentation for further information.
3783 The @code{aligned} attribute can also be used for functions
3784 (@pxref{Function Attributes}.)
3786 @item cleanup (@var{cleanup_function})
3787 @cindex @code{cleanup} attribute
3788 The @code{cleanup} attribute runs a function when the variable goes
3789 out of scope. This attribute can only be applied to auto function
3790 scope variables; it may not be applied to parameters or variables
3791 with static storage duration. The function must take one parameter,
3792 a pointer to a type compatible with the variable. The return value
3793 of the function (if any) is ignored.
3795 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3796 will be run during the stack unwinding that happens during the
3797 processing of the exception. Note that the @code{cleanup} attribute
3798 does not allow the exception to be caught, only to perform an action.
3799 It is undefined what happens if @var{cleanup_function} does not
3804 @cindex @code{common} attribute
3805 @cindex @code{nocommon} attribute
3808 The @code{common} attribute requests GCC to place a variable in
3809 ``common'' storage. The @code{nocommon} attribute requests the
3810 opposite---to allocate space for it directly.
3812 These attributes override the default chosen by the
3813 @option{-fno-common} and @option{-fcommon} flags respectively.
3816 @cindex @code{deprecated} attribute
3817 The @code{deprecated} attribute results in a warning if the variable
3818 is used anywhere in the source file. This is useful when identifying
3819 variables that are expected to be removed in a future version of a
3820 program. The warning also includes the location of the declaration
3821 of the deprecated variable, to enable users to easily find further
3822 information about why the variable is deprecated, or what they should
3823 do instead. Note that the warning only occurs for uses:
3826 extern int old_var __attribute__ ((deprecated));
3828 int new_fn () @{ return old_var; @}
3831 results in a warning on line 3 but not line 2.
3833 The @code{deprecated} attribute can also be used for functions and
3834 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3836 @item mode (@var{mode})
3837 @cindex @code{mode} attribute
3838 This attribute specifies the data type for the declaration---whichever
3839 type corresponds to the mode @var{mode}. This in effect lets you
3840 request an integer or floating point type according to its width.
3842 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3843 indicate the mode corresponding to a one-byte integer, @samp{word} or
3844 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3845 or @samp{__pointer__} for the mode used to represent pointers.
3848 @cindex @code{packed} attribute
3849 The @code{packed} attribute specifies that a variable or structure field
3850 should have the smallest possible alignment---one byte for a variable,
3851 and one bit for a field, unless you specify a larger value with the
3852 @code{aligned} attribute.
3854 Here is a structure in which the field @code{x} is packed, so that it
3855 immediately follows @code{a}:
3861 int x[2] __attribute__ ((packed));
3865 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
3866 @code{packed} attribute on bit-fields of type @code{char}. This has
3867 been fixed in GCC 4.4 but the change can lead to differences in the
3868 structure layout. See the documentation of
3869 @option{-Wpacked-bitfield-compat} for more information.
3871 @item section ("@var{section-name}")
3872 @cindex @code{section} variable attribute
3873 Normally, the compiler places the objects it generates in sections like
3874 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3875 or you need certain particular variables to appear in special sections,
3876 for example to map to special hardware. The @code{section}
3877 attribute specifies that a variable (or function) lives in a particular
3878 section. For example, this small program uses several specific section names:
3881 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3882 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3883 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3884 int init_data __attribute__ ((section ("INITDATA")));
3888 /* @r{Initialize stack pointer} */
3889 init_sp (stack + sizeof (stack));
3891 /* @r{Initialize initialized data} */
3892 memcpy (&init_data, &data, &edata - &data);
3894 /* @r{Turn on the serial ports} */
3901 Use the @code{section} attribute with
3902 @emph{global} variables and not @emph{local} variables,
3903 as shown in the example.
3905 You may use the @code{section} attribute with initialized or
3906 uninitialized global variables but the linker requires
3907 each object be defined once, with the exception that uninitialized
3908 variables tentatively go in the @code{common} (or @code{bss}) section
3909 and can be multiply ``defined''. Using the @code{section} attribute
3910 will change what section the variable goes into and may cause the
3911 linker to issue an error if an uninitialized variable has multiple
3912 definitions. You can force a variable to be initialized with the
3913 @option{-fno-common} flag or the @code{nocommon} attribute.
3915 Some file formats do not support arbitrary sections so the @code{section}
3916 attribute is not available on all platforms.
3917 If you need to map the entire contents of a module to a particular
3918 section, consider using the facilities of the linker instead.
3921 @cindex @code{shared} variable attribute
3922 On Microsoft Windows, in addition to putting variable definitions in a named
3923 section, the section can also be shared among all running copies of an
3924 executable or DLL@. For example, this small program defines shared data
3925 by putting it in a named section @code{shared} and marking the section
3929 int foo __attribute__((section ("shared"), shared)) = 0;
3934 /* @r{Read and write foo. All running
3935 copies see the same value.} */
3941 You may only use the @code{shared} attribute along with @code{section}
3942 attribute with a fully initialized global definition because of the way
3943 linkers work. See @code{section} attribute for more information.
3945 The @code{shared} attribute is only available on Microsoft Windows@.
3947 @item tls_model ("@var{tls_model}")
3948 @cindex @code{tls_model} attribute
3949 The @code{tls_model} attribute sets thread-local storage model
3950 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3951 overriding @option{-ftls-model=} command line switch on a per-variable
3953 The @var{tls_model} argument should be one of @code{global-dynamic},
3954 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3956 Not all targets support this attribute.
3959 This attribute, attached to a variable, means that the variable is meant
3960 to be possibly unused. GCC will not produce a warning for this
3964 This attribute, attached to a variable, means that the variable must be
3965 emitted even if it appears that the variable is not referenced.
3967 @item vector_size (@var{bytes})
3968 This attribute specifies the vector size for the variable, measured in
3969 bytes. For example, the declaration:
3972 int foo __attribute__ ((vector_size (16)));
3976 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3977 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3978 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3980 This attribute is only applicable to integral and float scalars,
3981 although arrays, pointers, and function return values are allowed in
3982 conjunction with this construct.
3984 Aggregates with this attribute are invalid, even if they are of the same
3985 size as a corresponding scalar. For example, the declaration:
3988 struct S @{ int a; @};
3989 struct S __attribute__ ((vector_size (16))) foo;
3993 is invalid even if the size of the structure is the same as the size of
3997 The @code{selectany} attribute causes an initialized global variable to
3998 have link-once semantics. When multiple definitions of the variable are
3999 encountered by the linker, the first is selected and the remainder are
4000 discarded. Following usage by the Microsoft compiler, the linker is told
4001 @emph{not} to warn about size or content differences of the multiple
4004 Although the primary usage of this attribute is for POD types, the
4005 attribute can also be applied to global C++ objects that are initialized
4006 by a constructor. In this case, the static initialization and destruction
4007 code for the object is emitted in each translation defining the object,
4008 but the calls to the constructor and destructor are protected by a
4009 link-once guard variable.
4011 The @code{selectany} attribute is only available on Microsoft Windows
4012 targets. You can use @code{__declspec (selectany)} as a synonym for
4013 @code{__attribute__ ((selectany))} for compatibility with other
4017 The @code{weak} attribute is described in @ref{Function Attributes}.
4020 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4023 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4027 @subsection Blackfin Variable Attributes
4029 Three attributes are currently defined for the Blackfin.
4035 @cindex @code{l1_data} variable attribute
4036 @cindex @code{l1_data_A} variable attribute
4037 @cindex @code{l1_data_B} variable attribute
4038 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4039 Variables with @code{l1_data} attribute will be put into the specific section
4040 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4041 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4042 attribute will be put into the specific section named @code{.l1.data.B}.
4045 @subsection M32R/D Variable Attributes
4047 One attribute is currently defined for the M32R/D@.
4050 @item model (@var{model-name})
4051 @cindex variable addressability on the M32R/D
4052 Use this attribute on the M32R/D to set the addressability of an object.
4053 The identifier @var{model-name} is one of @code{small}, @code{medium},
4054 or @code{large}, representing each of the code models.
4056 Small model objects live in the lower 16MB of memory (so that their
4057 addresses can be loaded with the @code{ld24} instruction).
4059 Medium and large model objects may live anywhere in the 32-bit address space
4060 (the compiler will generate @code{seth/add3} instructions to load their
4064 @anchor{i386 Variable Attributes}
4065 @subsection i386 Variable Attributes
4067 Two attributes are currently defined for i386 configurations:
4068 @code{ms_struct} and @code{gcc_struct}
4073 @cindex @code{ms_struct} attribute
4074 @cindex @code{gcc_struct} attribute
4076 If @code{packed} is used on a structure, or if bit-fields are used
4077 it may be that the Microsoft ABI packs them differently
4078 than GCC would normally pack them. Particularly when moving packed
4079 data between functions compiled with GCC and the native Microsoft compiler
4080 (either via function call or as data in a file), it may be necessary to access
4083 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4084 compilers to match the native Microsoft compiler.
4086 The Microsoft structure layout algorithm is fairly simple with the exception
4087 of the bitfield packing:
4089 The padding and alignment of members of structures and whether a bit field
4090 can straddle a storage-unit boundary
4093 @item Structure members are stored sequentially in the order in which they are
4094 declared: the first member has the lowest memory address and the last member
4097 @item Every data object has an alignment-requirement. The alignment-requirement
4098 for all data except structures, unions, and arrays is either the size of the
4099 object or the current packing size (specified with either the aligned attribute
4100 or the pack pragma), whichever is less. For structures, unions, and arrays,
4101 the alignment-requirement is the largest alignment-requirement of its members.
4102 Every object is allocated an offset so that:
4104 offset % alignment-requirement == 0
4106 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4107 unit if the integral types are the same size and if the next bit field fits
4108 into the current allocation unit without crossing the boundary imposed by the
4109 common alignment requirements of the bit fields.
4112 Handling of zero-length bitfields:
4114 MSVC interprets zero-length bitfields in the following ways:
4117 @item If a zero-length bitfield is inserted between two bitfields that would
4118 normally be coalesced, the bitfields will not be coalesced.
4125 unsigned long bf_1 : 12;
4127 unsigned long bf_2 : 12;
4131 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4132 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4134 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4135 alignment of the zero-length bitfield is greater than the member that follows it,
4136 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4156 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4157 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4158 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4161 Taking this into account, it is important to note the following:
4164 @item If a zero-length bitfield follows a normal bitfield, the type of the
4165 zero-length bitfield may affect the alignment of the structure as whole. For
4166 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4167 normal bitfield, and is of type short.
4169 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4170 still affect the alignment of the structure:
4180 Here, @code{t4} will take up 4 bytes.
4183 @item Zero-length bitfields following non-bitfield members are ignored:
4194 Here, @code{t5} will take up 2 bytes.
4198 @subsection PowerPC Variable Attributes
4200 Three attributes currently are defined for PowerPC configurations:
4201 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4203 For full documentation of the struct attributes please see the
4204 documentation in @ref{i386 Variable Attributes}.
4206 For documentation of @code{altivec} attribute please see the
4207 documentation in @ref{PowerPC Type Attributes}.
4209 @subsection SPU Variable Attributes
4211 The SPU supports the @code{spu_vector} attribute for variables. For
4212 documentation of this attribute please see the documentation in
4213 @ref{SPU Type Attributes}.
4215 @subsection Xstormy16 Variable Attributes
4217 One attribute is currently defined for xstormy16 configurations:
4222 @cindex @code{below100} attribute
4224 If a variable has the @code{below100} attribute (@code{BELOW100} is
4225 allowed also), GCC will place the variable in the first 0x100 bytes of
4226 memory and use special opcodes to access it. Such variables will be
4227 placed in either the @code{.bss_below100} section or the
4228 @code{.data_below100} section.
4232 @subsection AVR Variable Attributes
4236 @cindex @code{progmem} variable attribute
4237 The @code{progmem} attribute is used on the AVR to place data in the Program
4238 Memory address space. The AVR is a Harvard Architecture processor and data
4239 normally resides in the Data Memory address space.
4242 @node Type Attributes
4243 @section Specifying Attributes of Types
4244 @cindex attribute of types
4245 @cindex type attributes
4247 The keyword @code{__attribute__} allows you to specify special
4248 attributes of @code{struct} and @code{union} types when you define
4249 such types. This keyword is followed by an attribute specification
4250 inside double parentheses. Seven attributes are currently defined for
4251 types: @code{aligned}, @code{packed}, @code{transparent_union},
4252 @code{unused}, @code{deprecated}, @code{visibility}, and
4253 @code{may_alias}. Other attributes are defined for functions
4254 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4257 You may also specify any one of these attributes with @samp{__}
4258 preceding and following its keyword. This allows you to use these
4259 attributes in header files without being concerned about a possible
4260 macro of the same name. For example, you may use @code{__aligned__}
4261 instead of @code{aligned}.
4263 You may specify type attributes in an enum, struct or union type
4264 declaration or definition, or for other types in a @code{typedef}
4267 For an enum, struct or union type, you may specify attributes either
4268 between the enum, struct or union tag and the name of the type, or
4269 just past the closing curly brace of the @emph{definition}. The
4270 former syntax is preferred.
4272 @xref{Attribute Syntax}, for details of the exact syntax for using
4276 @cindex @code{aligned} attribute
4277 @item aligned (@var{alignment})
4278 This attribute specifies a minimum alignment (in bytes) for variables
4279 of the specified type. For example, the declarations:
4282 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4283 typedef int more_aligned_int __attribute__ ((aligned (8)));
4287 force the compiler to insure (as far as it can) that each variable whose
4288 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4289 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4290 variables of type @code{struct S} aligned to 8-byte boundaries allows
4291 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4292 store) instructions when copying one variable of type @code{struct S} to
4293 another, thus improving run-time efficiency.
4295 Note that the alignment of any given @code{struct} or @code{union} type
4296 is required by the ISO C standard to be at least a perfect multiple of
4297 the lowest common multiple of the alignments of all of the members of
4298 the @code{struct} or @code{union} in question. This means that you @emph{can}
4299 effectively adjust the alignment of a @code{struct} or @code{union}
4300 type by attaching an @code{aligned} attribute to any one of the members
4301 of such a type, but the notation illustrated in the example above is a
4302 more obvious, intuitive, and readable way to request the compiler to
4303 adjust the alignment of an entire @code{struct} or @code{union} type.
4305 As in the preceding example, you can explicitly specify the alignment
4306 (in bytes) that you wish the compiler to use for a given @code{struct}
4307 or @code{union} type. Alternatively, you can leave out the alignment factor
4308 and just ask the compiler to align a type to the maximum
4309 useful alignment for the target machine you are compiling for. For
4310 example, you could write:
4313 struct S @{ short f[3]; @} __attribute__ ((aligned));
4316 Whenever you leave out the alignment factor in an @code{aligned}
4317 attribute specification, the compiler automatically sets the alignment
4318 for the type to the largest alignment which is ever used for any data
4319 type on the target machine you are compiling for. Doing this can often
4320 make copy operations more efficient, because the compiler can use
4321 whatever instructions copy the biggest chunks of memory when performing
4322 copies to or from the variables which have types that you have aligned
4325 In the example above, if the size of each @code{short} is 2 bytes, then
4326 the size of the entire @code{struct S} type is 6 bytes. The smallest
4327 power of two which is greater than or equal to that is 8, so the
4328 compiler sets the alignment for the entire @code{struct S} type to 8
4331 Note that although you can ask the compiler to select a time-efficient
4332 alignment for a given type and then declare only individual stand-alone
4333 objects of that type, the compiler's ability to select a time-efficient
4334 alignment is primarily useful only when you plan to create arrays of
4335 variables having the relevant (efficiently aligned) type. If you
4336 declare or use arrays of variables of an efficiently-aligned type, then
4337 it is likely that your program will also be doing pointer arithmetic (or
4338 subscripting, which amounts to the same thing) on pointers to the
4339 relevant type, and the code that the compiler generates for these
4340 pointer arithmetic operations will often be more efficient for
4341 efficiently-aligned types than for other types.
4343 The @code{aligned} attribute can only increase the alignment; but you
4344 can decrease it by specifying @code{packed} as well. See below.
4346 Note that the effectiveness of @code{aligned} attributes may be limited
4347 by inherent limitations in your linker. On many systems, the linker is
4348 only able to arrange for variables to be aligned up to a certain maximum
4349 alignment. (For some linkers, the maximum supported alignment may
4350 be very very small.) If your linker is only able to align variables
4351 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4352 in an @code{__attribute__} will still only provide you with 8 byte
4353 alignment. See your linker documentation for further information.
4356 This attribute, attached to @code{struct} or @code{union} type
4357 definition, specifies that each member (other than zero-width bitfields)
4358 of the structure or union is placed to minimize the memory required. When
4359 attached to an @code{enum} definition, it indicates that the smallest
4360 integral type should be used.
4362 @opindex fshort-enums
4363 Specifying this attribute for @code{struct} and @code{union} types is
4364 equivalent to specifying the @code{packed} attribute on each of the
4365 structure or union members. Specifying the @option{-fshort-enums}
4366 flag on the line is equivalent to specifying the @code{packed}
4367 attribute on all @code{enum} definitions.
4369 In the following example @code{struct my_packed_struct}'s members are
4370 packed closely together, but the internal layout of its @code{s} member
4371 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4375 struct my_unpacked_struct
4381 struct __attribute__ ((__packed__)) my_packed_struct
4385 struct my_unpacked_struct s;
4389 You may only specify this attribute on the definition of a @code{enum},
4390 @code{struct} or @code{union}, not on a @code{typedef} which does not
4391 also define the enumerated type, structure or union.
4393 @item transparent_union
4394 This attribute, attached to a @code{union} type definition, indicates
4395 that any function parameter having that union type causes calls to that
4396 function to be treated in a special way.
4398 First, the argument corresponding to a transparent union type can be of
4399 any type in the union; no cast is required. Also, if the union contains
4400 a pointer type, the corresponding argument can be a null pointer
4401 constant or a void pointer expression; and if the union contains a void
4402 pointer type, the corresponding argument can be any pointer expression.
4403 If the union member type is a pointer, qualifiers like @code{const} on
4404 the referenced type must be respected, just as with normal pointer
4407 Second, the argument is passed to the function using the calling
4408 conventions of the first member of the transparent union, not the calling
4409 conventions of the union itself. All members of the union must have the
4410 same machine representation; this is necessary for this argument passing
4413 Transparent unions are designed for library functions that have multiple
4414 interfaces for compatibility reasons. For example, suppose the
4415 @code{wait} function must accept either a value of type @code{int *} to
4416 comply with Posix, or a value of type @code{union wait *} to comply with
4417 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4418 @code{wait} would accept both kinds of arguments, but it would also
4419 accept any other pointer type and this would make argument type checking
4420 less useful. Instead, @code{<sys/wait.h>} might define the interface
4424 typedef union __attribute__ ((__transparent_union__))
4428 @} wait_status_ptr_t;
4430 pid_t wait (wait_status_ptr_t);
4433 This interface allows either @code{int *} or @code{union wait *}
4434 arguments to be passed, using the @code{int *} calling convention.
4435 The program can call @code{wait} with arguments of either type:
4438 int w1 () @{ int w; return wait (&w); @}
4439 int w2 () @{ union wait w; return wait (&w); @}
4442 With this interface, @code{wait}'s implementation might look like this:
4445 pid_t wait (wait_status_ptr_t p)
4447 return waitpid (-1, p.__ip, 0);
4452 When attached to a type (including a @code{union} or a @code{struct}),
4453 this attribute means that variables of that type are meant to appear
4454 possibly unused. GCC will not produce a warning for any variables of
4455 that type, even if the variable appears to do nothing. This is often
4456 the case with lock or thread classes, which are usually defined and then
4457 not referenced, but contain constructors and destructors that have
4458 nontrivial bookkeeping functions.
4461 The @code{deprecated} attribute results in a warning if the type
4462 is used anywhere in the source file. This is useful when identifying
4463 types that are expected to be removed in a future version of a program.
4464 If possible, the warning also includes the location of the declaration
4465 of the deprecated type, to enable users to easily find further
4466 information about why the type is deprecated, or what they should do
4467 instead. Note that the warnings only occur for uses and then only
4468 if the type is being applied to an identifier that itself is not being
4469 declared as deprecated.
4472 typedef int T1 __attribute__ ((deprecated));
4476 typedef T1 T3 __attribute__ ((deprecated));
4477 T3 z __attribute__ ((deprecated));
4480 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4481 warning is issued for line 4 because T2 is not explicitly
4482 deprecated. Line 5 has no warning because T3 is explicitly
4483 deprecated. Similarly for line 6.
4485 The @code{deprecated} attribute can also be used for functions and
4486 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4489 Accesses through pointers to types with this attribute are not subject
4490 to type-based alias analysis, but are instead assumed to be able to alias
4491 any other type of objects. In the context of 6.5/7 an lvalue expression
4492 dereferencing such a pointer is treated like having a character type.
4493 See @option{-fstrict-aliasing} for more information on aliasing issues.
4494 This extension exists to support some vector APIs, in which pointers to
4495 one vector type are permitted to alias pointers to a different vector type.
4497 Note that an object of a type with this attribute does not have any
4503 typedef short __attribute__((__may_alias__)) short_a;
4509 short_a *b = (short_a *) &a;
4513 if (a == 0x12345678)
4520 If you replaced @code{short_a} with @code{short} in the variable
4521 declaration, the above program would abort when compiled with
4522 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4523 above in recent GCC versions.
4526 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4527 applied to class, struct, union and enum types. Unlike other type
4528 attributes, the attribute must appear between the initial keyword and
4529 the name of the type; it cannot appear after the body of the type.
4531 Note that the type visibility is applied to vague linkage entities
4532 associated with the class (vtable, typeinfo node, etc.). In
4533 particular, if a class is thrown as an exception in one shared object
4534 and caught in another, the class must have default visibility.
4535 Otherwise the two shared objects will be unable to use the same
4536 typeinfo node and exception handling will break.
4540 @subsection ARM Type Attributes
4542 On those ARM targets that support @code{dllimport} (such as Symbian
4543 OS), you can use the @code{notshared} attribute to indicate that the
4544 virtual table and other similar data for a class should not be
4545 exported from a DLL@. For example:
4548 class __declspec(notshared) C @{
4550 __declspec(dllimport) C();
4554 __declspec(dllexport)
4558 In this code, @code{C::C} is exported from the current DLL, but the
4559 virtual table for @code{C} is not exported. (You can use
4560 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4561 most Symbian OS code uses @code{__declspec}.)
4563 @anchor{i386 Type Attributes}
4564 @subsection i386 Type Attributes
4566 Two attributes are currently defined for i386 configurations:
4567 @code{ms_struct} and @code{gcc_struct}.
4573 @cindex @code{ms_struct}
4574 @cindex @code{gcc_struct}
4576 If @code{packed} is used on a structure, or if bit-fields are used
4577 it may be that the Microsoft ABI packs them differently
4578 than GCC would normally pack them. Particularly when moving packed
4579 data between functions compiled with GCC and the native Microsoft compiler
4580 (either via function call or as data in a file), it may be necessary to access
4583 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4584 compilers to match the native Microsoft compiler.
4587 To specify multiple attributes, separate them by commas within the
4588 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4591 @anchor{PowerPC Type Attributes}
4592 @subsection PowerPC Type Attributes
4594 Three attributes currently are defined for PowerPC configurations:
4595 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4597 For full documentation of the @code{ms_struct} and @code{gcc_struct}
4598 attributes please see the documentation in @ref{i386 Type Attributes}.
4600 The @code{altivec} attribute allows one to declare AltiVec vector data
4601 types supported by the AltiVec Programming Interface Manual. The
4602 attribute requires an argument to specify one of three vector types:
4603 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4604 and @code{bool__} (always followed by unsigned).
4607 __attribute__((altivec(vector__)))
4608 __attribute__((altivec(pixel__))) unsigned short
4609 __attribute__((altivec(bool__))) unsigned
4612 These attributes mainly are intended to support the @code{__vector},
4613 @code{__pixel}, and @code{__bool} AltiVec keywords.
4615 @anchor{SPU Type Attributes}
4616 @subsection SPU Type Attributes
4618 The SPU supports the @code{spu_vector} attribute for types. This attribute
4619 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
4620 Language Extensions Specification. It is intended to support the
4621 @code{__vector} keyword.
4625 @section An Inline Function is As Fast As a Macro
4626 @cindex inline functions
4627 @cindex integrating function code
4629 @cindex macros, inline alternative
4631 By declaring a function inline, you can direct GCC to make
4632 calls to that function faster. One way GCC can achieve this is to
4633 integrate that function's code into the code for its callers. This
4634 makes execution faster by eliminating the function-call overhead; in
4635 addition, if any of the actual argument values are constant, their
4636 known values may permit simplifications at compile time so that not
4637 all of the inline function's code needs to be included. The effect on
4638 code size is less predictable; object code may be larger or smaller
4639 with function inlining, depending on the particular case. You can
4640 also direct GCC to try to integrate all ``simple enough'' functions
4641 into their callers with the option @option{-finline-functions}.
4643 GCC implements three different semantics of declaring a function
4644 inline. One is available with @option{-std=gnu89} or
4645 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
4646 on all inline declarations, another when @option{-std=c99} or
4647 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
4648 is used when compiling C++.
4650 To declare a function inline, use the @code{inline} keyword in its
4651 declaration, like this:
4661 If you are writing a header file to be included in ISO C89 programs, write
4662 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4664 The three types of inlining behave similarly in two important cases:
4665 when the @code{inline} keyword is used on a @code{static} function,
4666 like the example above, and when a function is first declared without
4667 using the @code{inline} keyword and then is defined with
4668 @code{inline}, like this:
4671 extern int inc (int *a);
4679 In both of these common cases, the program behaves the same as if you
4680 had not used the @code{inline} keyword, except for its speed.
4682 @cindex inline functions, omission of
4683 @opindex fkeep-inline-functions
4684 When a function is both inline and @code{static}, if all calls to the
4685 function are integrated into the caller, and the function's address is
4686 never used, then the function's own assembler code is never referenced.
4687 In this case, GCC does not actually output assembler code for the
4688 function, unless you specify the option @option{-fkeep-inline-functions}.
4689 Some calls cannot be integrated for various reasons (in particular,
4690 calls that precede the function's definition cannot be integrated, and
4691 neither can recursive calls within the definition). If there is a
4692 nonintegrated call, then the function is compiled to assembler code as
4693 usual. The function must also be compiled as usual if the program
4694 refers to its address, because that can't be inlined.
4697 Note that certain usages in a function definition can make it unsuitable
4698 for inline substitution. Among these usages are: use of varargs, use of
4699 alloca, use of variable sized data types (@pxref{Variable Length}),
4700 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4701 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4702 will warn when a function marked @code{inline} could not be substituted,
4703 and will give the reason for the failure.
4705 @cindex automatic @code{inline} for C++ member fns
4706 @cindex @code{inline} automatic for C++ member fns
4707 @cindex member fns, automatically @code{inline}
4708 @cindex C++ member fns, automatically @code{inline}
4709 @opindex fno-default-inline
4710 As required by ISO C++, GCC considers member functions defined within
4711 the body of a class to be marked inline even if they are
4712 not explicitly declared with the @code{inline} keyword. You can
4713 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4714 Options,,Options Controlling C++ Dialect}.
4716 GCC does not inline any functions when not optimizing unless you specify
4717 the @samp{always_inline} attribute for the function, like this:
4720 /* @r{Prototype.} */
4721 inline void foo (const char) __attribute__((always_inline));
4724 The remainder of this section is specific to GNU C89 inlining.
4726 @cindex non-static inline function
4727 When an inline function is not @code{static}, then the compiler must assume
4728 that there may be calls from other source files; since a global symbol can
4729 be defined only once in any program, the function must not be defined in
4730 the other source files, so the calls therein cannot be integrated.
4731 Therefore, a non-@code{static} inline function is always compiled on its
4732 own in the usual fashion.
4734 If you specify both @code{inline} and @code{extern} in the function
4735 definition, then the definition is used only for inlining. In no case
4736 is the function compiled on its own, not even if you refer to its
4737 address explicitly. Such an address becomes an external reference, as
4738 if you had only declared the function, and had not defined it.
4740 This combination of @code{inline} and @code{extern} has almost the
4741 effect of a macro. The way to use it is to put a function definition in
4742 a header file with these keywords, and put another copy of the
4743 definition (lacking @code{inline} and @code{extern}) in a library file.
4744 The definition in the header file will cause most calls to the function
4745 to be inlined. If any uses of the function remain, they will refer to
4746 the single copy in the library.
4749 @section Assembler Instructions with C Expression Operands
4750 @cindex extended @code{asm}
4751 @cindex @code{asm} expressions
4752 @cindex assembler instructions
4755 In an assembler instruction using @code{asm}, you can specify the
4756 operands of the instruction using C expressions. This means you need not
4757 guess which registers or memory locations will contain the data you want
4760 You must specify an assembler instruction template much like what
4761 appears in a machine description, plus an operand constraint string for
4764 For example, here is how to use the 68881's @code{fsinx} instruction:
4767 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4771 Here @code{angle} is the C expression for the input operand while
4772 @code{result} is that of the output operand. Each has @samp{"f"} as its
4773 operand constraint, saying that a floating point register is required.
4774 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4775 output operands' constraints must use @samp{=}. The constraints use the
4776 same language used in the machine description (@pxref{Constraints}).
4778 Each operand is described by an operand-constraint string followed by
4779 the C expression in parentheses. A colon separates the assembler
4780 template from the first output operand and another separates the last
4781 output operand from the first input, if any. Commas separate the
4782 operands within each group. The total number of operands is currently
4783 limited to 30; this limitation may be lifted in some future version of
4786 If there are no output operands but there are input operands, you must
4787 place two consecutive colons surrounding the place where the output
4790 As of GCC version 3.1, it is also possible to specify input and output
4791 operands using symbolic names which can be referenced within the
4792 assembler code. These names are specified inside square brackets
4793 preceding the constraint string, and can be referenced inside the
4794 assembler code using @code{%[@var{name}]} instead of a percentage sign
4795 followed by the operand number. Using named operands the above example
4799 asm ("fsinx %[angle],%[output]"
4800 : [output] "=f" (result)
4801 : [angle] "f" (angle));
4805 Note that the symbolic operand names have no relation whatsoever to
4806 other C identifiers. You may use any name you like, even those of
4807 existing C symbols, but you must ensure that no two operands within the same
4808 assembler construct use the same symbolic name.
4810 Output operand expressions must be lvalues; the compiler can check this.
4811 The input operands need not be lvalues. The compiler cannot check
4812 whether the operands have data types that are reasonable for the
4813 instruction being executed. It does not parse the assembler instruction
4814 template and does not know what it means or even whether it is valid
4815 assembler input. The extended @code{asm} feature is most often used for
4816 machine instructions the compiler itself does not know exist. If
4817 the output expression cannot be directly addressed (for example, it is a
4818 bit-field), your constraint must allow a register. In that case, GCC
4819 will use the register as the output of the @code{asm}, and then store
4820 that register into the output.
4822 The ordinary output operands must be write-only; GCC will assume that
4823 the values in these operands before the instruction are dead and need
4824 not be generated. Extended asm supports input-output or read-write
4825 operands. Use the constraint character @samp{+} to indicate such an
4826 operand and list it with the output operands. You should only use
4827 read-write operands when the constraints for the operand (or the
4828 operand in which only some of the bits are to be changed) allow a
4831 You may, as an alternative, logically split its function into two
4832 separate operands, one input operand and one write-only output
4833 operand. The connection between them is expressed by constraints
4834 which say they need to be in the same location when the instruction
4835 executes. You can use the same C expression for both operands, or
4836 different expressions. For example, here we write the (fictitious)
4837 @samp{combine} instruction with @code{bar} as its read-only source
4838 operand and @code{foo} as its read-write destination:
4841 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4845 The constraint @samp{"0"} for operand 1 says that it must occupy the
4846 same location as operand 0. A number in constraint is allowed only in
4847 an input operand and it must refer to an output operand.
4849 Only a number in the constraint can guarantee that one operand will be in
4850 the same place as another. The mere fact that @code{foo} is the value
4851 of both operands is not enough to guarantee that they will be in the
4852 same place in the generated assembler code. The following would not
4856 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4859 Various optimizations or reloading could cause operands 0 and 1 to be in
4860 different registers; GCC knows no reason not to do so. For example, the
4861 compiler might find a copy of the value of @code{foo} in one register and
4862 use it for operand 1, but generate the output operand 0 in a different
4863 register (copying it afterward to @code{foo}'s own address). Of course,
4864 since the register for operand 1 is not even mentioned in the assembler
4865 code, the result will not work, but GCC can't tell that.
4867 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4868 the operand number for a matching constraint. For example:
4871 asm ("cmoveq %1,%2,%[result]"
4872 : [result] "=r"(result)
4873 : "r" (test), "r"(new), "[result]"(old));
4876 Sometimes you need to make an @code{asm} operand be a specific register,
4877 but there's no matching constraint letter for that register @emph{by
4878 itself}. To force the operand into that register, use a local variable
4879 for the operand and specify the register in the variable declaration.
4880 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4881 register constraint letter that matches the register:
4884 register int *p1 asm ("r0") = @dots{};
4885 register int *p2 asm ("r1") = @dots{};
4886 register int *result asm ("r0");
4887 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4890 @anchor{Example of asm with clobbered asm reg}
4891 In the above example, beware that a register that is call-clobbered by
4892 the target ABI will be overwritten by any function call in the
4893 assignment, including library calls for arithmetic operators.
4894 Also a register may be clobbered when generating some operations,
4895 like variable shift, memory copy or memory move on x86.
4896 Assuming it is a call-clobbered register, this may happen to @code{r0}
4897 above by the assignment to @code{p2}. If you have to use such a
4898 register, use temporary variables for expressions between the register
4903 register int *p1 asm ("r0") = @dots{};
4904 register int *p2 asm ("r1") = t1;
4905 register int *result asm ("r0");
4906 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4909 Some instructions clobber specific hard registers. To describe this,
4910 write a third colon after the input operands, followed by the names of
4911 the clobbered hard registers (given as strings). Here is a realistic
4912 example for the VAX:
4915 asm volatile ("movc3 %0,%1,%2"
4916 : /* @r{no outputs} */
4917 : "g" (from), "g" (to), "g" (count)
4918 : "r0", "r1", "r2", "r3", "r4", "r5");
4921 You may not write a clobber description in a way that overlaps with an
4922 input or output operand. For example, you may not have an operand
4923 describing a register class with one member if you mention that register
4924 in the clobber list. Variables declared to live in specific registers
4925 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4926 have no part mentioned in the clobber description.
4927 There is no way for you to specify that an input
4928 operand is modified without also specifying it as an output
4929 operand. Note that if all the output operands you specify are for this
4930 purpose (and hence unused), you will then also need to specify
4931 @code{volatile} for the @code{asm} construct, as described below, to
4932 prevent GCC from deleting the @code{asm} statement as unused.
4934 If you refer to a particular hardware register from the assembler code,
4935 you will probably have to list the register after the third colon to
4936 tell the compiler the register's value is modified. In some assemblers,
4937 the register names begin with @samp{%}; to produce one @samp{%} in the
4938 assembler code, you must write @samp{%%} in the input.
4940 If your assembler instruction can alter the condition code register, add
4941 @samp{cc} to the list of clobbered registers. GCC on some machines
4942 represents the condition codes as a specific hardware register;
4943 @samp{cc} serves to name this register. On other machines, the
4944 condition code is handled differently, and specifying @samp{cc} has no
4945 effect. But it is valid no matter what the machine.
4947 If your assembler instructions access memory in an unpredictable
4948 fashion, add @samp{memory} to the list of clobbered registers. This
4949 will cause GCC to not keep memory values cached in registers across the
4950 assembler instruction and not optimize stores or loads to that memory.
4951 You will also want to add the @code{volatile} keyword if the memory
4952 affected is not listed in the inputs or outputs of the @code{asm}, as
4953 the @samp{memory} clobber does not count as a side-effect of the
4954 @code{asm}. If you know how large the accessed memory is, you can add
4955 it as input or output but if this is not known, you should add
4956 @samp{memory}. As an example, if you access ten bytes of a string, you
4957 can use a memory input like:
4960 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4963 Note that in the following example the memory input is necessary,
4964 otherwise GCC might optimize the store to @code{x} away:
4971 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4972 "=&d" (r) : "a" (y), "m" (*y));
4977 You can put multiple assembler instructions together in a single
4978 @code{asm} template, separated by the characters normally used in assembly
4979 code for the system. A combination that works in most places is a newline
4980 to break the line, plus a tab character to move to the instruction field
4981 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4982 assembler allows semicolons as a line-breaking character. Note that some
4983 assembler dialects use semicolons to start a comment.
4984 The input operands are guaranteed not to use any of the clobbered
4985 registers, and neither will the output operands' addresses, so you can
4986 read and write the clobbered registers as many times as you like. Here
4987 is an example of multiple instructions in a template; it assumes the
4988 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4991 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4993 : "g" (from), "g" (to)
4997 Unless an output operand has the @samp{&} constraint modifier, GCC
4998 may allocate it in the same register as an unrelated input operand, on
4999 the assumption the inputs are consumed before the outputs are produced.
5000 This assumption may be false if the assembler code actually consists of
5001 more than one instruction. In such a case, use @samp{&} for each output
5002 operand that may not overlap an input. @xref{Modifiers}.
5004 If you want to test the condition code produced by an assembler
5005 instruction, you must include a branch and a label in the @code{asm}
5006 construct, as follows:
5009 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5015 This assumes your assembler supports local labels, as the GNU assembler
5016 and most Unix assemblers do.
5018 Speaking of labels, jumps from one @code{asm} to another are not
5019 supported. The compiler's optimizers do not know about these jumps, and
5020 therefore they cannot take account of them when deciding how to
5023 @cindex macros containing @code{asm}
5024 Usually the most convenient way to use these @code{asm} instructions is to
5025 encapsulate them in macros that look like functions. For example,
5029 (@{ double __value, __arg = (x); \
5030 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5035 Here the variable @code{__arg} is used to make sure that the instruction
5036 operates on a proper @code{double} value, and to accept only those
5037 arguments @code{x} which can convert automatically to a @code{double}.
5039 Another way to make sure the instruction operates on the correct data
5040 type is to use a cast in the @code{asm}. This is different from using a
5041 variable @code{__arg} in that it converts more different types. For
5042 example, if the desired type were @code{int}, casting the argument to
5043 @code{int} would accept a pointer with no complaint, while assigning the
5044 argument to an @code{int} variable named @code{__arg} would warn about
5045 using a pointer unless the caller explicitly casts it.
5047 If an @code{asm} has output operands, GCC assumes for optimization
5048 purposes the instruction has no side effects except to change the output
5049 operands. This does not mean instructions with a side effect cannot be
5050 used, but you must be careful, because the compiler may eliminate them
5051 if the output operands aren't used, or move them out of loops, or
5052 replace two with one if they constitute a common subexpression. Also,
5053 if your instruction does have a side effect on a variable that otherwise
5054 appears not to change, the old value of the variable may be reused later
5055 if it happens to be found in a register.
5057 You can prevent an @code{asm} instruction from being deleted
5058 by writing the keyword @code{volatile} after
5059 the @code{asm}. For example:
5062 #define get_and_set_priority(new) \
5064 asm volatile ("get_and_set_priority %0, %1" \
5065 : "=g" (__old) : "g" (new)); \
5070 The @code{volatile} keyword indicates that the instruction has
5071 important side-effects. GCC will not delete a volatile @code{asm} if
5072 it is reachable. (The instruction can still be deleted if GCC can
5073 prove that control-flow will never reach the location of the
5074 instruction.) Note that even a volatile @code{asm} instruction
5075 can be moved relative to other code, including across jump
5076 instructions. For example, on many targets there is a system
5077 register which can be set to control the rounding mode of
5078 floating point operations. You might try
5079 setting it with a volatile @code{asm}, like this PowerPC example:
5082 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5087 This will not work reliably, as the compiler may move the addition back
5088 before the volatile @code{asm}. To make it work you need to add an
5089 artificial dependency to the @code{asm} referencing a variable in the code
5090 you don't want moved, for example:
5093 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5097 Similarly, you can't expect a
5098 sequence of volatile @code{asm} instructions to remain perfectly
5099 consecutive. If you want consecutive output, use a single @code{asm}.
5100 Also, GCC will perform some optimizations across a volatile @code{asm}
5101 instruction; GCC does not ``forget everything'' when it encounters
5102 a volatile @code{asm} instruction the way some other compilers do.
5104 An @code{asm} instruction without any output operands will be treated
5105 identically to a volatile @code{asm} instruction.
5107 It is a natural idea to look for a way to give access to the condition
5108 code left by the assembler instruction. However, when we attempted to
5109 implement this, we found no way to make it work reliably. The problem
5110 is that output operands might need reloading, which would result in
5111 additional following ``store'' instructions. On most machines, these
5112 instructions would alter the condition code before there was time to
5113 test it. This problem doesn't arise for ordinary ``test'' and
5114 ``compare'' instructions because they don't have any output operands.
5116 For reasons similar to those described above, it is not possible to give
5117 an assembler instruction access to the condition code left by previous
5120 If you are writing a header file that should be includable in ISO C
5121 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5124 @subsection Size of an @code{asm}
5126 Some targets require that GCC track the size of each instruction used in
5127 order to generate correct code. Because the final length of an
5128 @code{asm} is only known by the assembler, GCC must make an estimate as
5129 to how big it will be. The estimate is formed by counting the number of
5130 statements in the pattern of the @code{asm} and multiplying that by the
5131 length of the longest instruction on that processor. Statements in the
5132 @code{asm} are identified by newline characters and whatever statement
5133 separator characters are supported by the assembler; on most processors
5134 this is the `@code{;}' character.
5136 Normally, GCC's estimate is perfectly adequate to ensure that correct
5137 code is generated, but it is possible to confuse the compiler if you use
5138 pseudo instructions or assembler macros that expand into multiple real
5139 instructions or if you use assembler directives that expand to more
5140 space in the object file than would be needed for a single instruction.
5141 If this happens then the assembler will produce a diagnostic saying that
5142 a label is unreachable.
5144 @subsection i386 floating point asm operands
5146 There are several rules on the usage of stack-like regs in
5147 asm_operands insns. These rules apply only to the operands that are
5152 Given a set of input regs that die in an asm_operands, it is
5153 necessary to know which are implicitly popped by the asm, and
5154 which must be explicitly popped by gcc.
5156 An input reg that is implicitly popped by the asm must be
5157 explicitly clobbered, unless it is constrained to match an
5161 For any input reg that is implicitly popped by an asm, it is
5162 necessary to know how to adjust the stack to compensate for the pop.
5163 If any non-popped input is closer to the top of the reg-stack than
5164 the implicitly popped reg, it would not be possible to know what the
5165 stack looked like---it's not clear how the rest of the stack ``slides
5168 All implicitly popped input regs must be closer to the top of
5169 the reg-stack than any input that is not implicitly popped.
5171 It is possible that if an input dies in an insn, reload might
5172 use the input reg for an output reload. Consider this example:
5175 asm ("foo" : "=t" (a) : "f" (b));
5178 This asm says that input B is not popped by the asm, and that
5179 the asm pushes a result onto the reg-stack, i.e., the stack is one
5180 deeper after the asm than it was before. But, it is possible that
5181 reload will think that it can use the same reg for both the input and
5182 the output, if input B dies in this insn.
5184 If any input operand uses the @code{f} constraint, all output reg
5185 constraints must use the @code{&} earlyclobber.
5187 The asm above would be written as
5190 asm ("foo" : "=&t" (a) : "f" (b));
5194 Some operands need to be in particular places on the stack. All
5195 output operands fall in this category---there is no other way to
5196 know which regs the outputs appear in unless the user indicates
5197 this in the constraints.
5199 Output operands must specifically indicate which reg an output
5200 appears in after an asm. @code{=f} is not allowed: the operand
5201 constraints must select a class with a single reg.
5204 Output operands may not be ``inserted'' between existing stack regs.
5205 Since no 387 opcode uses a read/write operand, all output operands
5206 are dead before the asm_operands, and are pushed by the asm_operands.
5207 It makes no sense to push anywhere but the top of the reg-stack.
5209 Output operands must start at the top of the reg-stack: output
5210 operands may not ``skip'' a reg.
5213 Some asm statements may need extra stack space for internal
5214 calculations. This can be guaranteed by clobbering stack registers
5215 unrelated to the inputs and outputs.
5219 Here are a couple of reasonable asms to want to write. This asm
5220 takes one input, which is internally popped, and produces two outputs.
5223 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
5226 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
5227 and replaces them with one output. The user must code the @code{st(1)}
5228 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
5231 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
5237 @section Controlling Names Used in Assembler Code
5238 @cindex assembler names for identifiers
5239 @cindex names used in assembler code
5240 @cindex identifiers, names in assembler code
5242 You can specify the name to be used in the assembler code for a C
5243 function or variable by writing the @code{asm} (or @code{__asm__})
5244 keyword after the declarator as follows:
5247 int foo asm ("myfoo") = 2;
5251 This specifies that the name to be used for the variable @code{foo} in
5252 the assembler code should be @samp{myfoo} rather than the usual
5255 On systems where an underscore is normally prepended to the name of a C
5256 function or variable, this feature allows you to define names for the
5257 linker that do not start with an underscore.
5259 It does not make sense to use this feature with a non-static local
5260 variable since such variables do not have assembler names. If you are
5261 trying to put the variable in a particular register, see @ref{Explicit
5262 Reg Vars}. GCC presently accepts such code with a warning, but will
5263 probably be changed to issue an error, rather than a warning, in the
5266 You cannot use @code{asm} in this way in a function @emph{definition}; but
5267 you can get the same effect by writing a declaration for the function
5268 before its definition and putting @code{asm} there, like this:
5271 extern func () asm ("FUNC");
5278 It is up to you to make sure that the assembler names you choose do not
5279 conflict with any other assembler symbols. Also, you must not use a
5280 register name; that would produce completely invalid assembler code. GCC
5281 does not as yet have the ability to store static variables in registers.
5282 Perhaps that will be added.
5284 @node Explicit Reg Vars
5285 @section Variables in Specified Registers
5286 @cindex explicit register variables
5287 @cindex variables in specified registers
5288 @cindex specified registers
5289 @cindex registers, global allocation
5291 GNU C allows you to put a few global variables into specified hardware
5292 registers. You can also specify the register in which an ordinary
5293 register variable should be allocated.
5297 Global register variables reserve registers throughout the program.
5298 This may be useful in programs such as programming language
5299 interpreters which have a couple of global variables that are accessed
5303 Local register variables in specific registers do not reserve the
5304 registers, except at the point where they are used as input or output
5305 operands in an @code{asm} statement and the @code{asm} statement itself is
5306 not deleted. The compiler's data flow analysis is capable of determining
5307 where the specified registers contain live values, and where they are
5308 available for other uses. Stores into local register variables may be deleted
5309 when they appear to be dead according to dataflow analysis. References
5310 to local register variables may be deleted or moved or simplified.
5312 These local variables are sometimes convenient for use with the extended
5313 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
5314 output of the assembler instruction directly into a particular register.
5315 (This will work provided the register you specify fits the constraints
5316 specified for that operand in the @code{asm}.)
5324 @node Global Reg Vars
5325 @subsection Defining Global Register Variables
5326 @cindex global register variables
5327 @cindex registers, global variables in
5329 You can define a global register variable in GNU C like this:
5332 register int *foo asm ("a5");
5336 Here @code{a5} is the name of the register which should be used. Choose a
5337 register which is normally saved and restored by function calls on your
5338 machine, so that library routines will not clobber it.
5340 Naturally the register name is cpu-dependent, so you would need to
5341 conditionalize your program according to cpu type. The register
5342 @code{a5} would be a good choice on a 68000 for a variable of pointer
5343 type. On machines with register windows, be sure to choose a ``global''
5344 register that is not affected magically by the function call mechanism.
5346 In addition, operating systems on one type of cpu may differ in how they
5347 name the registers; then you would need additional conditionals. For
5348 example, some 68000 operating systems call this register @code{%a5}.
5350 Eventually there may be a way of asking the compiler to choose a register
5351 automatically, but first we need to figure out how it should choose and
5352 how to enable you to guide the choice. No solution is evident.
5354 Defining a global register variable in a certain register reserves that
5355 register entirely for this use, at least within the current compilation.
5356 The register will not be allocated for any other purpose in the functions
5357 in the current compilation. The register will not be saved and restored by
5358 these functions. Stores into this register are never deleted even if they
5359 would appear to be dead, but references may be deleted or moved or
5362 It is not safe to access the global register variables from signal
5363 handlers, or from more than one thread of control, because the system
5364 library routines may temporarily use the register for other things (unless
5365 you recompile them specially for the task at hand).
5367 @cindex @code{qsort}, and global register variables
5368 It is not safe for one function that uses a global register variable to
5369 call another such function @code{foo} by way of a third function
5370 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
5371 different source file in which the variable wasn't declared). This is
5372 because @code{lose} might save the register and put some other value there.
5373 For example, you can't expect a global register variable to be available in
5374 the comparison-function that you pass to @code{qsort}, since @code{qsort}
5375 might have put something else in that register. (If you are prepared to
5376 recompile @code{qsort} with the same global register variable, you can
5377 solve this problem.)
5379 If you want to recompile @code{qsort} or other source files which do not
5380 actually use your global register variable, so that they will not use that
5381 register for any other purpose, then it suffices to specify the compiler
5382 option @option{-ffixed-@var{reg}}. You need not actually add a global
5383 register declaration to their source code.
5385 A function which can alter the value of a global register variable cannot
5386 safely be called from a function compiled without this variable, because it
5387 could clobber the value the caller expects to find there on return.
5388 Therefore, the function which is the entry point into the part of the
5389 program that uses the global register variable must explicitly save and
5390 restore the value which belongs to its caller.
5392 @cindex register variable after @code{longjmp}
5393 @cindex global register after @code{longjmp}
5394 @cindex value after @code{longjmp}
5397 On most machines, @code{longjmp} will restore to each global register
5398 variable the value it had at the time of the @code{setjmp}. On some
5399 machines, however, @code{longjmp} will not change the value of global
5400 register variables. To be portable, the function that called @code{setjmp}
5401 should make other arrangements to save the values of the global register
5402 variables, and to restore them in a @code{longjmp}. This way, the same
5403 thing will happen regardless of what @code{longjmp} does.
5405 All global register variable declarations must precede all function
5406 definitions. If such a declaration could appear after function
5407 definitions, the declaration would be too late to prevent the register from
5408 being used for other purposes in the preceding functions.
5410 Global register variables may not have initial values, because an
5411 executable file has no means to supply initial contents for a register.
5413 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5414 registers, but certain library functions, such as @code{getwd}, as well
5415 as the subroutines for division and remainder, modify g3 and g4. g1 and
5416 g2 are local temporaries.
5418 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5419 Of course, it will not do to use more than a few of those.
5421 @node Local Reg Vars
5422 @subsection Specifying Registers for Local Variables
5423 @cindex local variables, specifying registers
5424 @cindex specifying registers for local variables
5425 @cindex registers for local variables
5427 You can define a local register variable with a specified register
5431 register int *foo asm ("a5");
5435 Here @code{a5} is the name of the register which should be used. Note
5436 that this is the same syntax used for defining global register
5437 variables, but for a local variable it would appear within a function.
5439 Naturally the register name is cpu-dependent, but this is not a
5440 problem, since specific registers are most often useful with explicit
5441 assembler instructions (@pxref{Extended Asm}). Both of these things
5442 generally require that you conditionalize your program according to
5445 In addition, operating systems on one type of cpu may differ in how they
5446 name the registers; then you would need additional conditionals. For
5447 example, some 68000 operating systems call this register @code{%a5}.
5449 Defining such a register variable does not reserve the register; it
5450 remains available for other uses in places where flow control determines
5451 the variable's value is not live.
5453 This option does not guarantee that GCC will generate code that has
5454 this variable in the register you specify at all times. You may not
5455 code an explicit reference to this register in the @emph{assembler
5456 instruction template} part of an @code{asm} statement and assume it will
5457 always refer to this variable. However, using the variable as an
5458 @code{asm} @emph{operand} guarantees that the specified register is used
5461 Stores into local register variables may be deleted when they appear to be dead
5462 according to dataflow analysis. References to local register variables may
5463 be deleted or moved or simplified.
5465 As for global register variables, it's recommended that you choose a
5466 register which is normally saved and restored by function calls on
5467 your machine, so that library routines will not clobber it. A common
5468 pitfall is to initialize multiple call-clobbered registers with
5469 arbitrary expressions, where a function call or library call for an
5470 arithmetic operator will overwrite a register value from a previous
5471 assignment, for example @code{r0} below:
5473 register int *p1 asm ("r0") = @dots{};
5474 register int *p2 asm ("r1") = @dots{};
5476 In those cases, a solution is to use a temporary variable for
5477 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
5479 @node Alternate Keywords
5480 @section Alternate Keywords
5481 @cindex alternate keywords
5482 @cindex keywords, alternate
5484 @option{-ansi} and the various @option{-std} options disable certain
5485 keywords. This causes trouble when you want to use GNU C extensions, or
5486 a general-purpose header file that should be usable by all programs,
5487 including ISO C programs. The keywords @code{asm}, @code{typeof} and
5488 @code{inline} are not available in programs compiled with
5489 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
5490 program compiled with @option{-std=c99}). The ISO C99 keyword
5491 @code{restrict} is only available when @option{-std=gnu99} (which will
5492 eventually be the default) or @option{-std=c99} (or the equivalent
5493 @option{-std=iso9899:1999}) is used.
5495 The way to solve these problems is to put @samp{__} at the beginning and
5496 end of each problematical keyword. For example, use @code{__asm__}
5497 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
5499 Other C compilers won't accept these alternative keywords; if you want to
5500 compile with another compiler, you can define the alternate keywords as
5501 macros to replace them with the customary keywords. It looks like this:
5509 @findex __extension__
5511 @option{-pedantic} and other options cause warnings for many GNU C extensions.
5513 prevent such warnings within one expression by writing
5514 @code{__extension__} before the expression. @code{__extension__} has no
5515 effect aside from this.
5517 @node Incomplete Enums
5518 @section Incomplete @code{enum} Types
5520 You can define an @code{enum} tag without specifying its possible values.
5521 This results in an incomplete type, much like what you get if you write
5522 @code{struct foo} without describing the elements. A later declaration
5523 which does specify the possible values completes the type.
5525 You can't allocate variables or storage using the type while it is
5526 incomplete. However, you can work with pointers to that type.
5528 This extension may not be very useful, but it makes the handling of
5529 @code{enum} more consistent with the way @code{struct} and @code{union}
5532 This extension is not supported by GNU C++.
5534 @node Function Names
5535 @section Function Names as Strings
5536 @cindex @code{__func__} identifier
5537 @cindex @code{__FUNCTION__} identifier
5538 @cindex @code{__PRETTY_FUNCTION__} identifier
5540 GCC provides three magic variables which hold the name of the current
5541 function, as a string. The first of these is @code{__func__}, which
5542 is part of the C99 standard:
5544 The identifier @code{__func__} is implicitly declared by the translator
5545 as if, immediately following the opening brace of each function
5546 definition, the declaration
5549 static const char __func__[] = "function-name";
5553 appeared, where function-name is the name of the lexically-enclosing
5554 function. This name is the unadorned name of the function.
5556 @code{__FUNCTION__} is another name for @code{__func__}. Older
5557 versions of GCC recognize only this name. However, it is not
5558 standardized. For maximum portability, we recommend you use
5559 @code{__func__}, but provide a fallback definition with the
5563 #if __STDC_VERSION__ < 199901L
5565 # define __func__ __FUNCTION__
5567 # define __func__ "<unknown>"
5572 In C, @code{__PRETTY_FUNCTION__} is yet another name for
5573 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
5574 the type signature of the function as well as its bare name. For
5575 example, this program:
5579 extern int printf (char *, ...);
5586 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
5587 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
5605 __PRETTY_FUNCTION__ = void a::sub(int)
5608 These identifiers are not preprocessor macros. In GCC 3.3 and
5609 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
5610 were treated as string literals; they could be used to initialize
5611 @code{char} arrays, and they could be concatenated with other string
5612 literals. GCC 3.4 and later treat them as variables, like
5613 @code{__func__}. In C++, @code{__FUNCTION__} and
5614 @code{__PRETTY_FUNCTION__} have always been variables.
5616 @node Return Address
5617 @section Getting the Return or Frame Address of a Function
5619 These functions may be used to get information about the callers of a
5622 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
5623 This function returns the return address of the current function, or of
5624 one of its callers. The @var{level} argument is number of frames to
5625 scan up the call stack. A value of @code{0} yields the return address
5626 of the current function, a value of @code{1} yields the return address
5627 of the caller of the current function, and so forth. When inlining
5628 the expected behavior is that the function will return the address of
5629 the function that will be returned to. To work around this behavior use
5630 the @code{noinline} function attribute.
5632 The @var{level} argument must be a constant integer.
5634 On some machines it may be impossible to determine the return address of
5635 any function other than the current one; in such cases, or when the top
5636 of the stack has been reached, this function will return @code{0} or a
5637 random value. In addition, @code{__builtin_frame_address} may be used
5638 to determine if the top of the stack has been reached.
5640 This function should only be used with a nonzero argument for debugging
5644 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
5645 This function is similar to @code{__builtin_return_address}, but it
5646 returns the address of the function frame rather than the return address
5647 of the function. Calling @code{__builtin_frame_address} with a value of
5648 @code{0} yields the frame address of the current function, a value of
5649 @code{1} yields the frame address of the caller of the current function,
5652 The frame is the area on the stack which holds local variables and saved
5653 registers. The frame address is normally the address of the first word
5654 pushed on to the stack by the function. However, the exact definition
5655 depends upon the processor and the calling convention. If the processor
5656 has a dedicated frame pointer register, and the function has a frame,
5657 then @code{__builtin_frame_address} will return the value of the frame
5660 On some machines it may be impossible to determine the frame address of
5661 any function other than the current one; in such cases, or when the top
5662 of the stack has been reached, this function will return @code{0} if
5663 the first frame pointer is properly initialized by the startup code.
5665 This function should only be used with a nonzero argument for debugging
5669 @node Vector Extensions
5670 @section Using vector instructions through built-in functions
5672 On some targets, the instruction set contains SIMD vector instructions that
5673 operate on multiple values contained in one large register at the same time.
5674 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5677 The first step in using these extensions is to provide the necessary data
5678 types. This should be done using an appropriate @code{typedef}:
5681 typedef int v4si __attribute__ ((vector_size (16)));
5684 The @code{int} type specifies the base type, while the attribute specifies
5685 the vector size for the variable, measured in bytes. For example, the
5686 declaration above causes the compiler to set the mode for the @code{v4si}
5687 type to be 16 bytes wide and divided into @code{int} sized units. For
5688 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5689 corresponding mode of @code{foo} will be @acronym{V4SI}.
5691 The @code{vector_size} attribute is only applicable to integral and
5692 float scalars, although arrays, pointers, and function return values
5693 are allowed in conjunction with this construct.
5695 All the basic integer types can be used as base types, both as signed
5696 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5697 @code{long long}. In addition, @code{float} and @code{double} can be
5698 used to build floating-point vector types.
5700 Specifying a combination that is not valid for the current architecture
5701 will cause GCC to synthesize the instructions using a narrower mode.
5702 For example, if you specify a variable of type @code{V4SI} and your
5703 architecture does not allow for this specific SIMD type, GCC will
5704 produce code that uses 4 @code{SIs}.
5706 The types defined in this manner can be used with a subset of normal C
5707 operations. Currently, GCC will allow using the following operators
5708 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5710 The operations behave like C++ @code{valarrays}. Addition is defined as
5711 the addition of the corresponding elements of the operands. For
5712 example, in the code below, each of the 4 elements in @var{a} will be
5713 added to the corresponding 4 elements in @var{b} and the resulting
5714 vector will be stored in @var{c}.
5717 typedef int v4si __attribute__ ((vector_size (16)));
5724 Subtraction, multiplication, division, and the logical operations
5725 operate in a similar manner. Likewise, the result of using the unary
5726 minus or complement operators on a vector type is a vector whose
5727 elements are the negative or complemented values of the corresponding
5728 elements in the operand.
5730 You can declare variables and use them in function calls and returns, as
5731 well as in assignments and some casts. You can specify a vector type as
5732 a return type for a function. Vector types can also be used as function
5733 arguments. It is possible to cast from one vector type to another,
5734 provided they are of the same size (in fact, you can also cast vectors
5735 to and from other datatypes of the same size).
5737 You cannot operate between vectors of different lengths or different
5738 signedness without a cast.
5740 A port that supports hardware vector operations, usually provides a set
5741 of built-in functions that can be used to operate on vectors. For
5742 example, a function to add two vectors and multiply the result by a
5743 third could look like this:
5746 v4si f (v4si a, v4si b, v4si c)
5748 v4si tmp = __builtin_addv4si (a, b);
5749 return __builtin_mulv4si (tmp, c);
5756 @findex __builtin_offsetof
5758 GCC implements for both C and C++ a syntactic extension to implement
5759 the @code{offsetof} macro.
5763 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5765 offsetof_member_designator:
5767 | offsetof_member_designator "." @code{identifier}
5768 | offsetof_member_designator "[" @code{expr} "]"
5771 This extension is sufficient such that
5774 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5777 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5778 may be dependent. In either case, @var{member} may consist of a single
5779 identifier, or a sequence of member accesses and array references.
5781 @node Atomic Builtins
5782 @section Built-in functions for atomic memory access
5784 The following builtins are intended to be compatible with those described
5785 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5786 section 7.4. As such, they depart from the normal GCC practice of using
5787 the ``__builtin_'' prefix, and further that they are overloaded such that
5788 they work on multiple types.
5790 The definition given in the Intel documentation allows only for the use of
5791 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5792 counterparts. GCC will allow any integral scalar or pointer type that is
5793 1, 2, 4 or 8 bytes in length.
5795 Not all operations are supported by all target processors. If a particular
5796 operation cannot be implemented on the target processor, a warning will be
5797 generated and a call an external function will be generated. The external
5798 function will carry the same name as the builtin, with an additional suffix
5799 @samp{_@var{n}} where @var{n} is the size of the data type.
5801 @c ??? Should we have a mechanism to suppress this warning? This is almost
5802 @c useful for implementing the operation under the control of an external
5805 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5806 no memory operand will be moved across the operation, either forward or
5807 backward. Further, instructions will be issued as necessary to prevent the
5808 processor from speculating loads across the operation and from queuing stores
5809 after the operation.
5811 All of the routines are described in the Intel documentation to take
5812 ``an optional list of variables protected by the memory barrier''. It's
5813 not clear what is meant by that; it could mean that @emph{only} the
5814 following variables are protected, or it could mean that these variables
5815 should in addition be protected. At present GCC ignores this list and
5816 protects all variables which are globally accessible. If in the future
5817 we make some use of this list, an empty list will continue to mean all
5818 globally accessible variables.
5821 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5822 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5823 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5824 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5825 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5826 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5827 @findex __sync_fetch_and_add
5828 @findex __sync_fetch_and_sub
5829 @findex __sync_fetch_and_or
5830 @findex __sync_fetch_and_and
5831 @findex __sync_fetch_and_xor
5832 @findex __sync_fetch_and_nand
5833 These builtins perform the operation suggested by the name, and
5834 returns the value that had previously been in memory. That is,
5837 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5838 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
5841 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
5842 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
5844 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5845 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5846 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5847 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5848 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5849 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5850 @findex __sync_add_and_fetch
5851 @findex __sync_sub_and_fetch
5852 @findex __sync_or_and_fetch
5853 @findex __sync_and_and_fetch
5854 @findex __sync_xor_and_fetch
5855 @findex __sync_nand_and_fetch
5856 These builtins perform the operation suggested by the name, and
5857 return the new value. That is,
5860 @{ *ptr @var{op}= value; return *ptr; @}
5861 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
5864 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
5865 builtin as @code{*ptr = ~(*ptr & value)} instead of
5866 @code{*ptr = ~*ptr & value}.
5868 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5869 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5870 @findex __sync_bool_compare_and_swap
5871 @findex __sync_val_compare_and_swap
5872 These builtins perform an atomic compare and swap. That is, if the current
5873 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5876 The ``bool'' version returns true if the comparison is successful and
5877 @var{newval} was written. The ``val'' version returns the contents
5878 of @code{*@var{ptr}} before the operation.
5880 @item __sync_synchronize (...)
5881 @findex __sync_synchronize
5882 This builtin issues a full memory barrier.
5884 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5885 @findex __sync_lock_test_and_set
5886 This builtin, as described by Intel, is not a traditional test-and-set
5887 operation, but rather an atomic exchange operation. It writes @var{value}
5888 into @code{*@var{ptr}}, and returns the previous contents of
5891 Many targets have only minimal support for such locks, and do not support
5892 a full exchange operation. In this case, a target may support reduced
5893 functionality here by which the @emph{only} valid value to store is the
5894 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5895 is implementation defined.
5897 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5898 This means that references after the builtin cannot move to (or be
5899 speculated to) before the builtin, but previous memory stores may not
5900 be globally visible yet, and previous memory loads may not yet be
5903 @item void __sync_lock_release (@var{type} *ptr, ...)
5904 @findex __sync_lock_release
5905 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5906 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5908 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5909 This means that all previous memory stores are globally visible, and all
5910 previous memory loads have been satisfied, but following memory reads
5911 are not prevented from being speculated to before the barrier.
5914 @node Object Size Checking
5915 @section Object Size Checking Builtins
5916 @findex __builtin_object_size
5917 @findex __builtin___memcpy_chk
5918 @findex __builtin___mempcpy_chk
5919 @findex __builtin___memmove_chk
5920 @findex __builtin___memset_chk
5921 @findex __builtin___strcpy_chk
5922 @findex __builtin___stpcpy_chk
5923 @findex __builtin___strncpy_chk
5924 @findex __builtin___strcat_chk
5925 @findex __builtin___strncat_chk
5926 @findex __builtin___sprintf_chk
5927 @findex __builtin___snprintf_chk
5928 @findex __builtin___vsprintf_chk
5929 @findex __builtin___vsnprintf_chk
5930 @findex __builtin___printf_chk
5931 @findex __builtin___vprintf_chk
5932 @findex __builtin___fprintf_chk
5933 @findex __builtin___vfprintf_chk
5935 GCC implements a limited buffer overflow protection mechanism
5936 that can prevent some buffer overflow attacks.
5938 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5939 is a built-in construct that returns a constant number of bytes from
5940 @var{ptr} to the end of the object @var{ptr} pointer points to
5941 (if known at compile time). @code{__builtin_object_size} never evaluates
5942 its arguments for side-effects. If there are any side-effects in them, it
5943 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5944 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5945 point to and all of them are known at compile time, the returned number
5946 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5947 0 and minimum if nonzero. If it is not possible to determine which objects
5948 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5949 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5950 for @var{type} 2 or 3.
5952 @var{type} is an integer constant from 0 to 3. If the least significant
5953 bit is clear, objects are whole variables, if it is set, a closest
5954 surrounding subobject is considered the object a pointer points to.
5955 The second bit determines if maximum or minimum of remaining bytes
5959 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5960 char *p = &var.buf1[1], *q = &var.b;
5962 /* Here the object p points to is var. */
5963 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5964 /* The subobject p points to is var.buf1. */
5965 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5966 /* The object q points to is var. */
5967 assert (__builtin_object_size (q, 0)
5968 == (char *) (&var + 1) - (char *) &var.b);
5969 /* The subobject q points to is var.b. */
5970 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5974 There are built-in functions added for many common string operation
5975 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
5976 built-in is provided. This built-in has an additional last argument,
5977 which is the number of bytes remaining in object the @var{dest}
5978 argument points to or @code{(size_t) -1} if the size is not known.
5980 The built-in functions are optimized into the normal string functions
5981 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5982 it is known at compile time that the destination object will not
5983 be overflown. If the compiler can determine at compile time the
5984 object will be always overflown, it issues a warning.
5986 The intended use can be e.g.
5990 #define bos0(dest) __builtin_object_size (dest, 0)
5991 #define memcpy(dest, src, n) \
5992 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5996 /* It is unknown what object p points to, so this is optimized
5997 into plain memcpy - no checking is possible. */
5998 memcpy (p, "abcde", n);
5999 /* Destination is known and length too. It is known at compile
6000 time there will be no overflow. */
6001 memcpy (&buf[5], "abcde", 5);
6002 /* Destination is known, but the length is not known at compile time.
6003 This will result in __memcpy_chk call that can check for overflow
6005 memcpy (&buf[5], "abcde", n);
6006 /* Destination is known and it is known at compile time there will
6007 be overflow. There will be a warning and __memcpy_chk call that
6008 will abort the program at runtime. */
6009 memcpy (&buf[6], "abcde", 5);
6012 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
6013 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
6014 @code{strcat} and @code{strncat}.
6016 There are also checking built-in functions for formatted output functions.
6018 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
6019 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6020 const char *fmt, ...);
6021 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
6023 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6024 const char *fmt, va_list ap);
6027 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
6028 etc.@: functions and can contain implementation specific flags on what
6029 additional security measures the checking function might take, such as
6030 handling @code{%n} differently.
6032 The @var{os} argument is the object size @var{s} points to, like in the
6033 other built-in functions. There is a small difference in the behavior
6034 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
6035 optimized into the non-checking functions only if @var{flag} is 0, otherwise
6036 the checking function is called with @var{os} argument set to
6039 In addition to this, there are checking built-in functions
6040 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
6041 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
6042 These have just one additional argument, @var{flag}, right before
6043 format string @var{fmt}. If the compiler is able to optimize them to
6044 @code{fputc} etc.@: functions, it will, otherwise the checking function
6045 should be called and the @var{flag} argument passed to it.
6047 @node Other Builtins
6048 @section Other built-in functions provided by GCC
6049 @cindex built-in functions
6050 @findex __builtin_fpclassify
6051 @findex __builtin_isfinite
6052 @findex __builtin_isnormal
6053 @findex __builtin_isgreater
6054 @findex __builtin_isgreaterequal
6055 @findex __builtin_isinf_sign
6056 @findex __builtin_isless
6057 @findex __builtin_islessequal
6058 @findex __builtin_islessgreater
6059 @findex __builtin_isunordered
6060 @findex __builtin_powi
6061 @findex __builtin_powif
6062 @findex __builtin_powil
6220 @findex fprintf_unlocked
6222 @findex fputs_unlocked
6339 @findex printf_unlocked
6371 @findex significandf
6372 @findex significandl
6443 GCC provides a large number of built-in functions other than the ones
6444 mentioned above. Some of these are for internal use in the processing
6445 of exceptions or variable-length argument lists and will not be
6446 documented here because they may change from time to time; we do not
6447 recommend general use of these functions.
6449 The remaining functions are provided for optimization purposes.
6451 @opindex fno-builtin
6452 GCC includes built-in versions of many of the functions in the standard
6453 C library. The versions prefixed with @code{__builtin_} will always be
6454 treated as having the same meaning as the C library function even if you
6455 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
6456 Many of these functions are only optimized in certain cases; if they are
6457 not optimized in a particular case, a call to the library function will
6462 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
6463 @option{-std=c99}), the functions
6464 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
6465 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
6466 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
6467 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
6468 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
6469 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
6470 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
6471 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
6472 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
6473 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
6474 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
6475 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
6476 @code{signbitd64}, @code{signbitd128}, @code{significandf},
6477 @code{significandl}, @code{significand}, @code{sincosf},
6478 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
6479 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
6480 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
6481 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
6483 may be handled as built-in functions.
6484 All these functions have corresponding versions
6485 prefixed with @code{__builtin_}, which may be used even in strict C89
6488 The ISO C99 functions
6489 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
6490 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
6491 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
6492 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
6493 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
6494 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
6495 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
6496 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
6497 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
6498 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
6499 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
6500 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
6501 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
6502 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
6503 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
6504 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
6505 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
6506 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
6507 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
6508 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
6509 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
6510 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
6511 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
6512 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
6513 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
6514 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
6515 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
6516 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
6517 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
6518 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
6519 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
6520 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
6521 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
6522 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
6523 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
6524 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
6525 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
6526 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
6527 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
6528 are handled as built-in functions
6529 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6531 There are also built-in versions of the ISO C99 functions
6532 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
6533 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
6534 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
6535 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
6536 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
6537 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
6538 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
6539 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
6540 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
6541 that are recognized in any mode since ISO C90 reserves these names for
6542 the purpose to which ISO C99 puts them. All these functions have
6543 corresponding versions prefixed with @code{__builtin_}.
6545 The ISO C94 functions
6546 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
6547 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
6548 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
6550 are handled as built-in functions
6551 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6553 The ISO C90 functions
6554 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
6555 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
6556 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
6557 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
6558 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
6559 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
6560 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
6561 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
6562 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
6563 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
6564 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
6565 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
6566 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
6567 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
6568 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
6569 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
6570 are all recognized as built-in functions unless
6571 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
6572 is specified for an individual function). All of these functions have
6573 corresponding versions prefixed with @code{__builtin_}.
6575 GCC provides built-in versions of the ISO C99 floating point comparison
6576 macros that avoid raising exceptions for unordered operands. They have
6577 the same names as the standard macros ( @code{isgreater},
6578 @code{isgreaterequal}, @code{isless}, @code{islessequal},
6579 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
6580 prefixed. We intend for a library implementor to be able to simply
6581 @code{#define} each standard macro to its built-in equivalent.
6582 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
6583 @code{isinf_sign} and @code{isnormal} built-ins used with
6584 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
6585 builtins appear both with and without the @code{__builtin_} prefix.
6587 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
6589 You can use the built-in function @code{__builtin_types_compatible_p} to
6590 determine whether two types are the same.
6592 This built-in function returns 1 if the unqualified versions of the
6593 types @var{type1} and @var{type2} (which are types, not expressions) are
6594 compatible, 0 otherwise. The result of this built-in function can be
6595 used in integer constant expressions.
6597 This built-in function ignores top level qualifiers (e.g., @code{const},
6598 @code{volatile}). For example, @code{int} is equivalent to @code{const
6601 The type @code{int[]} and @code{int[5]} are compatible. On the other
6602 hand, @code{int} and @code{char *} are not compatible, even if the size
6603 of their types, on the particular architecture are the same. Also, the
6604 amount of pointer indirection is taken into account when determining
6605 similarity. Consequently, @code{short *} is not similar to
6606 @code{short **}. Furthermore, two types that are typedefed are
6607 considered compatible if their underlying types are compatible.
6609 An @code{enum} type is not considered to be compatible with another
6610 @code{enum} type even if both are compatible with the same integer
6611 type; this is what the C standard specifies.
6612 For example, @code{enum @{foo, bar@}} is not similar to
6613 @code{enum @{hot, dog@}}.
6615 You would typically use this function in code whose execution varies
6616 depending on the arguments' types. For example:
6621 typeof (x) tmp = (x); \
6622 if (__builtin_types_compatible_p (typeof (x), long double)) \
6623 tmp = foo_long_double (tmp); \
6624 else if (__builtin_types_compatible_p (typeof (x), double)) \
6625 tmp = foo_double (tmp); \
6626 else if (__builtin_types_compatible_p (typeof (x), float)) \
6627 tmp = foo_float (tmp); \
6634 @emph{Note:} This construct is only available for C@.
6638 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
6640 You can use the built-in function @code{__builtin_choose_expr} to
6641 evaluate code depending on the value of a constant expression. This
6642 built-in function returns @var{exp1} if @var{const_exp}, which is an
6643 integer constant expression, is nonzero. Otherwise it returns 0.
6645 This built-in function is analogous to the @samp{? :} operator in C,
6646 except that the expression returned has its type unaltered by promotion
6647 rules. Also, the built-in function does not evaluate the expression
6648 that was not chosen. For example, if @var{const_exp} evaluates to true,
6649 @var{exp2} is not evaluated even if it has side-effects.
6651 This built-in function can return an lvalue if the chosen argument is an
6654 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
6655 type. Similarly, if @var{exp2} is returned, its return type is the same
6662 __builtin_choose_expr ( \
6663 __builtin_types_compatible_p (typeof (x), double), \
6665 __builtin_choose_expr ( \
6666 __builtin_types_compatible_p (typeof (x), float), \
6668 /* @r{The void expression results in a compile-time error} \
6669 @r{when assigning the result to something.} */ \
6673 @emph{Note:} This construct is only available for C@. Furthermore, the
6674 unused expression (@var{exp1} or @var{exp2} depending on the value of
6675 @var{const_exp}) may still generate syntax errors. This may change in
6680 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6681 You can use the built-in function @code{__builtin_constant_p} to
6682 determine if a value is known to be constant at compile-time and hence
6683 that GCC can perform constant-folding on expressions involving that
6684 value. The argument of the function is the value to test. The function
6685 returns the integer 1 if the argument is known to be a compile-time
6686 constant and 0 if it is not known to be a compile-time constant. A
6687 return of 0 does not indicate that the value is @emph{not} a constant,
6688 but merely that GCC cannot prove it is a constant with the specified
6689 value of the @option{-O} option.
6691 You would typically use this function in an embedded application where
6692 memory was a critical resource. If you have some complex calculation,
6693 you may want it to be folded if it involves constants, but need to call
6694 a function if it does not. For example:
6697 #define Scale_Value(X) \
6698 (__builtin_constant_p (X) \
6699 ? ((X) * SCALE + OFFSET) : Scale (X))
6702 You may use this built-in function in either a macro or an inline
6703 function. However, if you use it in an inlined function and pass an
6704 argument of the function as the argument to the built-in, GCC will
6705 never return 1 when you call the inline function with a string constant
6706 or compound literal (@pxref{Compound Literals}) and will not return 1
6707 when you pass a constant numeric value to the inline function unless you
6708 specify the @option{-O} option.
6710 You may also use @code{__builtin_constant_p} in initializers for static
6711 data. For instance, you can write
6714 static const int table[] = @{
6715 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6721 This is an acceptable initializer even if @var{EXPRESSION} is not a
6722 constant expression, including the case where
6723 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
6724 folded to a constant but @var{EXPRESSION} contains operands that would
6725 not otherwize be permitted in a static initializer (for example,
6726 @code{0 && foo ()}). GCC must be more conservative about evaluating the
6727 built-in in this case, because it has no opportunity to perform
6730 Previous versions of GCC did not accept this built-in in data
6731 initializers. The earliest version where it is completely safe is
6735 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6736 @opindex fprofile-arcs
6737 You may use @code{__builtin_expect} to provide the compiler with
6738 branch prediction information. In general, you should prefer to
6739 use actual profile feedback for this (@option{-fprofile-arcs}), as
6740 programmers are notoriously bad at predicting how their programs
6741 actually perform. However, there are applications in which this
6742 data is hard to collect.
6744 The return value is the value of @var{exp}, which should be an integral
6745 expression. The semantics of the built-in are that it is expected that
6746 @var{exp} == @var{c}. For example:
6749 if (__builtin_expect (x, 0))
6754 would indicate that we do not expect to call @code{foo}, since
6755 we expect @code{x} to be zero. Since you are limited to integral
6756 expressions for @var{exp}, you should use constructions such as
6759 if (__builtin_expect (ptr != NULL, 1))
6764 when testing pointer or floating-point values.
6767 @deftypefn {Built-in Function} void __builtin_trap (void)
6768 This function causes the program to exit abnormally. GCC implements
6769 this function by using a target-dependent mechanism (such as
6770 intentionally executing an illegal instruction) or by calling
6771 @code{abort}. The mechanism used may vary from release to release so
6772 you should not rely on any particular implementation.
6775 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
6776 This function is used to flush the processor's instruction cache for
6777 the region of memory between @var{begin} inclusive and @var{end}
6778 exclusive. Some targets require that the instruction cache be
6779 flushed, after modifying memory containing code, in order to obtain
6780 deterministic behavior.
6782 If the target does not require instruction cache flushes,
6783 @code{__builtin___clear_cache} has no effect. Otherwise either
6784 instructions are emitted in-line to clear the instruction cache or a
6785 call to the @code{__clear_cache} function in libgcc is made.
6788 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6789 This function is used to minimize cache-miss latency by moving data into
6790 a cache before it is accessed.
6791 You can insert calls to @code{__builtin_prefetch} into code for which
6792 you know addresses of data in memory that is likely to be accessed soon.
6793 If the target supports them, data prefetch instructions will be generated.
6794 If the prefetch is done early enough before the access then the data will
6795 be in the cache by the time it is accessed.
6797 The value of @var{addr} is the address of the memory to prefetch.
6798 There are two optional arguments, @var{rw} and @var{locality}.
6799 The value of @var{rw} is a compile-time constant one or zero; one
6800 means that the prefetch is preparing for a write to the memory address
6801 and zero, the default, means that the prefetch is preparing for a read.
6802 The value @var{locality} must be a compile-time constant integer between
6803 zero and three. A value of zero means that the data has no temporal
6804 locality, so it need not be left in the cache after the access. A value
6805 of three means that the data has a high degree of temporal locality and
6806 should be left in all levels of cache possible. Values of one and two
6807 mean, respectively, a low or moderate degree of temporal locality. The
6811 for (i = 0; i < n; i++)
6814 __builtin_prefetch (&a[i+j], 1, 1);
6815 __builtin_prefetch (&b[i+j], 0, 1);
6820 Data prefetch does not generate faults if @var{addr} is invalid, but
6821 the address expression itself must be valid. For example, a prefetch
6822 of @code{p->next} will not fault if @code{p->next} is not a valid
6823 address, but evaluation will fault if @code{p} is not a valid address.
6825 If the target does not support data prefetch, the address expression
6826 is evaluated if it includes side effects but no other code is generated
6827 and GCC does not issue a warning.
6830 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6831 Returns a positive infinity, if supported by the floating-point format,
6832 else @code{DBL_MAX}. This function is suitable for implementing the
6833 ISO C macro @code{HUGE_VAL}.
6836 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6837 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6840 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6841 Similar to @code{__builtin_huge_val}, except the return
6842 type is @code{long double}.
6845 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
6846 This built-in implements the C99 fpclassify functionality. The first
6847 five int arguments should be the target library's notion of the
6848 possible FP classes and are used for return values. They must be
6849 constant values and they must appear in this order: @code{FP_NAN},
6850 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
6851 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
6852 to classify. GCC treats the last argument as type-generic, which
6853 means it does not do default promotion from float to double.
6856 @deftypefn {Built-in Function} double __builtin_inf (void)
6857 Similar to @code{__builtin_huge_val}, except a warning is generated
6858 if the target floating-point format does not support infinities.
6861 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6862 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6865 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6866 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6869 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6870 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6873 @deftypefn {Built-in Function} float __builtin_inff (void)
6874 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6875 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6878 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6879 Similar to @code{__builtin_inf}, except the return
6880 type is @code{long double}.
6883 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
6884 Similar to @code{isinf}, except the return value will be negative for
6885 an argument of @code{-Inf}. Note while the parameter list is an
6886 ellipsis, this function only accepts exactly one floating point
6887 argument. GCC treats this parameter as type-generic, which means it
6888 does not do default promotion from float to double.
6891 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6892 This is an implementation of the ISO C99 function @code{nan}.
6894 Since ISO C99 defines this function in terms of @code{strtod}, which we
6895 do not implement, a description of the parsing is in order. The string
6896 is parsed as by @code{strtol}; that is, the base is recognized by
6897 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6898 in the significand such that the least significant bit of the number
6899 is at the least significant bit of the significand. The number is
6900 truncated to fit the significand field provided. The significand is
6901 forced to be a quiet NaN@.
6903 This function, if given a string literal all of which would have been
6904 consumed by strtol, is evaluated early enough that it is considered a
6905 compile-time constant.
6908 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6909 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6912 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6913 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6916 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6917 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6920 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6921 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6924 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6925 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6928 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6929 Similar to @code{__builtin_nan}, except the significand is forced
6930 to be a signaling NaN@. The @code{nans} function is proposed by
6931 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6934 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6935 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6938 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6939 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6942 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6943 Returns one plus the index of the least significant 1-bit of @var{x}, or
6944 if @var{x} is zero, returns zero.
6947 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6948 Returns the number of leading 0-bits in @var{x}, starting at the most
6949 significant bit position. If @var{x} is 0, the result is undefined.
6952 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6953 Returns the number of trailing 0-bits in @var{x}, starting at the least
6954 significant bit position. If @var{x} is 0, the result is undefined.
6957 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6958 Returns the number of 1-bits in @var{x}.
6961 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6962 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6966 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6967 Similar to @code{__builtin_ffs}, except the argument type is
6968 @code{unsigned long}.
6971 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6972 Similar to @code{__builtin_clz}, except the argument type is
6973 @code{unsigned long}.
6976 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6977 Similar to @code{__builtin_ctz}, except the argument type is
6978 @code{unsigned long}.
6981 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6982 Similar to @code{__builtin_popcount}, except the argument type is
6983 @code{unsigned long}.
6986 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6987 Similar to @code{__builtin_parity}, except the argument type is
6988 @code{unsigned long}.
6991 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6992 Similar to @code{__builtin_ffs}, except the argument type is
6993 @code{unsigned long long}.
6996 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6997 Similar to @code{__builtin_clz}, except the argument type is
6998 @code{unsigned long long}.
7001 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
7002 Similar to @code{__builtin_ctz}, except the argument type is
7003 @code{unsigned long long}.
7006 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
7007 Similar to @code{__builtin_popcount}, except the argument type is
7008 @code{unsigned long long}.
7011 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
7012 Similar to @code{__builtin_parity}, except the argument type is
7013 @code{unsigned long long}.
7016 @deftypefn {Built-in Function} double __builtin_powi (double, int)
7017 Returns the first argument raised to the power of the second. Unlike the
7018 @code{pow} function no guarantees about precision and rounding are made.
7021 @deftypefn {Built-in Function} float __builtin_powif (float, int)
7022 Similar to @code{__builtin_powi}, except the argument and return types
7026 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
7027 Similar to @code{__builtin_powi}, except the argument and return types
7028 are @code{long double}.
7031 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
7032 Returns @var{x} with the order of the bytes reversed; for example,
7033 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
7037 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
7038 Similar to @code{__builtin_bswap32}, except the argument and return types
7042 @node Target Builtins
7043 @section Built-in Functions Specific to Particular Target Machines
7045 On some target machines, GCC supports many built-in functions specific
7046 to those machines. Generally these generate calls to specific machine
7047 instructions, but allow the compiler to schedule those calls.
7050 * Alpha Built-in Functions::
7051 * ARM iWMMXt Built-in Functions::
7052 * ARM NEON Intrinsics::
7053 * Blackfin Built-in Functions::
7054 * FR-V Built-in Functions::
7055 * X86 Built-in Functions::
7056 * MIPS DSP Built-in Functions::
7057 * MIPS Paired-Single Support::
7058 * MIPS Loongson Built-in Functions::
7059 * Other MIPS Built-in Functions::
7060 * picoChip Built-in Functions::
7061 * PowerPC AltiVec Built-in Functions::
7062 * SPARC VIS Built-in Functions::
7063 * SPU Built-in Functions::
7066 @node Alpha Built-in Functions
7067 @subsection Alpha Built-in Functions
7069 These built-in functions are available for the Alpha family of
7070 processors, depending on the command-line switches used.
7072 The following built-in functions are always available. They
7073 all generate the machine instruction that is part of the name.
7076 long __builtin_alpha_implver (void)
7077 long __builtin_alpha_rpcc (void)
7078 long __builtin_alpha_amask (long)
7079 long __builtin_alpha_cmpbge (long, long)
7080 long __builtin_alpha_extbl (long, long)
7081 long __builtin_alpha_extwl (long, long)
7082 long __builtin_alpha_extll (long, long)
7083 long __builtin_alpha_extql (long, long)
7084 long __builtin_alpha_extwh (long, long)
7085 long __builtin_alpha_extlh (long, long)
7086 long __builtin_alpha_extqh (long, long)
7087 long __builtin_alpha_insbl (long, long)
7088 long __builtin_alpha_inswl (long, long)
7089 long __builtin_alpha_insll (long, long)
7090 long __builtin_alpha_insql (long, long)
7091 long __builtin_alpha_inswh (long, long)
7092 long __builtin_alpha_inslh (long, long)
7093 long __builtin_alpha_insqh (long, long)
7094 long __builtin_alpha_mskbl (long, long)
7095 long __builtin_alpha_mskwl (long, long)
7096 long __builtin_alpha_mskll (long, long)
7097 long __builtin_alpha_mskql (long, long)
7098 long __builtin_alpha_mskwh (long, long)
7099 long __builtin_alpha_msklh (long, long)
7100 long __builtin_alpha_mskqh (long, long)
7101 long __builtin_alpha_umulh (long, long)
7102 long __builtin_alpha_zap (long, long)
7103 long __builtin_alpha_zapnot (long, long)
7106 The following built-in functions are always with @option{-mmax}
7107 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7108 later. They all generate the machine instruction that is part
7112 long __builtin_alpha_pklb (long)
7113 long __builtin_alpha_pkwb (long)
7114 long __builtin_alpha_unpkbl (long)
7115 long __builtin_alpha_unpkbw (long)
7116 long __builtin_alpha_minub8 (long, long)
7117 long __builtin_alpha_minsb8 (long, long)
7118 long __builtin_alpha_minuw4 (long, long)
7119 long __builtin_alpha_minsw4 (long, long)
7120 long __builtin_alpha_maxub8 (long, long)
7121 long __builtin_alpha_maxsb8 (long, long)
7122 long __builtin_alpha_maxuw4 (long, long)
7123 long __builtin_alpha_maxsw4 (long, long)
7124 long __builtin_alpha_perr (long, long)
7127 The following built-in functions are always with @option{-mcix}
7128 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
7129 later. They all generate the machine instruction that is part
7133 long __builtin_alpha_cttz (long)
7134 long __builtin_alpha_ctlz (long)
7135 long __builtin_alpha_ctpop (long)
7138 The following builtins are available on systems that use the OSF/1
7139 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
7140 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
7141 @code{rdval} and @code{wrval}.
7144 void *__builtin_thread_pointer (void)
7145 void __builtin_set_thread_pointer (void *)
7148 @node ARM iWMMXt Built-in Functions
7149 @subsection ARM iWMMXt Built-in Functions
7151 These built-in functions are available for the ARM family of
7152 processors when the @option{-mcpu=iwmmxt} switch is used:
7155 typedef int v2si __attribute__ ((vector_size (8)));
7156 typedef short v4hi __attribute__ ((vector_size (8)));
7157 typedef char v8qi __attribute__ ((vector_size (8)));
7159 int __builtin_arm_getwcx (int)
7160 void __builtin_arm_setwcx (int, int)
7161 int __builtin_arm_textrmsb (v8qi, int)
7162 int __builtin_arm_textrmsh (v4hi, int)
7163 int __builtin_arm_textrmsw (v2si, int)
7164 int __builtin_arm_textrmub (v8qi, int)
7165 int __builtin_arm_textrmuh (v4hi, int)
7166 int __builtin_arm_textrmuw (v2si, int)
7167 v8qi __builtin_arm_tinsrb (v8qi, int)
7168 v4hi __builtin_arm_tinsrh (v4hi, int)
7169 v2si __builtin_arm_tinsrw (v2si, int)
7170 long long __builtin_arm_tmia (long long, int, int)
7171 long long __builtin_arm_tmiabb (long long, int, int)
7172 long long __builtin_arm_tmiabt (long long, int, int)
7173 long long __builtin_arm_tmiaph (long long, int, int)
7174 long long __builtin_arm_tmiatb (long long, int, int)
7175 long long __builtin_arm_tmiatt (long long, int, int)
7176 int __builtin_arm_tmovmskb (v8qi)
7177 int __builtin_arm_tmovmskh (v4hi)
7178 int __builtin_arm_tmovmskw (v2si)
7179 long long __builtin_arm_waccb (v8qi)
7180 long long __builtin_arm_wacch (v4hi)
7181 long long __builtin_arm_waccw (v2si)
7182 v8qi __builtin_arm_waddb (v8qi, v8qi)
7183 v8qi __builtin_arm_waddbss (v8qi, v8qi)
7184 v8qi __builtin_arm_waddbus (v8qi, v8qi)
7185 v4hi __builtin_arm_waddh (v4hi, v4hi)
7186 v4hi __builtin_arm_waddhss (v4hi, v4hi)
7187 v4hi __builtin_arm_waddhus (v4hi, v4hi)
7188 v2si __builtin_arm_waddw (v2si, v2si)
7189 v2si __builtin_arm_waddwss (v2si, v2si)
7190 v2si __builtin_arm_waddwus (v2si, v2si)
7191 v8qi __builtin_arm_walign (v8qi, v8qi, int)
7192 long long __builtin_arm_wand(long long, long long)
7193 long long __builtin_arm_wandn (long long, long long)
7194 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
7195 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
7196 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
7197 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
7198 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
7199 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
7200 v2si __builtin_arm_wcmpeqw (v2si, v2si)
7201 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
7202 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
7203 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
7204 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
7205 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
7206 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
7207 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
7208 long long __builtin_arm_wmacsz (v4hi, v4hi)
7209 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
7210 long long __builtin_arm_wmacuz (v4hi, v4hi)
7211 v4hi __builtin_arm_wmadds (v4hi, v4hi)
7212 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
7213 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
7214 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
7215 v2si __builtin_arm_wmaxsw (v2si, v2si)
7216 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
7217 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
7218 v2si __builtin_arm_wmaxuw (v2si, v2si)
7219 v8qi __builtin_arm_wminsb (v8qi, v8qi)
7220 v4hi __builtin_arm_wminsh (v4hi, v4hi)
7221 v2si __builtin_arm_wminsw (v2si, v2si)
7222 v8qi __builtin_arm_wminub (v8qi, v8qi)
7223 v4hi __builtin_arm_wminuh (v4hi, v4hi)
7224 v2si __builtin_arm_wminuw (v2si, v2si)
7225 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
7226 v4hi __builtin_arm_wmulul (v4hi, v4hi)
7227 v4hi __builtin_arm_wmulum (v4hi, v4hi)
7228 long long __builtin_arm_wor (long long, long long)
7229 v2si __builtin_arm_wpackdss (long long, long long)
7230 v2si __builtin_arm_wpackdus (long long, long long)
7231 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
7232 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
7233 v4hi __builtin_arm_wpackwss (v2si, v2si)
7234 v4hi __builtin_arm_wpackwus (v2si, v2si)
7235 long long __builtin_arm_wrord (long long, long long)
7236 long long __builtin_arm_wrordi (long long, int)
7237 v4hi __builtin_arm_wrorh (v4hi, long long)
7238 v4hi __builtin_arm_wrorhi (v4hi, int)
7239 v2si __builtin_arm_wrorw (v2si, long long)
7240 v2si __builtin_arm_wrorwi (v2si, int)
7241 v2si __builtin_arm_wsadb (v8qi, v8qi)
7242 v2si __builtin_arm_wsadbz (v8qi, v8qi)
7243 v2si __builtin_arm_wsadh (v4hi, v4hi)
7244 v2si __builtin_arm_wsadhz (v4hi, v4hi)
7245 v4hi __builtin_arm_wshufh (v4hi, int)
7246 long long __builtin_arm_wslld (long long, long long)
7247 long long __builtin_arm_wslldi (long long, int)
7248 v4hi __builtin_arm_wsllh (v4hi, long long)
7249 v4hi __builtin_arm_wsllhi (v4hi, int)
7250 v2si __builtin_arm_wsllw (v2si, long long)
7251 v2si __builtin_arm_wsllwi (v2si, int)
7252 long long __builtin_arm_wsrad (long long, long long)
7253 long long __builtin_arm_wsradi (long long, int)
7254 v4hi __builtin_arm_wsrah (v4hi, long long)
7255 v4hi __builtin_arm_wsrahi (v4hi, int)
7256 v2si __builtin_arm_wsraw (v2si, long long)
7257 v2si __builtin_arm_wsrawi (v2si, int)
7258 long long __builtin_arm_wsrld (long long, long long)
7259 long long __builtin_arm_wsrldi (long long, int)
7260 v4hi __builtin_arm_wsrlh (v4hi, long long)
7261 v4hi __builtin_arm_wsrlhi (v4hi, int)
7262 v2si __builtin_arm_wsrlw (v2si, long long)
7263 v2si __builtin_arm_wsrlwi (v2si, int)
7264 v8qi __builtin_arm_wsubb (v8qi, v8qi)
7265 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
7266 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
7267 v4hi __builtin_arm_wsubh (v4hi, v4hi)
7268 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
7269 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
7270 v2si __builtin_arm_wsubw (v2si, v2si)
7271 v2si __builtin_arm_wsubwss (v2si, v2si)
7272 v2si __builtin_arm_wsubwus (v2si, v2si)
7273 v4hi __builtin_arm_wunpckehsb (v8qi)
7274 v2si __builtin_arm_wunpckehsh (v4hi)
7275 long long __builtin_arm_wunpckehsw (v2si)
7276 v4hi __builtin_arm_wunpckehub (v8qi)
7277 v2si __builtin_arm_wunpckehuh (v4hi)
7278 long long __builtin_arm_wunpckehuw (v2si)
7279 v4hi __builtin_arm_wunpckelsb (v8qi)
7280 v2si __builtin_arm_wunpckelsh (v4hi)
7281 long long __builtin_arm_wunpckelsw (v2si)
7282 v4hi __builtin_arm_wunpckelub (v8qi)
7283 v2si __builtin_arm_wunpckeluh (v4hi)
7284 long long __builtin_arm_wunpckeluw (v2si)
7285 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
7286 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
7287 v2si __builtin_arm_wunpckihw (v2si, v2si)
7288 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
7289 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
7290 v2si __builtin_arm_wunpckilw (v2si, v2si)
7291 long long __builtin_arm_wxor (long long, long long)
7292 long long __builtin_arm_wzero ()
7295 @node ARM NEON Intrinsics
7296 @subsection ARM NEON Intrinsics
7298 These built-in intrinsics for the ARM Advanced SIMD extension are available
7299 when the @option{-mfpu=neon} switch is used:
7301 @include arm-neon-intrinsics.texi
7303 @node Blackfin Built-in Functions
7304 @subsection Blackfin Built-in Functions
7306 Currently, there are two Blackfin-specific built-in functions. These are
7307 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
7308 using inline assembly; by using these built-in functions the compiler can
7309 automatically add workarounds for hardware errata involving these
7310 instructions. These functions are named as follows:
7313 void __builtin_bfin_csync (void)
7314 void __builtin_bfin_ssync (void)
7317 @node FR-V Built-in Functions
7318 @subsection FR-V Built-in Functions
7320 GCC provides many FR-V-specific built-in functions. In general,
7321 these functions are intended to be compatible with those described
7322 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
7323 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
7324 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
7325 pointer rather than by value.
7327 Most of the functions are named after specific FR-V instructions.
7328 Such functions are said to be ``directly mapped'' and are summarized
7329 here in tabular form.
7333 * Directly-mapped Integer Functions::
7334 * Directly-mapped Media Functions::
7335 * Raw read/write Functions::
7336 * Other Built-in Functions::
7339 @node Argument Types
7340 @subsubsection Argument Types
7342 The arguments to the built-in functions can be divided into three groups:
7343 register numbers, compile-time constants and run-time values. In order
7344 to make this classification clear at a glance, the arguments and return
7345 values are given the following pseudo types:
7347 @multitable @columnfractions .20 .30 .15 .35
7348 @item Pseudo type @tab Real C type @tab Constant? @tab Description
7349 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
7350 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
7351 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
7352 @item @code{uw2} @tab @code{unsigned long long} @tab No
7353 @tab an unsigned doubleword
7354 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
7355 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
7356 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
7357 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
7360 These pseudo types are not defined by GCC, they are simply a notational
7361 convenience used in this manual.
7363 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
7364 and @code{sw2} are evaluated at run time. They correspond to
7365 register operands in the underlying FR-V instructions.
7367 @code{const} arguments represent immediate operands in the underlying
7368 FR-V instructions. They must be compile-time constants.
7370 @code{acc} arguments are evaluated at compile time and specify the number
7371 of an accumulator register. For example, an @code{acc} argument of 2
7372 will select the ACC2 register.
7374 @code{iacc} arguments are similar to @code{acc} arguments but specify the
7375 number of an IACC register. See @pxref{Other Built-in Functions}
7378 @node Directly-mapped Integer Functions
7379 @subsubsection Directly-mapped Integer Functions
7381 The functions listed below map directly to FR-V I-type instructions.
7383 @multitable @columnfractions .45 .32 .23
7384 @item Function prototype @tab Example usage @tab Assembly output
7385 @item @code{sw1 __ADDSS (sw1, sw1)}
7386 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
7387 @tab @code{ADDSS @var{a},@var{b},@var{c}}
7388 @item @code{sw1 __SCAN (sw1, sw1)}
7389 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
7390 @tab @code{SCAN @var{a},@var{b},@var{c}}
7391 @item @code{sw1 __SCUTSS (sw1)}
7392 @tab @code{@var{b} = __SCUTSS (@var{a})}
7393 @tab @code{SCUTSS @var{a},@var{b}}
7394 @item @code{sw1 __SLASS (sw1, sw1)}
7395 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
7396 @tab @code{SLASS @var{a},@var{b},@var{c}}
7397 @item @code{void __SMASS (sw1, sw1)}
7398 @tab @code{__SMASS (@var{a}, @var{b})}
7399 @tab @code{SMASS @var{a},@var{b}}
7400 @item @code{void __SMSSS (sw1, sw1)}
7401 @tab @code{__SMSSS (@var{a}, @var{b})}
7402 @tab @code{SMSSS @var{a},@var{b}}
7403 @item @code{void __SMU (sw1, sw1)}
7404 @tab @code{__SMU (@var{a}, @var{b})}
7405 @tab @code{SMU @var{a},@var{b}}
7406 @item @code{sw2 __SMUL (sw1, sw1)}
7407 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
7408 @tab @code{SMUL @var{a},@var{b},@var{c}}
7409 @item @code{sw1 __SUBSS (sw1, sw1)}
7410 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
7411 @tab @code{SUBSS @var{a},@var{b},@var{c}}
7412 @item @code{uw2 __UMUL (uw1, uw1)}
7413 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
7414 @tab @code{UMUL @var{a},@var{b},@var{c}}
7417 @node Directly-mapped Media Functions
7418 @subsubsection Directly-mapped Media Functions
7420 The functions listed below map directly to FR-V M-type instructions.
7422 @multitable @columnfractions .45 .32 .23
7423 @item Function prototype @tab Example usage @tab Assembly output
7424 @item @code{uw1 __MABSHS (sw1)}
7425 @tab @code{@var{b} = __MABSHS (@var{a})}
7426 @tab @code{MABSHS @var{a},@var{b}}
7427 @item @code{void __MADDACCS (acc, acc)}
7428 @tab @code{__MADDACCS (@var{b}, @var{a})}
7429 @tab @code{MADDACCS @var{a},@var{b}}
7430 @item @code{sw1 __MADDHSS (sw1, sw1)}
7431 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
7432 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
7433 @item @code{uw1 __MADDHUS (uw1, uw1)}
7434 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
7435 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
7436 @item @code{uw1 __MAND (uw1, uw1)}
7437 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
7438 @tab @code{MAND @var{a},@var{b},@var{c}}
7439 @item @code{void __MASACCS (acc, acc)}
7440 @tab @code{__MASACCS (@var{b}, @var{a})}
7441 @tab @code{MASACCS @var{a},@var{b}}
7442 @item @code{uw1 __MAVEH (uw1, uw1)}
7443 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
7444 @tab @code{MAVEH @var{a},@var{b},@var{c}}
7445 @item @code{uw2 __MBTOH (uw1)}
7446 @tab @code{@var{b} = __MBTOH (@var{a})}
7447 @tab @code{MBTOH @var{a},@var{b}}
7448 @item @code{void __MBTOHE (uw1 *, uw1)}
7449 @tab @code{__MBTOHE (&@var{b}, @var{a})}
7450 @tab @code{MBTOHE @var{a},@var{b}}
7451 @item @code{void __MCLRACC (acc)}
7452 @tab @code{__MCLRACC (@var{a})}
7453 @tab @code{MCLRACC @var{a}}
7454 @item @code{void __MCLRACCA (void)}
7455 @tab @code{__MCLRACCA ()}
7456 @tab @code{MCLRACCA}
7457 @item @code{uw1 __Mcop1 (uw1, uw1)}
7458 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
7459 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
7460 @item @code{uw1 __Mcop2 (uw1, uw1)}
7461 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
7462 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
7463 @item @code{uw1 __MCPLHI (uw2, const)}
7464 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
7465 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
7466 @item @code{uw1 __MCPLI (uw2, const)}
7467 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
7468 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
7469 @item @code{void __MCPXIS (acc, sw1, sw1)}
7470 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
7471 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
7472 @item @code{void __MCPXIU (acc, uw1, uw1)}
7473 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
7474 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
7475 @item @code{void __MCPXRS (acc, sw1, sw1)}
7476 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
7477 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
7478 @item @code{void __MCPXRU (acc, uw1, uw1)}
7479 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
7480 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
7481 @item @code{uw1 __MCUT (acc, uw1)}
7482 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
7483 @tab @code{MCUT @var{a},@var{b},@var{c}}
7484 @item @code{uw1 __MCUTSS (acc, sw1)}
7485 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
7486 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
7487 @item @code{void __MDADDACCS (acc, acc)}
7488 @tab @code{__MDADDACCS (@var{b}, @var{a})}
7489 @tab @code{MDADDACCS @var{a},@var{b}}
7490 @item @code{void __MDASACCS (acc, acc)}
7491 @tab @code{__MDASACCS (@var{b}, @var{a})}
7492 @tab @code{MDASACCS @var{a},@var{b}}
7493 @item @code{uw2 __MDCUTSSI (acc, const)}
7494 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
7495 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
7496 @item @code{uw2 __MDPACKH (uw2, uw2)}
7497 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
7498 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
7499 @item @code{uw2 __MDROTLI (uw2, const)}
7500 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
7501 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
7502 @item @code{void __MDSUBACCS (acc, acc)}
7503 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
7504 @tab @code{MDSUBACCS @var{a},@var{b}}
7505 @item @code{void __MDUNPACKH (uw1 *, uw2)}
7506 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
7507 @tab @code{MDUNPACKH @var{a},@var{b}}
7508 @item @code{uw2 __MEXPDHD (uw1, const)}
7509 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
7510 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
7511 @item @code{uw1 __MEXPDHW (uw1, const)}
7512 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
7513 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
7514 @item @code{uw1 __MHDSETH (uw1, const)}
7515 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
7516 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
7517 @item @code{sw1 __MHDSETS (const)}
7518 @tab @code{@var{b} = __MHDSETS (@var{a})}
7519 @tab @code{MHDSETS #@var{a},@var{b}}
7520 @item @code{uw1 __MHSETHIH (uw1, const)}
7521 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
7522 @tab @code{MHSETHIH #@var{a},@var{b}}
7523 @item @code{sw1 __MHSETHIS (sw1, const)}
7524 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
7525 @tab @code{MHSETHIS #@var{a},@var{b}}
7526 @item @code{uw1 __MHSETLOH (uw1, const)}
7527 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
7528 @tab @code{MHSETLOH #@var{a},@var{b}}
7529 @item @code{sw1 __MHSETLOS (sw1, const)}
7530 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
7531 @tab @code{MHSETLOS #@var{a},@var{b}}
7532 @item @code{uw1 __MHTOB (uw2)}
7533 @tab @code{@var{b} = __MHTOB (@var{a})}
7534 @tab @code{MHTOB @var{a},@var{b}}
7535 @item @code{void __MMACHS (acc, sw1, sw1)}
7536 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
7537 @tab @code{MMACHS @var{a},@var{b},@var{c}}
7538 @item @code{void __MMACHU (acc, uw1, uw1)}
7539 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
7540 @tab @code{MMACHU @var{a},@var{b},@var{c}}
7541 @item @code{void __MMRDHS (acc, sw1, sw1)}
7542 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
7543 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
7544 @item @code{void __MMRDHU (acc, uw1, uw1)}
7545 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
7546 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
7547 @item @code{void __MMULHS (acc, sw1, sw1)}
7548 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
7549 @tab @code{MMULHS @var{a},@var{b},@var{c}}
7550 @item @code{void __MMULHU (acc, uw1, uw1)}
7551 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
7552 @tab @code{MMULHU @var{a},@var{b},@var{c}}
7553 @item @code{void __MMULXHS (acc, sw1, sw1)}
7554 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
7555 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
7556 @item @code{void __MMULXHU (acc, uw1, uw1)}
7557 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
7558 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
7559 @item @code{uw1 __MNOT (uw1)}
7560 @tab @code{@var{b} = __MNOT (@var{a})}
7561 @tab @code{MNOT @var{a},@var{b}}
7562 @item @code{uw1 __MOR (uw1, uw1)}
7563 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
7564 @tab @code{MOR @var{a},@var{b},@var{c}}
7565 @item @code{uw1 __MPACKH (uh, uh)}
7566 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
7567 @tab @code{MPACKH @var{a},@var{b},@var{c}}
7568 @item @code{sw2 __MQADDHSS (sw2, sw2)}
7569 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
7570 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
7571 @item @code{uw2 __MQADDHUS (uw2, uw2)}
7572 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
7573 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
7574 @item @code{void __MQCPXIS (acc, sw2, sw2)}
7575 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
7576 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
7577 @item @code{void __MQCPXIU (acc, uw2, uw2)}
7578 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
7579 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
7580 @item @code{void __MQCPXRS (acc, sw2, sw2)}
7581 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
7582 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
7583 @item @code{void __MQCPXRU (acc, uw2, uw2)}
7584 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
7585 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
7586 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
7587 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
7588 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
7589 @item @code{sw2 __MQLMTHS (sw2, sw2)}
7590 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
7591 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
7592 @item @code{void __MQMACHS (acc, sw2, sw2)}
7593 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
7594 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
7595 @item @code{void __MQMACHU (acc, uw2, uw2)}
7596 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
7597 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
7598 @item @code{void __MQMACXHS (acc, sw2, sw2)}
7599 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
7600 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
7601 @item @code{void __MQMULHS (acc, sw2, sw2)}
7602 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
7603 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
7604 @item @code{void __MQMULHU (acc, uw2, uw2)}
7605 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
7606 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
7607 @item @code{void __MQMULXHS (acc, sw2, sw2)}
7608 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
7609 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
7610 @item @code{void __MQMULXHU (acc, uw2, uw2)}
7611 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
7612 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
7613 @item @code{sw2 __MQSATHS (sw2, sw2)}
7614 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
7615 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
7616 @item @code{uw2 __MQSLLHI (uw2, int)}
7617 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
7618 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
7619 @item @code{sw2 __MQSRAHI (sw2, int)}
7620 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
7621 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
7622 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
7623 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
7624 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
7625 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
7626 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
7627 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
7628 @item @code{void __MQXMACHS (acc, sw2, sw2)}
7629 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
7630 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
7631 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
7632 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
7633 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
7634 @item @code{uw1 __MRDACC (acc)}
7635 @tab @code{@var{b} = __MRDACC (@var{a})}
7636 @tab @code{MRDACC @var{a},@var{b}}
7637 @item @code{uw1 __MRDACCG (acc)}
7638 @tab @code{@var{b} = __MRDACCG (@var{a})}
7639 @tab @code{MRDACCG @var{a},@var{b}}
7640 @item @code{uw1 __MROTLI (uw1, const)}
7641 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
7642 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
7643 @item @code{uw1 __MROTRI (uw1, const)}
7644 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
7645 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
7646 @item @code{sw1 __MSATHS (sw1, sw1)}
7647 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
7648 @tab @code{MSATHS @var{a},@var{b},@var{c}}
7649 @item @code{uw1 __MSATHU (uw1, uw1)}
7650 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
7651 @tab @code{MSATHU @var{a},@var{b},@var{c}}
7652 @item @code{uw1 __MSLLHI (uw1, const)}
7653 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
7654 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
7655 @item @code{sw1 __MSRAHI (sw1, const)}
7656 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
7657 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
7658 @item @code{uw1 __MSRLHI (uw1, const)}
7659 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
7660 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
7661 @item @code{void __MSUBACCS (acc, acc)}
7662 @tab @code{__MSUBACCS (@var{b}, @var{a})}
7663 @tab @code{MSUBACCS @var{a},@var{b}}
7664 @item @code{sw1 __MSUBHSS (sw1, sw1)}
7665 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
7666 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
7667 @item @code{uw1 __MSUBHUS (uw1, uw1)}
7668 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
7669 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
7670 @item @code{void __MTRAP (void)}
7671 @tab @code{__MTRAP ()}
7673 @item @code{uw2 __MUNPACKH (uw1)}
7674 @tab @code{@var{b} = __MUNPACKH (@var{a})}
7675 @tab @code{MUNPACKH @var{a},@var{b}}
7676 @item @code{uw1 __MWCUT (uw2, uw1)}
7677 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
7678 @tab @code{MWCUT @var{a},@var{b},@var{c}}
7679 @item @code{void __MWTACC (acc, uw1)}
7680 @tab @code{__MWTACC (@var{b}, @var{a})}
7681 @tab @code{MWTACC @var{a},@var{b}}
7682 @item @code{void __MWTACCG (acc, uw1)}
7683 @tab @code{__MWTACCG (@var{b}, @var{a})}
7684 @tab @code{MWTACCG @var{a},@var{b}}
7685 @item @code{uw1 __MXOR (uw1, uw1)}
7686 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
7687 @tab @code{MXOR @var{a},@var{b},@var{c}}
7690 @node Raw read/write Functions
7691 @subsubsection Raw read/write Functions
7693 This sections describes built-in functions related to read and write
7694 instructions to access memory. These functions generate
7695 @code{membar} instructions to flush the I/O load and stores where
7696 appropriate, as described in Fujitsu's manual described above.
7700 @item unsigned char __builtin_read8 (void *@var{data})
7701 @item unsigned short __builtin_read16 (void *@var{data})
7702 @item unsigned long __builtin_read32 (void *@var{data})
7703 @item unsigned long long __builtin_read64 (void *@var{data})
7705 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
7706 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
7707 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
7708 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
7711 @node Other Built-in Functions
7712 @subsubsection Other Built-in Functions
7714 This section describes built-in functions that are not named after
7715 a specific FR-V instruction.
7718 @item sw2 __IACCreadll (iacc @var{reg})
7719 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
7720 for future expansion and must be 0.
7722 @item sw1 __IACCreadl (iacc @var{reg})
7723 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
7724 Other values of @var{reg} are rejected as invalid.
7726 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
7727 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
7728 is reserved for future expansion and must be 0.
7730 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7731 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7732 is 1. Other values of @var{reg} are rejected as invalid.
7734 @item void __data_prefetch0 (const void *@var{x})
7735 Use the @code{dcpl} instruction to load the contents of address @var{x}
7736 into the data cache.
7738 @item void __data_prefetch (const void *@var{x})
7739 Use the @code{nldub} instruction to load the contents of address @var{x}
7740 into the data cache. The instruction will be issued in slot I1@.
7743 @node X86 Built-in Functions
7744 @subsection X86 Built-in Functions
7746 These built-in functions are available for the i386 and x86-64 family
7747 of computers, depending on the command-line switches used.
7749 Note that, if you specify command-line switches such as @option{-msse},
7750 the compiler could use the extended instruction sets even if the built-ins
7751 are not used explicitly in the program. For this reason, applications
7752 which perform runtime CPU detection must compile separate files for each
7753 supported architecture, using the appropriate flags. In particular,
7754 the file containing the CPU detection code should be compiled without
7757 The following machine modes are available for use with MMX built-in functions
7758 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7759 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7760 vector of eight 8-bit integers. Some of the built-in functions operate on
7761 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
7763 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7764 of two 32-bit floating point values.
7766 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7767 floating point values. Some instructions use a vector of four 32-bit
7768 integers, these use @code{V4SI}. Finally, some instructions operate on an
7769 entire vector register, interpreting it as a 128-bit integer, these use mode
7772 In 64-bit mode, the x86-64 family of processors uses additional built-in
7773 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7774 floating point and @code{TC} 128-bit complex floating point values.
7776 The following floating point built-in functions are available in 64-bit
7777 mode. All of them implement the function that is part of the name.
7780 __float128 __builtin_fabsq (__float128)
7781 __float128 __builtin_copysignq (__float128, __float128)
7784 The following floating point built-in functions are made available in the
7788 @item __float128 __builtin_infq (void)
7789 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7790 @findex __builtin_infq
7792 @item __float128 __builtin_huge_valq (void)
7793 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
7794 @findex __builtin_huge_valq
7797 The following built-in functions are made available by @option{-mmmx}.
7798 All of them generate the machine instruction that is part of the name.
7801 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7802 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7803 v2si __builtin_ia32_paddd (v2si, v2si)
7804 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7805 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7806 v2si __builtin_ia32_psubd (v2si, v2si)
7807 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7808 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7809 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7810 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7811 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7812 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7813 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7814 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7815 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7816 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7817 di __builtin_ia32_pand (di, di)
7818 di __builtin_ia32_pandn (di,di)
7819 di __builtin_ia32_por (di, di)
7820 di __builtin_ia32_pxor (di, di)
7821 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7822 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7823 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7824 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7825 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7826 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7827 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7828 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7829 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7830 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7831 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7832 v2si __builtin_ia32_punpckldq (v2si, v2si)
7833 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7834 v4hi __builtin_ia32_packssdw (v2si, v2si)
7835 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7837 v4hi __builtin_ia32_psllw (v4hi, v4hi)
7838 v2si __builtin_ia32_pslld (v2si, v2si)
7839 v1di __builtin_ia32_psllq (v1di, v1di)
7840 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
7841 v2si __builtin_ia32_psrld (v2si, v2si)
7842 v1di __builtin_ia32_psrlq (v1di, v1di)
7843 v4hi __builtin_ia32_psraw (v4hi, v4hi)
7844 v2si __builtin_ia32_psrad (v2si, v2si)
7845 v4hi __builtin_ia32_psllwi (v4hi, int)
7846 v2si __builtin_ia32_pslldi (v2si, int)
7847 v1di __builtin_ia32_psllqi (v1di, int)
7848 v4hi __builtin_ia32_psrlwi (v4hi, int)
7849 v2si __builtin_ia32_psrldi (v2si, int)
7850 v1di __builtin_ia32_psrlqi (v1di, int)
7851 v4hi __builtin_ia32_psrawi (v4hi, int)
7852 v2si __builtin_ia32_psradi (v2si, int)
7856 The following built-in functions are made available either with
7857 @option{-msse}, or with a combination of @option{-m3dnow} and
7858 @option{-march=athlon}. All of them generate the machine
7859 instruction that is part of the name.
7862 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7863 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7864 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7865 v1di __builtin_ia32_psadbw (v8qi, v8qi)
7866 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7867 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7868 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7869 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7870 int __builtin_ia32_pextrw (v4hi, int)
7871 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7872 int __builtin_ia32_pmovmskb (v8qi)
7873 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7874 void __builtin_ia32_movntq (di *, di)
7875 void __builtin_ia32_sfence (void)
7878 The following built-in functions are available when @option{-msse} is used.
7879 All of them generate the machine instruction that is part of the name.
7882 int __builtin_ia32_comieq (v4sf, v4sf)
7883 int __builtin_ia32_comineq (v4sf, v4sf)
7884 int __builtin_ia32_comilt (v4sf, v4sf)
7885 int __builtin_ia32_comile (v4sf, v4sf)
7886 int __builtin_ia32_comigt (v4sf, v4sf)
7887 int __builtin_ia32_comige (v4sf, v4sf)
7888 int __builtin_ia32_ucomieq (v4sf, v4sf)
7889 int __builtin_ia32_ucomineq (v4sf, v4sf)
7890 int __builtin_ia32_ucomilt (v4sf, v4sf)
7891 int __builtin_ia32_ucomile (v4sf, v4sf)
7892 int __builtin_ia32_ucomigt (v4sf, v4sf)
7893 int __builtin_ia32_ucomige (v4sf, v4sf)
7894 v4sf __builtin_ia32_addps (v4sf, v4sf)
7895 v4sf __builtin_ia32_subps (v4sf, v4sf)
7896 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7897 v4sf __builtin_ia32_divps (v4sf, v4sf)
7898 v4sf __builtin_ia32_addss (v4sf, v4sf)
7899 v4sf __builtin_ia32_subss (v4sf, v4sf)
7900 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7901 v4sf __builtin_ia32_divss (v4sf, v4sf)
7902 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7903 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7904 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7905 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7906 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7907 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7908 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7909 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7910 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7911 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7912 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7913 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7914 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7915 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7916 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7917 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7918 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7919 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7920 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7921 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7922 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7923 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7924 v4sf __builtin_ia32_minps (v4sf, v4sf)
7925 v4sf __builtin_ia32_minss (v4sf, v4sf)
7926 v4sf __builtin_ia32_andps (v4sf, v4sf)
7927 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7928 v4sf __builtin_ia32_orps (v4sf, v4sf)
7929 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7930 v4sf __builtin_ia32_movss (v4sf, v4sf)
7931 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7932 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7933 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7934 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7935 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7936 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7937 v2si __builtin_ia32_cvtps2pi (v4sf)
7938 int __builtin_ia32_cvtss2si (v4sf)
7939 v2si __builtin_ia32_cvttps2pi (v4sf)
7940 int __builtin_ia32_cvttss2si (v4sf)
7941 v4sf __builtin_ia32_rcpps (v4sf)
7942 v4sf __builtin_ia32_rsqrtps (v4sf)
7943 v4sf __builtin_ia32_sqrtps (v4sf)
7944 v4sf __builtin_ia32_rcpss (v4sf)
7945 v4sf __builtin_ia32_rsqrtss (v4sf)
7946 v4sf __builtin_ia32_sqrtss (v4sf)
7947 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7948 void __builtin_ia32_movntps (float *, v4sf)
7949 int __builtin_ia32_movmskps (v4sf)
7952 The following built-in functions are available when @option{-msse} is used.
7955 @item v4sf __builtin_ia32_loadaps (float *)
7956 Generates the @code{movaps} machine instruction as a load from memory.
7957 @item void __builtin_ia32_storeaps (float *, v4sf)
7958 Generates the @code{movaps} machine instruction as a store to memory.
7959 @item v4sf __builtin_ia32_loadups (float *)
7960 Generates the @code{movups} machine instruction as a load from memory.
7961 @item void __builtin_ia32_storeups (float *, v4sf)
7962 Generates the @code{movups} machine instruction as a store to memory.
7963 @item v4sf __builtin_ia32_loadsss (float *)
7964 Generates the @code{movss} machine instruction as a load from memory.
7965 @item void __builtin_ia32_storess (float *, v4sf)
7966 Generates the @code{movss} machine instruction as a store to memory.
7967 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
7968 Generates the @code{movhps} machine instruction as a load from memory.
7969 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
7970 Generates the @code{movlps} machine instruction as a load from memory
7971 @item void __builtin_ia32_storehps (v2sf *, v4sf)
7972 Generates the @code{movhps} machine instruction as a store to memory.
7973 @item void __builtin_ia32_storelps (v2sf *, v4sf)
7974 Generates the @code{movlps} machine instruction as a store to memory.
7977 The following built-in functions are available when @option{-msse2} is used.
7978 All of them generate the machine instruction that is part of the name.
7981 int __builtin_ia32_comisdeq (v2df, v2df)
7982 int __builtin_ia32_comisdlt (v2df, v2df)
7983 int __builtin_ia32_comisdle (v2df, v2df)
7984 int __builtin_ia32_comisdgt (v2df, v2df)
7985 int __builtin_ia32_comisdge (v2df, v2df)
7986 int __builtin_ia32_comisdneq (v2df, v2df)
7987 int __builtin_ia32_ucomisdeq (v2df, v2df)
7988 int __builtin_ia32_ucomisdlt (v2df, v2df)
7989 int __builtin_ia32_ucomisdle (v2df, v2df)
7990 int __builtin_ia32_ucomisdgt (v2df, v2df)
7991 int __builtin_ia32_ucomisdge (v2df, v2df)
7992 int __builtin_ia32_ucomisdneq (v2df, v2df)
7993 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7994 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7995 v2df __builtin_ia32_cmplepd (v2df, v2df)
7996 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7997 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7998 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7999 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
8000 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
8001 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
8002 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
8003 v2df __builtin_ia32_cmpngepd (v2df, v2df)
8004 v2df __builtin_ia32_cmpordpd (v2df, v2df)
8005 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
8006 v2df __builtin_ia32_cmpltsd (v2df, v2df)
8007 v2df __builtin_ia32_cmplesd (v2df, v2df)
8008 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
8009 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
8010 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
8011 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
8012 v2df __builtin_ia32_cmpordsd (v2df, v2df)
8013 v2di __builtin_ia32_paddq (v2di, v2di)
8014 v2di __builtin_ia32_psubq (v2di, v2di)
8015 v2df __builtin_ia32_addpd (v2df, v2df)
8016 v2df __builtin_ia32_subpd (v2df, v2df)
8017 v2df __builtin_ia32_mulpd (v2df, v2df)
8018 v2df __builtin_ia32_divpd (v2df, v2df)
8019 v2df __builtin_ia32_addsd (v2df, v2df)
8020 v2df __builtin_ia32_subsd (v2df, v2df)
8021 v2df __builtin_ia32_mulsd (v2df, v2df)
8022 v2df __builtin_ia32_divsd (v2df, v2df)
8023 v2df __builtin_ia32_minpd (v2df, v2df)
8024 v2df __builtin_ia32_maxpd (v2df, v2df)
8025 v2df __builtin_ia32_minsd (v2df, v2df)
8026 v2df __builtin_ia32_maxsd (v2df, v2df)
8027 v2df __builtin_ia32_andpd (v2df, v2df)
8028 v2df __builtin_ia32_andnpd (v2df, v2df)
8029 v2df __builtin_ia32_orpd (v2df, v2df)
8030 v2df __builtin_ia32_xorpd (v2df, v2df)
8031 v2df __builtin_ia32_movsd (v2df, v2df)
8032 v2df __builtin_ia32_unpckhpd (v2df, v2df)
8033 v2df __builtin_ia32_unpcklpd (v2df, v2df)
8034 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
8035 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
8036 v4si __builtin_ia32_paddd128 (v4si, v4si)
8037 v2di __builtin_ia32_paddq128 (v2di, v2di)
8038 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
8039 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
8040 v4si __builtin_ia32_psubd128 (v4si, v4si)
8041 v2di __builtin_ia32_psubq128 (v2di, v2di)
8042 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
8043 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
8044 v2di __builtin_ia32_pand128 (v2di, v2di)
8045 v2di __builtin_ia32_pandn128 (v2di, v2di)
8046 v2di __builtin_ia32_por128 (v2di, v2di)
8047 v2di __builtin_ia32_pxor128 (v2di, v2di)
8048 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
8049 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
8050 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
8051 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
8052 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
8053 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
8054 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
8055 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
8056 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
8057 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
8058 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
8059 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8060 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8061 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8062 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8063 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8064 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8065 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8066 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8067 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8068 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8069 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8070 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8071 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8072 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8073 v2df __builtin_ia32_loadupd (double *)
8074 void __builtin_ia32_storeupd (double *, v2df)
8075 v2df __builtin_ia32_loadhpd (v2df, double const *)
8076 v2df __builtin_ia32_loadlpd (v2df, double const *)
8077 int __builtin_ia32_movmskpd (v2df)
8078 int __builtin_ia32_pmovmskb128 (v16qi)
8079 void __builtin_ia32_movnti (int *, int)
8080 void __builtin_ia32_movntpd (double *, v2df)
8081 void __builtin_ia32_movntdq (v2df *, v2df)
8082 v4si __builtin_ia32_pshufd (v4si, int)
8083 v8hi __builtin_ia32_pshuflw (v8hi, int)
8084 v8hi __builtin_ia32_pshufhw (v8hi, int)
8085 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
8086 v2df __builtin_ia32_sqrtpd (v2df)
8087 v2df __builtin_ia32_sqrtsd (v2df)
8088 v2df __builtin_ia32_shufpd (v2df, v2df, int)
8089 v2df __builtin_ia32_cvtdq2pd (v4si)
8090 v4sf __builtin_ia32_cvtdq2ps (v4si)
8091 v4si __builtin_ia32_cvtpd2dq (v2df)
8092 v2si __builtin_ia32_cvtpd2pi (v2df)
8093 v4sf __builtin_ia32_cvtpd2ps (v2df)
8094 v4si __builtin_ia32_cvttpd2dq (v2df)
8095 v2si __builtin_ia32_cvttpd2pi (v2df)
8096 v2df __builtin_ia32_cvtpi2pd (v2si)
8097 int __builtin_ia32_cvtsd2si (v2df)
8098 int __builtin_ia32_cvttsd2si (v2df)
8099 long long __builtin_ia32_cvtsd2si64 (v2df)
8100 long long __builtin_ia32_cvttsd2si64 (v2df)
8101 v4si __builtin_ia32_cvtps2dq (v4sf)
8102 v2df __builtin_ia32_cvtps2pd (v4sf)
8103 v4si __builtin_ia32_cvttps2dq (v4sf)
8104 v2df __builtin_ia32_cvtsi2sd (v2df, int)
8105 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
8106 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
8107 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
8108 void __builtin_ia32_clflush (const void *)
8109 void __builtin_ia32_lfence (void)
8110 void __builtin_ia32_mfence (void)
8111 v16qi __builtin_ia32_loaddqu (const char *)
8112 void __builtin_ia32_storedqu (char *, v16qi)
8113 v1di __builtin_ia32_pmuludq (v2si, v2si)
8114 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
8115 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
8116 v4si __builtin_ia32_pslld128 (v4si, v4si)
8117 v2di __builtin_ia32_psllq128 (v2di, v2di)
8118 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
8119 v4si __builtin_ia32_psrld128 (v4si, v4si)
8120 v2di __builtin_ia32_psrlq128 (v2di, v2di)
8121 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
8122 v4si __builtin_ia32_psrad128 (v4si, v4si)
8123 v2di __builtin_ia32_pslldqi128 (v2di, int)
8124 v8hi __builtin_ia32_psllwi128 (v8hi, int)
8125 v4si __builtin_ia32_pslldi128 (v4si, int)
8126 v2di __builtin_ia32_psllqi128 (v2di, int)
8127 v2di __builtin_ia32_psrldqi128 (v2di, int)
8128 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
8129 v4si __builtin_ia32_psrldi128 (v4si, int)
8130 v2di __builtin_ia32_psrlqi128 (v2di, int)
8131 v8hi __builtin_ia32_psrawi128 (v8hi, int)
8132 v4si __builtin_ia32_psradi128 (v4si, int)
8133 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
8134 v2di __builtin_ia32_movq128 (v2di)
8137 The following built-in functions are available when @option{-msse3} is used.
8138 All of them generate the machine instruction that is part of the name.
8141 v2df __builtin_ia32_addsubpd (v2df, v2df)
8142 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
8143 v2df __builtin_ia32_haddpd (v2df, v2df)
8144 v4sf __builtin_ia32_haddps (v4sf, v4sf)
8145 v2df __builtin_ia32_hsubpd (v2df, v2df)
8146 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
8147 v16qi __builtin_ia32_lddqu (char const *)
8148 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
8149 v2df __builtin_ia32_movddup (v2df)
8150 v4sf __builtin_ia32_movshdup (v4sf)
8151 v4sf __builtin_ia32_movsldup (v4sf)
8152 void __builtin_ia32_mwait (unsigned int, unsigned int)
8155 The following built-in functions are available when @option{-msse3} is used.
8158 @item v2df __builtin_ia32_loadddup (double const *)
8159 Generates the @code{movddup} machine instruction as a load from memory.
8162 The following built-in functions are available when @option{-mssse3} is used.
8163 All of them generate the machine instruction that is part of the name
8167 v2si __builtin_ia32_phaddd (v2si, v2si)
8168 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
8169 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
8170 v2si __builtin_ia32_phsubd (v2si, v2si)
8171 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
8172 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
8173 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
8174 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
8175 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
8176 v8qi __builtin_ia32_psignb (v8qi, v8qi)
8177 v2si __builtin_ia32_psignd (v2si, v2si)
8178 v4hi __builtin_ia32_psignw (v4hi, v4hi)
8179 v1di __builtin_ia32_palignr (v1di, v1di, int)
8180 v8qi __builtin_ia32_pabsb (v8qi)
8181 v2si __builtin_ia32_pabsd (v2si)
8182 v4hi __builtin_ia32_pabsw (v4hi)
8185 The following built-in functions are available when @option{-mssse3} is used.
8186 All of them generate the machine instruction that is part of the name
8190 v4si __builtin_ia32_phaddd128 (v4si, v4si)
8191 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
8192 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
8193 v4si __builtin_ia32_phsubd128 (v4si, v4si)
8194 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
8195 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
8196 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
8197 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
8198 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
8199 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
8200 v4si __builtin_ia32_psignd128 (v4si, v4si)
8201 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
8202 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
8203 v16qi __builtin_ia32_pabsb128 (v16qi)
8204 v4si __builtin_ia32_pabsd128 (v4si)
8205 v8hi __builtin_ia32_pabsw128 (v8hi)
8208 The following built-in functions are available when @option{-msse4.1} is
8209 used. All of them generate the machine instruction that is part of the
8213 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
8214 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
8215 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
8216 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
8217 v2df __builtin_ia32_dppd (v2df, v2df, const int)
8218 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
8219 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
8220 v2di __builtin_ia32_movntdqa (v2di *);
8221 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
8222 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
8223 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
8224 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
8225 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
8226 v8hi __builtin_ia32_phminposuw128 (v8hi)
8227 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
8228 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
8229 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
8230 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
8231 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
8232 v4si __builtin_ia32_pminsd128 (v4si, v4si)
8233 v4si __builtin_ia32_pminud128 (v4si, v4si)
8234 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
8235 v4si __builtin_ia32_pmovsxbd128 (v16qi)
8236 v2di __builtin_ia32_pmovsxbq128 (v16qi)
8237 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
8238 v2di __builtin_ia32_pmovsxdq128 (v4si)
8239 v4si __builtin_ia32_pmovsxwd128 (v8hi)
8240 v2di __builtin_ia32_pmovsxwq128 (v8hi)
8241 v4si __builtin_ia32_pmovzxbd128 (v16qi)
8242 v2di __builtin_ia32_pmovzxbq128 (v16qi)
8243 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
8244 v2di __builtin_ia32_pmovzxdq128 (v4si)
8245 v4si __builtin_ia32_pmovzxwd128 (v8hi)
8246 v2di __builtin_ia32_pmovzxwq128 (v8hi)
8247 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
8248 v4si __builtin_ia32_pmulld128 (v4si, v4si)
8249 int __builtin_ia32_ptestc128 (v2di, v2di)
8250 int __builtin_ia32_ptestnzc128 (v2di, v2di)
8251 int __builtin_ia32_ptestz128 (v2di, v2di)
8252 v2df __builtin_ia32_roundpd (v2df, const int)
8253 v4sf __builtin_ia32_roundps (v4sf, const int)
8254 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
8255 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
8258 The following built-in functions are available when @option{-msse4.1} is
8262 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
8263 Generates the @code{insertps} machine instruction.
8264 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
8265 Generates the @code{pextrb} machine instruction.
8266 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
8267 Generates the @code{pinsrb} machine instruction.
8268 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
8269 Generates the @code{pinsrd} machine instruction.
8270 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
8271 Generates the @code{pinsrq} machine instruction in 64bit mode.
8274 The following built-in functions are changed to generate new SSE4.1
8275 instructions when @option{-msse4.1} is used.
8278 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
8279 Generates the @code{extractps} machine instruction.
8280 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
8281 Generates the @code{pextrd} machine instruction.
8282 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
8283 Generates the @code{pextrq} machine instruction in 64bit mode.
8286 The following built-in functions are available when @option{-msse4.2} is
8287 used. All of them generate the machine instruction that is part of the
8291 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
8292 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
8293 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
8294 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
8295 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
8296 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
8297 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
8298 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
8299 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
8300 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
8301 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
8302 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
8303 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
8304 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
8305 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
8308 The following built-in functions are available when @option{-msse4.2} is
8312 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
8313 Generates the @code{crc32b} machine instruction.
8314 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
8315 Generates the @code{crc32w} machine instruction.
8316 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
8317 Generates the @code{crc32l} machine instruction.
8318 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
8321 The following built-in functions are changed to generate new SSE4.2
8322 instructions when @option{-msse4.2} is used.
8325 @item int __builtin_popcount (unsigned int)
8326 Generates the @code{popcntl} machine instruction.
8327 @item int __builtin_popcountl (unsigned long)
8328 Generates the @code{popcntl} or @code{popcntq} machine instruction,
8329 depending on the size of @code{unsigned long}.
8330 @item int __builtin_popcountll (unsigned long long)
8331 Generates the @code{popcntq} machine instruction.
8334 The following built-in functions are available when @option{-mavx} is
8335 used. All of them generate the machine instruction that is part of the
8339 v4df __builtin_ia32_addpd256 (v4df,v4df)
8340 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
8341 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
8342 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
8343 v4df __builtin_ia32_andnpd256 (v4df,v4df)
8344 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
8345 v4df __builtin_ia32_andpd256 (v4df,v4df)
8346 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
8347 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
8348 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
8349 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
8350 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
8351 v2df __builtin_ia32_cmppd (v2df,v2df,int)
8352 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
8353 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
8354 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
8355 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
8356 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
8357 v4df __builtin_ia32_cvtdq2pd256 (v4si)
8358 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
8359 v4si __builtin_ia32_cvtpd2dq256 (v4df)
8360 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
8361 v8si __builtin_ia32_cvtps2dq256 (v8sf)
8362 v4df __builtin_ia32_cvtps2pd256 (v4sf)
8363 v4si __builtin_ia32_cvttpd2dq256 (v4df)
8364 v8si __builtin_ia32_cvttps2dq256 (v8sf)
8365 v4df __builtin_ia32_divpd256 (v4df,v4df)
8366 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
8367 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
8368 v4df __builtin_ia32_haddpd256 (v4df,v4df)
8369 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
8370 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
8371 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
8372 v32qi __builtin_ia32_lddqu256 (pcchar)
8373 v32qi __builtin_ia32_loaddqu256 (pcchar)
8374 v4df __builtin_ia32_loadupd256 (pcdouble)
8375 v8sf __builtin_ia32_loadups256 (pcfloat)
8376 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
8377 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
8378 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
8379 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
8380 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
8381 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
8382 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
8383 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
8384 v4df __builtin_ia32_maxpd256 (v4df,v4df)
8385 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
8386 v4df __builtin_ia32_minpd256 (v4df,v4df)
8387 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
8388 v4df __builtin_ia32_movddup256 (v4df)
8389 int __builtin_ia32_movmskpd256 (v4df)
8390 int __builtin_ia32_movmskps256 (v8sf)
8391 v8sf __builtin_ia32_movshdup256 (v8sf)
8392 v8sf __builtin_ia32_movsldup256 (v8sf)
8393 v4df __builtin_ia32_mulpd256 (v4df,v4df)
8394 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
8395 v4df __builtin_ia32_orpd256 (v4df,v4df)
8396 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
8397 v2df __builtin_ia32_pd_pd256 (v4df)
8398 v4df __builtin_ia32_pd256_pd (v2df)
8399 v4sf __builtin_ia32_ps_ps256 (v8sf)
8400 v8sf __builtin_ia32_ps256_ps (v4sf)
8401 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
8402 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
8403 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
8404 v8sf __builtin_ia32_rcpps256 (v8sf)
8405 v4df __builtin_ia32_roundpd256 (v4df,int)
8406 v8sf __builtin_ia32_roundps256 (v8sf,int)
8407 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
8408 v8sf __builtin_ia32_rsqrtps256 (v8sf)
8409 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
8410 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
8411 v4si __builtin_ia32_si_si256 (v8si)
8412 v8si __builtin_ia32_si256_si (v4si)
8413 v4df __builtin_ia32_sqrtpd256 (v4df)
8414 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
8415 v8sf __builtin_ia32_sqrtps256 (v8sf)
8416 void __builtin_ia32_storedqu256 (pchar,v32qi)
8417 void __builtin_ia32_storeupd256 (pdouble,v4df)
8418 void __builtin_ia32_storeups256 (pfloat,v8sf)
8419 v4df __builtin_ia32_subpd256 (v4df,v4df)
8420 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
8421 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
8422 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
8423 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
8424 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
8425 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
8426 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
8427 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
8428 v4sf __builtin_ia32_vbroadcastss (pcfloat)
8429 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
8430 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
8431 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
8432 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
8433 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
8434 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
8435 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
8436 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
8437 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
8438 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
8439 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
8440 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
8441 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
8442 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
8443 v2df __builtin_ia32_vpermilpd (v2df,int)
8444 v4df __builtin_ia32_vpermilpd256 (v4df,int)
8445 v4sf __builtin_ia32_vpermilps (v4sf,int)
8446 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
8447 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
8448 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
8449 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
8450 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
8451 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
8452 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
8453 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
8454 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
8455 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
8456 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
8457 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
8458 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
8459 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
8460 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
8461 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
8462 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
8463 void __builtin_ia32_vzeroall (void)
8464 void __builtin_ia32_vzeroupper (void)
8465 v4df __builtin_ia32_xorpd256 (v4df,v4df)
8466 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
8469 The following built-in functions are available when @option{-maes} is
8470 used. All of them generate the machine instruction that is part of the
8474 v2di __builtin_ia32_aesenc128 (v2di, v2di)
8475 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
8476 v2di __builtin_ia32_aesdec128 (v2di, v2di)
8477 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
8478 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
8479 v2di __builtin_ia32_aesimc128 (v2di)
8482 The following built-in function is available when @option{-mpclmul} is
8486 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
8487 Generates the @code{pclmulqdq} machine instruction.
8490 The following built-in functions are available when @option{-msse4a} is used.
8491 All of them generate the machine instruction that is part of the name.
8494 void __builtin_ia32_movntsd (double *, v2df)
8495 void __builtin_ia32_movntss (float *, v4sf)
8496 v2di __builtin_ia32_extrq (v2di, v16qi)
8497 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
8498 v2di __builtin_ia32_insertq (v2di, v2di)
8499 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
8502 The following built-in functions are available when @option{-msse5} is used.
8503 All of them generate the machine instruction that is part of the name
8507 v2df __builtin_ia32_comeqpd (v2df, v2df)
8508 v2df __builtin_ia32_comeqps (v2df, v2df)
8509 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
8510 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
8511 v2df __builtin_ia32_comfalsepd (v2df, v2df)
8512 v2df __builtin_ia32_comfalseps (v2df, v2df)
8513 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
8514 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
8515 v2df __builtin_ia32_comgepd (v2df, v2df)
8516 v2df __builtin_ia32_comgeps (v2df, v2df)
8517 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
8518 v4sf __builtin_ia32_comgess (v4sf, v4sf)
8519 v2df __builtin_ia32_comgtpd (v2df, v2df)
8520 v2df __builtin_ia32_comgtps (v2df, v2df)
8521 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
8522 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
8523 v2df __builtin_ia32_comlepd (v2df, v2df)
8524 v2df __builtin_ia32_comleps (v2df, v2df)
8525 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
8526 v4sf __builtin_ia32_comless (v4sf, v4sf)
8527 v2df __builtin_ia32_comltpd (v2df, v2df)
8528 v2df __builtin_ia32_comltps (v2df, v2df)
8529 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
8530 v4sf __builtin_ia32_comltss (v4sf, v4sf)
8531 v2df __builtin_ia32_comnepd (v2df, v2df)
8532 v2df __builtin_ia32_comneps (v2df, v2df)
8533 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
8534 v4sf __builtin_ia32_comness (v4sf, v4sf)
8535 v2df __builtin_ia32_comordpd (v2df, v2df)
8536 v2df __builtin_ia32_comordps (v2df, v2df)
8537 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
8538 v4sf __builtin_ia32_comordss (v4sf, v4sf)
8539 v2df __builtin_ia32_comtruepd (v2df, v2df)
8540 v2df __builtin_ia32_comtrueps (v2df, v2df)
8541 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
8542 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
8543 v2df __builtin_ia32_comueqpd (v2df, v2df)
8544 v2df __builtin_ia32_comueqps (v2df, v2df)
8545 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
8546 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
8547 v2df __builtin_ia32_comugepd (v2df, v2df)
8548 v2df __builtin_ia32_comugeps (v2df, v2df)
8549 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
8550 v4sf __builtin_ia32_comugess (v4sf, v4sf)
8551 v2df __builtin_ia32_comugtpd (v2df, v2df)
8552 v2df __builtin_ia32_comugtps (v2df, v2df)
8553 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
8554 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
8555 v2df __builtin_ia32_comulepd (v2df, v2df)
8556 v2df __builtin_ia32_comuleps (v2df, v2df)
8557 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
8558 v4sf __builtin_ia32_comuless (v4sf, v4sf)
8559 v2df __builtin_ia32_comultpd (v2df, v2df)
8560 v2df __builtin_ia32_comultps (v2df, v2df)
8561 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
8562 v4sf __builtin_ia32_comultss (v4sf, v4sf)
8563 v2df __builtin_ia32_comunepd (v2df, v2df)
8564 v2df __builtin_ia32_comuneps (v2df, v2df)
8565 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
8566 v4sf __builtin_ia32_comuness (v4sf, v4sf)
8567 v2df __builtin_ia32_comunordpd (v2df, v2df)
8568 v2df __builtin_ia32_comunordps (v2df, v2df)
8569 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
8570 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
8571 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
8572 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
8573 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
8574 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
8575 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
8576 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
8577 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
8578 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
8579 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
8580 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
8581 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
8582 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
8583 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
8584 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
8585 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
8586 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
8587 v2df __builtin_ia32_frczpd (v2df)
8588 v4sf __builtin_ia32_frczps (v4sf)
8589 v2df __builtin_ia32_frczsd (v2df, v2df)
8590 v4sf __builtin_ia32_frczss (v4sf, v4sf)
8591 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
8592 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
8593 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
8594 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
8595 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
8596 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
8597 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
8598 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
8599 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8600 v4si __builtin_ia32_pcomeqd (v4si, v4si)
8601 v2di __builtin_ia32_pcomeqq (v2di, v2di)
8602 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
8603 v4si __builtin_ia32_pcomequd (v4si, v4si)
8604 v2di __builtin_ia32_pcomequq (v2di, v2di)
8605 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
8606 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8607 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
8608 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
8609 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
8610 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
8611 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
8612 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
8613 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
8614 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
8615 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
8616 v4si __builtin_ia32_pcomged (v4si, v4si)
8617 v2di __builtin_ia32_pcomgeq (v2di, v2di)
8618 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
8619 v4si __builtin_ia32_pcomgeud (v4si, v4si)
8620 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
8621 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
8622 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
8623 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
8624 v4si __builtin_ia32_pcomgtd (v4si, v4si)
8625 v2di __builtin_ia32_pcomgtq (v2di, v2di)
8626 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
8627 v4si __builtin_ia32_pcomgtud (v4si, v4si)
8628 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
8629 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
8630 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
8631 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
8632 v4si __builtin_ia32_pcomled (v4si, v4si)
8633 v2di __builtin_ia32_pcomleq (v2di, v2di)
8634 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
8635 v4si __builtin_ia32_pcomleud (v4si, v4si)
8636 v2di __builtin_ia32_pcomleuq (v2di, v2di)
8637 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
8638 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
8639 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
8640 v4si __builtin_ia32_pcomltd (v4si, v4si)
8641 v2di __builtin_ia32_pcomltq (v2di, v2di)
8642 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
8643 v4si __builtin_ia32_pcomltud (v4si, v4si)
8644 v2di __builtin_ia32_pcomltuq (v2di, v2di)
8645 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
8646 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
8647 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
8648 v4si __builtin_ia32_pcomned (v4si, v4si)
8649 v2di __builtin_ia32_pcomneq (v2di, v2di)
8650 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
8651 v4si __builtin_ia32_pcomneud (v4si, v4si)
8652 v2di __builtin_ia32_pcomneuq (v2di, v2di)
8653 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
8654 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
8655 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
8656 v4si __builtin_ia32_pcomtrued (v4si, v4si)
8657 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
8658 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
8659 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
8660 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
8661 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
8662 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
8663 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
8664 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
8665 v4si __builtin_ia32_phaddbd (v16qi)
8666 v2di __builtin_ia32_phaddbq (v16qi)
8667 v8hi __builtin_ia32_phaddbw (v16qi)
8668 v2di __builtin_ia32_phadddq (v4si)
8669 v4si __builtin_ia32_phaddubd (v16qi)
8670 v2di __builtin_ia32_phaddubq (v16qi)
8671 v8hi __builtin_ia32_phaddubw (v16qi)
8672 v2di __builtin_ia32_phaddudq (v4si)
8673 v4si __builtin_ia32_phadduwd (v8hi)
8674 v2di __builtin_ia32_phadduwq (v8hi)
8675 v4si __builtin_ia32_phaddwd (v8hi)
8676 v2di __builtin_ia32_phaddwq (v8hi)
8677 v8hi __builtin_ia32_phsubbw (v16qi)
8678 v2di __builtin_ia32_phsubdq (v4si)
8679 v4si __builtin_ia32_phsubwd (v8hi)
8680 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
8681 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
8682 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
8683 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
8684 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
8685 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
8686 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
8687 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
8688 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
8689 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
8690 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
8691 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
8692 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
8693 v16qi __builtin_ia32_protb (v16qi, v16qi)
8694 v4si __builtin_ia32_protd (v4si, v4si)
8695 v2di __builtin_ia32_protq (v2di, v2di)
8696 v8hi __builtin_ia32_protw (v8hi, v8hi)
8697 v16qi __builtin_ia32_pshab (v16qi, v16qi)
8698 v4si __builtin_ia32_pshad (v4si, v4si)
8699 v2di __builtin_ia32_pshaq (v2di, v2di)
8700 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
8701 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
8702 v4si __builtin_ia32_pshld (v4si, v4si)
8703 v2di __builtin_ia32_pshlq (v2di, v2di)
8704 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
8707 The following builtin-in functions are available when @option{-msse5}
8708 is used. The second argument must be an integer constant and generate
8709 the machine instruction that is part of the name with the @samp{_imm}
8713 v16qi __builtin_ia32_protb_imm (v16qi, int)
8714 v4si __builtin_ia32_protd_imm (v4si, int)
8715 v2di __builtin_ia32_protq_imm (v2di, int)
8716 v8hi __builtin_ia32_protw_imm (v8hi, int)
8719 The following built-in functions are available when @option{-m3dnow} is used.
8720 All of them generate the machine instruction that is part of the name.
8723 void __builtin_ia32_femms (void)
8724 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
8725 v2si __builtin_ia32_pf2id (v2sf)
8726 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
8727 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
8728 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
8729 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
8730 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
8731 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
8732 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
8733 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
8734 v2sf __builtin_ia32_pfrcp (v2sf)
8735 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
8736 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
8737 v2sf __builtin_ia32_pfrsqrt (v2sf)
8738 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
8739 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
8740 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
8741 v2sf __builtin_ia32_pi2fd (v2si)
8742 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
8745 The following built-in functions are available when both @option{-m3dnow}
8746 and @option{-march=athlon} are used. All of them generate the machine
8747 instruction that is part of the name.
8750 v2si __builtin_ia32_pf2iw (v2sf)
8751 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
8752 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
8753 v2sf __builtin_ia32_pi2fw (v2si)
8754 v2sf __builtin_ia32_pswapdsf (v2sf)
8755 v2si __builtin_ia32_pswapdsi (v2si)
8758 @node MIPS DSP Built-in Functions
8759 @subsection MIPS DSP Built-in Functions
8761 The MIPS DSP Application-Specific Extension (ASE) includes new
8762 instructions that are designed to improve the performance of DSP and
8763 media applications. It provides instructions that operate on packed
8764 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
8766 GCC supports MIPS DSP operations using both the generic
8767 vector extensions (@pxref{Vector Extensions}) and a collection of
8768 MIPS-specific built-in functions. Both kinds of support are
8769 enabled by the @option{-mdsp} command-line option.
8771 Revision 2 of the ASE was introduced in the second half of 2006.
8772 This revision adds extra instructions to the original ASE, but is
8773 otherwise backwards-compatible with it. You can select revision 2
8774 using the command-line option @option{-mdspr2}; this option implies
8777 The SCOUNT and POS bits of the DSP control register are global. The
8778 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
8779 POS bits. During optimization, the compiler will not delete these
8780 instructions and it will not delete calls to functions containing
8783 At present, GCC only provides support for operations on 32-bit
8784 vectors. The vector type associated with 8-bit integer data is
8785 usually called @code{v4i8}, the vector type associated with Q7
8786 is usually called @code{v4q7}, the vector type associated with 16-bit
8787 integer data is usually called @code{v2i16}, and the vector type
8788 associated with Q15 is usually called @code{v2q15}. They can be
8789 defined in C as follows:
8792 typedef signed char v4i8 __attribute__ ((vector_size(4)));
8793 typedef signed char v4q7 __attribute__ ((vector_size(4)));
8794 typedef short v2i16 __attribute__ ((vector_size(4)));
8795 typedef short v2q15 __attribute__ ((vector_size(4)));
8798 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
8799 initialized in the same way as aggregates. For example:
8802 v4i8 a = @{1, 2, 3, 4@};
8804 b = (v4i8) @{5, 6, 7, 8@};
8806 v2q15 c = @{0x0fcb, 0x3a75@};
8808 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
8811 @emph{Note:} The CPU's endianness determines the order in which values
8812 are packed. On little-endian targets, the first value is the least
8813 significant and the last value is the most significant. The opposite
8814 order applies to big-endian targets. For example, the code above will
8815 set the lowest byte of @code{a} to @code{1} on little-endian targets
8816 and @code{4} on big-endian targets.
8818 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
8819 representation. As shown in this example, the integer representation
8820 of a Q7 value can be obtained by multiplying the fractional value by
8821 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
8822 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
8825 The table below lists the @code{v4i8} and @code{v2q15} operations for which
8826 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
8827 and @code{c} and @code{d} are @code{v2q15} values.
8829 @multitable @columnfractions .50 .50
8830 @item C code @tab MIPS instruction
8831 @item @code{a + b} @tab @code{addu.qb}
8832 @item @code{c + d} @tab @code{addq.ph}
8833 @item @code{a - b} @tab @code{subu.qb}
8834 @item @code{c - d} @tab @code{subq.ph}
8837 The table below lists the @code{v2i16} operation for which
8838 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
8839 @code{v2i16} values.
8841 @multitable @columnfractions .50 .50
8842 @item C code @tab MIPS instruction
8843 @item @code{e * f} @tab @code{mul.ph}
8846 It is easier to describe the DSP built-in functions if we first define
8847 the following types:
8852 typedef unsigned int ui32;
8853 typedef long long a64;
8856 @code{q31} and @code{i32} are actually the same as @code{int}, but we
8857 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
8858 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
8859 @code{long long}, but we use @code{a64} to indicate values that will
8860 be placed in one of the four DSP accumulators (@code{$ac0},
8861 @code{$ac1}, @code{$ac2} or @code{$ac3}).
8863 Also, some built-in functions prefer or require immediate numbers as
8864 parameters, because the corresponding DSP instructions accept both immediate
8865 numbers and register operands, or accept immediate numbers only. The
8866 immediate parameters are listed as follows.
8875 imm_n32_31: -32 to 31.
8876 imm_n512_511: -512 to 511.
8879 The following built-in functions map directly to a particular MIPS DSP
8880 instruction. Please refer to the architecture specification
8881 for details on what each instruction does.
8884 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
8885 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
8886 q31 __builtin_mips_addq_s_w (q31, q31)
8887 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
8888 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
8889 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
8890 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
8891 q31 __builtin_mips_subq_s_w (q31, q31)
8892 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
8893 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
8894 i32 __builtin_mips_addsc (i32, i32)
8895 i32 __builtin_mips_addwc (i32, i32)
8896 i32 __builtin_mips_modsub (i32, i32)
8897 i32 __builtin_mips_raddu_w_qb (v4i8)
8898 v2q15 __builtin_mips_absq_s_ph (v2q15)
8899 q31 __builtin_mips_absq_s_w (q31)
8900 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
8901 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
8902 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
8903 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
8904 q31 __builtin_mips_preceq_w_phl (v2q15)
8905 q31 __builtin_mips_preceq_w_phr (v2q15)
8906 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
8907 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
8908 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
8909 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
8910 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
8911 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
8912 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
8913 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
8914 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
8915 v4i8 __builtin_mips_shll_qb (v4i8, i32)
8916 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
8917 v2q15 __builtin_mips_shll_ph (v2q15, i32)
8918 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
8919 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
8920 q31 __builtin_mips_shll_s_w (q31, imm0_31)
8921 q31 __builtin_mips_shll_s_w (q31, i32)
8922 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
8923 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
8924 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
8925 v2q15 __builtin_mips_shra_ph (v2q15, i32)
8926 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
8927 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
8928 q31 __builtin_mips_shra_r_w (q31, imm0_31)
8929 q31 __builtin_mips_shra_r_w (q31, i32)
8930 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
8931 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
8932 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
8933 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
8934 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
8935 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
8936 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
8937 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
8938 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
8939 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
8940 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
8941 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
8942 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
8943 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
8944 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
8945 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
8946 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
8947 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
8948 i32 __builtin_mips_bitrev (i32)
8949 i32 __builtin_mips_insv (i32, i32)
8950 v4i8 __builtin_mips_repl_qb (imm0_255)
8951 v4i8 __builtin_mips_repl_qb (i32)
8952 v2q15 __builtin_mips_repl_ph (imm_n512_511)
8953 v2q15 __builtin_mips_repl_ph (i32)
8954 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
8955 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
8956 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
8957 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
8958 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
8959 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
8960 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
8961 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
8962 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
8963 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
8964 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
8965 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
8966 i32 __builtin_mips_extr_w (a64, imm0_31)
8967 i32 __builtin_mips_extr_w (a64, i32)
8968 i32 __builtin_mips_extr_r_w (a64, imm0_31)
8969 i32 __builtin_mips_extr_s_h (a64, i32)
8970 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
8971 i32 __builtin_mips_extr_rs_w (a64, i32)
8972 i32 __builtin_mips_extr_s_h (a64, imm0_31)
8973 i32 __builtin_mips_extr_r_w (a64, i32)
8974 i32 __builtin_mips_extp (a64, imm0_31)
8975 i32 __builtin_mips_extp (a64, i32)
8976 i32 __builtin_mips_extpdp (a64, imm0_31)
8977 i32 __builtin_mips_extpdp (a64, i32)
8978 a64 __builtin_mips_shilo (a64, imm_n32_31)
8979 a64 __builtin_mips_shilo (a64, i32)
8980 a64 __builtin_mips_mthlip (a64, i32)
8981 void __builtin_mips_wrdsp (i32, imm0_63)
8982 i32 __builtin_mips_rddsp (imm0_63)
8983 i32 __builtin_mips_lbux (void *, i32)
8984 i32 __builtin_mips_lhx (void *, i32)
8985 i32 __builtin_mips_lwx (void *, i32)
8986 i32 __builtin_mips_bposge32 (void)
8989 The following built-in functions map directly to a particular MIPS DSP REV 2
8990 instruction. Please refer to the architecture specification
8991 for details on what each instruction does.
8994 v4q7 __builtin_mips_absq_s_qb (v4q7);
8995 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
8996 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
8997 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
8998 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
8999 i32 __builtin_mips_append (i32, i32, imm0_31);
9000 i32 __builtin_mips_balign (i32, i32, imm0_3);
9001 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
9002 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
9003 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
9004 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
9005 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
9006 a64 __builtin_mips_madd (a64, i32, i32);
9007 a64 __builtin_mips_maddu (a64, ui32, ui32);
9008 a64 __builtin_mips_msub (a64, i32, i32);
9009 a64 __builtin_mips_msubu (a64, ui32, ui32);
9010 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
9011 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
9012 q31 __builtin_mips_mulq_rs_w (q31, q31);
9013 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
9014 q31 __builtin_mips_mulq_s_w (q31, q31);
9015 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
9016 a64 __builtin_mips_mult (i32, i32);
9017 a64 __builtin_mips_multu (ui32, ui32);
9018 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
9019 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
9020 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
9021 i32 __builtin_mips_prepend (i32, i32, imm0_31);
9022 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
9023 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
9024 v4i8 __builtin_mips_shra_qb (v4i8, i32);
9025 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
9026 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
9027 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
9028 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
9029 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
9030 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
9031 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
9032 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
9033 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
9034 q31 __builtin_mips_addqh_w (q31, q31);
9035 q31 __builtin_mips_addqh_r_w (q31, q31);
9036 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
9037 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
9038 q31 __builtin_mips_subqh_w (q31, q31);
9039 q31 __builtin_mips_subqh_r_w (q31, q31);
9040 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
9041 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
9042 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
9043 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
9044 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
9045 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
9049 @node MIPS Paired-Single Support
9050 @subsection MIPS Paired-Single Support
9052 The MIPS64 architecture includes a number of instructions that
9053 operate on pairs of single-precision floating-point values.
9054 Each pair is packed into a 64-bit floating-point register,
9055 with one element being designated the ``upper half'' and
9056 the other being designated the ``lower half''.
9058 GCC supports paired-single operations using both the generic
9059 vector extensions (@pxref{Vector Extensions}) and a collection of
9060 MIPS-specific built-in functions. Both kinds of support are
9061 enabled by the @option{-mpaired-single} command-line option.
9063 The vector type associated with paired-single values is usually
9064 called @code{v2sf}. It can be defined in C as follows:
9067 typedef float v2sf __attribute__ ((vector_size (8)));
9070 @code{v2sf} values are initialized in the same way as aggregates.
9074 v2sf a = @{1.5, 9.1@};
9077 b = (v2sf) @{e, f@};
9080 @emph{Note:} The CPU's endianness determines which value is stored in
9081 the upper half of a register and which value is stored in the lower half.
9082 On little-endian targets, the first value is the lower one and the second
9083 value is the upper one. The opposite order applies to big-endian targets.
9084 For example, the code above will set the lower half of @code{a} to
9085 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
9087 @node MIPS Loongson Built-in Functions
9088 @subsection MIPS Loongson Built-in Functions
9090 GCC provides intrinsics to access the SIMD instructions provided by the
9091 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
9092 available after inclusion of the @code{loongson.h} header file,
9093 operate on the following 64-bit vector types:
9096 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
9097 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
9098 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
9099 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
9100 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
9101 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
9104 The intrinsics provided are listed below; each is named after the
9105 machine instruction to which it corresponds, with suffixes added as
9106 appropriate to distinguish intrinsics that expand to the same machine
9107 instruction yet have different argument types. Refer to the architecture
9108 documentation for a description of the functionality of each
9112 int16x4_t packsswh (int32x2_t s, int32x2_t t);
9113 int8x8_t packsshb (int16x4_t s, int16x4_t t);
9114 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
9115 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
9116 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
9117 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
9118 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
9119 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
9120 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
9121 uint64_t paddd_u (uint64_t s, uint64_t t);
9122 int64_t paddd_s (int64_t s, int64_t t);
9123 int16x4_t paddsh (int16x4_t s, int16x4_t t);
9124 int8x8_t paddsb (int8x8_t s, int8x8_t t);
9125 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
9126 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
9127 uint64_t pandn_ud (uint64_t s, uint64_t t);
9128 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
9129 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
9130 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
9131 int64_t pandn_sd (int64_t s, int64_t t);
9132 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
9133 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
9134 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
9135 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
9136 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
9137 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
9138 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
9139 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
9140 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
9141 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
9142 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
9143 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
9144 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
9145 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
9146 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
9147 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
9148 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
9149 uint16x4_t pextrh_u (uint16x4_t s, int field);
9150 int16x4_t pextrh_s (int16x4_t s, int field);
9151 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
9152 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
9153 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
9154 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
9155 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
9156 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
9157 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
9158 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
9159 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
9160 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
9161 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
9162 int16x4_t pminsh (int16x4_t s, int16x4_t t);
9163 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
9164 uint8x8_t pmovmskb_u (uint8x8_t s);
9165 int8x8_t pmovmskb_s (int8x8_t s);
9166 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
9167 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
9168 int16x4_t pmullh (int16x4_t s, int16x4_t t);
9169 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
9170 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
9171 uint16x4_t biadd (uint8x8_t s);
9172 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
9173 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
9174 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
9175 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
9176 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
9177 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
9178 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
9179 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
9180 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
9181 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
9182 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
9183 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
9184 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
9185 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
9186 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
9187 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
9188 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
9189 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
9190 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
9191 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
9192 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
9193 uint64_t psubd_u (uint64_t s, uint64_t t);
9194 int64_t psubd_s (int64_t s, int64_t t);
9195 int16x4_t psubsh (int16x4_t s, int16x4_t t);
9196 int8x8_t psubsb (int8x8_t s, int8x8_t t);
9197 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
9198 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
9199 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
9200 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
9201 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
9202 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
9203 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
9204 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
9205 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
9206 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
9207 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
9208 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
9209 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
9210 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
9214 * Paired-Single Arithmetic::
9215 * Paired-Single Built-in Functions::
9216 * MIPS-3D Built-in Functions::
9219 @node Paired-Single Arithmetic
9220 @subsubsection Paired-Single Arithmetic
9222 The table below lists the @code{v2sf} operations for which hardware
9223 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
9224 values and @code{x} is an integral value.
9226 @multitable @columnfractions .50 .50
9227 @item C code @tab MIPS instruction
9228 @item @code{a + b} @tab @code{add.ps}
9229 @item @code{a - b} @tab @code{sub.ps}
9230 @item @code{-a} @tab @code{neg.ps}
9231 @item @code{a * b} @tab @code{mul.ps}
9232 @item @code{a * b + c} @tab @code{madd.ps}
9233 @item @code{a * b - c} @tab @code{msub.ps}
9234 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
9235 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
9236 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
9239 Note that the multiply-accumulate instructions can be disabled
9240 using the command-line option @code{-mno-fused-madd}.
9242 @node Paired-Single Built-in Functions
9243 @subsubsection Paired-Single Built-in Functions
9245 The following paired-single functions map directly to a particular
9246 MIPS instruction. Please refer to the architecture specification
9247 for details on what each instruction does.
9250 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
9251 Pair lower lower (@code{pll.ps}).
9253 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
9254 Pair upper lower (@code{pul.ps}).
9256 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
9257 Pair lower upper (@code{plu.ps}).
9259 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
9260 Pair upper upper (@code{puu.ps}).
9262 @item v2sf __builtin_mips_cvt_ps_s (float, float)
9263 Convert pair to paired single (@code{cvt.ps.s}).
9265 @item float __builtin_mips_cvt_s_pl (v2sf)
9266 Convert pair lower to single (@code{cvt.s.pl}).
9268 @item float __builtin_mips_cvt_s_pu (v2sf)
9269 Convert pair upper to single (@code{cvt.s.pu}).
9271 @item v2sf __builtin_mips_abs_ps (v2sf)
9272 Absolute value (@code{abs.ps}).
9274 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
9275 Align variable (@code{alnv.ps}).
9277 @emph{Note:} The value of the third parameter must be 0 or 4
9278 modulo 8, otherwise the result will be unpredictable. Please read the
9279 instruction description for details.
9282 The following multi-instruction functions are also available.
9283 In each case, @var{cond} can be any of the 16 floating-point conditions:
9284 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9285 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
9286 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9289 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9290 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9291 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
9292 @code{movt.ps}/@code{movf.ps}).
9294 The @code{movt} functions return the value @var{x} computed by:
9297 c.@var{cond}.ps @var{cc},@var{a},@var{b}
9298 mov.ps @var{x},@var{c}
9299 movt.ps @var{x},@var{d},@var{cc}
9302 The @code{movf} functions are similar but use @code{movf.ps} instead
9305 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9306 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9307 Comparison of two paired-single values (@code{c.@var{cond}.ps},
9308 @code{bc1t}/@code{bc1f}).
9310 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9311 and return either the upper or lower half of the result. For example:
9315 if (__builtin_mips_upper_c_eq_ps (a, b))
9316 upper_halves_are_equal ();
9318 upper_halves_are_unequal ();
9320 if (__builtin_mips_lower_c_eq_ps (a, b))
9321 lower_halves_are_equal ();
9323 lower_halves_are_unequal ();
9327 @node MIPS-3D Built-in Functions
9328 @subsubsection MIPS-3D Built-in Functions
9330 The MIPS-3D Application-Specific Extension (ASE) includes additional
9331 paired-single instructions that are designed to improve the performance
9332 of 3D graphics operations. Support for these instructions is controlled
9333 by the @option{-mips3d} command-line option.
9335 The functions listed below map directly to a particular MIPS-3D
9336 instruction. Please refer to the architecture specification for
9337 more details on what each instruction does.
9340 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
9341 Reduction add (@code{addr.ps}).
9343 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
9344 Reduction multiply (@code{mulr.ps}).
9346 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
9347 Convert paired single to paired word (@code{cvt.pw.ps}).
9349 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
9350 Convert paired word to paired single (@code{cvt.ps.pw}).
9352 @item float __builtin_mips_recip1_s (float)
9353 @itemx double __builtin_mips_recip1_d (double)
9354 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
9355 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
9357 @item float __builtin_mips_recip2_s (float, float)
9358 @itemx double __builtin_mips_recip2_d (double, double)
9359 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
9360 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
9362 @item float __builtin_mips_rsqrt1_s (float)
9363 @itemx double __builtin_mips_rsqrt1_d (double)
9364 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
9365 Reduced precision reciprocal square root (sequence step 1)
9366 (@code{rsqrt1.@var{fmt}}).
9368 @item float __builtin_mips_rsqrt2_s (float, float)
9369 @itemx double __builtin_mips_rsqrt2_d (double, double)
9370 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
9371 Reduced precision reciprocal square root (sequence step 2)
9372 (@code{rsqrt2.@var{fmt}}).
9375 The following multi-instruction functions are also available.
9376 In each case, @var{cond} can be any of the 16 floating-point conditions:
9377 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9378 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
9379 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9382 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
9383 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
9384 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
9385 @code{bc1t}/@code{bc1f}).
9387 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
9388 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
9393 if (__builtin_mips_cabs_eq_s (a, b))
9399 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9400 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9401 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
9402 @code{bc1t}/@code{bc1f}).
9404 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
9405 and return either the upper or lower half of the result. For example:
9409 if (__builtin_mips_upper_cabs_eq_ps (a, b))
9410 upper_halves_are_equal ();
9412 upper_halves_are_unequal ();
9414 if (__builtin_mips_lower_cabs_eq_ps (a, b))
9415 lower_halves_are_equal ();
9417 lower_halves_are_unequal ();
9420 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9421 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9422 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
9423 @code{movt.ps}/@code{movf.ps}).
9425 The @code{movt} functions return the value @var{x} computed by:
9428 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
9429 mov.ps @var{x},@var{c}
9430 movt.ps @var{x},@var{d},@var{cc}
9433 The @code{movf} functions are similar but use @code{movf.ps} instead
9436 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9437 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9438 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9439 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9440 Comparison of two paired-single values
9441 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9442 @code{bc1any2t}/@code{bc1any2f}).
9444 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9445 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
9446 result is true and the @code{all} forms return true if both results are true.
9451 if (__builtin_mips_any_c_eq_ps (a, b))
9456 if (__builtin_mips_all_c_eq_ps (a, b))
9462 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9463 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9464 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9465 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9466 Comparison of four paired-single values
9467 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9468 @code{bc1any4t}/@code{bc1any4f}).
9470 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
9471 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
9472 The @code{any} forms return true if any of the four results are true
9473 and the @code{all} forms return true if all four results are true.
9478 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
9483 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
9490 @node picoChip Built-in Functions
9491 @subsection picoChip Built-in Functions
9493 GCC provides an interface to selected machine instructions from the
9494 picoChip instruction set.
9497 @item int __builtin_sbc (int @var{value})
9498 Sign bit count. Return the number of consecutive bits in @var{value}
9499 which have the same value as the sign-bit. The result is the number of
9500 leading sign bits minus one, giving the number of redundant sign bits in
9503 @item int __builtin_byteswap (int @var{value})
9504 Byte swap. Return the result of swapping the upper and lower bytes of
9507 @item int __builtin_brev (int @var{value})
9508 Bit reversal. Return the result of reversing the bits in
9509 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
9512 @item int __builtin_adds (int @var{x}, int @var{y})
9513 Saturating addition. Return the result of adding @var{x} and @var{y},
9514 storing the value 32767 if the result overflows.
9516 @item int __builtin_subs (int @var{x}, int @var{y})
9517 Saturating subtraction. Return the result of subtracting @var{y} from
9518 @var{x}, storing the value -32768 if the result overflows.
9520 @item void __builtin_halt (void)
9521 Halt. The processor will stop execution. This built-in is useful for
9522 implementing assertions.
9526 @node Other MIPS Built-in Functions
9527 @subsection Other MIPS Built-in Functions
9529 GCC provides other MIPS-specific built-in functions:
9532 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
9533 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
9534 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
9535 when this function is available.
9538 @node PowerPC AltiVec Built-in Functions
9539 @subsection PowerPC AltiVec Built-in Functions
9541 GCC provides an interface for the PowerPC family of processors to access
9542 the AltiVec operations described in Motorola's AltiVec Programming
9543 Interface Manual. The interface is made available by including
9544 @code{<altivec.h>} and using @option{-maltivec} and
9545 @option{-mabi=altivec}. The interface supports the following vector
9549 vector unsigned char
9553 vector unsigned short
9564 GCC's implementation of the high-level language interface available from
9565 C and C++ code differs from Motorola's documentation in several ways.
9570 A vector constant is a list of constant expressions within curly braces.
9573 A vector initializer requires no cast if the vector constant is of the
9574 same type as the variable it is initializing.
9577 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9578 vector type is the default signedness of the base type. The default
9579 varies depending on the operating system, so a portable program should
9580 always specify the signedness.
9583 Compiling with @option{-maltivec} adds keywords @code{__vector},
9584 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
9585 @code{bool}. When compiling ISO C, the context-sensitive substitution
9586 of the keywords @code{vector}, @code{pixel} and @code{bool} is
9587 disabled. To use them, you must include @code{<altivec.h>} instead.
9590 GCC allows using a @code{typedef} name as the type specifier for a
9594 For C, overloaded functions are implemented with macros so the following
9598 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9601 Since @code{vec_add} is a macro, the vector constant in the example
9602 is treated as four separate arguments. Wrap the entire argument in
9603 parentheses for this to work.
9606 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
9607 Internally, GCC uses built-in functions to achieve the functionality in
9608 the aforementioned header file, but they are not supported and are
9609 subject to change without notice.
9611 The following interfaces are supported for the generic and specific
9612 AltiVec operations and the AltiVec predicates. In cases where there
9613 is a direct mapping between generic and specific operations, only the
9614 generic names are shown here, although the specific operations can also
9617 Arguments that are documented as @code{const int} require literal
9618 integral values within the range required for that operation.
9621 vector signed char vec_abs (vector signed char);
9622 vector signed short vec_abs (vector signed short);
9623 vector signed int vec_abs (vector signed int);
9624 vector float vec_abs (vector float);
9626 vector signed char vec_abss (vector signed char);
9627 vector signed short vec_abss (vector signed short);
9628 vector signed int vec_abss (vector signed int);
9630 vector signed char vec_add (vector bool char, vector signed char);
9631 vector signed char vec_add (vector signed char, vector bool char);
9632 vector signed char vec_add (vector signed char, vector signed char);
9633 vector unsigned char vec_add (vector bool char, vector unsigned char);
9634 vector unsigned char vec_add (vector unsigned char, vector bool char);
9635 vector unsigned char vec_add (vector unsigned char,
9636 vector unsigned char);
9637 vector signed short vec_add (vector bool short, vector signed short);
9638 vector signed short vec_add (vector signed short, vector bool short);
9639 vector signed short vec_add (vector signed short, vector signed short);
9640 vector unsigned short vec_add (vector bool short,
9641 vector unsigned short);
9642 vector unsigned short vec_add (vector unsigned short,
9644 vector unsigned short vec_add (vector unsigned short,
9645 vector unsigned short);
9646 vector signed int vec_add (vector bool int, vector signed int);
9647 vector signed int vec_add (vector signed int, vector bool int);
9648 vector signed int vec_add (vector signed int, vector signed int);
9649 vector unsigned int vec_add (vector bool int, vector unsigned int);
9650 vector unsigned int vec_add (vector unsigned int, vector bool int);
9651 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
9652 vector float vec_add (vector float, vector float);
9654 vector float vec_vaddfp (vector float, vector float);
9656 vector signed int vec_vadduwm (vector bool int, vector signed int);
9657 vector signed int vec_vadduwm (vector signed int, vector bool int);
9658 vector signed int vec_vadduwm (vector signed int, vector signed int);
9659 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
9660 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
9661 vector unsigned int vec_vadduwm (vector unsigned int,
9662 vector unsigned int);
9664 vector signed short vec_vadduhm (vector bool short,
9665 vector signed short);
9666 vector signed short vec_vadduhm (vector signed short,
9668 vector signed short vec_vadduhm (vector signed short,
9669 vector signed short);
9670 vector unsigned short vec_vadduhm (vector bool short,
9671 vector unsigned short);
9672 vector unsigned short vec_vadduhm (vector unsigned short,
9674 vector unsigned short vec_vadduhm (vector unsigned short,
9675 vector unsigned short);
9677 vector signed char vec_vaddubm (vector bool char, vector signed char);
9678 vector signed char vec_vaddubm (vector signed char, vector bool char);
9679 vector signed char vec_vaddubm (vector signed char, vector signed char);
9680 vector unsigned char vec_vaddubm (vector bool char,
9681 vector unsigned char);
9682 vector unsigned char vec_vaddubm (vector unsigned char,
9684 vector unsigned char vec_vaddubm (vector unsigned char,
9685 vector unsigned char);
9687 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
9689 vector unsigned char vec_adds (vector bool char, vector unsigned char);
9690 vector unsigned char vec_adds (vector unsigned char, vector bool char);
9691 vector unsigned char vec_adds (vector unsigned char,
9692 vector unsigned char);
9693 vector signed char vec_adds (vector bool char, vector signed char);
9694 vector signed char vec_adds (vector signed char, vector bool char);
9695 vector signed char vec_adds (vector signed char, vector signed char);
9696 vector unsigned short vec_adds (vector bool short,
9697 vector unsigned short);
9698 vector unsigned short vec_adds (vector unsigned short,
9700 vector unsigned short vec_adds (vector unsigned short,
9701 vector unsigned short);
9702 vector signed short vec_adds (vector bool short, vector signed short);
9703 vector signed short vec_adds (vector signed short, vector bool short);
9704 vector signed short vec_adds (vector signed short, vector signed short);
9705 vector unsigned int vec_adds (vector bool int, vector unsigned int);
9706 vector unsigned int vec_adds (vector unsigned int, vector bool int);
9707 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
9708 vector signed int vec_adds (vector bool int, vector signed int);
9709 vector signed int vec_adds (vector signed int, vector bool int);
9710 vector signed int vec_adds (vector signed int, vector signed int);
9712 vector signed int vec_vaddsws (vector bool int, vector signed int);
9713 vector signed int vec_vaddsws (vector signed int, vector bool int);
9714 vector signed int vec_vaddsws (vector signed int, vector signed int);
9716 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
9717 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
9718 vector unsigned int vec_vadduws (vector unsigned int,
9719 vector unsigned int);
9721 vector signed short vec_vaddshs (vector bool short,
9722 vector signed short);
9723 vector signed short vec_vaddshs (vector signed short,
9725 vector signed short vec_vaddshs (vector signed short,
9726 vector signed short);
9728 vector unsigned short vec_vadduhs (vector bool short,
9729 vector unsigned short);
9730 vector unsigned short vec_vadduhs (vector unsigned short,
9732 vector unsigned short vec_vadduhs (vector unsigned short,
9733 vector unsigned short);
9735 vector signed char vec_vaddsbs (vector bool char, vector signed char);
9736 vector signed char vec_vaddsbs (vector signed char, vector bool char);
9737 vector signed char vec_vaddsbs (vector signed char, vector signed char);
9739 vector unsigned char vec_vaddubs (vector bool char,
9740 vector unsigned char);
9741 vector unsigned char vec_vaddubs (vector unsigned char,
9743 vector unsigned char vec_vaddubs (vector unsigned char,
9744 vector unsigned char);
9746 vector float vec_and (vector float, vector float);
9747 vector float vec_and (vector float, vector bool int);
9748 vector float vec_and (vector bool int, vector float);
9749 vector bool int vec_and (vector bool int, vector bool int);
9750 vector signed int vec_and (vector bool int, vector signed int);
9751 vector signed int vec_and (vector signed int, vector bool int);
9752 vector signed int vec_and (vector signed int, vector signed int);
9753 vector unsigned int vec_and (vector bool int, vector unsigned int);
9754 vector unsigned int vec_and (vector unsigned int, vector bool int);
9755 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
9756 vector bool short vec_and (vector bool short, vector bool short);
9757 vector signed short vec_and (vector bool short, vector signed short);
9758 vector signed short vec_and (vector signed short, vector bool short);
9759 vector signed short vec_and (vector signed short, vector signed short);
9760 vector unsigned short vec_and (vector bool short,
9761 vector unsigned short);
9762 vector unsigned short vec_and (vector unsigned short,
9764 vector unsigned short vec_and (vector unsigned short,
9765 vector unsigned short);
9766 vector signed char vec_and (vector bool char, vector signed char);
9767 vector bool char vec_and (vector bool char, vector bool char);
9768 vector signed char vec_and (vector signed char, vector bool char);
9769 vector signed char vec_and (vector signed char, vector signed char);
9770 vector unsigned char vec_and (vector bool char, vector unsigned char);
9771 vector unsigned char vec_and (vector unsigned char, vector bool char);
9772 vector unsigned char vec_and (vector unsigned char,
9773 vector unsigned char);
9775 vector float vec_andc (vector float, vector float);
9776 vector float vec_andc (vector float, vector bool int);
9777 vector float vec_andc (vector bool int, vector float);
9778 vector bool int vec_andc (vector bool int, vector bool int);
9779 vector signed int vec_andc (vector bool int, vector signed int);
9780 vector signed int vec_andc (vector signed int, vector bool int);
9781 vector signed int vec_andc (vector signed int, vector signed int);
9782 vector unsigned int vec_andc (vector bool int, vector unsigned int);
9783 vector unsigned int vec_andc (vector unsigned int, vector bool int);
9784 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
9785 vector bool short vec_andc (vector bool short, vector bool short);
9786 vector signed short vec_andc (vector bool short, vector signed short);
9787 vector signed short vec_andc (vector signed short, vector bool short);
9788 vector signed short vec_andc (vector signed short, vector signed short);
9789 vector unsigned short vec_andc (vector bool short,
9790 vector unsigned short);
9791 vector unsigned short vec_andc (vector unsigned short,
9793 vector unsigned short vec_andc (vector unsigned short,
9794 vector unsigned short);
9795 vector signed char vec_andc (vector bool char, vector signed char);
9796 vector bool char vec_andc (vector bool char, vector bool char);
9797 vector signed char vec_andc (vector signed char, vector bool char);
9798 vector signed char vec_andc (vector signed char, vector signed char);
9799 vector unsigned char vec_andc (vector bool char, vector unsigned char);
9800 vector unsigned char vec_andc (vector unsigned char, vector bool char);
9801 vector unsigned char vec_andc (vector unsigned char,
9802 vector unsigned char);
9804 vector unsigned char vec_avg (vector unsigned char,
9805 vector unsigned char);
9806 vector signed char vec_avg (vector signed char, vector signed char);
9807 vector unsigned short vec_avg (vector unsigned short,
9808 vector unsigned short);
9809 vector signed short vec_avg (vector signed short, vector signed short);
9810 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
9811 vector signed int vec_avg (vector signed int, vector signed int);
9813 vector signed int vec_vavgsw (vector signed int, vector signed int);
9815 vector unsigned int vec_vavguw (vector unsigned int,
9816 vector unsigned int);
9818 vector signed short vec_vavgsh (vector signed short,
9819 vector signed short);
9821 vector unsigned short vec_vavguh (vector unsigned short,
9822 vector unsigned short);
9824 vector signed char vec_vavgsb (vector signed char, vector signed char);
9826 vector unsigned char vec_vavgub (vector unsigned char,
9827 vector unsigned char);
9829 vector float vec_ceil (vector float);
9831 vector signed int vec_cmpb (vector float, vector float);
9833 vector bool char vec_cmpeq (vector signed char, vector signed char);
9834 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
9835 vector bool short vec_cmpeq (vector signed short, vector signed short);
9836 vector bool short vec_cmpeq (vector unsigned short,
9837 vector unsigned short);
9838 vector bool int vec_cmpeq (vector signed int, vector signed int);
9839 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
9840 vector bool int vec_cmpeq (vector float, vector float);
9842 vector bool int vec_vcmpeqfp (vector float, vector float);
9844 vector bool int vec_vcmpequw (vector signed int, vector signed int);
9845 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
9847 vector bool short vec_vcmpequh (vector signed short,
9848 vector signed short);
9849 vector bool short vec_vcmpequh (vector unsigned short,
9850 vector unsigned short);
9852 vector bool char vec_vcmpequb (vector signed char, vector signed char);
9853 vector bool char vec_vcmpequb (vector unsigned char,
9854 vector unsigned char);
9856 vector bool int vec_cmpge (vector float, vector float);
9858 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
9859 vector bool char vec_cmpgt (vector signed char, vector signed char);
9860 vector bool short vec_cmpgt (vector unsigned short,
9861 vector unsigned short);
9862 vector bool short vec_cmpgt (vector signed short, vector signed short);
9863 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
9864 vector bool int vec_cmpgt (vector signed int, vector signed int);
9865 vector bool int vec_cmpgt (vector float, vector float);
9867 vector bool int vec_vcmpgtfp (vector float, vector float);
9869 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
9871 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
9873 vector bool short vec_vcmpgtsh (vector signed short,
9874 vector signed short);
9876 vector bool short vec_vcmpgtuh (vector unsigned short,
9877 vector unsigned short);
9879 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
9881 vector bool char vec_vcmpgtub (vector unsigned char,
9882 vector unsigned char);
9884 vector bool int vec_cmple (vector float, vector float);
9886 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
9887 vector bool char vec_cmplt (vector signed char, vector signed char);
9888 vector bool short vec_cmplt (vector unsigned short,
9889 vector unsigned short);
9890 vector bool short vec_cmplt (vector signed short, vector signed short);
9891 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
9892 vector bool int vec_cmplt (vector signed int, vector signed int);
9893 vector bool int vec_cmplt (vector float, vector float);
9895 vector float vec_ctf (vector unsigned int, const int);
9896 vector float vec_ctf (vector signed int, const int);
9898 vector float vec_vcfsx (vector signed int, const int);
9900 vector float vec_vcfux (vector unsigned int, const int);
9902 vector signed int vec_cts (vector float, const int);
9904 vector unsigned int vec_ctu (vector float, const int);
9906 void vec_dss (const int);
9908 void vec_dssall (void);
9910 void vec_dst (const vector unsigned char *, int, const int);
9911 void vec_dst (const vector signed char *, int, const int);
9912 void vec_dst (const vector bool char *, int, const int);
9913 void vec_dst (const vector unsigned short *, int, const int);
9914 void vec_dst (const vector signed short *, int, const int);
9915 void vec_dst (const vector bool short *, int, const int);
9916 void vec_dst (const vector pixel *, int, const int);
9917 void vec_dst (const vector unsigned int *, int, const int);
9918 void vec_dst (const vector signed int *, int, const int);
9919 void vec_dst (const vector bool int *, int, const int);
9920 void vec_dst (const vector float *, int, const int);
9921 void vec_dst (const unsigned char *, int, const int);
9922 void vec_dst (const signed char *, int, const int);
9923 void vec_dst (const unsigned short *, int, const int);
9924 void vec_dst (const short *, int, const int);
9925 void vec_dst (const unsigned int *, int, const int);
9926 void vec_dst (const int *, int, const int);
9927 void vec_dst (const unsigned long *, int, const int);
9928 void vec_dst (const long *, int, const int);
9929 void vec_dst (const float *, int, const int);
9931 void vec_dstst (const vector unsigned char *, int, const int);
9932 void vec_dstst (const vector signed char *, int, const int);
9933 void vec_dstst (const vector bool char *, int, const int);
9934 void vec_dstst (const vector unsigned short *, int, const int);
9935 void vec_dstst (const vector signed short *, int, const int);
9936 void vec_dstst (const vector bool short *, int, const int);
9937 void vec_dstst (const vector pixel *, int, const int);
9938 void vec_dstst (const vector unsigned int *, int, const int);
9939 void vec_dstst (const vector signed int *, int, const int);
9940 void vec_dstst (const vector bool int *, int, const int);
9941 void vec_dstst (const vector float *, int, const int);
9942 void vec_dstst (const unsigned char *, int, const int);
9943 void vec_dstst (const signed char *, int, const int);
9944 void vec_dstst (const unsigned short *, int, const int);
9945 void vec_dstst (const short *, int, const int);
9946 void vec_dstst (const unsigned int *, int, const int);
9947 void vec_dstst (const int *, int, const int);
9948 void vec_dstst (const unsigned long *, int, const int);
9949 void vec_dstst (const long *, int, const int);
9950 void vec_dstst (const float *, int, const int);
9952 void vec_dststt (const vector unsigned char *, int, const int);
9953 void vec_dststt (const vector signed char *, int, const int);
9954 void vec_dststt (const vector bool char *, int, const int);
9955 void vec_dststt (const vector unsigned short *, int, const int);
9956 void vec_dststt (const vector signed short *, int, const int);
9957 void vec_dststt (const vector bool short *, int, const int);
9958 void vec_dststt (const vector pixel *, int, const int);
9959 void vec_dststt (const vector unsigned int *, int, const int);
9960 void vec_dststt (const vector signed int *, int, const int);
9961 void vec_dststt (const vector bool int *, int, const int);
9962 void vec_dststt (const vector float *, int, const int);
9963 void vec_dststt (const unsigned char *, int, const int);
9964 void vec_dststt (const signed char *, int, const int);
9965 void vec_dststt (const unsigned short *, int, const int);
9966 void vec_dststt (const short *, int, const int);
9967 void vec_dststt (const unsigned int *, int, const int);
9968 void vec_dststt (const int *, int, const int);
9969 void vec_dststt (const unsigned long *, int, const int);
9970 void vec_dststt (const long *, int, const int);
9971 void vec_dststt (const float *, int, const int);
9973 void vec_dstt (const vector unsigned char *, int, const int);
9974 void vec_dstt (const vector signed char *, int, const int);
9975 void vec_dstt (const vector bool char *, int, const int);
9976 void vec_dstt (const vector unsigned short *, int, const int);
9977 void vec_dstt (const vector signed short *, int, const int);
9978 void vec_dstt (const vector bool short *, int, const int);
9979 void vec_dstt (const vector pixel *, int, const int);
9980 void vec_dstt (const vector unsigned int *, int, const int);
9981 void vec_dstt (const vector signed int *, int, const int);
9982 void vec_dstt (const vector bool int *, int, const int);
9983 void vec_dstt (const vector float *, int, const int);
9984 void vec_dstt (const unsigned char *, int, const int);
9985 void vec_dstt (const signed char *, int, const int);
9986 void vec_dstt (const unsigned short *, int, const int);
9987 void vec_dstt (const short *, int, const int);
9988 void vec_dstt (const unsigned int *, int, const int);
9989 void vec_dstt (const int *, int, const int);
9990 void vec_dstt (const unsigned long *, int, const int);
9991 void vec_dstt (const long *, int, const int);
9992 void vec_dstt (const float *, int, const int);
9994 vector float vec_expte (vector float);
9996 vector float vec_floor (vector float);
9998 vector float vec_ld (int, const vector float *);
9999 vector float vec_ld (int, const float *);
10000 vector bool int vec_ld (int, const vector bool int *);
10001 vector signed int vec_ld (int, const vector signed int *);
10002 vector signed int vec_ld (int, const int *);
10003 vector signed int vec_ld (int, const long *);
10004 vector unsigned int vec_ld (int, const vector unsigned int *);
10005 vector unsigned int vec_ld (int, const unsigned int *);
10006 vector unsigned int vec_ld (int, const unsigned long *);
10007 vector bool short vec_ld (int, const vector bool short *);
10008 vector pixel vec_ld (int, const vector pixel *);
10009 vector signed short vec_ld (int, const vector signed short *);
10010 vector signed short vec_ld (int, const short *);
10011 vector unsigned short vec_ld (int, const vector unsigned short *);
10012 vector unsigned short vec_ld (int, const unsigned short *);
10013 vector bool char vec_ld (int, const vector bool char *);
10014 vector signed char vec_ld (int, const vector signed char *);
10015 vector signed char vec_ld (int, const signed char *);
10016 vector unsigned char vec_ld (int, const vector unsigned char *);
10017 vector unsigned char vec_ld (int, const unsigned char *);
10019 vector signed char vec_lde (int, const signed char *);
10020 vector unsigned char vec_lde (int, const unsigned char *);
10021 vector signed short vec_lde (int, const short *);
10022 vector unsigned short vec_lde (int, const unsigned short *);
10023 vector float vec_lde (int, const float *);
10024 vector signed int vec_lde (int, const int *);
10025 vector unsigned int vec_lde (int, const unsigned int *);
10026 vector signed int vec_lde (int, const long *);
10027 vector unsigned int vec_lde (int, const unsigned long *);
10029 vector float vec_lvewx (int, float *);
10030 vector signed int vec_lvewx (int, int *);
10031 vector unsigned int vec_lvewx (int, unsigned int *);
10032 vector signed int vec_lvewx (int, long *);
10033 vector unsigned int vec_lvewx (int, unsigned long *);
10035 vector signed short vec_lvehx (int, short *);
10036 vector unsigned short vec_lvehx (int, unsigned short *);
10038 vector signed char vec_lvebx (int, char *);
10039 vector unsigned char vec_lvebx (int, unsigned char *);
10041 vector float vec_ldl (int, const vector float *);
10042 vector float vec_ldl (int, const float *);
10043 vector bool int vec_ldl (int, const vector bool int *);
10044 vector signed int vec_ldl (int, const vector signed int *);
10045 vector signed int vec_ldl (int, const int *);
10046 vector signed int vec_ldl (int, const long *);
10047 vector unsigned int vec_ldl (int, const vector unsigned int *);
10048 vector unsigned int vec_ldl (int, const unsigned int *);
10049 vector unsigned int vec_ldl (int, const unsigned long *);
10050 vector bool short vec_ldl (int, const vector bool short *);
10051 vector pixel vec_ldl (int, const vector pixel *);
10052 vector signed short vec_ldl (int, const vector signed short *);
10053 vector signed short vec_ldl (int, const short *);
10054 vector unsigned short vec_ldl (int, const vector unsigned short *);
10055 vector unsigned short vec_ldl (int, const unsigned short *);
10056 vector bool char vec_ldl (int, const vector bool char *);
10057 vector signed char vec_ldl (int, const vector signed char *);
10058 vector signed char vec_ldl (int, const signed char *);
10059 vector unsigned char vec_ldl (int, const vector unsigned char *);
10060 vector unsigned char vec_ldl (int, const unsigned char *);
10062 vector float vec_loge (vector float);
10064 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
10065 vector unsigned char vec_lvsl (int, const volatile signed char *);
10066 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
10067 vector unsigned char vec_lvsl (int, const volatile short *);
10068 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
10069 vector unsigned char vec_lvsl (int, const volatile int *);
10070 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
10071 vector unsigned char vec_lvsl (int, const volatile long *);
10072 vector unsigned char vec_lvsl (int, const volatile float *);
10074 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
10075 vector unsigned char vec_lvsr (int, const volatile signed char *);
10076 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
10077 vector unsigned char vec_lvsr (int, const volatile short *);
10078 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
10079 vector unsigned char vec_lvsr (int, const volatile int *);
10080 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
10081 vector unsigned char vec_lvsr (int, const volatile long *);
10082 vector unsigned char vec_lvsr (int, const volatile float *);
10084 vector float vec_madd (vector float, vector float, vector float);
10086 vector signed short vec_madds (vector signed short,
10087 vector signed short,
10088 vector signed short);
10090 vector unsigned char vec_max (vector bool char, vector unsigned char);
10091 vector unsigned char vec_max (vector unsigned char, vector bool char);
10092 vector unsigned char vec_max (vector unsigned char,
10093 vector unsigned char);
10094 vector signed char vec_max (vector bool char, vector signed char);
10095 vector signed char vec_max (vector signed char, vector bool char);
10096 vector signed char vec_max (vector signed char, vector signed char);
10097 vector unsigned short vec_max (vector bool short,
10098 vector unsigned short);
10099 vector unsigned short vec_max (vector unsigned short,
10100 vector bool short);
10101 vector unsigned short vec_max (vector unsigned short,
10102 vector unsigned short);
10103 vector signed short vec_max (vector bool short, vector signed short);
10104 vector signed short vec_max (vector signed short, vector bool short);
10105 vector signed short vec_max (vector signed short, vector signed short);
10106 vector unsigned int vec_max (vector bool int, vector unsigned int);
10107 vector unsigned int vec_max (vector unsigned int, vector bool int);
10108 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
10109 vector signed int vec_max (vector bool int, vector signed int);
10110 vector signed int vec_max (vector signed int, vector bool int);
10111 vector signed int vec_max (vector signed int, vector signed int);
10112 vector float vec_max (vector float, vector float);
10114 vector float vec_vmaxfp (vector float, vector float);
10116 vector signed int vec_vmaxsw (vector bool int, vector signed int);
10117 vector signed int vec_vmaxsw (vector signed int, vector bool int);
10118 vector signed int vec_vmaxsw (vector signed int, vector signed int);
10120 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
10121 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
10122 vector unsigned int vec_vmaxuw (vector unsigned int,
10123 vector unsigned int);
10125 vector signed short vec_vmaxsh (vector bool short, vector signed short);
10126 vector signed short vec_vmaxsh (vector signed short, vector bool short);
10127 vector signed short vec_vmaxsh (vector signed short,
10128 vector signed short);
10130 vector unsigned short vec_vmaxuh (vector bool short,
10131 vector unsigned short);
10132 vector unsigned short vec_vmaxuh (vector unsigned short,
10133 vector bool short);
10134 vector unsigned short vec_vmaxuh (vector unsigned short,
10135 vector unsigned short);
10137 vector signed char vec_vmaxsb (vector bool char, vector signed char);
10138 vector signed char vec_vmaxsb (vector signed char, vector bool char);
10139 vector signed char vec_vmaxsb (vector signed char, vector signed char);
10141 vector unsigned char vec_vmaxub (vector bool char,
10142 vector unsigned char);
10143 vector unsigned char vec_vmaxub (vector unsigned char,
10145 vector unsigned char vec_vmaxub (vector unsigned char,
10146 vector unsigned char);
10148 vector bool char vec_mergeh (vector bool char, vector bool char);
10149 vector signed char vec_mergeh (vector signed char, vector signed char);
10150 vector unsigned char vec_mergeh (vector unsigned char,
10151 vector unsigned char);
10152 vector bool short vec_mergeh (vector bool short, vector bool short);
10153 vector pixel vec_mergeh (vector pixel, vector pixel);
10154 vector signed short vec_mergeh (vector signed short,
10155 vector signed short);
10156 vector unsigned short vec_mergeh (vector unsigned short,
10157 vector unsigned short);
10158 vector float vec_mergeh (vector float, vector float);
10159 vector bool int vec_mergeh (vector bool int, vector bool int);
10160 vector signed int vec_mergeh (vector signed int, vector signed int);
10161 vector unsigned int vec_mergeh (vector unsigned int,
10162 vector unsigned int);
10164 vector float vec_vmrghw (vector float, vector float);
10165 vector bool int vec_vmrghw (vector bool int, vector bool int);
10166 vector signed int vec_vmrghw (vector signed int, vector signed int);
10167 vector unsigned int vec_vmrghw (vector unsigned int,
10168 vector unsigned int);
10170 vector bool short vec_vmrghh (vector bool short, vector bool short);
10171 vector signed short vec_vmrghh (vector signed short,
10172 vector signed short);
10173 vector unsigned short vec_vmrghh (vector unsigned short,
10174 vector unsigned short);
10175 vector pixel vec_vmrghh (vector pixel, vector pixel);
10177 vector bool char vec_vmrghb (vector bool char, vector bool char);
10178 vector signed char vec_vmrghb (vector signed char, vector signed char);
10179 vector unsigned char vec_vmrghb (vector unsigned char,
10180 vector unsigned char);
10182 vector bool char vec_mergel (vector bool char, vector bool char);
10183 vector signed char vec_mergel (vector signed char, vector signed char);
10184 vector unsigned char vec_mergel (vector unsigned char,
10185 vector unsigned char);
10186 vector bool short vec_mergel (vector bool short, vector bool short);
10187 vector pixel vec_mergel (vector pixel, vector pixel);
10188 vector signed short vec_mergel (vector signed short,
10189 vector signed short);
10190 vector unsigned short vec_mergel (vector unsigned short,
10191 vector unsigned short);
10192 vector float vec_mergel (vector float, vector float);
10193 vector bool int vec_mergel (vector bool int, vector bool int);
10194 vector signed int vec_mergel (vector signed int, vector signed int);
10195 vector unsigned int vec_mergel (vector unsigned int,
10196 vector unsigned int);
10198 vector float vec_vmrglw (vector float, vector float);
10199 vector signed int vec_vmrglw (vector signed int, vector signed int);
10200 vector unsigned int vec_vmrglw (vector unsigned int,
10201 vector unsigned int);
10202 vector bool int vec_vmrglw (vector bool int, vector bool int);
10204 vector bool short vec_vmrglh (vector bool short, vector bool short);
10205 vector signed short vec_vmrglh (vector signed short,
10206 vector signed short);
10207 vector unsigned short vec_vmrglh (vector unsigned short,
10208 vector unsigned short);
10209 vector pixel vec_vmrglh (vector pixel, vector pixel);
10211 vector bool char vec_vmrglb (vector bool char, vector bool char);
10212 vector signed char vec_vmrglb (vector signed char, vector signed char);
10213 vector unsigned char vec_vmrglb (vector unsigned char,
10214 vector unsigned char);
10216 vector unsigned short vec_mfvscr (void);
10218 vector unsigned char vec_min (vector bool char, vector unsigned char);
10219 vector unsigned char vec_min (vector unsigned char, vector bool char);
10220 vector unsigned char vec_min (vector unsigned char,
10221 vector unsigned char);
10222 vector signed char vec_min (vector bool char, vector signed char);
10223 vector signed char vec_min (vector signed char, vector bool char);
10224 vector signed char vec_min (vector signed char, vector signed char);
10225 vector unsigned short vec_min (vector bool short,
10226 vector unsigned short);
10227 vector unsigned short vec_min (vector unsigned short,
10228 vector bool short);
10229 vector unsigned short vec_min (vector unsigned short,
10230 vector unsigned short);
10231 vector signed short vec_min (vector bool short, vector signed short);
10232 vector signed short vec_min (vector signed short, vector bool short);
10233 vector signed short vec_min (vector signed short, vector signed short);
10234 vector unsigned int vec_min (vector bool int, vector unsigned int);
10235 vector unsigned int vec_min (vector unsigned int, vector bool int);
10236 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
10237 vector signed int vec_min (vector bool int, vector signed int);
10238 vector signed int vec_min (vector signed int, vector bool int);
10239 vector signed int vec_min (vector signed int, vector signed int);
10240 vector float vec_min (vector float, vector float);
10242 vector float vec_vminfp (vector float, vector float);
10244 vector signed int vec_vminsw (vector bool int, vector signed int);
10245 vector signed int vec_vminsw (vector signed int, vector bool int);
10246 vector signed int vec_vminsw (vector signed int, vector signed int);
10248 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
10249 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
10250 vector unsigned int vec_vminuw (vector unsigned int,
10251 vector unsigned int);
10253 vector signed short vec_vminsh (vector bool short, vector signed short);
10254 vector signed short vec_vminsh (vector signed short, vector bool short);
10255 vector signed short vec_vminsh (vector signed short,
10256 vector signed short);
10258 vector unsigned short vec_vminuh (vector bool short,
10259 vector unsigned short);
10260 vector unsigned short vec_vminuh (vector unsigned short,
10261 vector bool short);
10262 vector unsigned short vec_vminuh (vector unsigned short,
10263 vector unsigned short);
10265 vector signed char vec_vminsb (vector bool char, vector signed char);
10266 vector signed char vec_vminsb (vector signed char, vector bool char);
10267 vector signed char vec_vminsb (vector signed char, vector signed char);
10269 vector unsigned char vec_vminub (vector bool char,
10270 vector unsigned char);
10271 vector unsigned char vec_vminub (vector unsigned char,
10273 vector unsigned char vec_vminub (vector unsigned char,
10274 vector unsigned char);
10276 vector signed short vec_mladd (vector signed short,
10277 vector signed short,
10278 vector signed short);
10279 vector signed short vec_mladd (vector signed short,
10280 vector unsigned short,
10281 vector unsigned short);
10282 vector signed short vec_mladd (vector unsigned short,
10283 vector signed short,
10284 vector signed short);
10285 vector unsigned short vec_mladd (vector unsigned short,
10286 vector unsigned short,
10287 vector unsigned short);
10289 vector signed short vec_mradds (vector signed short,
10290 vector signed short,
10291 vector signed short);
10293 vector unsigned int vec_msum (vector unsigned char,
10294 vector unsigned char,
10295 vector unsigned int);
10296 vector signed int vec_msum (vector signed char,
10297 vector unsigned char,
10298 vector signed int);
10299 vector unsigned int vec_msum (vector unsigned short,
10300 vector unsigned short,
10301 vector unsigned int);
10302 vector signed int vec_msum (vector signed short,
10303 vector signed short,
10304 vector signed int);
10306 vector signed int vec_vmsumshm (vector signed short,
10307 vector signed short,
10308 vector signed int);
10310 vector unsigned int vec_vmsumuhm (vector unsigned short,
10311 vector unsigned short,
10312 vector unsigned int);
10314 vector signed int vec_vmsummbm (vector signed char,
10315 vector unsigned char,
10316 vector signed int);
10318 vector unsigned int vec_vmsumubm (vector unsigned char,
10319 vector unsigned char,
10320 vector unsigned int);
10322 vector unsigned int vec_msums (vector unsigned short,
10323 vector unsigned short,
10324 vector unsigned int);
10325 vector signed int vec_msums (vector signed short,
10326 vector signed short,
10327 vector signed int);
10329 vector signed int vec_vmsumshs (vector signed short,
10330 vector signed short,
10331 vector signed int);
10333 vector unsigned int vec_vmsumuhs (vector unsigned short,
10334 vector unsigned short,
10335 vector unsigned int);
10337 void vec_mtvscr (vector signed int);
10338 void vec_mtvscr (vector unsigned int);
10339 void vec_mtvscr (vector bool int);
10340 void vec_mtvscr (vector signed short);
10341 void vec_mtvscr (vector unsigned short);
10342 void vec_mtvscr (vector bool short);
10343 void vec_mtvscr (vector pixel);
10344 void vec_mtvscr (vector signed char);
10345 void vec_mtvscr (vector unsigned char);
10346 void vec_mtvscr (vector bool char);
10348 vector unsigned short vec_mule (vector unsigned char,
10349 vector unsigned char);
10350 vector signed short vec_mule (vector signed char,
10351 vector signed char);
10352 vector unsigned int vec_mule (vector unsigned short,
10353 vector unsigned short);
10354 vector signed int vec_mule (vector signed short, vector signed short);
10356 vector signed int vec_vmulesh (vector signed short,
10357 vector signed short);
10359 vector unsigned int vec_vmuleuh (vector unsigned short,
10360 vector unsigned short);
10362 vector signed short vec_vmulesb (vector signed char,
10363 vector signed char);
10365 vector unsigned short vec_vmuleub (vector unsigned char,
10366 vector unsigned char);
10368 vector unsigned short vec_mulo (vector unsigned char,
10369 vector unsigned char);
10370 vector signed short vec_mulo (vector signed char, vector signed char);
10371 vector unsigned int vec_mulo (vector unsigned short,
10372 vector unsigned short);
10373 vector signed int vec_mulo (vector signed short, vector signed short);
10375 vector signed int vec_vmulosh (vector signed short,
10376 vector signed short);
10378 vector unsigned int vec_vmulouh (vector unsigned short,
10379 vector unsigned short);
10381 vector signed short vec_vmulosb (vector signed char,
10382 vector signed char);
10384 vector unsigned short vec_vmuloub (vector unsigned char,
10385 vector unsigned char);
10387 vector float vec_nmsub (vector float, vector float, vector float);
10389 vector float vec_nor (vector float, vector float);
10390 vector signed int vec_nor (vector signed int, vector signed int);
10391 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
10392 vector bool int vec_nor (vector bool int, vector bool int);
10393 vector signed short vec_nor (vector signed short, vector signed short);
10394 vector unsigned short vec_nor (vector unsigned short,
10395 vector unsigned short);
10396 vector bool short vec_nor (vector bool short, vector bool short);
10397 vector signed char vec_nor (vector signed char, vector signed char);
10398 vector unsigned char vec_nor (vector unsigned char,
10399 vector unsigned char);
10400 vector bool char vec_nor (vector bool char, vector bool char);
10402 vector float vec_or (vector float, vector float);
10403 vector float vec_or (vector float, vector bool int);
10404 vector float vec_or (vector bool int, vector float);
10405 vector bool int vec_or (vector bool int, vector bool int);
10406 vector signed int vec_or (vector bool int, vector signed int);
10407 vector signed int vec_or (vector signed int, vector bool int);
10408 vector signed int vec_or (vector signed int, vector signed int);
10409 vector unsigned int vec_or (vector bool int, vector unsigned int);
10410 vector unsigned int vec_or (vector unsigned int, vector bool int);
10411 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
10412 vector bool short vec_or (vector bool short, vector bool short);
10413 vector signed short vec_or (vector bool short, vector signed short);
10414 vector signed short vec_or (vector signed short, vector bool short);
10415 vector signed short vec_or (vector signed short, vector signed short);
10416 vector unsigned short vec_or (vector bool short, vector unsigned short);
10417 vector unsigned short vec_or (vector unsigned short, vector bool short);
10418 vector unsigned short vec_or (vector unsigned short,
10419 vector unsigned short);
10420 vector signed char vec_or (vector bool char, vector signed char);
10421 vector bool char vec_or (vector bool char, vector bool char);
10422 vector signed char vec_or (vector signed char, vector bool char);
10423 vector signed char vec_or (vector signed char, vector signed char);
10424 vector unsigned char vec_or (vector bool char, vector unsigned char);
10425 vector unsigned char vec_or (vector unsigned char, vector bool char);
10426 vector unsigned char vec_or (vector unsigned char,
10427 vector unsigned char);
10429 vector signed char vec_pack (vector signed short, vector signed short);
10430 vector unsigned char vec_pack (vector unsigned short,
10431 vector unsigned short);
10432 vector bool char vec_pack (vector bool short, vector bool short);
10433 vector signed short vec_pack (vector signed int, vector signed int);
10434 vector unsigned short vec_pack (vector unsigned int,
10435 vector unsigned int);
10436 vector bool short vec_pack (vector bool int, vector bool int);
10438 vector bool short vec_vpkuwum (vector bool int, vector bool int);
10439 vector signed short vec_vpkuwum (vector signed int, vector signed int);
10440 vector unsigned short vec_vpkuwum (vector unsigned int,
10441 vector unsigned int);
10443 vector bool char vec_vpkuhum (vector bool short, vector bool short);
10444 vector signed char vec_vpkuhum (vector signed short,
10445 vector signed short);
10446 vector unsigned char vec_vpkuhum (vector unsigned short,
10447 vector unsigned short);
10449 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
10451 vector unsigned char vec_packs (vector unsigned short,
10452 vector unsigned short);
10453 vector signed char vec_packs (vector signed short, vector signed short);
10454 vector unsigned short vec_packs (vector unsigned int,
10455 vector unsigned int);
10456 vector signed short vec_packs (vector signed int, vector signed int);
10458 vector signed short vec_vpkswss (vector signed int, vector signed int);
10460 vector unsigned short vec_vpkuwus (vector unsigned int,
10461 vector unsigned int);
10463 vector signed char vec_vpkshss (vector signed short,
10464 vector signed short);
10466 vector unsigned char vec_vpkuhus (vector unsigned short,
10467 vector unsigned short);
10469 vector unsigned char vec_packsu (vector unsigned short,
10470 vector unsigned short);
10471 vector unsigned char vec_packsu (vector signed short,
10472 vector signed short);
10473 vector unsigned short vec_packsu (vector unsigned int,
10474 vector unsigned int);
10475 vector unsigned short vec_packsu (vector signed int, vector signed int);
10477 vector unsigned short vec_vpkswus (vector signed int,
10478 vector signed int);
10480 vector unsigned char vec_vpkshus (vector signed short,
10481 vector signed short);
10483 vector float vec_perm (vector float,
10485 vector unsigned char);
10486 vector signed int vec_perm (vector signed int,
10488 vector unsigned char);
10489 vector unsigned int vec_perm (vector unsigned int,
10490 vector unsigned int,
10491 vector unsigned char);
10492 vector bool int vec_perm (vector bool int,
10494 vector unsigned char);
10495 vector signed short vec_perm (vector signed short,
10496 vector signed short,
10497 vector unsigned char);
10498 vector unsigned short vec_perm (vector unsigned short,
10499 vector unsigned short,
10500 vector unsigned char);
10501 vector bool short vec_perm (vector bool short,
10503 vector unsigned char);
10504 vector pixel vec_perm (vector pixel,
10506 vector unsigned char);
10507 vector signed char vec_perm (vector signed char,
10508 vector signed char,
10509 vector unsigned char);
10510 vector unsigned char vec_perm (vector unsigned char,
10511 vector unsigned char,
10512 vector unsigned char);
10513 vector bool char vec_perm (vector bool char,
10515 vector unsigned char);
10517 vector float vec_re (vector float);
10519 vector signed char vec_rl (vector signed char,
10520 vector unsigned char);
10521 vector unsigned char vec_rl (vector unsigned char,
10522 vector unsigned char);
10523 vector signed short vec_rl (vector signed short, vector unsigned short);
10524 vector unsigned short vec_rl (vector unsigned short,
10525 vector unsigned short);
10526 vector signed int vec_rl (vector signed int, vector unsigned int);
10527 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
10529 vector signed int vec_vrlw (vector signed int, vector unsigned int);
10530 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
10532 vector signed short vec_vrlh (vector signed short,
10533 vector unsigned short);
10534 vector unsigned short vec_vrlh (vector unsigned short,
10535 vector unsigned short);
10537 vector signed char vec_vrlb (vector signed char, vector unsigned char);
10538 vector unsigned char vec_vrlb (vector unsigned char,
10539 vector unsigned char);
10541 vector float vec_round (vector float);
10543 vector float vec_rsqrte (vector float);
10545 vector float vec_sel (vector float, vector float, vector bool int);
10546 vector float vec_sel (vector float, vector float, vector unsigned int);
10547 vector signed int vec_sel (vector signed int,
10550 vector signed int vec_sel (vector signed int,
10552 vector unsigned int);
10553 vector unsigned int vec_sel (vector unsigned int,
10554 vector unsigned int,
10556 vector unsigned int vec_sel (vector unsigned int,
10557 vector unsigned int,
10558 vector unsigned int);
10559 vector bool int vec_sel (vector bool int,
10562 vector bool int vec_sel (vector bool int,
10564 vector unsigned int);
10565 vector signed short vec_sel (vector signed short,
10566 vector signed short,
10567 vector bool short);
10568 vector signed short vec_sel (vector signed short,
10569 vector signed short,
10570 vector unsigned short);
10571 vector unsigned short vec_sel (vector unsigned short,
10572 vector unsigned short,
10573 vector bool short);
10574 vector unsigned short vec_sel (vector unsigned short,
10575 vector unsigned short,
10576 vector unsigned short);
10577 vector bool short vec_sel (vector bool short,
10579 vector bool short);
10580 vector bool short vec_sel (vector bool short,
10582 vector unsigned short);
10583 vector signed char vec_sel (vector signed char,
10584 vector signed char,
10586 vector signed char vec_sel (vector signed char,
10587 vector signed char,
10588 vector unsigned char);
10589 vector unsigned char vec_sel (vector unsigned char,
10590 vector unsigned char,
10592 vector unsigned char vec_sel (vector unsigned char,
10593 vector unsigned char,
10594 vector unsigned char);
10595 vector bool char vec_sel (vector bool char,
10598 vector bool char vec_sel (vector bool char,
10600 vector unsigned char);
10602 vector signed char vec_sl (vector signed char,
10603 vector unsigned char);
10604 vector unsigned char vec_sl (vector unsigned char,
10605 vector unsigned char);
10606 vector signed short vec_sl (vector signed short, vector unsigned short);
10607 vector unsigned short vec_sl (vector unsigned short,
10608 vector unsigned short);
10609 vector signed int vec_sl (vector signed int, vector unsigned int);
10610 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
10612 vector signed int vec_vslw (vector signed int, vector unsigned int);
10613 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
10615 vector signed short vec_vslh (vector signed short,
10616 vector unsigned short);
10617 vector unsigned short vec_vslh (vector unsigned short,
10618 vector unsigned short);
10620 vector signed char vec_vslb (vector signed char, vector unsigned char);
10621 vector unsigned char vec_vslb (vector unsigned char,
10622 vector unsigned char);
10624 vector float vec_sld (vector float, vector float, const int);
10625 vector signed int vec_sld (vector signed int,
10628 vector unsigned int vec_sld (vector unsigned int,
10629 vector unsigned int,
10631 vector bool int vec_sld (vector bool int,
10634 vector signed short vec_sld (vector signed short,
10635 vector signed short,
10637 vector unsigned short vec_sld (vector unsigned short,
10638 vector unsigned short,
10640 vector bool short vec_sld (vector bool short,
10643 vector pixel vec_sld (vector pixel,
10646 vector signed char vec_sld (vector signed char,
10647 vector signed char,
10649 vector unsigned char vec_sld (vector unsigned char,
10650 vector unsigned char,
10652 vector bool char vec_sld (vector bool char,
10656 vector signed int vec_sll (vector signed int,
10657 vector unsigned int);
10658 vector signed int vec_sll (vector signed int,
10659 vector unsigned short);
10660 vector signed int vec_sll (vector signed int,
10661 vector unsigned char);
10662 vector unsigned int vec_sll (vector unsigned int,
10663 vector unsigned int);
10664 vector unsigned int vec_sll (vector unsigned int,
10665 vector unsigned short);
10666 vector unsigned int vec_sll (vector unsigned int,
10667 vector unsigned char);
10668 vector bool int vec_sll (vector bool int,
10669 vector unsigned int);
10670 vector bool int vec_sll (vector bool int,
10671 vector unsigned short);
10672 vector bool int vec_sll (vector bool int,
10673 vector unsigned char);
10674 vector signed short vec_sll (vector signed short,
10675 vector unsigned int);
10676 vector signed short vec_sll (vector signed short,
10677 vector unsigned short);
10678 vector signed short vec_sll (vector signed short,
10679 vector unsigned char);
10680 vector unsigned short vec_sll (vector unsigned short,
10681 vector unsigned int);
10682 vector unsigned short vec_sll (vector unsigned short,
10683 vector unsigned short);
10684 vector unsigned short vec_sll (vector unsigned short,
10685 vector unsigned char);
10686 vector bool short vec_sll (vector bool short, vector unsigned int);
10687 vector bool short vec_sll (vector bool short, vector unsigned short);
10688 vector bool short vec_sll (vector bool short, vector unsigned char);
10689 vector pixel vec_sll (vector pixel, vector unsigned int);
10690 vector pixel vec_sll (vector pixel, vector unsigned short);
10691 vector pixel vec_sll (vector pixel, vector unsigned char);
10692 vector signed char vec_sll (vector signed char, vector unsigned int);
10693 vector signed char vec_sll (vector signed char, vector unsigned short);
10694 vector signed char vec_sll (vector signed char, vector unsigned char);
10695 vector unsigned char vec_sll (vector unsigned char,
10696 vector unsigned int);
10697 vector unsigned char vec_sll (vector unsigned char,
10698 vector unsigned short);
10699 vector unsigned char vec_sll (vector unsigned char,
10700 vector unsigned char);
10701 vector bool char vec_sll (vector bool char, vector unsigned int);
10702 vector bool char vec_sll (vector bool char, vector unsigned short);
10703 vector bool char vec_sll (vector bool char, vector unsigned char);
10705 vector float vec_slo (vector float, vector signed char);
10706 vector float vec_slo (vector float, vector unsigned char);
10707 vector signed int vec_slo (vector signed int, vector signed char);
10708 vector signed int vec_slo (vector signed int, vector unsigned char);
10709 vector unsigned int vec_slo (vector unsigned int, vector signed char);
10710 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
10711 vector signed short vec_slo (vector signed short, vector signed char);
10712 vector signed short vec_slo (vector signed short, vector unsigned char);
10713 vector unsigned short vec_slo (vector unsigned short,
10714 vector signed char);
10715 vector unsigned short vec_slo (vector unsigned short,
10716 vector unsigned char);
10717 vector pixel vec_slo (vector pixel, vector signed char);
10718 vector pixel vec_slo (vector pixel, vector unsigned char);
10719 vector signed char vec_slo (vector signed char, vector signed char);
10720 vector signed char vec_slo (vector signed char, vector unsigned char);
10721 vector unsigned char vec_slo (vector unsigned char, vector signed char);
10722 vector unsigned char vec_slo (vector unsigned char,
10723 vector unsigned char);
10725 vector signed char vec_splat (vector signed char, const int);
10726 vector unsigned char vec_splat (vector unsigned char, const int);
10727 vector bool char vec_splat (vector bool char, const int);
10728 vector signed short vec_splat (vector signed short, const int);
10729 vector unsigned short vec_splat (vector unsigned short, const int);
10730 vector bool short vec_splat (vector bool short, const int);
10731 vector pixel vec_splat (vector pixel, const int);
10732 vector float vec_splat (vector float, const int);
10733 vector signed int vec_splat (vector signed int, const int);
10734 vector unsigned int vec_splat (vector unsigned int, const int);
10735 vector bool int vec_splat (vector bool int, const int);
10737 vector float vec_vspltw (vector float, const int);
10738 vector signed int vec_vspltw (vector signed int, const int);
10739 vector unsigned int vec_vspltw (vector unsigned int, const int);
10740 vector bool int vec_vspltw (vector bool int, const int);
10742 vector bool short vec_vsplth (vector bool short, const int);
10743 vector signed short vec_vsplth (vector signed short, const int);
10744 vector unsigned short vec_vsplth (vector unsigned short, const int);
10745 vector pixel vec_vsplth (vector pixel, const int);
10747 vector signed char vec_vspltb (vector signed char, const int);
10748 vector unsigned char vec_vspltb (vector unsigned char, const int);
10749 vector bool char vec_vspltb (vector bool char, const int);
10751 vector signed char vec_splat_s8 (const int);
10753 vector signed short vec_splat_s16 (const int);
10755 vector signed int vec_splat_s32 (const int);
10757 vector unsigned char vec_splat_u8 (const int);
10759 vector unsigned short vec_splat_u16 (const int);
10761 vector unsigned int vec_splat_u32 (const int);
10763 vector signed char vec_sr (vector signed char, vector unsigned char);
10764 vector unsigned char vec_sr (vector unsigned char,
10765 vector unsigned char);
10766 vector signed short vec_sr (vector signed short,
10767 vector unsigned short);
10768 vector unsigned short vec_sr (vector unsigned short,
10769 vector unsigned short);
10770 vector signed int vec_sr (vector signed int, vector unsigned int);
10771 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
10773 vector signed int vec_vsrw (vector signed int, vector unsigned int);
10774 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
10776 vector signed short vec_vsrh (vector signed short,
10777 vector unsigned short);
10778 vector unsigned short vec_vsrh (vector unsigned short,
10779 vector unsigned short);
10781 vector signed char vec_vsrb (vector signed char, vector unsigned char);
10782 vector unsigned char vec_vsrb (vector unsigned char,
10783 vector unsigned char);
10785 vector signed char vec_sra (vector signed char, vector unsigned char);
10786 vector unsigned char vec_sra (vector unsigned char,
10787 vector unsigned char);
10788 vector signed short vec_sra (vector signed short,
10789 vector unsigned short);
10790 vector unsigned short vec_sra (vector unsigned short,
10791 vector unsigned short);
10792 vector signed int vec_sra (vector signed int, vector unsigned int);
10793 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
10795 vector signed int vec_vsraw (vector signed int, vector unsigned int);
10796 vector unsigned int vec_vsraw (vector unsigned int,
10797 vector unsigned int);
10799 vector signed short vec_vsrah (vector signed short,
10800 vector unsigned short);
10801 vector unsigned short vec_vsrah (vector unsigned short,
10802 vector unsigned short);
10804 vector signed char vec_vsrab (vector signed char, vector unsigned char);
10805 vector unsigned char vec_vsrab (vector unsigned char,
10806 vector unsigned char);
10808 vector signed int vec_srl (vector signed int, vector unsigned int);
10809 vector signed int vec_srl (vector signed int, vector unsigned short);
10810 vector signed int vec_srl (vector signed int, vector unsigned char);
10811 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
10812 vector unsigned int vec_srl (vector unsigned int,
10813 vector unsigned short);
10814 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
10815 vector bool int vec_srl (vector bool int, vector unsigned int);
10816 vector bool int vec_srl (vector bool int, vector unsigned short);
10817 vector bool int vec_srl (vector bool int, vector unsigned char);
10818 vector signed short vec_srl (vector signed short, vector unsigned int);
10819 vector signed short vec_srl (vector signed short,
10820 vector unsigned short);
10821 vector signed short vec_srl (vector signed short, vector unsigned char);
10822 vector unsigned short vec_srl (vector unsigned short,
10823 vector unsigned int);
10824 vector unsigned short vec_srl (vector unsigned short,
10825 vector unsigned short);
10826 vector unsigned short vec_srl (vector unsigned short,
10827 vector unsigned char);
10828 vector bool short vec_srl (vector bool short, vector unsigned int);
10829 vector bool short vec_srl (vector bool short, vector unsigned short);
10830 vector bool short vec_srl (vector bool short, vector unsigned char);
10831 vector pixel vec_srl (vector pixel, vector unsigned int);
10832 vector pixel vec_srl (vector pixel, vector unsigned short);
10833 vector pixel vec_srl (vector pixel, vector unsigned char);
10834 vector signed char vec_srl (vector signed char, vector unsigned int);
10835 vector signed char vec_srl (vector signed char, vector unsigned short);
10836 vector signed char vec_srl (vector signed char, vector unsigned char);
10837 vector unsigned char vec_srl (vector unsigned char,
10838 vector unsigned int);
10839 vector unsigned char vec_srl (vector unsigned char,
10840 vector unsigned short);
10841 vector unsigned char vec_srl (vector unsigned char,
10842 vector unsigned char);
10843 vector bool char vec_srl (vector bool char, vector unsigned int);
10844 vector bool char vec_srl (vector bool char, vector unsigned short);
10845 vector bool char vec_srl (vector bool char, vector unsigned char);
10847 vector float vec_sro (vector float, vector signed char);
10848 vector float vec_sro (vector float, vector unsigned char);
10849 vector signed int vec_sro (vector signed int, vector signed char);
10850 vector signed int vec_sro (vector signed int, vector unsigned char);
10851 vector unsigned int vec_sro (vector unsigned int, vector signed char);
10852 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
10853 vector signed short vec_sro (vector signed short, vector signed char);
10854 vector signed short vec_sro (vector signed short, vector unsigned char);
10855 vector unsigned short vec_sro (vector unsigned short,
10856 vector signed char);
10857 vector unsigned short vec_sro (vector unsigned short,
10858 vector unsigned char);
10859 vector pixel vec_sro (vector pixel, vector signed char);
10860 vector pixel vec_sro (vector pixel, vector unsigned char);
10861 vector signed char vec_sro (vector signed char, vector signed char);
10862 vector signed char vec_sro (vector signed char, vector unsigned char);
10863 vector unsigned char vec_sro (vector unsigned char, vector signed char);
10864 vector unsigned char vec_sro (vector unsigned char,
10865 vector unsigned char);
10867 void vec_st (vector float, int, vector float *);
10868 void vec_st (vector float, int, float *);
10869 void vec_st (vector signed int, int, vector signed int *);
10870 void vec_st (vector signed int, int, int *);
10871 void vec_st (vector unsigned int, int, vector unsigned int *);
10872 void vec_st (vector unsigned int, int, unsigned int *);
10873 void vec_st (vector bool int, int, vector bool int *);
10874 void vec_st (vector bool int, int, unsigned int *);
10875 void vec_st (vector bool int, int, int *);
10876 void vec_st (vector signed short, int, vector signed short *);
10877 void vec_st (vector signed short, int, short *);
10878 void vec_st (vector unsigned short, int, vector unsigned short *);
10879 void vec_st (vector unsigned short, int, unsigned short *);
10880 void vec_st (vector bool short, int, vector bool short *);
10881 void vec_st (vector bool short, int, unsigned short *);
10882 void vec_st (vector pixel, int, vector pixel *);
10883 void vec_st (vector pixel, int, unsigned short *);
10884 void vec_st (vector pixel, int, short *);
10885 void vec_st (vector bool short, int, short *);
10886 void vec_st (vector signed char, int, vector signed char *);
10887 void vec_st (vector signed char, int, signed char *);
10888 void vec_st (vector unsigned char, int, vector unsigned char *);
10889 void vec_st (vector unsigned char, int, unsigned char *);
10890 void vec_st (vector bool char, int, vector bool char *);
10891 void vec_st (vector bool char, int, unsigned char *);
10892 void vec_st (vector bool char, int, signed char *);
10894 void vec_ste (vector signed char, int, signed char *);
10895 void vec_ste (vector unsigned char, int, unsigned char *);
10896 void vec_ste (vector bool char, int, signed char *);
10897 void vec_ste (vector bool char, int, unsigned char *);
10898 void vec_ste (vector signed short, int, short *);
10899 void vec_ste (vector unsigned short, int, unsigned short *);
10900 void vec_ste (vector bool short, int, short *);
10901 void vec_ste (vector bool short, int, unsigned short *);
10902 void vec_ste (vector pixel, int, short *);
10903 void vec_ste (vector pixel, int, unsigned short *);
10904 void vec_ste (vector float, int, float *);
10905 void vec_ste (vector signed int, int, int *);
10906 void vec_ste (vector unsigned int, int, unsigned int *);
10907 void vec_ste (vector bool int, int, int *);
10908 void vec_ste (vector bool int, int, unsigned int *);
10910 void vec_stvewx (vector float, int, float *);
10911 void vec_stvewx (vector signed int, int, int *);
10912 void vec_stvewx (vector unsigned int, int, unsigned int *);
10913 void vec_stvewx (vector bool int, int, int *);
10914 void vec_stvewx (vector bool int, int, unsigned int *);
10916 void vec_stvehx (vector signed short, int, short *);
10917 void vec_stvehx (vector unsigned short, int, unsigned short *);
10918 void vec_stvehx (vector bool short, int, short *);
10919 void vec_stvehx (vector bool short, int, unsigned short *);
10920 void vec_stvehx (vector pixel, int, short *);
10921 void vec_stvehx (vector pixel, int, unsigned short *);
10923 void vec_stvebx (vector signed char, int, signed char *);
10924 void vec_stvebx (vector unsigned char, int, unsigned char *);
10925 void vec_stvebx (vector bool char, int, signed char *);
10926 void vec_stvebx (vector bool char, int, unsigned char *);
10928 void vec_stl (vector float, int, vector float *);
10929 void vec_stl (vector float, int, float *);
10930 void vec_stl (vector signed int, int, vector signed int *);
10931 void vec_stl (vector signed int, int, int *);
10932 void vec_stl (vector unsigned int, int, vector unsigned int *);
10933 void vec_stl (vector unsigned int, int, unsigned int *);
10934 void vec_stl (vector bool int, int, vector bool int *);
10935 void vec_stl (vector bool int, int, unsigned int *);
10936 void vec_stl (vector bool int, int, int *);
10937 void vec_stl (vector signed short, int, vector signed short *);
10938 void vec_stl (vector signed short, int, short *);
10939 void vec_stl (vector unsigned short, int, vector unsigned short *);
10940 void vec_stl (vector unsigned short, int, unsigned short *);
10941 void vec_stl (vector bool short, int, vector bool short *);
10942 void vec_stl (vector bool short, int, unsigned short *);
10943 void vec_stl (vector bool short, int, short *);
10944 void vec_stl (vector pixel, int, vector pixel *);
10945 void vec_stl (vector pixel, int, unsigned short *);
10946 void vec_stl (vector pixel, int, short *);
10947 void vec_stl (vector signed char, int, vector signed char *);
10948 void vec_stl (vector signed char, int, signed char *);
10949 void vec_stl (vector unsigned char, int, vector unsigned char *);
10950 void vec_stl (vector unsigned char, int, unsigned char *);
10951 void vec_stl (vector bool char, int, vector bool char *);
10952 void vec_stl (vector bool char, int, unsigned char *);
10953 void vec_stl (vector bool char, int, signed char *);
10955 vector signed char vec_sub (vector bool char, vector signed char);
10956 vector signed char vec_sub (vector signed char, vector bool char);
10957 vector signed char vec_sub (vector signed char, vector signed char);
10958 vector unsigned char vec_sub (vector bool char, vector unsigned char);
10959 vector unsigned char vec_sub (vector unsigned char, vector bool char);
10960 vector unsigned char vec_sub (vector unsigned char,
10961 vector unsigned char);
10962 vector signed short vec_sub (vector bool short, vector signed short);
10963 vector signed short vec_sub (vector signed short, vector bool short);
10964 vector signed short vec_sub (vector signed short, vector signed short);
10965 vector unsigned short vec_sub (vector bool short,
10966 vector unsigned short);
10967 vector unsigned short vec_sub (vector unsigned short,
10968 vector bool short);
10969 vector unsigned short vec_sub (vector unsigned short,
10970 vector unsigned short);
10971 vector signed int vec_sub (vector bool int, vector signed int);
10972 vector signed int vec_sub (vector signed int, vector bool int);
10973 vector signed int vec_sub (vector signed int, vector signed int);
10974 vector unsigned int vec_sub (vector bool int, vector unsigned int);
10975 vector unsigned int vec_sub (vector unsigned int, vector bool int);
10976 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
10977 vector float vec_sub (vector float, vector float);
10979 vector float vec_vsubfp (vector float, vector float);
10981 vector signed int vec_vsubuwm (vector bool int, vector signed int);
10982 vector signed int vec_vsubuwm (vector signed int, vector bool int);
10983 vector signed int vec_vsubuwm (vector signed int, vector signed int);
10984 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
10985 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
10986 vector unsigned int vec_vsubuwm (vector unsigned int,
10987 vector unsigned int);
10989 vector signed short vec_vsubuhm (vector bool short,
10990 vector signed short);
10991 vector signed short vec_vsubuhm (vector signed short,
10992 vector bool short);
10993 vector signed short vec_vsubuhm (vector signed short,
10994 vector signed short);
10995 vector unsigned short vec_vsubuhm (vector bool short,
10996 vector unsigned short);
10997 vector unsigned short vec_vsubuhm (vector unsigned short,
10998 vector bool short);
10999 vector unsigned short vec_vsubuhm (vector unsigned short,
11000 vector unsigned short);
11002 vector signed char vec_vsububm (vector bool char, vector signed char);
11003 vector signed char vec_vsububm (vector signed char, vector bool char);
11004 vector signed char vec_vsububm (vector signed char, vector signed char);
11005 vector unsigned char vec_vsububm (vector bool char,
11006 vector unsigned char);
11007 vector unsigned char vec_vsububm (vector unsigned char,
11009 vector unsigned char vec_vsububm (vector unsigned char,
11010 vector unsigned char);
11012 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
11014 vector unsigned char vec_subs (vector bool char, vector unsigned char);
11015 vector unsigned char vec_subs (vector unsigned char, vector bool char);
11016 vector unsigned char vec_subs (vector unsigned char,
11017 vector unsigned char);
11018 vector signed char vec_subs (vector bool char, vector signed char);
11019 vector signed char vec_subs (vector signed char, vector bool char);
11020 vector signed char vec_subs (vector signed char, vector signed char);
11021 vector unsigned short vec_subs (vector bool short,
11022 vector unsigned short);
11023 vector unsigned short vec_subs (vector unsigned short,
11024 vector bool short);
11025 vector unsigned short vec_subs (vector unsigned short,
11026 vector unsigned short);
11027 vector signed short vec_subs (vector bool short, vector signed short);
11028 vector signed short vec_subs (vector signed short, vector bool short);
11029 vector signed short vec_subs (vector signed short, vector signed short);
11030 vector unsigned int vec_subs (vector bool int, vector unsigned int);
11031 vector unsigned int vec_subs (vector unsigned int, vector bool int);
11032 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
11033 vector signed int vec_subs (vector bool int, vector signed int);
11034 vector signed int vec_subs (vector signed int, vector bool int);
11035 vector signed int vec_subs (vector signed int, vector signed int);
11037 vector signed int vec_vsubsws (vector bool int, vector signed int);
11038 vector signed int vec_vsubsws (vector signed int, vector bool int);
11039 vector signed int vec_vsubsws (vector signed int, vector signed int);
11041 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
11042 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
11043 vector unsigned int vec_vsubuws (vector unsigned int,
11044 vector unsigned int);
11046 vector signed short vec_vsubshs (vector bool short,
11047 vector signed short);
11048 vector signed short vec_vsubshs (vector signed short,
11049 vector bool short);
11050 vector signed short vec_vsubshs (vector signed short,
11051 vector signed short);
11053 vector unsigned short vec_vsubuhs (vector bool short,
11054 vector unsigned short);
11055 vector unsigned short vec_vsubuhs (vector unsigned short,
11056 vector bool short);
11057 vector unsigned short vec_vsubuhs (vector unsigned short,
11058 vector unsigned short);
11060 vector signed char vec_vsubsbs (vector bool char, vector signed char);
11061 vector signed char vec_vsubsbs (vector signed char, vector bool char);
11062 vector signed char vec_vsubsbs (vector signed char, vector signed char);
11064 vector unsigned char vec_vsububs (vector bool char,
11065 vector unsigned char);
11066 vector unsigned char vec_vsububs (vector unsigned char,
11068 vector unsigned char vec_vsububs (vector unsigned char,
11069 vector unsigned char);
11071 vector unsigned int vec_sum4s (vector unsigned char,
11072 vector unsigned int);
11073 vector signed int vec_sum4s (vector signed char, vector signed int);
11074 vector signed int vec_sum4s (vector signed short, vector signed int);
11076 vector signed int vec_vsum4shs (vector signed short, vector signed int);
11078 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
11080 vector unsigned int vec_vsum4ubs (vector unsigned char,
11081 vector unsigned int);
11083 vector signed int vec_sum2s (vector signed int, vector signed int);
11085 vector signed int vec_sums (vector signed int, vector signed int);
11087 vector float vec_trunc (vector float);
11089 vector signed short vec_unpackh (vector signed char);
11090 vector bool short vec_unpackh (vector bool char);
11091 vector signed int vec_unpackh (vector signed short);
11092 vector bool int vec_unpackh (vector bool short);
11093 vector unsigned int vec_unpackh (vector pixel);
11095 vector bool int vec_vupkhsh (vector bool short);
11096 vector signed int vec_vupkhsh (vector signed short);
11098 vector unsigned int vec_vupkhpx (vector pixel);
11100 vector bool short vec_vupkhsb (vector bool char);
11101 vector signed short vec_vupkhsb (vector signed char);
11103 vector signed short vec_unpackl (vector signed char);
11104 vector bool short vec_unpackl (vector bool char);
11105 vector unsigned int vec_unpackl (vector pixel);
11106 vector signed int vec_unpackl (vector signed short);
11107 vector bool int vec_unpackl (vector bool short);
11109 vector unsigned int vec_vupklpx (vector pixel);
11111 vector bool int vec_vupklsh (vector bool short);
11112 vector signed int vec_vupklsh (vector signed short);
11114 vector bool short vec_vupklsb (vector bool char);
11115 vector signed short vec_vupklsb (vector signed char);
11117 vector float vec_xor (vector float, vector float);
11118 vector float vec_xor (vector float, vector bool int);
11119 vector float vec_xor (vector bool int, vector float);
11120 vector bool int vec_xor (vector bool int, vector bool int);
11121 vector signed int vec_xor (vector bool int, vector signed int);
11122 vector signed int vec_xor (vector signed int, vector bool int);
11123 vector signed int vec_xor (vector signed int, vector signed int);
11124 vector unsigned int vec_xor (vector bool int, vector unsigned int);
11125 vector unsigned int vec_xor (vector unsigned int, vector bool int);
11126 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
11127 vector bool short vec_xor (vector bool short, vector bool short);
11128 vector signed short vec_xor (vector bool short, vector signed short);
11129 vector signed short vec_xor (vector signed short, vector bool short);
11130 vector signed short vec_xor (vector signed short, vector signed short);
11131 vector unsigned short vec_xor (vector bool short,
11132 vector unsigned short);
11133 vector unsigned short vec_xor (vector unsigned short,
11134 vector bool short);
11135 vector unsigned short vec_xor (vector unsigned short,
11136 vector unsigned short);
11137 vector signed char vec_xor (vector bool char, vector signed char);
11138 vector bool char vec_xor (vector bool char, vector bool char);
11139 vector signed char vec_xor (vector signed char, vector bool char);
11140 vector signed char vec_xor (vector signed char, vector signed char);
11141 vector unsigned char vec_xor (vector bool char, vector unsigned char);
11142 vector unsigned char vec_xor (vector unsigned char, vector bool char);
11143 vector unsigned char vec_xor (vector unsigned char,
11144 vector unsigned char);
11146 int vec_all_eq (vector signed char, vector bool char);
11147 int vec_all_eq (vector signed char, vector signed char);
11148 int vec_all_eq (vector unsigned char, vector bool char);
11149 int vec_all_eq (vector unsigned char, vector unsigned char);
11150 int vec_all_eq (vector bool char, vector bool char);
11151 int vec_all_eq (vector bool char, vector unsigned char);
11152 int vec_all_eq (vector bool char, vector signed char);
11153 int vec_all_eq (vector signed short, vector bool short);
11154 int vec_all_eq (vector signed short, vector signed short);
11155 int vec_all_eq (vector unsigned short, vector bool short);
11156 int vec_all_eq (vector unsigned short, vector unsigned short);
11157 int vec_all_eq (vector bool short, vector bool short);
11158 int vec_all_eq (vector bool short, vector unsigned short);
11159 int vec_all_eq (vector bool short, vector signed short);
11160 int vec_all_eq (vector pixel, vector pixel);
11161 int vec_all_eq (vector signed int, vector bool int);
11162 int vec_all_eq (vector signed int, vector signed int);
11163 int vec_all_eq (vector unsigned int, vector bool int);
11164 int vec_all_eq (vector unsigned int, vector unsigned int);
11165 int vec_all_eq (vector bool int, vector bool int);
11166 int vec_all_eq (vector bool int, vector unsigned int);
11167 int vec_all_eq (vector bool int, vector signed int);
11168 int vec_all_eq (vector float, vector float);
11170 int vec_all_ge (vector bool char, vector unsigned char);
11171 int vec_all_ge (vector unsigned char, vector bool char);
11172 int vec_all_ge (vector unsigned char, vector unsigned char);
11173 int vec_all_ge (vector bool char, vector signed char);
11174 int vec_all_ge (vector signed char, vector bool char);
11175 int vec_all_ge (vector signed char, vector signed char);
11176 int vec_all_ge (vector bool short, vector unsigned short);
11177 int vec_all_ge (vector unsigned short, vector bool short);
11178 int vec_all_ge (vector unsigned short, vector unsigned short);
11179 int vec_all_ge (vector signed short, vector signed short);
11180 int vec_all_ge (vector bool short, vector signed short);
11181 int vec_all_ge (vector signed short, vector bool short);
11182 int vec_all_ge (vector bool int, vector unsigned int);
11183 int vec_all_ge (vector unsigned int, vector bool int);
11184 int vec_all_ge (vector unsigned int, vector unsigned int);
11185 int vec_all_ge (vector bool int, vector signed int);
11186 int vec_all_ge (vector signed int, vector bool int);
11187 int vec_all_ge (vector signed int, vector signed int);
11188 int vec_all_ge (vector float, vector float);
11190 int vec_all_gt (vector bool char, vector unsigned char);
11191 int vec_all_gt (vector unsigned char, vector bool char);
11192 int vec_all_gt (vector unsigned char, vector unsigned char);
11193 int vec_all_gt (vector bool char, vector signed char);
11194 int vec_all_gt (vector signed char, vector bool char);
11195 int vec_all_gt (vector signed char, vector signed char);
11196 int vec_all_gt (vector bool short, vector unsigned short);
11197 int vec_all_gt (vector unsigned short, vector bool short);
11198 int vec_all_gt (vector unsigned short, vector unsigned short);
11199 int vec_all_gt (vector bool short, vector signed short);
11200 int vec_all_gt (vector signed short, vector bool short);
11201 int vec_all_gt (vector signed short, vector signed short);
11202 int vec_all_gt (vector bool int, vector unsigned int);
11203 int vec_all_gt (vector unsigned int, vector bool int);
11204 int vec_all_gt (vector unsigned int, vector unsigned int);
11205 int vec_all_gt (vector bool int, vector signed int);
11206 int vec_all_gt (vector signed int, vector bool int);
11207 int vec_all_gt (vector signed int, vector signed int);
11208 int vec_all_gt (vector float, vector float);
11210 int vec_all_in (vector float, vector float);
11212 int vec_all_le (vector bool char, vector unsigned char);
11213 int vec_all_le (vector unsigned char, vector bool char);
11214 int vec_all_le (vector unsigned char, vector unsigned char);
11215 int vec_all_le (vector bool char, vector signed char);
11216 int vec_all_le (vector signed char, vector bool char);
11217 int vec_all_le (vector signed char, vector signed char);
11218 int vec_all_le (vector bool short, vector unsigned short);
11219 int vec_all_le (vector unsigned short, vector bool short);
11220 int vec_all_le (vector unsigned short, vector unsigned short);
11221 int vec_all_le (vector bool short, vector signed short);
11222 int vec_all_le (vector signed short, vector bool short);
11223 int vec_all_le (vector signed short, vector signed short);
11224 int vec_all_le (vector bool int, vector unsigned int);
11225 int vec_all_le (vector unsigned int, vector bool int);
11226 int vec_all_le (vector unsigned int, vector unsigned int);
11227 int vec_all_le (vector bool int, vector signed int);
11228 int vec_all_le (vector signed int, vector bool int);
11229 int vec_all_le (vector signed int, vector signed int);
11230 int vec_all_le (vector float, vector float);
11232 int vec_all_lt (vector bool char, vector unsigned char);
11233 int vec_all_lt (vector unsigned char, vector bool char);
11234 int vec_all_lt (vector unsigned char, vector unsigned char);
11235 int vec_all_lt (vector bool char, vector signed char);
11236 int vec_all_lt (vector signed char, vector bool char);
11237 int vec_all_lt (vector signed char, vector signed char);
11238 int vec_all_lt (vector bool short, vector unsigned short);
11239 int vec_all_lt (vector unsigned short, vector bool short);
11240 int vec_all_lt (vector unsigned short, vector unsigned short);
11241 int vec_all_lt (vector bool short, vector signed short);
11242 int vec_all_lt (vector signed short, vector bool short);
11243 int vec_all_lt (vector signed short, vector signed short);
11244 int vec_all_lt (vector bool int, vector unsigned int);
11245 int vec_all_lt (vector unsigned int, vector bool int);
11246 int vec_all_lt (vector unsigned int, vector unsigned int);
11247 int vec_all_lt (vector bool int, vector signed int);
11248 int vec_all_lt (vector signed int, vector bool int);
11249 int vec_all_lt (vector signed int, vector signed int);
11250 int vec_all_lt (vector float, vector float);
11252 int vec_all_nan (vector float);
11254 int vec_all_ne (vector signed char, vector bool char);
11255 int vec_all_ne (vector signed char, vector signed char);
11256 int vec_all_ne (vector unsigned char, vector bool char);
11257 int vec_all_ne (vector unsigned char, vector unsigned char);
11258 int vec_all_ne (vector bool char, vector bool char);
11259 int vec_all_ne (vector bool char, vector unsigned char);
11260 int vec_all_ne (vector bool char, vector signed char);
11261 int vec_all_ne (vector signed short, vector bool short);
11262 int vec_all_ne (vector signed short, vector signed short);
11263 int vec_all_ne (vector unsigned short, vector bool short);
11264 int vec_all_ne (vector unsigned short, vector unsigned short);
11265 int vec_all_ne (vector bool short, vector bool short);
11266 int vec_all_ne (vector bool short, vector unsigned short);
11267 int vec_all_ne (vector bool short, vector signed short);
11268 int vec_all_ne (vector pixel, vector pixel);
11269 int vec_all_ne (vector signed int, vector bool int);
11270 int vec_all_ne (vector signed int, vector signed int);
11271 int vec_all_ne (vector unsigned int, vector bool int);
11272 int vec_all_ne (vector unsigned int, vector unsigned int);
11273 int vec_all_ne (vector bool int, vector bool int);
11274 int vec_all_ne (vector bool int, vector unsigned int);
11275 int vec_all_ne (vector bool int, vector signed int);
11276 int vec_all_ne (vector float, vector float);
11278 int vec_all_nge (vector float, vector float);
11280 int vec_all_ngt (vector float, vector float);
11282 int vec_all_nle (vector float, vector float);
11284 int vec_all_nlt (vector float, vector float);
11286 int vec_all_numeric (vector float);
11288 int vec_any_eq (vector signed char, vector bool char);
11289 int vec_any_eq (vector signed char, vector signed char);
11290 int vec_any_eq (vector unsigned char, vector bool char);
11291 int vec_any_eq (vector unsigned char, vector unsigned char);
11292 int vec_any_eq (vector bool char, vector bool char);
11293 int vec_any_eq (vector bool char, vector unsigned char);
11294 int vec_any_eq (vector bool char, vector signed char);
11295 int vec_any_eq (vector signed short, vector bool short);
11296 int vec_any_eq (vector signed short, vector signed short);
11297 int vec_any_eq (vector unsigned short, vector bool short);
11298 int vec_any_eq (vector unsigned short, vector unsigned short);
11299 int vec_any_eq (vector bool short, vector bool short);
11300 int vec_any_eq (vector bool short, vector unsigned short);
11301 int vec_any_eq (vector bool short, vector signed short);
11302 int vec_any_eq (vector pixel, vector pixel);
11303 int vec_any_eq (vector signed int, vector bool int);
11304 int vec_any_eq (vector signed int, vector signed int);
11305 int vec_any_eq (vector unsigned int, vector bool int);
11306 int vec_any_eq (vector unsigned int, vector unsigned int);
11307 int vec_any_eq (vector bool int, vector bool int);
11308 int vec_any_eq (vector bool int, vector unsigned int);
11309 int vec_any_eq (vector bool int, vector signed int);
11310 int vec_any_eq (vector float, vector float);
11312 int vec_any_ge (vector signed char, vector bool char);
11313 int vec_any_ge (vector unsigned char, vector bool char);
11314 int vec_any_ge (vector unsigned char, vector unsigned char);
11315 int vec_any_ge (vector signed char, vector signed char);
11316 int vec_any_ge (vector bool char, vector unsigned char);
11317 int vec_any_ge (vector bool char, vector signed char);
11318 int vec_any_ge (vector unsigned short, vector bool short);
11319 int vec_any_ge (vector unsigned short, vector unsigned short);
11320 int vec_any_ge (vector signed short, vector signed short);
11321 int vec_any_ge (vector signed short, vector bool short);
11322 int vec_any_ge (vector bool short, vector unsigned short);
11323 int vec_any_ge (vector bool short, vector signed short);
11324 int vec_any_ge (vector signed int, vector bool int);
11325 int vec_any_ge (vector unsigned int, vector bool int);
11326 int vec_any_ge (vector unsigned int, vector unsigned int);
11327 int vec_any_ge (vector signed int, vector signed int);
11328 int vec_any_ge (vector bool int, vector unsigned int);
11329 int vec_any_ge (vector bool int, vector signed int);
11330 int vec_any_ge (vector float, vector float);
11332 int vec_any_gt (vector bool char, vector unsigned char);
11333 int vec_any_gt (vector unsigned char, vector bool char);
11334 int vec_any_gt (vector unsigned char, vector unsigned char);
11335 int vec_any_gt (vector bool char, vector signed char);
11336 int vec_any_gt (vector signed char, vector bool char);
11337 int vec_any_gt (vector signed char, vector signed char);
11338 int vec_any_gt (vector bool short, vector unsigned short);
11339 int vec_any_gt (vector unsigned short, vector bool short);
11340 int vec_any_gt (vector unsigned short, vector unsigned short);
11341 int vec_any_gt (vector bool short, vector signed short);
11342 int vec_any_gt (vector signed short, vector bool short);
11343 int vec_any_gt (vector signed short, vector signed short);
11344 int vec_any_gt (vector bool int, vector unsigned int);
11345 int vec_any_gt (vector unsigned int, vector bool int);
11346 int vec_any_gt (vector unsigned int, vector unsigned int);
11347 int vec_any_gt (vector bool int, vector signed int);
11348 int vec_any_gt (vector signed int, vector bool int);
11349 int vec_any_gt (vector signed int, vector signed int);
11350 int vec_any_gt (vector float, vector float);
11352 int vec_any_le (vector bool char, vector unsigned char);
11353 int vec_any_le (vector unsigned char, vector bool char);
11354 int vec_any_le (vector unsigned char, vector unsigned char);
11355 int vec_any_le (vector bool char, vector signed char);
11356 int vec_any_le (vector signed char, vector bool char);
11357 int vec_any_le (vector signed char, vector signed char);
11358 int vec_any_le (vector bool short, vector unsigned short);
11359 int vec_any_le (vector unsigned short, vector bool short);
11360 int vec_any_le (vector unsigned short, vector unsigned short);
11361 int vec_any_le (vector bool short, vector signed short);
11362 int vec_any_le (vector signed short, vector bool short);
11363 int vec_any_le (vector signed short, vector signed short);
11364 int vec_any_le (vector bool int, vector unsigned int);
11365 int vec_any_le (vector unsigned int, vector bool int);
11366 int vec_any_le (vector unsigned int, vector unsigned int);
11367 int vec_any_le (vector bool int, vector signed int);
11368 int vec_any_le (vector signed int, vector bool int);
11369 int vec_any_le (vector signed int, vector signed int);
11370 int vec_any_le (vector float, vector float);
11372 int vec_any_lt (vector bool char, vector unsigned char);
11373 int vec_any_lt (vector unsigned char, vector bool char);
11374 int vec_any_lt (vector unsigned char, vector unsigned char);
11375 int vec_any_lt (vector bool char, vector signed char);
11376 int vec_any_lt (vector signed char, vector bool char);
11377 int vec_any_lt (vector signed char, vector signed char);
11378 int vec_any_lt (vector bool short, vector unsigned short);
11379 int vec_any_lt (vector unsigned short, vector bool short);
11380 int vec_any_lt (vector unsigned short, vector unsigned short);
11381 int vec_any_lt (vector bool short, vector signed short);
11382 int vec_any_lt (vector signed short, vector bool short);
11383 int vec_any_lt (vector signed short, vector signed short);
11384 int vec_any_lt (vector bool int, vector unsigned int);
11385 int vec_any_lt (vector unsigned int, vector bool int);
11386 int vec_any_lt (vector unsigned int, vector unsigned int);
11387 int vec_any_lt (vector bool int, vector signed int);
11388 int vec_any_lt (vector signed int, vector bool int);
11389 int vec_any_lt (vector signed int, vector signed int);
11390 int vec_any_lt (vector float, vector float);
11392 int vec_any_nan (vector float);
11394 int vec_any_ne (vector signed char, vector bool char);
11395 int vec_any_ne (vector signed char, vector signed char);
11396 int vec_any_ne (vector unsigned char, vector bool char);
11397 int vec_any_ne (vector unsigned char, vector unsigned char);
11398 int vec_any_ne (vector bool char, vector bool char);
11399 int vec_any_ne (vector bool char, vector unsigned char);
11400 int vec_any_ne (vector bool char, vector signed char);
11401 int vec_any_ne (vector signed short, vector bool short);
11402 int vec_any_ne (vector signed short, vector signed short);
11403 int vec_any_ne (vector unsigned short, vector bool short);
11404 int vec_any_ne (vector unsigned short, vector unsigned short);
11405 int vec_any_ne (vector bool short, vector bool short);
11406 int vec_any_ne (vector bool short, vector unsigned short);
11407 int vec_any_ne (vector bool short, vector signed short);
11408 int vec_any_ne (vector pixel, vector pixel);
11409 int vec_any_ne (vector signed int, vector bool int);
11410 int vec_any_ne (vector signed int, vector signed int);
11411 int vec_any_ne (vector unsigned int, vector bool int);
11412 int vec_any_ne (vector unsigned int, vector unsigned int);
11413 int vec_any_ne (vector bool int, vector bool int);
11414 int vec_any_ne (vector bool int, vector unsigned int);
11415 int vec_any_ne (vector bool int, vector signed int);
11416 int vec_any_ne (vector float, vector float);
11418 int vec_any_nge (vector float, vector float);
11420 int vec_any_ngt (vector float, vector float);
11422 int vec_any_nle (vector float, vector float);
11424 int vec_any_nlt (vector float, vector float);
11426 int vec_any_numeric (vector float);
11428 int vec_any_out (vector float, vector float);
11431 @node SPARC VIS Built-in Functions
11432 @subsection SPARC VIS Built-in Functions
11434 GCC supports SIMD operations on the SPARC using both the generic vector
11435 extensions (@pxref{Vector Extensions}) as well as built-in functions for
11436 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
11437 switch, the VIS extension is exposed as the following built-in functions:
11440 typedef int v2si __attribute__ ((vector_size (8)));
11441 typedef short v4hi __attribute__ ((vector_size (8)));
11442 typedef short v2hi __attribute__ ((vector_size (4)));
11443 typedef char v8qi __attribute__ ((vector_size (8)));
11444 typedef char v4qi __attribute__ ((vector_size (4)));
11446 void * __builtin_vis_alignaddr (void *, long);
11447 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
11448 v2si __builtin_vis_faligndatav2si (v2si, v2si);
11449 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
11450 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
11452 v4hi __builtin_vis_fexpand (v4qi);
11454 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
11455 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
11456 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
11457 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
11458 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
11459 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
11460 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
11462 v4qi __builtin_vis_fpack16 (v4hi);
11463 v8qi __builtin_vis_fpack32 (v2si, v2si);
11464 v2hi __builtin_vis_fpackfix (v2si);
11465 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
11467 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
11470 @node SPU Built-in Functions
11471 @subsection SPU Built-in Functions
11473 GCC provides extensions for the SPU processor as described in the
11474 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
11475 found at @uref{http://cell.scei.co.jp/} or
11476 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
11477 implementation differs in several ways.
11482 The optional extension of specifying vector constants in parentheses is
11486 A vector initializer requires no cast if the vector constant is of the
11487 same type as the variable it is initializing.
11490 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11491 vector type is the default signedness of the base type. The default
11492 varies depending on the operating system, so a portable program should
11493 always specify the signedness.
11496 By default, the keyword @code{__vector} is added. The macro
11497 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
11501 GCC allows using a @code{typedef} name as the type specifier for a
11505 For C, overloaded functions are implemented with macros so the following
11509 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11512 Since @code{spu_add} is a macro, the vector constant in the example
11513 is treated as four separate arguments. Wrap the entire argument in
11514 parentheses for this to work.
11517 The extended version of @code{__builtin_expect} is not supported.
11521 @emph{Note:} Only the interface described in the aforementioned
11522 specification is supported. Internally, GCC uses built-in functions to
11523 implement the required functionality, but these are not supported and
11524 are subject to change without notice.
11526 @node Target Format Checks
11527 @section Format Checks Specific to Particular Target Machines
11529 For some target machines, GCC supports additional options to the
11531 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
11534 * Solaris Format Checks::
11537 @node Solaris Format Checks
11538 @subsection Solaris Format Checks
11540 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
11541 check. @code{cmn_err} accepts a subset of the standard @code{printf}
11542 conversions, and the two-argument @code{%b} conversion for displaying
11543 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
11546 @section Pragmas Accepted by GCC
11550 GCC supports several types of pragmas, primarily in order to compile
11551 code originally written for other compilers. Note that in general
11552 we do not recommend the use of pragmas; @xref{Function Attributes},
11553 for further explanation.
11558 * RS/6000 and PowerPC Pragmas::
11560 * Solaris Pragmas::
11561 * Symbol-Renaming Pragmas::
11562 * Structure-Packing Pragmas::
11564 * Diagnostic Pragmas::
11565 * Visibility Pragmas::
11566 * Push/Pop Macro Pragmas::
11567 * Function Specific Option Pragmas::
11571 @subsection ARM Pragmas
11573 The ARM target defines pragmas for controlling the default addition of
11574 @code{long_call} and @code{short_call} attributes to functions.
11575 @xref{Function Attributes}, for information about the effects of these
11580 @cindex pragma, long_calls
11581 Set all subsequent functions to have the @code{long_call} attribute.
11583 @item no_long_calls
11584 @cindex pragma, no_long_calls
11585 Set all subsequent functions to have the @code{short_call} attribute.
11587 @item long_calls_off
11588 @cindex pragma, long_calls_off
11589 Do not affect the @code{long_call} or @code{short_call} attributes of
11590 subsequent functions.
11594 @subsection M32C Pragmas
11597 @item memregs @var{number}
11598 @cindex pragma, memregs
11599 Overrides the command line option @code{-memregs=} for the current
11600 file. Use with care! This pragma must be before any function in the
11601 file, and mixing different memregs values in different objects may
11602 make them incompatible. This pragma is useful when a
11603 performance-critical function uses a memreg for temporary values,
11604 as it may allow you to reduce the number of memregs used.
11608 @node RS/6000 and PowerPC Pragmas
11609 @subsection RS/6000 and PowerPC Pragmas
11611 The RS/6000 and PowerPC targets define one pragma for controlling
11612 whether or not the @code{longcall} attribute is added to function
11613 declarations by default. This pragma overrides the @option{-mlongcall}
11614 option, but not the @code{longcall} and @code{shortcall} attributes.
11615 @xref{RS/6000 and PowerPC Options}, for more information about when long
11616 calls are and are not necessary.
11620 @cindex pragma, longcall
11621 Apply the @code{longcall} attribute to all subsequent function
11625 Do not apply the @code{longcall} attribute to subsequent function
11629 @c Describe h8300 pragmas here.
11630 @c Describe sh pragmas here.
11631 @c Describe v850 pragmas here.
11633 @node Darwin Pragmas
11634 @subsection Darwin Pragmas
11636 The following pragmas are available for all architectures running the
11637 Darwin operating system. These are useful for compatibility with other
11641 @item mark @var{tokens}@dots{}
11642 @cindex pragma, mark
11643 This pragma is accepted, but has no effect.
11645 @item options align=@var{alignment}
11646 @cindex pragma, options align
11647 This pragma sets the alignment of fields in structures. The values of
11648 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
11649 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
11650 properly; to restore the previous setting, use @code{reset} for the
11653 @item segment @var{tokens}@dots{}
11654 @cindex pragma, segment
11655 This pragma is accepted, but has no effect.
11657 @item unused (@var{var} [, @var{var}]@dots{})
11658 @cindex pragma, unused
11659 This pragma declares variables to be possibly unused. GCC will not
11660 produce warnings for the listed variables. The effect is similar to
11661 that of the @code{unused} attribute, except that this pragma may appear
11662 anywhere within the variables' scopes.
11665 @node Solaris Pragmas
11666 @subsection Solaris Pragmas
11668 The Solaris target supports @code{#pragma redefine_extname}
11669 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
11670 @code{#pragma} directives for compatibility with the system compiler.
11673 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
11674 @cindex pragma, align
11676 Increase the minimum alignment of each @var{variable} to @var{alignment}.
11677 This is the same as GCC's @code{aligned} attribute @pxref{Variable
11678 Attributes}). Macro expansion occurs on the arguments to this pragma
11679 when compiling C and Objective-C@. It does not currently occur when
11680 compiling C++, but this is a bug which may be fixed in a future
11683 @item fini (@var{function} [, @var{function}]...)
11684 @cindex pragma, fini
11686 This pragma causes each listed @var{function} to be called after
11687 main, or during shared module unloading, by adding a call to the
11688 @code{.fini} section.
11690 @item init (@var{function} [, @var{function}]...)
11691 @cindex pragma, init
11693 This pragma causes each listed @var{function} to be called during
11694 initialization (before @code{main}) or during shared module loading, by
11695 adding a call to the @code{.init} section.
11699 @node Symbol-Renaming Pragmas
11700 @subsection Symbol-Renaming Pragmas
11702 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
11703 supports two @code{#pragma} directives which change the name used in
11704 assembly for a given declaration. These pragmas are only available on
11705 platforms whose system headers need them. To get this effect on all
11706 platforms supported by GCC, use the asm labels extension (@pxref{Asm
11710 @item redefine_extname @var{oldname} @var{newname}
11711 @cindex pragma, redefine_extname
11713 This pragma gives the C function @var{oldname} the assembly symbol
11714 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
11715 will be defined if this pragma is available (currently only on
11718 @item extern_prefix @var{string}
11719 @cindex pragma, extern_prefix
11721 This pragma causes all subsequent external function and variable
11722 declarations to have @var{string} prepended to their assembly symbols.
11723 This effect may be terminated with another @code{extern_prefix} pragma
11724 whose argument is an empty string. The preprocessor macro
11725 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
11726 available (currently only on Tru64 UNIX)@.
11729 These pragmas and the asm labels extension interact in a complicated
11730 manner. Here are some corner cases you may want to be aware of.
11733 @item Both pragmas silently apply only to declarations with external
11734 linkage. Asm labels do not have this restriction.
11736 @item In C++, both pragmas silently apply only to declarations with
11737 ``C'' linkage. Again, asm labels do not have this restriction.
11739 @item If any of the three ways of changing the assembly name of a
11740 declaration is applied to a declaration whose assembly name has
11741 already been determined (either by a previous use of one of these
11742 features, or because the compiler needed the assembly name in order to
11743 generate code), and the new name is different, a warning issues and
11744 the name does not change.
11746 @item The @var{oldname} used by @code{#pragma redefine_extname} is
11747 always the C-language name.
11749 @item If @code{#pragma extern_prefix} is in effect, and a declaration
11750 occurs with an asm label attached, the prefix is silently ignored for
11753 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
11754 apply to the same declaration, whichever triggered first wins, and a
11755 warning issues if they contradict each other. (We would like to have
11756 @code{#pragma redefine_extname} always win, for consistency with asm
11757 labels, but if @code{#pragma extern_prefix} triggers first we have no
11758 way of knowing that that happened.)
11761 @node Structure-Packing Pragmas
11762 @subsection Structure-Packing Pragmas
11764 For compatibility with Microsoft Windows compilers, GCC supports a
11765 set of @code{#pragma} directives which change the maximum alignment of
11766 members of structures (other than zero-width bitfields), unions, and
11767 classes subsequently defined. The @var{n} value below always is required
11768 to be a small power of two and specifies the new alignment in bytes.
11771 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
11772 @item @code{#pragma pack()} sets the alignment to the one that was in
11773 effect when compilation started (see also command line option
11774 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
11775 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
11776 setting on an internal stack and then optionally sets the new alignment.
11777 @item @code{#pragma pack(pop)} restores the alignment setting to the one
11778 saved at the top of the internal stack (and removes that stack entry).
11779 Note that @code{#pragma pack([@var{n}])} does not influence this internal
11780 stack; thus it is possible to have @code{#pragma pack(push)} followed by
11781 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
11782 @code{#pragma pack(pop)}.
11785 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
11786 @code{#pragma} which lays out a structure as the documented
11787 @code{__attribute__ ((ms_struct))}.
11789 @item @code{#pragma ms_struct on} turns on the layout for structures
11791 @item @code{#pragma ms_struct off} turns off the layout for structures
11793 @item @code{#pragma ms_struct reset} goes back to the default layout.
11797 @subsection Weak Pragmas
11799 For compatibility with SVR4, GCC supports a set of @code{#pragma}
11800 directives for declaring symbols to be weak, and defining weak
11804 @item #pragma weak @var{symbol}
11805 @cindex pragma, weak
11806 This pragma declares @var{symbol} to be weak, as if the declaration
11807 had the attribute of the same name. The pragma may appear before
11808 or after the declaration of @var{symbol}, but must appear before
11809 either its first use or its definition. It is not an error for
11810 @var{symbol} to never be defined at all.
11812 @item #pragma weak @var{symbol1} = @var{symbol2}
11813 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
11814 It is an error if @var{symbol2} is not defined in the current
11818 @node Diagnostic Pragmas
11819 @subsection Diagnostic Pragmas
11821 GCC allows the user to selectively enable or disable certain types of
11822 diagnostics, and change the kind of the diagnostic. For example, a
11823 project's policy might require that all sources compile with
11824 @option{-Werror} but certain files might have exceptions allowing
11825 specific types of warnings. Or, a project might selectively enable
11826 diagnostics and treat them as errors depending on which preprocessor
11827 macros are defined.
11830 @item #pragma GCC diagnostic @var{kind} @var{option}
11831 @cindex pragma, diagnostic
11833 Modifies the disposition of a diagnostic. Note that not all
11834 diagnostics are modifiable; at the moment only warnings (normally
11835 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
11836 Use @option{-fdiagnostics-show-option} to determine which diagnostics
11837 are controllable and which option controls them.
11839 @var{kind} is @samp{error} to treat this diagnostic as an error,
11840 @samp{warning} to treat it like a warning (even if @option{-Werror} is
11841 in effect), or @samp{ignored} if the diagnostic is to be ignored.
11842 @var{option} is a double quoted string which matches the command line
11846 #pragma GCC diagnostic warning "-Wformat"
11847 #pragma GCC diagnostic error "-Wformat"
11848 #pragma GCC diagnostic ignored "-Wformat"
11851 Note that these pragmas override any command line options. Also,
11852 while it is syntactically valid to put these pragmas anywhere in your
11853 sources, the only supported location for them is before any data or
11854 functions are defined. Doing otherwise may result in unpredictable
11855 results depending on how the optimizer manages your sources. If the
11856 same option is listed multiple times, the last one specified is the
11857 one that is in effect. This pragma is not intended to be a general
11858 purpose replacement for command line options, but for implementing
11859 strict control over project policies.
11863 GCC also offers a simple mechanism for printing messages during
11867 @item #pragma message @var{string}
11868 @cindex pragma, diagnostic
11870 Prints @var{string} as a compiler message on compilation. The message
11871 is informational only, and is neither a compilation warning nor an error.
11874 #pragma message "Compiling " __FILE__ "..."
11877 @var{string} may be parenthesized, and is printed with location
11878 information. For example,
11881 #define DO_PRAGMA(x) _Pragma (#x)
11882 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
11884 TODO(Remember to fix this)
11887 prints @samp{/tmp/file.c:4: note: #pragma message:
11888 TODO - Remember to fix this}.
11892 @node Visibility Pragmas
11893 @subsection Visibility Pragmas
11896 @item #pragma GCC visibility push(@var{visibility})
11897 @itemx #pragma GCC visibility pop
11898 @cindex pragma, visibility
11900 This pragma allows the user to set the visibility for multiple
11901 declarations without having to give each a visibility attribute
11902 @xref{Function Attributes}, for more information about visibility and
11903 the attribute syntax.
11905 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
11906 declarations. Class members and template specializations are not
11907 affected; if you want to override the visibility for a particular
11908 member or instantiation, you must use an attribute.
11913 @node Push/Pop Macro Pragmas
11914 @subsection Push/Pop Macro Pragmas
11916 For compatibility with Microsoft Windows compilers, GCC supports
11917 @samp{#pragma push_macro(@var{"macro_name"})}
11918 and @samp{#pragma pop_macro(@var{"macro_name"})}.
11921 @item #pragma push_macro(@var{"macro_name"})
11922 @cindex pragma, push_macro
11923 This pragma saves the value of the macro named as @var{macro_name} to
11924 the top of the stack for this macro.
11926 @item #pragma pop_macro(@var{"macro_name"})
11927 @cindex pragma, pop_macro
11928 This pragma sets the value of the macro named as @var{macro_name} to
11929 the value on top of the stack for this macro. If the stack for
11930 @var{macro_name} is empty, the value of the macro remains unchanged.
11937 #pragma push_macro("X")
11940 #pragma pop_macro("X")
11944 In this example, the definition of X as 1 is saved by @code{#pragma
11945 push_macro} and restored by @code{#pragma pop_macro}.
11947 @node Function Specific Option Pragmas
11948 @subsection Function Specific Option Pragmas
11951 @item #pragma GCC target (@var{"string"}...)
11952 @cindex pragma GCC target
11954 This pragma allows you to set target specific options for functions
11955 defined later in the source file. One or more strings can be
11956 specified. Each function that is defined after this point will be as
11957 if @code{attribute((target("STRING")))} was specified for that
11958 function. The parenthesis around the options is optional.
11959 @xref{Function Attributes}, for more information about the
11960 @code{target} attribute and the attribute syntax.
11962 The @samp{#pragma GCC target} pragma is not implemented in GCC
11963 versions earlier than 4.4, and is currently only implemented for the
11964 386 and x86_64 backends.
11968 @item #pragma GCC optimize (@var{"string"}...)
11969 @cindex pragma GCC optimize
11971 This pragma allows you to set global optimization options for functions
11972 defined later in the source file. One or more strings can be
11973 specified. Each function that is defined after this point will be as
11974 if @code{attribute((optimize("STRING")))} was specified for that
11975 function. The parenthesis around the options is optional.
11976 @xref{Function Attributes}, for more information about the
11977 @code{optimize} attribute and the attribute syntax.
11979 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
11980 versions earlier than 4.4.
11984 @item #pragma GCC push_options
11985 @itemx #pragma GCC pop_options
11986 @cindex pragma GCC push_options
11987 @cindex pragma GCC pop_options
11989 These pragmas maintain a stack of the current target and optimization
11990 options. It is intended for include files where you temporarily want
11991 to switch to using a different @samp{#pragma GCC target} or
11992 @samp{#pragma GCC optimize} and then to pop back to the previous
11995 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
11996 pragmas are not implemented in GCC versions earlier than 4.4.
12000 @item #pragma GCC reset_options
12001 @cindex pragma GCC reset_options
12003 This pragma clears the current @code{#pragma GCC target} and
12004 @code{#pragma GCC optimize} to use the default switches as specified
12005 on the command line.
12007 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
12008 versions earlier than 4.4.
12011 @node Unnamed Fields
12012 @section Unnamed struct/union fields within structs/unions
12016 For compatibility with other compilers, GCC allows you to define
12017 a structure or union that contains, as fields, structures and unions
12018 without names. For example:
12031 In this example, the user would be able to access members of the unnamed
12032 union with code like @samp{foo.b}. Note that only unnamed structs and
12033 unions are allowed, you may not have, for example, an unnamed
12036 You must never create such structures that cause ambiguous field definitions.
12037 For example, this structure:
12048 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
12049 Such constructs are not supported and must be avoided. In the future,
12050 such constructs may be detected and treated as compilation errors.
12052 @opindex fms-extensions
12053 Unless @option{-fms-extensions} is used, the unnamed field must be a
12054 structure or union definition without a tag (for example, @samp{struct
12055 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
12056 also be a definition with a tag such as @samp{struct foo @{ int a;
12057 @};}, a reference to a previously defined structure or union such as
12058 @samp{struct foo;}, or a reference to a @code{typedef} name for a
12059 previously defined structure or union type.
12062 @section Thread-Local Storage
12063 @cindex Thread-Local Storage
12064 @cindex @acronym{TLS}
12067 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
12068 are allocated such that there is one instance of the variable per extant
12069 thread. The run-time model GCC uses to implement this originates
12070 in the IA-64 processor-specific ABI, but has since been migrated
12071 to other processors as well. It requires significant support from
12072 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
12073 system libraries (@file{libc.so} and @file{libpthread.so}), so it
12074 is not available everywhere.
12076 At the user level, the extension is visible with a new storage
12077 class keyword: @code{__thread}. For example:
12081 extern __thread struct state s;
12082 static __thread char *p;
12085 The @code{__thread} specifier may be used alone, with the @code{extern}
12086 or @code{static} specifiers, but with no other storage class specifier.
12087 When used with @code{extern} or @code{static}, @code{__thread} must appear
12088 immediately after the other storage class specifier.
12090 The @code{__thread} specifier may be applied to any global, file-scoped
12091 static, function-scoped static, or static data member of a class. It may
12092 not be applied to block-scoped automatic or non-static data member.
12094 When the address-of operator is applied to a thread-local variable, it is
12095 evaluated at run-time and returns the address of the current thread's
12096 instance of that variable. An address so obtained may be used by any
12097 thread. When a thread terminates, any pointers to thread-local variables
12098 in that thread become invalid.
12100 No static initialization may refer to the address of a thread-local variable.
12102 In C++, if an initializer is present for a thread-local variable, it must
12103 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
12106 See @uref{http://people.redhat.com/drepper/tls.pdf,
12107 ELF Handling For Thread-Local Storage} for a detailed explanation of
12108 the four thread-local storage addressing models, and how the run-time
12109 is expected to function.
12112 * C99 Thread-Local Edits::
12113 * C++98 Thread-Local Edits::
12116 @node C99 Thread-Local Edits
12117 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
12119 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
12120 that document the exact semantics of the language extension.
12124 @cite{5.1.2 Execution environments}
12126 Add new text after paragraph 1
12129 Within either execution environment, a @dfn{thread} is a flow of
12130 control within a program. It is implementation defined whether
12131 or not there may be more than one thread associated with a program.
12132 It is implementation defined how threads beyond the first are
12133 created, the name and type of the function called at thread
12134 startup, and how threads may be terminated. However, objects
12135 with thread storage duration shall be initialized before thread
12140 @cite{6.2.4 Storage durations of objects}
12142 Add new text before paragraph 3
12145 An object whose identifier is declared with the storage-class
12146 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
12147 Its lifetime is the entire execution of the thread, and its
12148 stored value is initialized only once, prior to thread startup.
12152 @cite{6.4.1 Keywords}
12154 Add @code{__thread}.
12157 @cite{6.7.1 Storage-class specifiers}
12159 Add @code{__thread} to the list of storage class specifiers in
12162 Change paragraph 2 to
12165 With the exception of @code{__thread}, at most one storage-class
12166 specifier may be given [@dots{}]. The @code{__thread} specifier may
12167 be used alone, or immediately following @code{extern} or
12171 Add new text after paragraph 6
12174 The declaration of an identifier for a variable that has
12175 block scope that specifies @code{__thread} shall also
12176 specify either @code{extern} or @code{static}.
12178 The @code{__thread} specifier shall be used only with
12183 @node C++98 Thread-Local Edits
12184 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
12186 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
12187 that document the exact semantics of the language extension.
12191 @b{[intro.execution]}
12193 New text after paragraph 4
12196 A @dfn{thread} is a flow of control within the abstract machine.
12197 It is implementation defined whether or not there may be more than
12201 New text after paragraph 7
12204 It is unspecified whether additional action must be taken to
12205 ensure when and whether side effects are visible to other threads.
12211 Add @code{__thread}.
12214 @b{[basic.start.main]}
12216 Add after paragraph 5
12219 The thread that begins execution at the @code{main} function is called
12220 the @dfn{main thread}. It is implementation defined how functions
12221 beginning threads other than the main thread are designated or typed.
12222 A function so designated, as well as the @code{main} function, is called
12223 a @dfn{thread startup function}. It is implementation defined what
12224 happens if a thread startup function returns. It is implementation
12225 defined what happens to other threads when any thread calls @code{exit}.
12229 @b{[basic.start.init]}
12231 Add after paragraph 4
12234 The storage for an object of thread storage duration shall be
12235 statically initialized before the first statement of the thread startup
12236 function. An object of thread storage duration shall not require
12237 dynamic initialization.
12241 @b{[basic.start.term]}
12243 Add after paragraph 3
12246 The type of an object with thread storage duration shall not have a
12247 non-trivial destructor, nor shall it be an array type whose elements
12248 (directly or indirectly) have non-trivial destructors.
12254 Add ``thread storage duration'' to the list in paragraph 1.
12259 Thread, static, and automatic storage durations are associated with
12260 objects introduced by declarations [@dots{}].
12263 Add @code{__thread} to the list of specifiers in paragraph 3.
12266 @b{[basic.stc.thread]}
12268 New section before @b{[basic.stc.static]}
12271 The keyword @code{__thread} applied to a non-local object gives the
12272 object thread storage duration.
12274 A local variable or class data member declared both @code{static}
12275 and @code{__thread} gives the variable or member thread storage
12280 @b{[basic.stc.static]}
12285 All objects which have neither thread storage duration, dynamic
12286 storage duration nor are local [@dots{}].
12292 Add @code{__thread} to the list in paragraph 1.
12297 With the exception of @code{__thread}, at most one
12298 @var{storage-class-specifier} shall appear in a given
12299 @var{decl-specifier-seq}. The @code{__thread} specifier may
12300 be used alone, or immediately following the @code{extern} or
12301 @code{static} specifiers. [@dots{}]
12304 Add after paragraph 5
12307 The @code{__thread} specifier can be applied only to the names of objects
12308 and to anonymous unions.
12314 Add after paragraph 6
12317 Non-@code{static} members shall not be @code{__thread}.
12321 @node Binary constants
12322 @section Binary constants using the @samp{0b} prefix
12323 @cindex Binary constants using the @samp{0b} prefix
12325 Integer constants can be written as binary constants, consisting of a
12326 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
12327 @samp{0B}. This is particularly useful in environments that operate a
12328 lot on the bit-level (like microcontrollers).
12330 The following statements are identical:
12339 The type of these constants follows the same rules as for octal or
12340 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
12343 @node C++ Extensions
12344 @chapter Extensions to the C++ Language
12345 @cindex extensions, C++ language
12346 @cindex C++ language extensions
12348 The GNU compiler provides these extensions to the C++ language (and you
12349 can also use most of the C language extensions in your C++ programs). If you
12350 want to write code that checks whether these features are available, you can
12351 test for the GNU compiler the same way as for C programs: check for a
12352 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
12353 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
12354 Predefined Macros,cpp,The GNU C Preprocessor}).
12357 * Volatiles:: What constitutes an access to a volatile object.
12358 * Restricted Pointers:: C99 restricted pointers and references.
12359 * Vague Linkage:: Where G++ puts inlines, vtables and such.
12360 * C++ Interface:: You can use a single C++ header file for both
12361 declarations and definitions.
12362 * Template Instantiation:: Methods for ensuring that exactly one copy of
12363 each needed template instantiation is emitted.
12364 * Bound member functions:: You can extract a function pointer to the
12365 method denoted by a @samp{->*} or @samp{.*} expression.
12366 * C++ Attributes:: Variable, function, and type attributes for C++ only.
12367 * Namespace Association:: Strong using-directives for namespace association.
12368 * Type Traits:: Compiler support for type traits
12369 * Java Exceptions:: Tweaking exception handling to work with Java.
12370 * Deprecated Features:: Things will disappear from g++.
12371 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
12375 @section When is a Volatile Object Accessed?
12376 @cindex accessing volatiles
12377 @cindex volatile read
12378 @cindex volatile write
12379 @cindex volatile access
12381 Both the C and C++ standard have the concept of volatile objects. These
12382 are normally accessed by pointers and used for accessing hardware. The
12383 standards encourage compilers to refrain from optimizations concerning
12384 accesses to volatile objects. The C standard leaves it implementation
12385 defined as to what constitutes a volatile access. The C++ standard omits
12386 to specify this, except to say that C++ should behave in a similar manner
12387 to C with respect to volatiles, where possible. The minimum either
12388 standard specifies is that at a sequence point all previous accesses to
12389 volatile objects have stabilized and no subsequent accesses have
12390 occurred. Thus an implementation is free to reorder and combine
12391 volatile accesses which occur between sequence points, but cannot do so
12392 for accesses across a sequence point. The use of volatiles does not
12393 allow you to violate the restriction on updating objects multiple times
12394 within a sequence point.
12396 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
12398 The behavior differs slightly between C and C++ in the non-obvious cases:
12401 volatile int *src = @var{somevalue};
12405 With C, such expressions are rvalues, and GCC interprets this either as a
12406 read of the volatile object being pointed to or only as request to evaluate
12407 the side-effects. The C++ standard specifies that such expressions do not
12408 undergo lvalue to rvalue conversion, and that the type of the dereferenced
12409 object may be incomplete. The C++ standard does not specify explicitly
12410 that it is this lvalue to rvalue conversion which may be responsible for
12411 causing an access. However, there is reason to believe that it is,
12412 because otherwise certain simple expressions become undefined. However,
12413 because it would surprise most programmers, G++ treats dereferencing a
12414 pointer to volatile object of complete type when the value is unused as
12415 GCC would do for an equivalent type in C@. When the object has incomplete
12416 type, G++ issues a warning; if you wish to force an error, you must
12417 force a conversion to rvalue with, for instance, a static cast.
12419 When using a reference to volatile, G++ does not treat equivalent
12420 expressions as accesses to volatiles, but instead issues a warning that
12421 no volatile is accessed. The rationale for this is that otherwise it
12422 becomes difficult to determine where volatile access occur, and not
12423 possible to ignore the return value from functions returning volatile
12424 references. Again, if you wish to force a read, cast the reference to
12427 @node Restricted Pointers
12428 @section Restricting Pointer Aliasing
12429 @cindex restricted pointers
12430 @cindex restricted references
12431 @cindex restricted this pointer
12433 As with the C front end, G++ understands the C99 feature of restricted pointers,
12434 specified with the @code{__restrict__}, or @code{__restrict} type
12435 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
12436 language flag, @code{restrict} is not a keyword in C++.
12438 In addition to allowing restricted pointers, you can specify restricted
12439 references, which indicate that the reference is not aliased in the local
12443 void fn (int *__restrict__ rptr, int &__restrict__ rref)
12450 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
12451 @var{rref} refers to a (different) unaliased integer.
12453 You may also specify whether a member function's @var{this} pointer is
12454 unaliased by using @code{__restrict__} as a member function qualifier.
12457 void T::fn () __restrict__
12464 Within the body of @code{T::fn}, @var{this} will have the effective
12465 definition @code{T *__restrict__ const this}. Notice that the
12466 interpretation of a @code{__restrict__} member function qualifier is
12467 different to that of @code{const} or @code{volatile} qualifier, in that it
12468 is applied to the pointer rather than the object. This is consistent with
12469 other compilers which implement restricted pointers.
12471 As with all outermost parameter qualifiers, @code{__restrict__} is
12472 ignored in function definition matching. This means you only need to
12473 specify @code{__restrict__} in a function definition, rather than
12474 in a function prototype as well.
12476 @node Vague Linkage
12477 @section Vague Linkage
12478 @cindex vague linkage
12480 There are several constructs in C++ which require space in the object
12481 file but are not clearly tied to a single translation unit. We say that
12482 these constructs have ``vague linkage''. Typically such constructs are
12483 emitted wherever they are needed, though sometimes we can be more
12487 @item Inline Functions
12488 Inline functions are typically defined in a header file which can be
12489 included in many different compilations. Hopefully they can usually be
12490 inlined, but sometimes an out-of-line copy is necessary, if the address
12491 of the function is taken or if inlining fails. In general, we emit an
12492 out-of-line copy in all translation units where one is needed. As an
12493 exception, we only emit inline virtual functions with the vtable, since
12494 it will always require a copy.
12496 Local static variables and string constants used in an inline function
12497 are also considered to have vague linkage, since they must be shared
12498 between all inlined and out-of-line instances of the function.
12502 C++ virtual functions are implemented in most compilers using a lookup
12503 table, known as a vtable. The vtable contains pointers to the virtual
12504 functions provided by a class, and each object of the class contains a
12505 pointer to its vtable (or vtables, in some multiple-inheritance
12506 situations). If the class declares any non-inline, non-pure virtual
12507 functions, the first one is chosen as the ``key method'' for the class,
12508 and the vtable is only emitted in the translation unit where the key
12511 @emph{Note:} If the chosen key method is later defined as inline, the
12512 vtable will still be emitted in every translation unit which defines it.
12513 Make sure that any inline virtuals are declared inline in the class
12514 body, even if they are not defined there.
12516 @item type_info objects
12519 C++ requires information about types to be written out in order to
12520 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
12521 For polymorphic classes (classes with virtual functions), the type_info
12522 object is written out along with the vtable so that @samp{dynamic_cast}
12523 can determine the dynamic type of a class object at runtime. For all
12524 other types, we write out the type_info object when it is used: when
12525 applying @samp{typeid} to an expression, throwing an object, or
12526 referring to a type in a catch clause or exception specification.
12528 @item Template Instantiations
12529 Most everything in this section also applies to template instantiations,
12530 but there are other options as well.
12531 @xref{Template Instantiation,,Where's the Template?}.
12535 When used with GNU ld version 2.8 or later on an ELF system such as
12536 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
12537 these constructs will be discarded at link time. This is known as
12540 On targets that don't support COMDAT, but do support weak symbols, GCC
12541 will use them. This way one copy will override all the others, but
12542 the unused copies will still take up space in the executable.
12544 For targets which do not support either COMDAT or weak symbols,
12545 most entities with vague linkage will be emitted as local symbols to
12546 avoid duplicate definition errors from the linker. This will not happen
12547 for local statics in inlines, however, as having multiple copies will
12548 almost certainly break things.
12550 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
12551 another way to control placement of these constructs.
12553 @node C++ Interface
12554 @section #pragma interface and implementation
12556 @cindex interface and implementation headers, C++
12557 @cindex C++ interface and implementation headers
12558 @cindex pragmas, interface and implementation
12560 @code{#pragma interface} and @code{#pragma implementation} provide the
12561 user with a way of explicitly directing the compiler to emit entities
12562 with vague linkage (and debugging information) in a particular
12565 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
12566 most cases, because of COMDAT support and the ``key method'' heuristic
12567 mentioned in @ref{Vague Linkage}. Using them can actually cause your
12568 program to grow due to unnecessary out-of-line copies of inline
12569 functions. Currently (3.4) the only benefit of these
12570 @code{#pragma}s is reduced duplication of debugging information, and
12571 that should be addressed soon on DWARF 2 targets with the use of
12575 @item #pragma interface
12576 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
12577 @kindex #pragma interface
12578 Use this directive in @emph{header files} that define object classes, to save
12579 space in most of the object files that use those classes. Normally,
12580 local copies of certain information (backup copies of inline member
12581 functions, debugging information, and the internal tables that implement
12582 virtual functions) must be kept in each object file that includes class
12583 definitions. You can use this pragma to avoid such duplication. When a
12584 header file containing @samp{#pragma interface} is included in a
12585 compilation, this auxiliary information will not be generated (unless
12586 the main input source file itself uses @samp{#pragma implementation}).
12587 Instead, the object files will contain references to be resolved at link
12590 The second form of this directive is useful for the case where you have
12591 multiple headers with the same name in different directories. If you
12592 use this form, you must specify the same string to @samp{#pragma
12595 @item #pragma implementation
12596 @itemx #pragma implementation "@var{objects}.h"
12597 @kindex #pragma implementation
12598 Use this pragma in a @emph{main input file}, when you want full output from
12599 included header files to be generated (and made globally visible). The
12600 included header file, in turn, should use @samp{#pragma interface}.
12601 Backup copies of inline member functions, debugging information, and the
12602 internal tables used to implement virtual functions are all generated in
12603 implementation files.
12605 @cindex implied @code{#pragma implementation}
12606 @cindex @code{#pragma implementation}, implied
12607 @cindex naming convention, implementation headers
12608 If you use @samp{#pragma implementation} with no argument, it applies to
12609 an include file with the same basename@footnote{A file's @dfn{basename}
12610 was the name stripped of all leading path information and of trailing
12611 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
12612 file. For example, in @file{allclass.cc}, giving just
12613 @samp{#pragma implementation}
12614 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
12616 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
12617 an implementation file whenever you would include it from
12618 @file{allclass.cc} even if you never specified @samp{#pragma
12619 implementation}. This was deemed to be more trouble than it was worth,
12620 however, and disabled.
12622 Use the string argument if you want a single implementation file to
12623 include code from multiple header files. (You must also use
12624 @samp{#include} to include the header file; @samp{#pragma
12625 implementation} only specifies how to use the file---it doesn't actually
12628 There is no way to split up the contents of a single header file into
12629 multiple implementation files.
12632 @cindex inlining and C++ pragmas
12633 @cindex C++ pragmas, effect on inlining
12634 @cindex pragmas in C++, effect on inlining
12635 @samp{#pragma implementation} and @samp{#pragma interface} also have an
12636 effect on function inlining.
12638 If you define a class in a header file marked with @samp{#pragma
12639 interface}, the effect on an inline function defined in that class is
12640 similar to an explicit @code{extern} declaration---the compiler emits
12641 no code at all to define an independent version of the function. Its
12642 definition is used only for inlining with its callers.
12644 @opindex fno-implement-inlines
12645 Conversely, when you include the same header file in a main source file
12646 that declares it as @samp{#pragma implementation}, the compiler emits
12647 code for the function itself; this defines a version of the function
12648 that can be found via pointers (or by callers compiled without
12649 inlining). If all calls to the function can be inlined, you can avoid
12650 emitting the function by compiling with @option{-fno-implement-inlines}.
12651 If any calls were not inlined, you will get linker errors.
12653 @node Template Instantiation
12654 @section Where's the Template?
12655 @cindex template instantiation
12657 C++ templates are the first language feature to require more
12658 intelligence from the environment than one usually finds on a UNIX
12659 system. Somehow the compiler and linker have to make sure that each
12660 template instance occurs exactly once in the executable if it is needed,
12661 and not at all otherwise. There are two basic approaches to this
12662 problem, which are referred to as the Borland model and the Cfront model.
12665 @item Borland model
12666 Borland C++ solved the template instantiation problem by adding the code
12667 equivalent of common blocks to their linker; the compiler emits template
12668 instances in each translation unit that uses them, and the linker
12669 collapses them together. The advantage of this model is that the linker
12670 only has to consider the object files themselves; there is no external
12671 complexity to worry about. This disadvantage is that compilation time
12672 is increased because the template code is being compiled repeatedly.
12673 Code written for this model tends to include definitions of all
12674 templates in the header file, since they must be seen to be
12678 The AT&T C++ translator, Cfront, solved the template instantiation
12679 problem by creating the notion of a template repository, an
12680 automatically maintained place where template instances are stored. A
12681 more modern version of the repository works as follows: As individual
12682 object files are built, the compiler places any template definitions and
12683 instantiations encountered in the repository. At link time, the link
12684 wrapper adds in the objects in the repository and compiles any needed
12685 instances that were not previously emitted. The advantages of this
12686 model are more optimal compilation speed and the ability to use the
12687 system linker; to implement the Borland model a compiler vendor also
12688 needs to replace the linker. The disadvantages are vastly increased
12689 complexity, and thus potential for error; for some code this can be
12690 just as transparent, but in practice it can been very difficult to build
12691 multiple programs in one directory and one program in multiple
12692 directories. Code written for this model tends to separate definitions
12693 of non-inline member templates into a separate file, which should be
12694 compiled separately.
12697 When used with GNU ld version 2.8 or later on an ELF system such as
12698 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
12699 Borland model. On other systems, G++ implements neither automatic
12702 A future version of G++ will support a hybrid model whereby the compiler
12703 will emit any instantiations for which the template definition is
12704 included in the compile, and store template definitions and
12705 instantiation context information into the object file for the rest.
12706 The link wrapper will extract that information as necessary and invoke
12707 the compiler to produce the remaining instantiations. The linker will
12708 then combine duplicate instantiations.
12710 In the mean time, you have the following options for dealing with
12711 template instantiations:
12716 Compile your template-using code with @option{-frepo}. The compiler will
12717 generate files with the extension @samp{.rpo} listing all of the
12718 template instantiations used in the corresponding object files which
12719 could be instantiated there; the link wrapper, @samp{collect2}, will
12720 then update the @samp{.rpo} files to tell the compiler where to place
12721 those instantiations and rebuild any affected object files. The
12722 link-time overhead is negligible after the first pass, as the compiler
12723 will continue to place the instantiations in the same files.
12725 This is your best option for application code written for the Borland
12726 model, as it will just work. Code written for the Cfront model will
12727 need to be modified so that the template definitions are available at
12728 one or more points of instantiation; usually this is as simple as adding
12729 @code{#include <tmethods.cc>} to the end of each template header.
12731 For library code, if you want the library to provide all of the template
12732 instantiations it needs, just try to link all of its object files
12733 together; the link will fail, but cause the instantiations to be
12734 generated as a side effect. Be warned, however, that this may cause
12735 conflicts if multiple libraries try to provide the same instantiations.
12736 For greater control, use explicit instantiation as described in the next
12740 @opindex fno-implicit-templates
12741 Compile your code with @option{-fno-implicit-templates} to disable the
12742 implicit generation of template instances, and explicitly instantiate
12743 all the ones you use. This approach requires more knowledge of exactly
12744 which instances you need than do the others, but it's less
12745 mysterious and allows greater control. You can scatter the explicit
12746 instantiations throughout your program, perhaps putting them in the
12747 translation units where the instances are used or the translation units
12748 that define the templates themselves; you can put all of the explicit
12749 instantiations you need into one big file; or you can create small files
12756 template class Foo<int>;
12757 template ostream& operator <<
12758 (ostream&, const Foo<int>&);
12761 for each of the instances you need, and create a template instantiation
12762 library from those.
12764 If you are using Cfront-model code, you can probably get away with not
12765 using @option{-fno-implicit-templates} when compiling files that don't
12766 @samp{#include} the member template definitions.
12768 If you use one big file to do the instantiations, you may want to
12769 compile it without @option{-fno-implicit-templates} so you get all of the
12770 instances required by your explicit instantiations (but not by any
12771 other files) without having to specify them as well.
12773 G++ has extended the template instantiation syntax given in the ISO
12774 standard to allow forward declaration of explicit instantiations
12775 (with @code{extern}), instantiation of the compiler support data for a
12776 template class (i.e.@: the vtable) without instantiating any of its
12777 members (with @code{inline}), and instantiation of only the static data
12778 members of a template class, without the support data or member
12779 functions (with (@code{static}):
12782 extern template int max (int, int);
12783 inline template class Foo<int>;
12784 static template class Foo<int>;
12788 Do nothing. Pretend G++ does implement automatic instantiation
12789 management. Code written for the Borland model will work fine, but
12790 each translation unit will contain instances of each of the templates it
12791 uses. In a large program, this can lead to an unacceptable amount of code
12795 @node Bound member functions
12796 @section Extracting the function pointer from a bound pointer to member function
12798 @cindex pointer to member function
12799 @cindex bound pointer to member function
12801 In C++, pointer to member functions (PMFs) are implemented using a wide
12802 pointer of sorts to handle all the possible call mechanisms; the PMF
12803 needs to store information about how to adjust the @samp{this} pointer,
12804 and if the function pointed to is virtual, where to find the vtable, and
12805 where in the vtable to look for the member function. If you are using
12806 PMFs in an inner loop, you should really reconsider that decision. If
12807 that is not an option, you can extract the pointer to the function that
12808 would be called for a given object/PMF pair and call it directly inside
12809 the inner loop, to save a bit of time.
12811 Note that you will still be paying the penalty for the call through a
12812 function pointer; on most modern architectures, such a call defeats the
12813 branch prediction features of the CPU@. This is also true of normal
12814 virtual function calls.
12816 The syntax for this extension is
12820 extern int (A::*fp)();
12821 typedef int (*fptr)(A *);
12823 fptr p = (fptr)(a.*fp);
12826 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
12827 no object is needed to obtain the address of the function. They can be
12828 converted to function pointers directly:
12831 fptr p1 = (fptr)(&A::foo);
12834 @opindex Wno-pmf-conversions
12835 You must specify @option{-Wno-pmf-conversions} to use this extension.
12837 @node C++ Attributes
12838 @section C++-Specific Variable, Function, and Type Attributes
12840 Some attributes only make sense for C++ programs.
12843 @item init_priority (@var{priority})
12844 @cindex init_priority attribute
12847 In Standard C++, objects defined at namespace scope are guaranteed to be
12848 initialized in an order in strict accordance with that of their definitions
12849 @emph{in a given translation unit}. No guarantee is made for initializations
12850 across translation units. However, GNU C++ allows users to control the
12851 order of initialization of objects defined at namespace scope with the
12852 @code{init_priority} attribute by specifying a relative @var{priority},
12853 a constant integral expression currently bounded between 101 and 65535
12854 inclusive. Lower numbers indicate a higher priority.
12856 In the following example, @code{A} would normally be created before
12857 @code{B}, but the @code{init_priority} attribute has reversed that order:
12860 Some_Class A __attribute__ ((init_priority (2000)));
12861 Some_Class B __attribute__ ((init_priority (543)));
12865 Note that the particular values of @var{priority} do not matter; only their
12868 @item java_interface
12869 @cindex java_interface attribute
12871 This type attribute informs C++ that the class is a Java interface. It may
12872 only be applied to classes declared within an @code{extern "Java"} block.
12873 Calls to methods declared in this interface will be dispatched using GCJ's
12874 interface table mechanism, instead of regular virtual table dispatch.
12878 See also @ref{Namespace Association}.
12880 @node Namespace Association
12881 @section Namespace Association
12883 @strong{Caution:} The semantics of this extension are not fully
12884 defined. Users should refrain from using this extension as its
12885 semantics may change subtly over time. It is possible that this
12886 extension will be removed in future versions of G++.
12888 A using-directive with @code{__attribute ((strong))} is stronger
12889 than a normal using-directive in two ways:
12893 Templates from the used namespace can be specialized and explicitly
12894 instantiated as though they were members of the using namespace.
12897 The using namespace is considered an associated namespace of all
12898 templates in the used namespace for purposes of argument-dependent
12902 The used namespace must be nested within the using namespace so that
12903 normal unqualified lookup works properly.
12905 This is useful for composing a namespace transparently from
12906 implementation namespaces. For example:
12911 template <class T> struct A @{ @};
12913 using namespace debug __attribute ((__strong__));
12914 template <> struct A<int> @{ @}; // @r{ok to specialize}
12916 template <class T> void f (A<T>);
12921 f (std::A<float>()); // @r{lookup finds} std::f
12927 @section Type Traits
12929 The C++ front-end implements syntactic extensions that allow to
12930 determine at compile time various characteristics of a type (or of a
12934 @item __has_nothrow_assign (type)
12935 If @code{type} is const qualified or is a reference type then the trait is
12936 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
12937 is true, else if @code{type} is a cv class or union type with copy assignment
12938 operators that are known not to throw an exception then the trait is true,
12939 else it is false. Requires: @code{type} shall be a complete type, an array
12940 type of unknown bound, or is a @code{void} type.
12942 @item __has_nothrow_copy (type)
12943 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
12944 @code{type} is a cv class or union type with copy constructors that
12945 are known not to throw an exception then the trait is true, else it is false.
12946 Requires: @code{type} shall be a complete type, an array type of
12947 unknown bound, or is a @code{void} type.
12949 @item __has_nothrow_constructor (type)
12950 If @code{__has_trivial_constructor (type)} is true then the trait is
12951 true, else if @code{type} is a cv class or union type (or array
12952 thereof) with a default constructor that is known not to throw an
12953 exception then the trait is true, else it is false. Requires:
12954 @code{type} shall be a complete type, an array type of unknown bound,
12955 or is a @code{void} type.
12957 @item __has_trivial_assign (type)
12958 If @code{type} is const qualified or is a reference type then the trait is
12959 false. Otherwise if @code{__is_pod (type)} is true then the trait is
12960 true, else if @code{type} is a cv class or union type with a trivial
12961 copy assignment ([class.copy]) then the trait is true, else it is
12962 false. Requires: @code{type} shall be a complete type, an array type
12963 of unknown bound, or is a @code{void} type.
12965 @item __has_trivial_copy (type)
12966 If @code{__is_pod (type)} is true or @code{type} is a reference type
12967 then the trait is true, else if @code{type} is a cv class or union type
12968 with a trivial copy constructor ([class.copy]) then the trait
12969 is true, else it is false. Requires: @code{type} shall be a complete
12970 type, an array type of unknown bound, or is a @code{void} type.
12972 @item __has_trivial_constructor (type)
12973 If @code{__is_pod (type)} is true then the trait is true, else if
12974 @code{type} is a cv class or union type (or array thereof) with a
12975 trivial default constructor ([class.ctor]) then the trait is true,
12976 else it is false. Requires: @code{type} shall be a complete type, an
12977 array type of unknown bound, or is a @code{void} type.
12979 @item __has_trivial_destructor (type)
12980 If @code{__is_pod (type)} is true or @code{type} is a reference type then
12981 the trait is true, else if @code{type} is a cv class or union type (or
12982 array thereof) with a trivial destructor ([class.dtor]) then the trait
12983 is true, else it is false. Requires: @code{type} shall be a complete
12984 type, an array type of unknown bound, or is a @code{void} type.
12986 @item __has_virtual_destructor (type)
12987 If @code{type} is a class type with a virtual destructor
12988 ([class.dtor]) then the trait is true, else it is false. Requires:
12989 @code{type} shall be a complete type, an array type of unknown bound,
12990 or is a @code{void} type.
12992 @item __is_abstract (type)
12993 If @code{type} is an abstract class ([class.abstract]) then the trait
12994 is true, else it is false. Requires: @code{type} shall be a complete
12995 type, an array type of unknown bound, or is a @code{void} type.
12997 @item __is_base_of (base_type, derived_type)
12998 If @code{base_type} is a base class of @code{derived_type}
12999 ([class.derived]) then the trait is true, otherwise it is false.
13000 Top-level cv qualifications of @code{base_type} and
13001 @code{derived_type} are ignored. For the purposes of this trait, a
13002 class type is considered is own base. Requires: if @code{__is_class
13003 (base_type)} and @code{__is_class (derived_type)} are true and
13004 @code{base_type} and @code{derived_type} are not the same type
13005 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
13006 type. Diagnostic is produced if this requirement is not met.
13008 @item __is_class (type)
13009 If @code{type} is a cv class type, and not a union type
13010 ([basic.compound]) the trait is true, else it is false.
13012 @item __is_empty (type)
13013 If @code{__is_class (type)} is false then the trait is false.
13014 Otherwise @code{type} is considered empty if and only if: @code{type}
13015 has no non-static data members, or all non-static data members, if
13016 any, are bit-fields of length 0, and @code{type} has no virtual
13017 members, and @code{type} has no virtual base classes, and @code{type}
13018 has no base classes @code{base_type} for which
13019 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
13020 be a complete type, an array type of unknown bound, or is a
13023 @item __is_enum (type)
13024 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
13025 true, else it is false.
13027 @item __is_pod (type)
13028 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
13029 else it is false. Requires: @code{type} shall be a complete type,
13030 an array type of unknown bound, or is a @code{void} type.
13032 @item __is_polymorphic (type)
13033 If @code{type} is a polymorphic class ([class.virtual]) then the trait
13034 is true, else it is false. Requires: @code{type} shall be a complete
13035 type, an array type of unknown bound, or is a @code{void} type.
13037 @item __is_union (type)
13038 If @code{type} is a cv union type ([basic.compound]) the trait is
13039 true, else it is false.
13043 @node Java Exceptions
13044 @section Java Exceptions
13046 The Java language uses a slightly different exception handling model
13047 from C++. Normally, GNU C++ will automatically detect when you are
13048 writing C++ code that uses Java exceptions, and handle them
13049 appropriately. However, if C++ code only needs to execute destructors
13050 when Java exceptions are thrown through it, GCC will guess incorrectly.
13051 Sample problematic code is:
13054 struct S @{ ~S(); @};
13055 extern void bar(); // @r{is written in Java, and may throw exceptions}
13064 The usual effect of an incorrect guess is a link failure, complaining of
13065 a missing routine called @samp{__gxx_personality_v0}.
13067 You can inform the compiler that Java exceptions are to be used in a
13068 translation unit, irrespective of what it might think, by writing
13069 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
13070 @samp{#pragma} must appear before any functions that throw or catch
13071 exceptions, or run destructors when exceptions are thrown through them.
13073 You cannot mix Java and C++ exceptions in the same translation unit. It
13074 is believed to be safe to throw a C++ exception from one file through
13075 another file compiled for the Java exception model, or vice versa, but
13076 there may be bugs in this area.
13078 @node Deprecated Features
13079 @section Deprecated Features
13081 In the past, the GNU C++ compiler was extended to experiment with new
13082 features, at a time when the C++ language was still evolving. Now that
13083 the C++ standard is complete, some of those features are superseded by
13084 superior alternatives. Using the old features might cause a warning in
13085 some cases that the feature will be dropped in the future. In other
13086 cases, the feature might be gone already.
13088 While the list below is not exhaustive, it documents some of the options
13089 that are now deprecated:
13092 @item -fexternal-templates
13093 @itemx -falt-external-templates
13094 These are two of the many ways for G++ to implement template
13095 instantiation. @xref{Template Instantiation}. The C++ standard clearly
13096 defines how template definitions have to be organized across
13097 implementation units. G++ has an implicit instantiation mechanism that
13098 should work just fine for standard-conforming code.
13100 @item -fstrict-prototype
13101 @itemx -fno-strict-prototype
13102 Previously it was possible to use an empty prototype parameter list to
13103 indicate an unspecified number of parameters (like C), rather than no
13104 parameters, as C++ demands. This feature has been removed, except where
13105 it is required for backwards compatibility. @xref{Backwards Compatibility}.
13108 G++ allows a virtual function returning @samp{void *} to be overridden
13109 by one returning a different pointer type. This extension to the
13110 covariant return type rules is now deprecated and will be removed from a
13113 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
13114 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
13115 and are now removed from G++. Code using these operators should be
13116 modified to use @code{std::min} and @code{std::max} instead.
13118 The named return value extension has been deprecated, and is now
13121 The use of initializer lists with new expressions has been deprecated,
13122 and is now removed from G++.
13124 Floating and complex non-type template parameters have been deprecated,
13125 and are now removed from G++.
13127 The implicit typename extension has been deprecated and is now
13130 The use of default arguments in function pointers, function typedefs
13131 and other places where they are not permitted by the standard is
13132 deprecated and will be removed from a future version of G++.
13134 G++ allows floating-point literals to appear in integral constant expressions,
13135 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
13136 This extension is deprecated and will be removed from a future version.
13138 G++ allows static data members of const floating-point type to be declared
13139 with an initializer in a class definition. The standard only allows
13140 initializers for static members of const integral types and const
13141 enumeration types so this extension has been deprecated and will be removed
13142 from a future version.
13144 @node Backwards Compatibility
13145 @section Backwards Compatibility
13146 @cindex Backwards Compatibility
13147 @cindex ARM [Annotated C++ Reference Manual]
13149 Now that there is a definitive ISO standard C++, G++ has a specification
13150 to adhere to. The C++ language evolved over time, and features that
13151 used to be acceptable in previous drafts of the standard, such as the ARM
13152 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
13153 compilation of C++ written to such drafts, G++ contains some backwards
13154 compatibilities. @emph{All such backwards compatibility features are
13155 liable to disappear in future versions of G++.} They should be considered
13156 deprecated. @xref{Deprecated Features}.
13160 If a variable is declared at for scope, it used to remain in scope until
13161 the end of the scope which contained the for statement (rather than just
13162 within the for scope). G++ retains this, but issues a warning, if such a
13163 variable is accessed outside the for scope.
13165 @item Implicit C language
13166 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
13167 scope to set the language. On such systems, all header files are
13168 implicitly scoped inside a C language scope. Also, an empty prototype
13169 @code{()} will be treated as an unspecified number of arguments, rather
13170 than no arguments, as C++ demands.