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 * Half-Precision:: Half-Precision Floating Point.
39 * Decimal Float:: Decimal Floating Types.
40 * Hex Floats:: Hexadecimal floating-point constants.
41 * Fixed-Point:: Fixed-Point Types.
42 * Zero Length:: Zero-length arrays.
43 * Variable Length:: Arrays whose length is computed at run time.
44 * Empty Structures:: Structures with no members.
45 * Variadic Macros:: Macros with a variable number of arguments.
46 * Escaped Newlines:: Slightly looser rules for escaped newlines.
47 * Subscripting:: Any array can be subscripted, even if not an lvalue.
48 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
49 * Initializers:: Non-constant initializers.
50 * Compound Literals:: Compound literals give structures, unions
52 * Designated Inits:: Labeling elements of initializers.
53 * Cast to Union:: Casting to union type from any member of the union.
54 * Case Ranges:: `case 1 ... 9' and such.
55 * Mixed Declarations:: Mixing declarations and code.
56 * Function Attributes:: Declaring that functions have no side effects,
57 or that they can never return.
58 * Attribute Syntax:: Formal syntax for attributes.
59 * Function Prototypes:: Prototype declarations and old-style definitions.
60 * C++ Comments:: C++ comments are recognized.
61 * Dollar Signs:: Dollar sign is allowed in identifiers.
62 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
63 * Variable Attributes:: Specifying attributes of variables.
64 * Type Attributes:: Specifying attributes of types.
65 * Alignment:: Inquiring about the alignment of a type or variable.
66 * Inline:: Defining inline functions (as fast as macros).
67 * Extended Asm:: Assembler instructions with C expressions as operands.
68 (With them you can define ``built-in'' functions.)
69 * Constraints:: Constraints for asm operands
70 * Asm Labels:: Specifying the assembler name to use for a C symbol.
71 * Explicit Reg Vars:: Defining variables residing in specified registers.
72 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
73 * Incomplete Enums:: @code{enum foo;}, with details to follow.
74 * Function Names:: Printable strings which are the name of the current
76 * Return Address:: Getting the return or frame address of a function.
77 * Vector Extensions:: Using vector instructions through built-in functions.
78 * Offsetof:: Special syntax for implementing @code{offsetof}.
79 * Atomic Builtins:: Built-in functions for atomic memory access.
80 * Object Size Checking:: Built-in functions for limited buffer overflow
82 * Other Builtins:: Other built-in functions.
83 * Target Builtins:: Built-in functions specific to particular targets.
84 * Target Format Checks:: Format checks specific to particular targets.
85 * Pragmas:: Pragmas accepted by GCC.
86 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
87 * Thread-Local:: Per-thread variables.
88 * Binary constants:: Binary constants using the @samp{0b} prefix.
92 @section Statements and Declarations in Expressions
93 @cindex statements inside expressions
94 @cindex declarations inside expressions
95 @cindex expressions containing statements
96 @cindex macros, statements in expressions
98 @c the above section title wrapped and causes an underfull hbox.. i
99 @c changed it from "within" to "in". --mew 4feb93
100 A compound statement enclosed in parentheses may appear as an expression
101 in GNU C@. This allows you to use loops, switches, and local variables
102 within an expression.
104 Recall that a compound statement is a sequence of statements surrounded
105 by braces; in this construct, parentheses go around the braces. For
109 (@{ int y = foo (); int z;
116 is a valid (though slightly more complex than necessary) expression
117 for the absolute value of @code{foo ()}.
119 The last thing in the compound statement should be an expression
120 followed by a semicolon; the value of this subexpression serves as the
121 value of the entire construct. (If you use some other kind of statement
122 last within the braces, the construct has type @code{void}, and thus
123 effectively no value.)
125 This feature is especially useful in making macro definitions ``safe'' (so
126 that they evaluate each operand exactly once). For example, the
127 ``maximum'' function is commonly defined as a macro in standard C as
131 #define max(a,b) ((a) > (b) ? (a) : (b))
135 @cindex side effects, macro argument
136 But this definition computes either @var{a} or @var{b} twice, with bad
137 results if the operand has side effects. In GNU C, if you know the
138 type of the operands (here taken as @code{int}), you can define
139 the macro safely as follows:
142 #define maxint(a,b) \
143 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
146 Embedded statements are not allowed in constant expressions, such as
147 the value of an enumeration constant, the width of a bit-field, or
148 the initial value of a static variable.
150 If you don't know the type of the operand, you can still do this, but you
151 must use @code{typeof} (@pxref{Typeof}).
153 In G++, the result value of a statement expression undergoes array and
154 function pointer decay, and is returned by value to the enclosing
155 expression. For instance, if @code{A} is a class, then
164 will construct a temporary @code{A} object to hold the result of the
165 statement expression, and that will be used to invoke @code{Foo}.
166 Therefore the @code{this} pointer observed by @code{Foo} will not be the
169 Any temporaries created within a statement within a statement expression
170 will be destroyed at the statement's end. This makes statement
171 expressions inside macros slightly different from function calls. In
172 the latter case temporaries introduced during argument evaluation will
173 be destroyed at the end of the statement that includes the function
174 call. In the statement expression case they will be destroyed during
175 the statement expression. For instance,
178 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
179 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
189 will have different places where temporaries are destroyed. For the
190 @code{macro} case, the temporary @code{X} will be destroyed just after
191 the initialization of @code{b}. In the @code{function} case that
192 temporary will be destroyed when the function returns.
194 These considerations mean that it is probably a bad idea to use
195 statement-expressions of this form in header files that are designed to
196 work with C++. (Note that some versions of the GNU C Library contained
197 header files using statement-expression that lead to precisely this
200 Jumping into a statement expression with @code{goto} or using a
201 @code{switch} statement outside the statement expression with a
202 @code{case} or @code{default} label inside the statement expression is
203 not permitted. Jumping into a statement expression with a computed
204 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
205 Jumping out of a statement expression is permitted, but if the
206 statement expression is part of a larger expression then it is
207 unspecified which other subexpressions of that expression have been
208 evaluated except where the language definition requires certain
209 subexpressions to be evaluated before or after the statement
210 expression. In any case, as with a function call the evaluation of a
211 statement expression is not interleaved with the evaluation of other
212 parts of the containing expression. For example,
215 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
219 will call @code{foo} and @code{bar1} and will not call @code{baz} but
220 may or may not call @code{bar2}. If @code{bar2} is called, it will be
221 called after @code{foo} and before @code{bar1}
224 @section Locally Declared Labels
226 @cindex macros, local labels
228 GCC allows you to declare @dfn{local labels} in any nested block
229 scope. A local label is just like an ordinary label, but you can
230 only reference it (with a @code{goto} statement, or by taking its
231 address) within the block in which it was declared.
233 A local label declaration looks like this:
236 __label__ @var{label};
243 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
246 Local label declarations must come at the beginning of the block,
247 before any ordinary declarations or statements.
249 The label declaration defines the label @emph{name}, but does not define
250 the label itself. You must do this in the usual way, with
251 @code{@var{label}:}, within the statements of the statement expression.
253 The local label feature is useful for complex macros. If a macro
254 contains nested loops, a @code{goto} can be useful for breaking out of
255 them. However, an ordinary label whose scope is the whole function
256 cannot be used: if the macro can be expanded several times in one
257 function, the label will be multiply defined in that function. A
258 local label avoids this problem. For example:
261 #define SEARCH(value, array, target) \
264 typeof (target) _SEARCH_target = (target); \
265 typeof (*(array)) *_SEARCH_array = (array); \
268 for (i = 0; i < max; i++) \
269 for (j = 0; j < max; j++) \
270 if (_SEARCH_array[i][j] == _SEARCH_target) \
271 @{ (value) = i; goto found; @} \
277 This could also be written using a statement-expression:
280 #define SEARCH(array, target) \
283 typeof (target) _SEARCH_target = (target); \
284 typeof (*(array)) *_SEARCH_array = (array); \
287 for (i = 0; i < max; i++) \
288 for (j = 0; j < max; j++) \
289 if (_SEARCH_array[i][j] == _SEARCH_target) \
290 @{ value = i; goto found; @} \
297 Local label declarations also make the labels they declare visible to
298 nested functions, if there are any. @xref{Nested Functions}, for details.
300 @node Labels as Values
301 @section Labels as Values
302 @cindex labels as values
303 @cindex computed gotos
304 @cindex goto with computed label
305 @cindex address of a label
307 You can get the address of a label defined in the current function
308 (or a containing function) with the unary operator @samp{&&}. The
309 value has type @code{void *}. This value is a constant and can be used
310 wherever a constant of that type is valid. For example:
318 To use these values, you need to be able to jump to one. This is done
319 with the computed goto statement@footnote{The analogous feature in
320 Fortran is called an assigned goto, but that name seems inappropriate in
321 C, where one can do more than simply store label addresses in label
322 variables.}, @code{goto *@var{exp};}. For example,
329 Any expression of type @code{void *} is allowed.
331 One way of using these constants is in initializing a static array that
332 will serve as a jump table:
335 static void *array[] = @{ &&foo, &&bar, &&hack @};
338 Then you can select a label with indexing, like this:
345 Note that this does not check whether the subscript is in bounds---array
346 indexing in C never does that.
348 Such an array of label values serves a purpose much like that of the
349 @code{switch} statement. The @code{switch} statement is cleaner, so
350 use that rather than an array unless the problem does not fit a
351 @code{switch} statement very well.
353 Another use of label values is in an interpreter for threaded code.
354 The labels within the interpreter function can be stored in the
355 threaded code for super-fast dispatching.
357 You may not use this mechanism to jump to code in a different function.
358 If you do that, totally unpredictable things will happen. The best way to
359 avoid this is to store the label address only in automatic variables and
360 never pass it as an argument.
362 An alternate way to write the above example is
365 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
367 goto *(&&foo + array[i]);
371 This is more friendly to code living in shared libraries, as it reduces
372 the number of dynamic relocations that are needed, and by consequence,
373 allows the data to be read-only.
375 The @code{&&foo} expressions for the same label might have different values
376 if the containing function is inlined or cloned. If a program relies on
377 them being always the same, @code{__attribute__((__noinline__))} should
378 be used to prevent inlining. If @code{&&foo} is used
379 in a static variable initializer, inlining is forbidden.
381 @node Nested Functions
382 @section Nested Functions
383 @cindex nested functions
384 @cindex downward funargs
387 A @dfn{nested function} is a function defined inside another function.
388 (Nested functions are not supported for GNU C++.) The nested function's
389 name is local to the block where it is defined. For example, here we
390 define a nested function named @code{square}, and call it twice:
394 foo (double a, double b)
396 double square (double z) @{ return z * z; @}
398 return square (a) + square (b);
403 The nested function can access all the variables of the containing
404 function that are visible at the point of its definition. This is
405 called @dfn{lexical scoping}. For example, here we show a nested
406 function which uses an inherited variable named @code{offset}:
410 bar (int *array, int offset, int size)
412 int access (int *array, int index)
413 @{ return array[index + offset]; @}
416 for (i = 0; i < size; i++)
417 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
422 Nested function definitions are permitted within functions in the places
423 where variable definitions are allowed; that is, in any block, mixed
424 with the other declarations and statements in the block.
426 It is possible to call the nested function from outside the scope of its
427 name by storing its address or passing the address to another function:
430 hack (int *array, int size)
432 void store (int index, int value)
433 @{ array[index] = value; @}
435 intermediate (store, size);
439 Here, the function @code{intermediate} receives the address of
440 @code{store} as an argument. If @code{intermediate} calls @code{store},
441 the arguments given to @code{store} are used to store into @code{array}.
442 But this technique works only so long as the containing function
443 (@code{hack}, in this example) does not exit.
445 If you try to call the nested function through its address after the
446 containing function has exited, all hell will break loose. If you try
447 to call it after a containing scope level has exited, and if it refers
448 to some of the variables that are no longer in scope, you may be lucky,
449 but it's not wise to take the risk. If, however, the nested function
450 does not refer to anything that has gone out of scope, you should be
453 GCC implements taking the address of a nested function using a technique
454 called @dfn{trampolines}. A paper describing them is available as
457 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
459 A nested function can jump to a label inherited from a containing
460 function, provided the label was explicitly declared in the containing
461 function (@pxref{Local Labels}). Such a jump returns instantly to the
462 containing function, exiting the nested function which did the
463 @code{goto} and any intermediate functions as well. Here is an example:
467 bar (int *array, int offset, int size)
470 int access (int *array, int index)
474 return array[index + offset];
478 for (i = 0; i < size; i++)
479 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
483 /* @r{Control comes here from @code{access}
484 if it detects an error.} */
491 A nested function always has no linkage. Declaring one with
492 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
493 before its definition, use @code{auto} (which is otherwise meaningless
494 for function declarations).
497 bar (int *array, int offset, int size)
500 auto int access (int *, int);
502 int access (int *array, int index)
506 return array[index + offset];
512 @node Constructing Calls
513 @section Constructing Function Calls
514 @cindex constructing calls
515 @cindex forwarding calls
517 Using the built-in functions described below, you can record
518 the arguments a function received, and call another function
519 with the same arguments, without knowing the number or types
522 You can also record the return value of that function call,
523 and later return that value, without knowing what data type
524 the function tried to return (as long as your caller expects
527 However, these built-in functions may interact badly with some
528 sophisticated features or other extensions of the language. It
529 is, therefore, not recommended to use them outside very simple
530 functions acting as mere forwarders for their arguments.
532 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
533 This built-in function returns a pointer to data
534 describing how to perform a call with the same arguments as were passed
535 to the current function.
537 The function saves the arg pointer register, structure value address,
538 and all registers that might be used to pass arguments to a function
539 into a block of memory allocated on the stack. Then it returns the
540 address of that block.
543 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
544 This built-in function invokes @var{function}
545 with a copy of the parameters described by @var{arguments}
548 The value of @var{arguments} should be the value returned by
549 @code{__builtin_apply_args}. The argument @var{size} specifies the size
550 of the stack argument data, in bytes.
552 This function returns a pointer to data describing
553 how to return whatever value was returned by @var{function}. The data
554 is saved in a block of memory allocated on the stack.
556 It is not always simple to compute the proper value for @var{size}. The
557 value is used by @code{__builtin_apply} to compute the amount of data
558 that should be pushed on the stack and copied from the incoming argument
562 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
563 This built-in function returns the value described by @var{result} from
564 the containing function. You should specify, for @var{result}, a value
565 returned by @code{__builtin_apply}.
568 @deftypefn {Built-in Function} __builtin_va_arg_pack ()
569 This built-in function represents all anonymous arguments of an inline
570 function. It can be used only in inline functions which will be always
571 inlined, never compiled as a separate function, such as those using
572 @code{__attribute__ ((__always_inline__))} or
573 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
574 It must be only passed as last argument to some other function
575 with variable arguments. This is useful for writing small wrapper
576 inlines for variable argument functions, when using preprocessor
577 macros is undesirable. For example:
579 extern int myprintf (FILE *f, const char *format, ...);
580 extern inline __attribute__ ((__gnu_inline__)) int
581 myprintf (FILE *f, const char *format, ...)
583 int r = fprintf (f, "myprintf: ");
586 int s = fprintf (f, format, __builtin_va_arg_pack ());
594 @deftypefn {Built-in Function} __builtin_va_arg_pack_len ()
595 This built-in function returns the number of anonymous arguments of
596 an inline function. It can be used only in inline functions which
597 will be always inlined, never compiled as a separate function, such
598 as those using @code{__attribute__ ((__always_inline__))} or
599 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
600 For example following will do link or runtime checking of open
601 arguments for optimized code:
604 extern inline __attribute__((__gnu_inline__)) int
605 myopen (const char *path, int oflag, ...)
607 if (__builtin_va_arg_pack_len () > 1)
608 warn_open_too_many_arguments ();
610 if (__builtin_constant_p (oflag))
612 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
614 warn_open_missing_mode ();
615 return __open_2 (path, oflag);
617 return open (path, oflag, __builtin_va_arg_pack ());
620 if (__builtin_va_arg_pack_len () < 1)
621 return __open_2 (path, oflag);
623 return open (path, oflag, __builtin_va_arg_pack ());
630 @section Referring to a Type with @code{typeof}
633 @cindex macros, types of arguments
635 Another way to refer to the type of an expression is with @code{typeof}.
636 The syntax of using of this keyword looks like @code{sizeof}, but the
637 construct acts semantically like a type name defined with @code{typedef}.
639 There are two ways of writing the argument to @code{typeof}: with an
640 expression or with a type. Here is an example with an expression:
647 This assumes that @code{x} is an array of pointers to functions;
648 the type described is that of the values of the functions.
650 Here is an example with a typename as the argument:
657 Here the type described is that of pointers to @code{int}.
659 If you are writing a header file that must work when included in ISO C
660 programs, write @code{__typeof__} instead of @code{typeof}.
661 @xref{Alternate Keywords}.
663 A @code{typeof}-construct can be used anywhere a typedef name could be
664 used. For example, you can use it in a declaration, in a cast, or inside
665 of @code{sizeof} or @code{typeof}.
667 The operand of @code{typeof} is evaluated for its side effects if and
668 only if it is an expression of variably modified type or the name of
671 @code{typeof} is often useful in conjunction with the
672 statements-within-expressions feature. Here is how the two together can
673 be used to define a safe ``maximum'' macro that operates on any
674 arithmetic type and evaluates each of its arguments exactly once:
678 (@{ typeof (a) _a = (a); \
679 typeof (b) _b = (b); \
680 _a > _b ? _a : _b; @})
683 @cindex underscores in variables in macros
684 @cindex @samp{_} in variables in macros
685 @cindex local variables in macros
686 @cindex variables, local, in macros
687 @cindex macros, local variables in
689 The reason for using names that start with underscores for the local
690 variables is to avoid conflicts with variable names that occur within the
691 expressions that are substituted for @code{a} and @code{b}. Eventually we
692 hope to design a new form of declaration syntax that allows you to declare
693 variables whose scopes start only after their initializers; this will be a
694 more reliable way to prevent such conflicts.
697 Some more examples of the use of @code{typeof}:
701 This declares @code{y} with the type of what @code{x} points to.
708 This declares @code{y} as an array of such values.
715 This declares @code{y} as an array of pointers to characters:
718 typeof (typeof (char *)[4]) y;
722 It is equivalent to the following traditional C declaration:
728 To see the meaning of the declaration using @code{typeof}, and why it
729 might be a useful way to write, rewrite it with these macros:
732 #define pointer(T) typeof(T *)
733 #define array(T, N) typeof(T [N])
737 Now the declaration can be rewritten this way:
740 array (pointer (char), 4) y;
744 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
745 pointers to @code{char}.
748 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
749 a more limited extension which permitted one to write
752 typedef @var{T} = @var{expr};
756 with the effect of declaring @var{T} to have the type of the expression
757 @var{expr}. This extension does not work with GCC 3 (versions between
758 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
759 relies on it should be rewritten to use @code{typeof}:
762 typedef typeof(@var{expr}) @var{T};
766 This will work with all versions of GCC@.
769 @section Conditionals with Omitted Operands
770 @cindex conditional expressions, extensions
771 @cindex omitted middle-operands
772 @cindex middle-operands, omitted
773 @cindex extensions, @code{?:}
774 @cindex @code{?:} extensions
776 The middle operand in a conditional expression may be omitted. Then
777 if the first operand is nonzero, its value is the value of the conditional
780 Therefore, the expression
787 has the value of @code{x} if that is nonzero; otherwise, the value of
790 This example is perfectly equivalent to
796 @cindex side effect in ?:
797 @cindex ?: side effect
799 In this simple case, the ability to omit the middle operand is not
800 especially useful. When it becomes useful is when the first operand does,
801 or may (if it is a macro argument), contain a side effect. Then repeating
802 the operand in the middle would perform the side effect twice. Omitting
803 the middle operand uses the value already computed without the undesirable
804 effects of recomputing it.
807 @section Double-Word Integers
808 @cindex @code{long long} data types
809 @cindex double-word arithmetic
810 @cindex multiprecision arithmetic
811 @cindex @code{LL} integer suffix
812 @cindex @code{ULL} integer suffix
814 ISO C99 supports data types for integers that are at least 64 bits wide,
815 and as an extension GCC supports them in C89 mode and in C++.
816 Simply write @code{long long int} for a signed integer, or
817 @code{unsigned long long int} for an unsigned integer. To make an
818 integer constant of type @code{long long int}, add the suffix @samp{LL}
819 to the integer. To make an integer constant of type @code{unsigned long
820 long int}, add the suffix @samp{ULL} to the integer.
822 You can use these types in arithmetic like any other integer types.
823 Addition, subtraction, and bitwise boolean operations on these types
824 are open-coded on all types of machines. Multiplication is open-coded
825 if the machine supports fullword-to-doubleword a widening multiply
826 instruction. Division and shifts are open-coded only on machines that
827 provide special support. The operations that are not open-coded use
828 special library routines that come with GCC@.
830 There may be pitfalls when you use @code{long long} types for function
831 arguments, unless you declare function prototypes. If a function
832 expects type @code{int} for its argument, and you pass a value of type
833 @code{long long int}, confusion will result because the caller and the
834 subroutine will disagree about the number of bytes for the argument.
835 Likewise, if the function expects @code{long long int} and you pass
836 @code{int}. The best way to avoid such problems is to use prototypes.
839 @section Complex Numbers
840 @cindex complex numbers
841 @cindex @code{_Complex} keyword
842 @cindex @code{__complex__} keyword
844 ISO C99 supports complex floating data types, and as an extension GCC
845 supports them in C89 mode and in C++, and supports complex integer data
846 types which are not part of ISO C99. You can declare complex types
847 using the keyword @code{_Complex}. As an extension, the older GNU
848 keyword @code{__complex__} is also supported.
850 For example, @samp{_Complex double x;} declares @code{x} as a
851 variable whose real part and imaginary part are both of type
852 @code{double}. @samp{_Complex short int y;} declares @code{y} to
853 have real and imaginary parts of type @code{short int}; this is not
854 likely to be useful, but it shows that the set of complex types is
857 To write a constant with a complex data type, use the suffix @samp{i} or
858 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
859 has type @code{_Complex float} and @code{3i} has type
860 @code{_Complex int}. Such a constant always has a pure imaginary
861 value, but you can form any complex value you like by adding one to a
862 real constant. This is a GNU extension; if you have an ISO C99
863 conforming C library (such as GNU libc), and want to construct complex
864 constants of floating type, you should include @code{<complex.h>} and
865 use the macros @code{I} or @code{_Complex_I} instead.
867 @cindex @code{__real__} keyword
868 @cindex @code{__imag__} keyword
869 To extract the real part of a complex-valued expression @var{exp}, write
870 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
871 extract the imaginary part. This is a GNU extension; for values of
872 floating type, you should use the ISO C99 functions @code{crealf},
873 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
874 @code{cimagl}, declared in @code{<complex.h>} and also provided as
875 built-in functions by GCC@.
877 @cindex complex conjugation
878 The operator @samp{~} performs complex conjugation when used on a value
879 with a complex type. This is a GNU extension; for values of
880 floating type, you should use the ISO C99 functions @code{conjf},
881 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
882 provided as built-in functions by GCC@.
884 GCC can allocate complex automatic variables in a noncontiguous
885 fashion; it's even possible for the real part to be in a register while
886 the imaginary part is on the stack (or vice-versa). Only the DWARF2
887 debug info format can represent this, so use of DWARF2 is recommended.
888 If you are using the stabs debug info format, GCC describes a noncontiguous
889 complex variable as if it were two separate variables of noncomplex type.
890 If the variable's actual name is @code{foo}, the two fictitious
891 variables are named @code{foo$real} and @code{foo$imag}. You can
892 examine and set these two fictitious variables with your debugger.
895 @section Additional Floating Types
896 @cindex additional floating types
897 @cindex @code{__float80} data type
898 @cindex @code{__float128} data type
899 @cindex @code{w} floating point suffix
900 @cindex @code{q} floating point suffix
901 @cindex @code{W} floating point suffix
902 @cindex @code{Q} floating point suffix
904 As an extension, the GNU C compiler supports additional floating
905 types, @code{__float80} and @code{__float128} to support 80bit
906 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
907 Support for additional types includes the arithmetic operators:
908 add, subtract, multiply, divide; unary arithmetic operators;
909 relational operators; equality operators; and conversions to and from
910 integer and other floating types. Use a suffix @samp{w} or @samp{W}
911 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
912 for @code{_float128}. You can declare complex types using the
913 corresponding internal complex type, @code{XCmode} for @code{__float80}
914 type and @code{TCmode} for @code{__float128} type:
917 typedef _Complex float __attribute__((mode(TC))) _Complex128;
918 typedef _Complex float __attribute__((mode(XC))) _Complex80;
921 Not all targets support additional floating point types. @code{__float80}
922 is supported on i386, x86_64 and ia64 targets and target @code{__float128}
923 is supported on x86_64 and ia64 targets.
926 @section Half-Precision Floating Point
927 @cindex half-precision floating point
928 @cindex @code{__fp16} data type
930 On ARM targets, GCC supports half-precision (16-bit) floating point via
931 the @code{__fp16} type. You must enable this type explicitly
932 with the @option{-mfp16-format} command-line option in order to use it.
934 ARM supports two incompatible representations for half-precision
935 floating-point values. You must choose one of the representations and
936 use it consistently in your program.
938 Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
939 This format can represent normalized values in the range of @math{2^{-14}} to 65504.
940 There are 11 bits of significand precision, approximately 3
943 Specifying @option{-mfp16-format=alternative} selects the ARM
944 alternative format. This representation is similar to the IEEE
945 format, but does not support infinities or NaNs. Instead, the range
946 of exponents is extended, so that this format can represent normalized
947 values in the range of @math{2^{-14}} to 131008.
949 The @code{__fp16} type is a storage format only. For purposes
950 of arithmetic and other operations, @code{__fp16} values in C or C++
951 expressions are automatically promoted to @code{float}. In addition,
952 you cannot declare a function with a return value or parameters
953 of type @code{__fp16}.
955 Note that conversions from @code{double} to @code{__fp16}
956 involve an intermediate conversion to @code{float}. Because
957 of rounding, this can sometimes produce a different result than a
960 ARM provides hardware support for conversions between
961 @code{__fp16} and @code{float} values
962 as an extension to VFP and NEON (Advanced SIMD). GCC generates
963 code using the instructions provided by this extension if you compile
964 with the options @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
965 in addition to the @option{-mfp16-format} option to select
966 a half-precision format.
968 Language-level support for the @code{__fp16} data type is
969 independent of whether GCC generates code using hardware floating-point
970 instructions. In cases where hardware support is not specified, GCC
971 implements conversions between @code{__fp16} and @code{float} values
975 @section Decimal Floating Types
976 @cindex decimal floating types
977 @cindex @code{_Decimal32} data type
978 @cindex @code{_Decimal64} data type
979 @cindex @code{_Decimal128} data type
980 @cindex @code{df} integer suffix
981 @cindex @code{dd} integer suffix
982 @cindex @code{dl} integer suffix
983 @cindex @code{DF} integer suffix
984 @cindex @code{DD} integer suffix
985 @cindex @code{DL} integer suffix
987 As an extension, the GNU C compiler supports decimal floating types as
988 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
989 floating types in GCC will evolve as the draft technical report changes.
990 Calling conventions for any target might also change. Not all targets
991 support decimal floating types.
993 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
994 @code{_Decimal128}. They use a radix of ten, unlike the floating types
995 @code{float}, @code{double}, and @code{long double} whose radix is not
996 specified by the C standard but is usually two.
998 Support for decimal floating types includes the arithmetic operators
999 add, subtract, multiply, divide; unary arithmetic operators;
1000 relational operators; equality operators; and conversions to and from
1001 integer and other floating types. Use a suffix @samp{df} or
1002 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
1003 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
1006 GCC support of decimal float as specified by the draft technical report
1011 When the value of a decimal floating type cannot be represented in the
1012 integer type to which it is being converted, the result is undefined
1013 rather than the result value specified by the draft technical report.
1016 GCC does not provide the C library functionality associated with
1017 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
1018 @file{wchar.h}, which must come from a separate C library implementation.
1019 Because of this the GNU C compiler does not define macro
1020 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
1021 the technical report.
1024 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
1025 are supported by the DWARF2 debug information format.
1031 ISO C99 supports floating-point numbers written not only in the usual
1032 decimal notation, such as @code{1.55e1}, but also numbers such as
1033 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
1034 supports this in C89 mode (except in some cases when strictly
1035 conforming) and in C++. In that format the
1036 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1037 mandatory. The exponent is a decimal number that indicates the power of
1038 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
1045 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1046 is the same as @code{1.55e1}.
1048 Unlike for floating-point numbers in the decimal notation the exponent
1049 is always required in the hexadecimal notation. Otherwise the compiler
1050 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1051 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1052 extension for floating-point constants of type @code{float}.
1055 @section Fixed-Point Types
1056 @cindex fixed-point types
1057 @cindex @code{_Fract} data type
1058 @cindex @code{_Accum} data type
1059 @cindex @code{_Sat} data type
1060 @cindex @code{hr} fixed-suffix
1061 @cindex @code{r} fixed-suffix
1062 @cindex @code{lr} fixed-suffix
1063 @cindex @code{llr} fixed-suffix
1064 @cindex @code{uhr} fixed-suffix
1065 @cindex @code{ur} fixed-suffix
1066 @cindex @code{ulr} fixed-suffix
1067 @cindex @code{ullr} fixed-suffix
1068 @cindex @code{hk} fixed-suffix
1069 @cindex @code{k} fixed-suffix
1070 @cindex @code{lk} fixed-suffix
1071 @cindex @code{llk} fixed-suffix
1072 @cindex @code{uhk} fixed-suffix
1073 @cindex @code{uk} fixed-suffix
1074 @cindex @code{ulk} fixed-suffix
1075 @cindex @code{ullk} fixed-suffix
1076 @cindex @code{HR} fixed-suffix
1077 @cindex @code{R} fixed-suffix
1078 @cindex @code{LR} fixed-suffix
1079 @cindex @code{LLR} fixed-suffix
1080 @cindex @code{UHR} fixed-suffix
1081 @cindex @code{UR} fixed-suffix
1082 @cindex @code{ULR} fixed-suffix
1083 @cindex @code{ULLR} fixed-suffix
1084 @cindex @code{HK} fixed-suffix
1085 @cindex @code{K} fixed-suffix
1086 @cindex @code{LK} fixed-suffix
1087 @cindex @code{LLK} fixed-suffix
1088 @cindex @code{UHK} fixed-suffix
1089 @cindex @code{UK} fixed-suffix
1090 @cindex @code{ULK} fixed-suffix
1091 @cindex @code{ULLK} fixed-suffix
1093 As an extension, the GNU C compiler supports fixed-point types as
1094 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1095 types in GCC will evolve as the draft technical report changes.
1096 Calling conventions for any target might also change. Not all targets
1097 support fixed-point types.
1099 The fixed-point types are
1100 @code{short _Fract},
1103 @code{long long _Fract},
1104 @code{unsigned short _Fract},
1105 @code{unsigned _Fract},
1106 @code{unsigned long _Fract},
1107 @code{unsigned long long _Fract},
1108 @code{_Sat short _Fract},
1110 @code{_Sat long _Fract},
1111 @code{_Sat long long _Fract},
1112 @code{_Sat unsigned short _Fract},
1113 @code{_Sat unsigned _Fract},
1114 @code{_Sat unsigned long _Fract},
1115 @code{_Sat unsigned long long _Fract},
1116 @code{short _Accum},
1119 @code{long long _Accum},
1120 @code{unsigned short _Accum},
1121 @code{unsigned _Accum},
1122 @code{unsigned long _Accum},
1123 @code{unsigned long long _Accum},
1124 @code{_Sat short _Accum},
1126 @code{_Sat long _Accum},
1127 @code{_Sat long long _Accum},
1128 @code{_Sat unsigned short _Accum},
1129 @code{_Sat unsigned _Accum},
1130 @code{_Sat unsigned long _Accum},
1131 @code{_Sat unsigned long long _Accum}.
1133 Fixed-point data values contain fractional and optional integral parts.
1134 The format of fixed-point data varies and depends on the target machine.
1136 Support for fixed-point types includes:
1139 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1141 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1143 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1145 binary shift operators (@code{<<}, @code{>>})
1147 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1149 equality operators (@code{==}, @code{!=})
1151 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1152 @code{<<=}, @code{>>=})
1154 conversions to and from integer, floating-point, or fixed-point types
1157 Use a suffix in a fixed-point literal constant:
1159 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1160 @code{_Sat short _Fract}
1161 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1162 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1163 @code{_Sat long _Fract}
1164 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1165 @code{_Sat long long _Fract}
1166 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1167 @code{_Sat unsigned short _Fract}
1168 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1169 @code{_Sat unsigned _Fract}
1170 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1171 @code{_Sat unsigned long _Fract}
1172 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1173 and @code{_Sat unsigned long long _Fract}
1174 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1175 @code{_Sat short _Accum}
1176 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1177 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1178 @code{_Sat long _Accum}
1179 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1180 @code{_Sat long long _Accum}
1181 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1182 @code{_Sat unsigned short _Accum}
1183 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1184 @code{_Sat unsigned _Accum}
1185 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1186 @code{_Sat unsigned long _Accum}
1187 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1188 and @code{_Sat unsigned long long _Accum}
1191 GCC support of fixed-point types as specified by the draft technical report
1196 Pragmas to control overflow and rounding behaviors are not implemented.
1199 Fixed-point types are supported by the DWARF2 debug information format.
1202 @section Arrays of Length Zero
1203 @cindex arrays of length zero
1204 @cindex zero-length arrays
1205 @cindex length-zero arrays
1206 @cindex flexible array members
1208 Zero-length arrays are allowed in GNU C@. They are very useful as the
1209 last element of a structure which is really a header for a variable-length
1218 struct line *thisline = (struct line *)
1219 malloc (sizeof (struct line) + this_length);
1220 thisline->length = this_length;
1223 In ISO C90, you would have to give @code{contents} a length of 1, which
1224 means either you waste space or complicate the argument to @code{malloc}.
1226 In ISO C99, you would use a @dfn{flexible array member}, which is
1227 slightly different in syntax and semantics:
1231 Flexible array members are written as @code{contents[]} without
1235 Flexible array members have incomplete type, and so the @code{sizeof}
1236 operator may not be applied. As a quirk of the original implementation
1237 of zero-length arrays, @code{sizeof} evaluates to zero.
1240 Flexible array members may only appear as the last member of a
1241 @code{struct} that is otherwise non-empty.
1244 A structure containing a flexible array member, or a union containing
1245 such a structure (possibly recursively), may not be a member of a
1246 structure or an element of an array. (However, these uses are
1247 permitted by GCC as extensions.)
1250 GCC versions before 3.0 allowed zero-length arrays to be statically
1251 initialized, as if they were flexible arrays. In addition to those
1252 cases that were useful, it also allowed initializations in situations
1253 that would corrupt later data. Non-empty initialization of zero-length
1254 arrays is now treated like any case where there are more initializer
1255 elements than the array holds, in that a suitable warning about "excess
1256 elements in array" is given, and the excess elements (all of them, in
1257 this case) are ignored.
1259 Instead GCC allows static initialization of flexible array members.
1260 This is equivalent to defining a new structure containing the original
1261 structure followed by an array of sufficient size to contain the data.
1262 I.e.@: in the following, @code{f1} is constructed as if it were declared
1268 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1271 struct f1 f1; int data[3];
1272 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1276 The convenience of this extension is that @code{f1} has the desired
1277 type, eliminating the need to consistently refer to @code{f2.f1}.
1279 This has symmetry with normal static arrays, in that an array of
1280 unknown size is also written with @code{[]}.
1282 Of course, this extension only makes sense if the extra data comes at
1283 the end of a top-level object, as otherwise we would be overwriting
1284 data at subsequent offsets. To avoid undue complication and confusion
1285 with initialization of deeply nested arrays, we simply disallow any
1286 non-empty initialization except when the structure is the top-level
1287 object. For example:
1290 struct foo @{ int x; int y[]; @};
1291 struct bar @{ struct foo z; @};
1293 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1294 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1295 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1296 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1299 @node Empty Structures
1300 @section Structures With No Members
1301 @cindex empty structures
1302 @cindex zero-size structures
1304 GCC permits a C structure to have no members:
1311 The structure will have size zero. In C++, empty structures are part
1312 of the language. G++ treats empty structures as if they had a single
1313 member of type @code{char}.
1315 @node Variable Length
1316 @section Arrays of Variable Length
1317 @cindex variable-length arrays
1318 @cindex arrays of variable length
1321 Variable-length automatic arrays are allowed in ISO C99, and as an
1322 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1323 implementation of variable-length arrays does not yet conform in detail
1324 to the ISO C99 standard.) These arrays are
1325 declared like any other automatic arrays, but with a length that is not
1326 a constant expression. The storage is allocated at the point of
1327 declaration and deallocated when the brace-level is exited. For
1332 concat_fopen (char *s1, char *s2, char *mode)
1334 char str[strlen (s1) + strlen (s2) + 1];
1337 return fopen (str, mode);
1341 @cindex scope of a variable length array
1342 @cindex variable-length array scope
1343 @cindex deallocating variable length arrays
1344 Jumping or breaking out of the scope of the array name deallocates the
1345 storage. Jumping into the scope is not allowed; you get an error
1348 @cindex @code{alloca} vs variable-length arrays
1349 You can use the function @code{alloca} to get an effect much like
1350 variable-length arrays. The function @code{alloca} is available in
1351 many other C implementations (but not in all). On the other hand,
1352 variable-length arrays are more elegant.
1354 There are other differences between these two methods. Space allocated
1355 with @code{alloca} exists until the containing @emph{function} returns.
1356 The space for a variable-length array is deallocated as soon as the array
1357 name's scope ends. (If you use both variable-length arrays and
1358 @code{alloca} in the same function, deallocation of a variable-length array
1359 will also deallocate anything more recently allocated with @code{alloca}.)
1361 You can also use variable-length arrays as arguments to functions:
1365 tester (int len, char data[len][len])
1371 The length of an array is computed once when the storage is allocated
1372 and is remembered for the scope of the array in case you access it with
1375 If you want to pass the array first and the length afterward, you can
1376 use a forward declaration in the parameter list---another GNU extension.
1380 tester (int len; char data[len][len], int len)
1386 @cindex parameter forward declaration
1387 The @samp{int len} before the semicolon is a @dfn{parameter forward
1388 declaration}, and it serves the purpose of making the name @code{len}
1389 known when the declaration of @code{data} is parsed.
1391 You can write any number of such parameter forward declarations in the
1392 parameter list. They can be separated by commas or semicolons, but the
1393 last one must end with a semicolon, which is followed by the ``real''
1394 parameter declarations. Each forward declaration must match a ``real''
1395 declaration in parameter name and data type. ISO C99 does not support
1396 parameter forward declarations.
1398 @node Variadic Macros
1399 @section Macros with a Variable Number of Arguments.
1400 @cindex variable number of arguments
1401 @cindex macro with variable arguments
1402 @cindex rest argument (in macro)
1403 @cindex variadic macros
1405 In the ISO C standard of 1999, a macro can be declared to accept a
1406 variable number of arguments much as a function can. The syntax for
1407 defining the macro is similar to that of a function. Here is an
1411 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1414 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1415 such a macro, it represents the zero or more tokens until the closing
1416 parenthesis that ends the invocation, including any commas. This set of
1417 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1418 wherever it appears. See the CPP manual for more information.
1420 GCC has long supported variadic macros, and used a different syntax that
1421 allowed you to give a name to the variable arguments just like any other
1422 argument. Here is an example:
1425 #define debug(format, args...) fprintf (stderr, format, args)
1428 This is in all ways equivalent to the ISO C example above, but arguably
1429 more readable and descriptive.
1431 GNU CPP has two further variadic macro extensions, and permits them to
1432 be used with either of the above forms of macro definition.
1434 In standard C, you are not allowed to leave the variable argument out
1435 entirely; but you are allowed to pass an empty argument. For example,
1436 this invocation is invalid in ISO C, because there is no comma after
1443 GNU CPP permits you to completely omit the variable arguments in this
1444 way. In the above examples, the compiler would complain, though since
1445 the expansion of the macro still has the extra comma after the format
1448 To help solve this problem, CPP behaves specially for variable arguments
1449 used with the token paste operator, @samp{##}. If instead you write
1452 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1455 and if the variable arguments are omitted or empty, the @samp{##}
1456 operator causes the preprocessor to remove the comma before it. If you
1457 do provide some variable arguments in your macro invocation, GNU CPP
1458 does not complain about the paste operation and instead places the
1459 variable arguments after the comma. Just like any other pasted macro
1460 argument, these arguments are not macro expanded.
1462 @node Escaped Newlines
1463 @section Slightly Looser Rules for Escaped Newlines
1464 @cindex escaped newlines
1465 @cindex newlines (escaped)
1467 Recently, the preprocessor has relaxed its treatment of escaped
1468 newlines. Previously, the newline had to immediately follow a
1469 backslash. The current implementation allows whitespace in the form
1470 of spaces, horizontal and vertical tabs, and form feeds between the
1471 backslash and the subsequent newline. The preprocessor issues a
1472 warning, but treats it as a valid escaped newline and combines the two
1473 lines to form a single logical line. This works within comments and
1474 tokens, as well as between tokens. Comments are @emph{not} treated as
1475 whitespace for the purposes of this relaxation, since they have not
1476 yet been replaced with spaces.
1479 @section Non-Lvalue Arrays May Have Subscripts
1480 @cindex subscripting
1481 @cindex arrays, non-lvalue
1483 @cindex subscripting and function values
1484 In ISO C99, arrays that are not lvalues still decay to pointers, and
1485 may be subscripted, although they may not be modified or used after
1486 the next sequence point and the unary @samp{&} operator may not be
1487 applied to them. As an extension, GCC allows such arrays to be
1488 subscripted in C89 mode, though otherwise they do not decay to
1489 pointers outside C99 mode. For example,
1490 this is valid in GNU C though not valid in C89:
1494 struct foo @{int a[4];@};
1500 return f().a[index];
1506 @section Arithmetic on @code{void}- and Function-Pointers
1507 @cindex void pointers, arithmetic
1508 @cindex void, size of pointer to
1509 @cindex function pointers, arithmetic
1510 @cindex function, size of pointer to
1512 In GNU C, addition and subtraction operations are supported on pointers to
1513 @code{void} and on pointers to functions. This is done by treating the
1514 size of a @code{void} or of a function as 1.
1516 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1517 and on function types, and returns 1.
1519 @opindex Wpointer-arith
1520 The option @option{-Wpointer-arith} requests a warning if these extensions
1524 @section Non-Constant Initializers
1525 @cindex initializers, non-constant
1526 @cindex non-constant initializers
1528 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1529 automatic variable are not required to be constant expressions in GNU C@.
1530 Here is an example of an initializer with run-time varying elements:
1533 foo (float f, float g)
1535 float beat_freqs[2] = @{ f-g, f+g @};
1540 @node Compound Literals
1541 @section Compound Literals
1542 @cindex constructor expressions
1543 @cindex initializations in expressions
1544 @cindex structures, constructor expression
1545 @cindex expressions, constructor
1546 @cindex compound literals
1547 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1549 ISO C99 supports compound literals. A compound literal looks like
1550 a cast containing an initializer. Its value is an object of the
1551 type specified in the cast, containing the elements specified in
1552 the initializer; it is an lvalue. As an extension, GCC supports
1553 compound literals in C89 mode and in C++.
1555 Usually, the specified type is a structure. Assume that
1556 @code{struct foo} and @code{structure} are declared as shown:
1559 struct foo @{int a; char b[2];@} structure;
1563 Here is an example of constructing a @code{struct foo} with a compound literal:
1566 structure = ((struct foo) @{x + y, 'a', 0@});
1570 This is equivalent to writing the following:
1574 struct foo temp = @{x + y, 'a', 0@};
1579 You can also construct an array. If all the elements of the compound literal
1580 are (made up of) simple constant expressions, suitable for use in
1581 initializers of objects of static storage duration, then the compound
1582 literal can be coerced to a pointer to its first element and used in
1583 such an initializer, as shown here:
1586 char **foo = (char *[]) @{ "x", "y", "z" @};
1589 Compound literals for scalar types and union types are is
1590 also allowed, but then the compound literal is equivalent
1593 As a GNU extension, GCC allows initialization of objects with static storage
1594 duration by compound literals (which is not possible in ISO C99, because
1595 the initializer is not a constant).
1596 It is handled as if the object was initialized only with the bracket
1597 enclosed list if the types of the compound literal and the object match.
1598 The initializer list of the compound literal must be constant.
1599 If the object being initialized has array type of unknown size, the size is
1600 determined by compound literal size.
1603 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1604 static int y[] = (int []) @{1, 2, 3@};
1605 static int z[] = (int [3]) @{1@};
1609 The above lines are equivalent to the following:
1611 static struct foo x = @{1, 'a', 'b'@};
1612 static int y[] = @{1, 2, 3@};
1613 static int z[] = @{1, 0, 0@};
1616 @node Designated Inits
1617 @section Designated Initializers
1618 @cindex initializers with labeled elements
1619 @cindex labeled elements in initializers
1620 @cindex case labels in initializers
1621 @cindex designated initializers
1623 Standard C89 requires the elements of an initializer to appear in a fixed
1624 order, the same as the order of the elements in the array or structure
1627 In ISO C99 you can give the elements in any order, specifying the array
1628 indices or structure field names they apply to, and GNU C allows this as
1629 an extension in C89 mode as well. This extension is not
1630 implemented in GNU C++.
1632 To specify an array index, write
1633 @samp{[@var{index}] =} before the element value. For example,
1636 int a[6] = @{ [4] = 29, [2] = 15 @};
1643 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1647 The index values must be constant expressions, even if the array being
1648 initialized is automatic.
1650 An alternative syntax for this which has been obsolete since GCC 2.5 but
1651 GCC still accepts is to write @samp{[@var{index}]} before the element
1652 value, with no @samp{=}.
1654 To initialize a range of elements to the same value, write
1655 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1656 extension. For example,
1659 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1663 If the value in it has side-effects, the side-effects will happen only once,
1664 not for each initialized field by the range initializer.
1667 Note that the length of the array is the highest value specified
1670 In a structure initializer, specify the name of a field to initialize
1671 with @samp{.@var{fieldname} =} before the element value. For example,
1672 given the following structure,
1675 struct point @{ int x, y; @};
1679 the following initialization
1682 struct point p = @{ .y = yvalue, .x = xvalue @};
1689 struct point p = @{ xvalue, yvalue @};
1692 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1693 @samp{@var{fieldname}:}, as shown here:
1696 struct point p = @{ y: yvalue, x: xvalue @};
1700 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1701 @dfn{designator}. You can also use a designator (or the obsolete colon
1702 syntax) when initializing a union, to specify which element of the union
1703 should be used. For example,
1706 union foo @{ int i; double d; @};
1708 union foo f = @{ .d = 4 @};
1712 will convert 4 to a @code{double} to store it in the union using
1713 the second element. By contrast, casting 4 to type @code{union foo}
1714 would store it into the union as the integer @code{i}, since it is
1715 an integer. (@xref{Cast to Union}.)
1717 You can combine this technique of naming elements with ordinary C
1718 initialization of successive elements. Each initializer element that
1719 does not have a designator applies to the next consecutive element of the
1720 array or structure. For example,
1723 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1730 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1733 Labeling the elements of an array initializer is especially useful
1734 when the indices are characters or belong to an @code{enum} type.
1739 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1740 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1743 @cindex designator lists
1744 You can also write a series of @samp{.@var{fieldname}} and
1745 @samp{[@var{index}]} designators before an @samp{=} to specify a
1746 nested subobject to initialize; the list is taken relative to the
1747 subobject corresponding to the closest surrounding brace pair. For
1748 example, with the @samp{struct point} declaration above:
1751 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1755 If the same field is initialized multiple times, it will have value from
1756 the last initialization. If any such overridden initialization has
1757 side-effect, it is unspecified whether the side-effect happens or not.
1758 Currently, GCC will discard them and issue a warning.
1761 @section Case Ranges
1763 @cindex ranges in case statements
1765 You can specify a range of consecutive values in a single @code{case} label,
1769 case @var{low} ... @var{high}:
1773 This has the same effect as the proper number of individual @code{case}
1774 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1776 This feature is especially useful for ranges of ASCII character codes:
1782 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1783 it may be parsed wrong when you use it with integer values. For example,
1798 @section Cast to a Union Type
1799 @cindex cast to a union
1800 @cindex union, casting to a
1802 A cast to union type is similar to other casts, except that the type
1803 specified is a union type. You can specify the type either with
1804 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1805 a constructor though, not a cast, and hence does not yield an lvalue like
1806 normal casts. (@xref{Compound Literals}.)
1808 The types that may be cast to the union type are those of the members
1809 of the union. Thus, given the following union and variables:
1812 union foo @{ int i; double d; @};
1818 both @code{x} and @code{y} can be cast to type @code{union foo}.
1820 Using the cast as the right-hand side of an assignment to a variable of
1821 union type is equivalent to storing in a member of the union:
1826 u = (union foo) x @equiv{} u.i = x
1827 u = (union foo) y @equiv{} u.d = y
1830 You can also use the union cast as a function argument:
1833 void hack (union foo);
1835 hack ((union foo) x);
1838 @node Mixed Declarations
1839 @section Mixed Declarations and Code
1840 @cindex mixed declarations and code
1841 @cindex declarations, mixed with code
1842 @cindex code, mixed with declarations
1844 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1845 within compound statements. As an extension, GCC also allows this in
1846 C89 mode. For example, you could do:
1855 Each identifier is visible from where it is declared until the end of
1856 the enclosing block.
1858 @node Function Attributes
1859 @section Declaring Attributes of Functions
1860 @cindex function attributes
1861 @cindex declaring attributes of functions
1862 @cindex functions that never return
1863 @cindex functions that return more than once
1864 @cindex functions that have no side effects
1865 @cindex functions in arbitrary sections
1866 @cindex functions that behave like malloc
1867 @cindex @code{volatile} applied to function
1868 @cindex @code{const} applied to function
1869 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1870 @cindex functions with non-null pointer arguments
1871 @cindex functions that are passed arguments in registers on the 386
1872 @cindex functions that pop the argument stack on the 386
1873 @cindex functions that do not pop the argument stack on the 386
1874 @cindex functions that have different compilation options on the 386
1875 @cindex functions that have different optimization options
1877 In GNU C, you declare certain things about functions called in your program
1878 which help the compiler optimize function calls and check your code more
1881 The keyword @code{__attribute__} allows you to specify special
1882 attributes when making a declaration. This keyword is followed by an
1883 attribute specification inside double parentheses. The following
1884 attributes are currently defined for functions on all targets:
1885 @code{aligned}, @code{alloc_size}, @code{noreturn},
1886 @code{returns_twice}, @code{noinline}, @code{always_inline},
1887 @code{flatten}, @code{pure}, @code{const}, @code{nothrow},
1888 @code{sentinel}, @code{format}, @code{format_arg},
1889 @code{no_instrument_function}, @code{section}, @code{constructor},
1890 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1891 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1892 @code{nonnull}, @code{gnu_inline}, @code{externally_visible},
1893 @code{hot}, @code{cold}, @code{artificial}, @code{error}
1895 Several other attributes are defined for functions on particular
1896 target systems. Other attributes, including @code{section} are
1897 supported for variables declarations (@pxref{Variable Attributes}) and
1898 for types (@pxref{Type Attributes}).
1900 You may also specify attributes with @samp{__} preceding and following
1901 each keyword. This allows you to use them in header files without
1902 being concerned about a possible macro of the same name. For example,
1903 you may use @code{__noreturn__} instead of @code{noreturn}.
1905 @xref{Attribute Syntax}, for details of the exact syntax for using
1909 @c Keep this table alphabetized by attribute name. Treat _ as space.
1911 @item alias ("@var{target}")
1912 @cindex @code{alias} attribute
1913 The @code{alias} attribute causes the declaration to be emitted as an
1914 alias for another symbol, which must be specified. For instance,
1917 void __f () @{ /* @r{Do something.} */; @}
1918 void f () __attribute__ ((weak, alias ("__f")));
1921 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1922 mangled name for the target must be used. It is an error if @samp{__f}
1923 is not defined in the same translation unit.
1925 Not all target machines support this attribute.
1927 @item aligned (@var{alignment})
1928 @cindex @code{aligned} attribute
1929 This attribute specifies a minimum alignment for the function,
1932 You cannot use this attribute to decrease the alignment of a function,
1933 only to increase it. However, when you explicitly specify a function
1934 alignment this will override the effect of the
1935 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1938 Note that the effectiveness of @code{aligned} attributes may be
1939 limited by inherent limitations in your linker. On many systems, the
1940 linker is only able to arrange for functions to be aligned up to a
1941 certain maximum alignment. (For some linkers, the maximum supported
1942 alignment may be very very small.) See your linker documentation for
1943 further information.
1945 The @code{aligned} attribute can also be used for variables and fields
1946 (@pxref{Variable Attributes}.)
1949 @cindex @code{alloc_size} attribute
1950 The @code{alloc_size} attribute is used to tell the compiler that the
1951 function return value points to memory, where the size is given by
1952 one or two of the functions parameters. GCC uses this
1953 information to improve the correctness of @code{__builtin_object_size}.
1955 The function parameter(s) denoting the allocated size are specified by
1956 one or two integer arguments supplied to the attribute. The allocated size
1957 is either the value of the single function argument specified or the product
1958 of the two function arguments specified. Argument numbering starts at
1964 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1965 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
1968 declares that my_calloc will return memory of the size given by
1969 the product of parameter 1 and 2 and that my_realloc will return memory
1970 of the size given by parameter 2.
1973 @cindex @code{always_inline} function attribute
1974 Generally, functions are not inlined unless optimization is specified.
1975 For functions declared inline, this attribute inlines the function even
1976 if no optimization level was specified.
1979 @cindex @code{gnu_inline} function attribute
1980 This attribute should be used with a function which is also declared
1981 with the @code{inline} keyword. It directs GCC to treat the function
1982 as if it were defined in gnu89 mode even when compiling in C99 or
1985 If the function is declared @code{extern}, then this definition of the
1986 function is used only for inlining. In no case is the function
1987 compiled as a standalone function, not even if you take its address
1988 explicitly. Such an address becomes an external reference, as if you
1989 had only declared the function, and had not defined it. This has
1990 almost the effect of a macro. The way to use this is to put a
1991 function definition in a header file with this attribute, and put
1992 another copy of the function, without @code{extern}, in a library
1993 file. The definition in the header file will cause most calls to the
1994 function to be inlined. If any uses of the function remain, they will
1995 refer to the single copy in the library. Note that the two
1996 definitions of the functions need not be precisely the same, although
1997 if they do not have the same effect your program may behave oddly.
1999 In C, if the function is neither @code{extern} nor @code{static}, then
2000 the function is compiled as a standalone function, as well as being
2001 inlined where possible.
2003 This is how GCC traditionally handled functions declared
2004 @code{inline}. Since ISO C99 specifies a different semantics for
2005 @code{inline}, this function attribute is provided as a transition
2006 measure and as a useful feature in its own right. This attribute is
2007 available in GCC 4.1.3 and later. It is available if either of the
2008 preprocessor macros @code{__GNUC_GNU_INLINE__} or
2009 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
2010 Function is As Fast As a Macro}.
2012 In C++, this attribute does not depend on @code{extern} in any way,
2013 but it still requires the @code{inline} keyword to enable its special
2017 @cindex @code{artificial} function attribute
2018 This attribute is useful for small inline wrappers which if possible
2019 should appear during debugging as a unit, depending on the debug
2020 info format it will either mean marking the function as artificial
2021 or using the caller location for all instructions within the inlined
2025 @cindex @code{flatten} function attribute
2026 Generally, inlining into a function is limited. For a function marked with
2027 this attribute, every call inside this function will be inlined, if possible.
2028 Whether the function itself is considered for inlining depends on its size and
2029 the current inlining parameters.
2031 @item error ("@var{message}")
2032 @cindex @code{error} function attribute
2033 If this attribute is used on a function declaration and a call to such a function
2034 is not eliminated through dead code elimination or other optimizations, an error
2035 which will include @var{message} will be diagnosed. This is useful
2036 for compile time checking, especially together with @code{__builtin_constant_p}
2037 and inline functions where checking the inline function arguments is not
2038 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
2039 While it is possible to leave the function undefined and thus invoke
2040 a link failure, when using this attribute the problem will be diagnosed
2041 earlier and with exact location of the call even in presence of inline
2042 functions or when not emitting debugging information.
2044 @item warning ("@var{message}")
2045 @cindex @code{warning} function attribute
2046 If this attribute is used on a function declaration and a call to such a function
2047 is not eliminated through dead code elimination or other optimizations, a warning
2048 which will include @var{message} will be diagnosed. This is useful
2049 for compile time checking, especially together with @code{__builtin_constant_p}
2050 and inline functions. While it is possible to define the function with
2051 a message in @code{.gnu.warning*} section, when using this attribute the problem
2052 will be diagnosed earlier and with exact location of the call even in presence
2053 of inline functions or when not emitting debugging information.
2056 @cindex functions that do pop the argument stack on the 386
2058 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2059 assume that the calling function will pop off the stack space used to
2060 pass arguments. This is
2061 useful to override the effects of the @option{-mrtd} switch.
2064 @cindex @code{const} function attribute
2065 Many functions do not examine any values except their arguments, and
2066 have no effects except the return value. Basically this is just slightly
2067 more strict class than the @code{pure} attribute below, since function is not
2068 allowed to read global memory.
2070 @cindex pointer arguments
2071 Note that a function that has pointer arguments and examines the data
2072 pointed to must @emph{not} be declared @code{const}. Likewise, a
2073 function that calls a non-@code{const} function usually must not be
2074 @code{const}. It does not make sense for a @code{const} function to
2077 The attribute @code{const} is not implemented in GCC versions earlier
2078 than 2.5. An alternative way to declare that a function has no side
2079 effects, which works in the current version and in some older versions,
2083 typedef int intfn ();
2085 extern const intfn square;
2088 This approach does not work in GNU C++ from 2.6.0 on, since the language
2089 specifies that the @samp{const} must be attached to the return value.
2093 @itemx constructor (@var{priority})
2094 @itemx destructor (@var{priority})
2095 @cindex @code{constructor} function attribute
2096 @cindex @code{destructor} function attribute
2097 The @code{constructor} attribute causes the function to be called
2098 automatically before execution enters @code{main ()}. Similarly, the
2099 @code{destructor} attribute causes the function to be called
2100 automatically after @code{main ()} has completed or @code{exit ()} has
2101 been called. Functions with these attributes are useful for
2102 initializing data that will be used implicitly during the execution of
2105 You may provide an optional integer priority to control the order in
2106 which constructor and destructor functions are run. A constructor
2107 with a smaller priority number runs before a constructor with a larger
2108 priority number; the opposite relationship holds for destructors. So,
2109 if you have a constructor that allocates a resource and a destructor
2110 that deallocates the same resource, both functions typically have the
2111 same priority. The priorities for constructor and destructor
2112 functions are the same as those specified for namespace-scope C++
2113 objects (@pxref{C++ Attributes}).
2115 These attributes are not currently implemented for Objective-C@.
2118 @itemx deprecated (@var{msg})
2119 @cindex @code{deprecated} attribute.
2120 The @code{deprecated} attribute results in a warning if the function
2121 is used anywhere in the source file. This is useful when identifying
2122 functions that are expected to be removed in a future version of a
2123 program. The warning also includes the location of the declaration
2124 of the deprecated function, to enable users to easily find further
2125 information about why the function is deprecated, or what they should
2126 do instead. Note that the warnings only occurs for uses:
2129 int old_fn () __attribute__ ((deprecated));
2131 int (*fn_ptr)() = old_fn;
2134 results in a warning on line 3 but not line 2. The optional msg
2135 argument, which must be a string, will be printed in the warning if
2138 The @code{deprecated} attribute can also be used for variables and
2139 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2142 @cindex @code{__declspec(dllexport)}
2143 On Microsoft Windows targets and Symbian OS targets the
2144 @code{dllexport} attribute causes the compiler to provide a global
2145 pointer to a pointer in a DLL, so that it can be referenced with the
2146 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2147 name is formed by combining @code{_imp__} and the function or variable
2150 You can use @code{__declspec(dllexport)} as a synonym for
2151 @code{__attribute__ ((dllexport))} for compatibility with other
2154 On systems that support the @code{visibility} attribute, this
2155 attribute also implies ``default'' visibility. It is an error to
2156 explicitly specify any other visibility.
2158 Currently, the @code{dllexport} attribute is ignored for inlined
2159 functions, unless the @option{-fkeep-inline-functions} flag has been
2160 used. The attribute is also ignored for undefined symbols.
2162 When applied to C++ classes, the attribute marks defined non-inlined
2163 member functions and static data members as exports. Static consts
2164 initialized in-class are not marked unless they are also defined
2167 For Microsoft Windows targets there are alternative methods for
2168 including the symbol in the DLL's export table such as using a
2169 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2170 the @option{--export-all} linker flag.
2173 @cindex @code{__declspec(dllimport)}
2174 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2175 attribute causes the compiler to reference a function or variable via
2176 a global pointer to a pointer that is set up by the DLL exporting the
2177 symbol. The attribute implies @code{extern}. On Microsoft Windows
2178 targets, the pointer name is formed by combining @code{_imp__} and the
2179 function or variable name.
2181 You can use @code{__declspec(dllimport)} as a synonym for
2182 @code{__attribute__ ((dllimport))} for compatibility with other
2185 On systems that support the @code{visibility} attribute, this
2186 attribute also implies ``default'' visibility. It is an error to
2187 explicitly specify any other visibility.
2189 Currently, the attribute is ignored for inlined functions. If the
2190 attribute is applied to a symbol @emph{definition}, an error is reported.
2191 If a symbol previously declared @code{dllimport} is later defined, the
2192 attribute is ignored in subsequent references, and a warning is emitted.
2193 The attribute is also overridden by a subsequent declaration as
2196 When applied to C++ classes, the attribute marks non-inlined
2197 member functions and static data members as imports. However, the
2198 attribute is ignored for virtual methods to allow creation of vtables
2201 On the SH Symbian OS target the @code{dllimport} attribute also has
2202 another affect---it can cause the vtable and run-time type information
2203 for a class to be exported. This happens when the class has a
2204 dllimport'ed constructor or a non-inline, non-pure virtual function
2205 and, for either of those two conditions, the class also has an inline
2206 constructor or destructor and has a key function that is defined in
2207 the current translation unit.
2209 For Microsoft Windows based targets the use of the @code{dllimport}
2210 attribute on functions is not necessary, but provides a small
2211 performance benefit by eliminating a thunk in the DLL@. The use of the
2212 @code{dllimport} attribute on imported variables was required on older
2213 versions of the GNU linker, but can now be avoided by passing the
2214 @option{--enable-auto-import} switch to the GNU linker. As with
2215 functions, using the attribute for a variable eliminates a thunk in
2218 One drawback to using this attribute is that a pointer to a
2219 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2220 address. However, a pointer to a @emph{function} with the
2221 @code{dllimport} attribute can be used as a constant initializer; in
2222 this case, the address of a stub function in the import lib is
2223 referenced. On Microsoft Windows targets, the attribute can be disabled
2224 for functions by setting the @option{-mnop-fun-dllimport} flag.
2227 @cindex eight bit data on the H8/300, H8/300H, and H8S
2228 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2229 variable should be placed into the eight bit data section.
2230 The compiler will generate more efficient code for certain operations
2231 on data in the eight bit data area. Note the eight bit data area is limited to
2234 You must use GAS and GLD from GNU binutils version 2.7 or later for
2235 this attribute to work correctly.
2237 @item exception_handler
2238 @cindex exception handler functions on the Blackfin processor
2239 Use this attribute on the Blackfin to indicate that the specified function
2240 is an exception handler. The compiler will generate function entry and
2241 exit sequences suitable for use in an exception handler when this
2242 attribute is present.
2244 @item externally_visible
2245 @cindex @code{externally_visible} attribute.
2246 This attribute, attached to a global variable or function, nullifies
2247 the effect of the @option{-fwhole-program} command-line option, so the
2248 object remains visible outside the current compilation unit.
2251 @cindex functions which handle memory bank switching
2252 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2253 use a calling convention that takes care of switching memory banks when
2254 entering and leaving a function. This calling convention is also the
2255 default when using the @option{-mlong-calls} option.
2257 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2258 to call and return from a function.
2260 On 68HC11 the compiler will generate a sequence of instructions
2261 to invoke a board-specific routine to switch the memory bank and call the
2262 real function. The board-specific routine simulates a @code{call}.
2263 At the end of a function, it will jump to a board-specific routine
2264 instead of using @code{rts}. The board-specific return routine simulates
2268 @cindex functions that pop the argument stack on the 386
2269 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2270 pass the first argument (if of integral type) in the register ECX and
2271 the second argument (if of integral type) in the register EDX@. Subsequent
2272 and other typed arguments are passed on the stack. The called function will
2273 pop the arguments off the stack. If the number of arguments is variable all
2274 arguments are pushed on the stack.
2276 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2277 @cindex @code{format} function attribute
2279 The @code{format} attribute specifies that a function takes @code{printf},
2280 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2281 should be type-checked against a format string. For example, the
2286 my_printf (void *my_object, const char *my_format, ...)
2287 __attribute__ ((format (printf, 2, 3)));
2291 causes the compiler to check the arguments in calls to @code{my_printf}
2292 for consistency with the @code{printf} style format string argument
2295 The parameter @var{archetype} determines how the format string is
2296 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2297 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2298 @code{strfmon}. (You can also use @code{__printf__},
2299 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2300 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2301 @code{ms_strftime} are also present.
2302 @var{archtype} values such as @code{printf} refer to the formats accepted
2303 by the system's C run-time library, while @code{gnu_} values always refer
2304 to the formats accepted by the GNU C Library. On Microsoft Windows
2305 targets, @code{ms_} values refer to the formats accepted by the
2306 @file{msvcrt.dll} library.
2307 The parameter @var{string-index}
2308 specifies which argument is the format string argument (starting
2309 from 1), while @var{first-to-check} is the number of the first
2310 argument to check against the format string. For functions
2311 where the arguments are not available to be checked (such as
2312 @code{vprintf}), specify the third parameter as zero. In this case the
2313 compiler only checks the format string for consistency. For
2314 @code{strftime} formats, the third parameter is required to be zero.
2315 Since non-static C++ methods have an implicit @code{this} argument, the
2316 arguments of such methods should be counted from two, not one, when
2317 giving values for @var{string-index} and @var{first-to-check}.
2319 In the example above, the format string (@code{my_format}) is the second
2320 argument of the function @code{my_print}, and the arguments to check
2321 start with the third argument, so the correct parameters for the format
2322 attribute are 2 and 3.
2324 @opindex ffreestanding
2325 @opindex fno-builtin
2326 The @code{format} attribute allows you to identify your own functions
2327 which take format strings as arguments, so that GCC can check the
2328 calls to these functions for errors. The compiler always (unless
2329 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2330 for the standard library functions @code{printf}, @code{fprintf},
2331 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2332 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2333 warnings are requested (using @option{-Wformat}), so there is no need to
2334 modify the header file @file{stdio.h}. In C99 mode, the functions
2335 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2336 @code{vsscanf} are also checked. Except in strictly conforming C
2337 standard modes, the X/Open function @code{strfmon} is also checked as
2338 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2339 @xref{C Dialect Options,,Options Controlling C Dialect}.
2341 The target may provide additional types of format checks.
2342 @xref{Target Format Checks,,Format Checks Specific to Particular
2345 @item format_arg (@var{string-index})
2346 @cindex @code{format_arg} function attribute
2347 @opindex Wformat-nonliteral
2348 The @code{format_arg} attribute specifies that a function takes a format
2349 string for a @code{printf}, @code{scanf}, @code{strftime} or
2350 @code{strfmon} style function and modifies it (for example, to translate
2351 it into another language), so the result can be passed to a
2352 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2353 function (with the remaining arguments to the format function the same
2354 as they would have been for the unmodified string). For example, the
2359 my_dgettext (char *my_domain, const char *my_format)
2360 __attribute__ ((format_arg (2)));
2364 causes the compiler to check the arguments in calls to a @code{printf},
2365 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2366 format string argument is a call to the @code{my_dgettext} function, for
2367 consistency with the format string argument @code{my_format}. If the
2368 @code{format_arg} attribute had not been specified, all the compiler
2369 could tell in such calls to format functions would be that the format
2370 string argument is not constant; this would generate a warning when
2371 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2372 without the attribute.
2374 The parameter @var{string-index} specifies which argument is the format
2375 string argument (starting from one). Since non-static C++ methods have
2376 an implicit @code{this} argument, the arguments of such methods should
2377 be counted from two.
2379 The @code{format-arg} attribute allows you to identify your own
2380 functions which modify format strings, so that GCC can check the
2381 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2382 type function whose operands are a call to one of your own function.
2383 The compiler always treats @code{gettext}, @code{dgettext}, and
2384 @code{dcgettext} in this manner except when strict ISO C support is
2385 requested by @option{-ansi} or an appropriate @option{-std} option, or
2386 @option{-ffreestanding} or @option{-fno-builtin}
2387 is used. @xref{C Dialect Options,,Options
2388 Controlling C Dialect}.
2390 @item function_vector
2391 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2392 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2393 function should be called through the function vector. Calling a
2394 function through the function vector will reduce code size, however;
2395 the function vector has a limited size (maximum 128 entries on the H8/300
2396 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2398 In SH2A target, this attribute declares a function to be called using the
2399 TBR relative addressing mode. The argument to this attribute is the entry
2400 number of the same function in a vector table containing all the TBR
2401 relative addressable functions. For the successful jump, register TBR
2402 should contain the start address of this TBR relative vector table.
2403 In the startup routine of the user application, user needs to care of this
2404 TBR register initialization. The TBR relative vector table can have at
2405 max 256 function entries. The jumps to these functions will be generated
2406 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2407 You must use GAS and GLD from GNU binutils version 2.7 or later for
2408 this attribute to work correctly.
2410 Please refer the example of M16C target, to see the use of this
2411 attribute while declaring a function,
2413 In an application, for a function being called once, this attribute will
2414 save at least 8 bytes of code; and if other successive calls are being
2415 made to the same function, it will save 2 bytes of code per each of these
2418 On M16C/M32C targets, the @code{function_vector} attribute declares a
2419 special page subroutine call function. Use of this attribute reduces
2420 the code size by 2 bytes for each call generated to the
2421 subroutine. The argument to the attribute is the vector number entry
2422 from the special page vector table which contains the 16 low-order
2423 bits of the subroutine's entry address. Each vector table has special
2424 page number (18 to 255) which are used in @code{jsrs} instruction.
2425 Jump addresses of the routines are generated by adding 0x0F0000 (in
2426 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2427 byte addresses set in the vector table. Therefore you need to ensure
2428 that all the special page vector routines should get mapped within the
2429 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2432 In the following example 2 bytes will be saved for each call to
2433 function @code{foo}.
2436 void foo (void) __attribute__((function_vector(0x18)));
2447 If functions are defined in one file and are called in another file,
2448 then be sure to write this declaration in both files.
2450 This attribute is ignored for R8C target.
2453 @cindex interrupt handler functions
2454 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k, MIPS
2455 and Xstormy16 ports to indicate that the specified function is an
2456 interrupt handler. The compiler will generate function entry and exit
2457 sequences suitable for use in an interrupt handler when this attribute
2460 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2461 SH processors can be specified via the @code{interrupt_handler} attribute.
2463 Note, on the AVR, interrupts will be enabled inside the function.
2465 Note, for the ARM, you can specify the kind of interrupt to be handled by
2466 adding an optional parameter to the interrupt attribute like this:
2469 void f () __attribute__ ((interrupt ("IRQ")));
2472 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2474 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2475 may be called with a word aligned stack pointer.
2477 On MIPS targets, you can use the following attributes to modify the behavior
2478 of an interrupt handler:
2480 @item use_shadow_register_set
2481 @cindex @code{use_shadow_register_set} attribute
2482 Assume that the handler uses a shadow register set, instead of
2483 the main general-purpose registers.
2485 @item keep_interrupts_masked
2486 @cindex @code{keep_interrupts_masked} attribute
2487 Keep interrupts masked for the whole function. Without this attribute,
2488 GCC tries to reenable interrupts for as much of the function as it can.
2490 @item use_debug_exception_return
2491 @cindex @code{use_debug_exception_return} attribute
2492 Return using the @code{deret} instruction. Interrupt handlers that don't
2493 have this attribute return using @code{eret} instead.
2496 You can use any combination of these attributes, as shown below:
2498 void __attribute__ ((interrupt)) v0 ();
2499 void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
2500 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
2501 void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
2502 void __attribute__ ((interrupt, use_shadow_register_set,
2503 keep_interrupts_masked)) v4 ();
2504 void __attribute__ ((interrupt, use_shadow_register_set,
2505 use_debug_exception_return)) v5 ();
2506 void __attribute__ ((interrupt, keep_interrupts_masked,
2507 use_debug_exception_return)) v6 ();
2508 void __attribute__ ((interrupt, use_shadow_register_set,
2509 keep_interrupts_masked,
2510 use_debug_exception_return)) v7 ();
2513 @item interrupt_handler
2514 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2515 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2516 indicate that the specified function is an interrupt handler. The compiler
2517 will generate function entry and exit sequences suitable for use in an
2518 interrupt handler when this attribute is present.
2520 @item interrupt_thread
2521 @cindex interrupt thread functions on fido
2522 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2523 that the specified function is an interrupt handler that is designed
2524 to run as a thread. The compiler omits generate prologue/epilogue
2525 sequences and replaces the return instruction with a @code{sleep}
2526 instruction. This attribute is available only on fido.
2529 @cindex interrupt service routines on ARM
2530 Use this attribute on ARM to write Interrupt Service Routines. This is an
2531 alias to the @code{interrupt} attribute above.
2534 @cindex User stack pointer in interrupts on the Blackfin
2535 When used together with @code{interrupt_handler}, @code{exception_handler}
2536 or @code{nmi_handler}, code will be generated to load the stack pointer
2537 from the USP register in the function prologue.
2540 @cindex @code{l1_text} function attribute
2541 This attribute specifies a function to be placed into L1 Instruction
2542 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2543 With @option{-mfdpic}, function calls with a such function as the callee
2544 or caller will use inlined PLT.
2546 @item long_call/short_call
2547 @cindex indirect calls on ARM
2548 This attribute specifies how a particular function is called on
2549 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2550 command line switch and @code{#pragma long_calls} settings. The
2551 @code{long_call} attribute indicates that the function might be far
2552 away from the call site and require a different (more expensive)
2553 calling sequence. The @code{short_call} attribute always places
2554 the offset to the function from the call site into the @samp{BL}
2555 instruction directly.
2557 @item longcall/shortcall
2558 @cindex functions called via pointer on the RS/6000 and PowerPC
2559 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2560 indicates that the function might be far away from the call site and
2561 require a different (more expensive) calling sequence. The
2562 @code{shortcall} attribute indicates that the function is always close
2563 enough for the shorter calling sequence to be used. These attributes
2564 override both the @option{-mlongcall} switch and, on the RS/6000 and
2565 PowerPC, the @code{#pragma longcall} setting.
2567 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2568 calls are necessary.
2570 @item long_call/near/far
2571 @cindex indirect calls on MIPS
2572 These attributes specify how a particular function is called on MIPS@.
2573 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2574 command-line switch. The @code{long_call} and @code{far} attributes are
2575 synonyms, and cause the compiler to always call
2576 the function by first loading its address into a register, and then using
2577 the contents of that register. The @code{near} attribute has the opposite
2578 effect; it specifies that non-PIC calls should be made using the more
2579 efficient @code{jal} instruction.
2582 @cindex @code{malloc} attribute
2583 The @code{malloc} attribute is used to tell the compiler that a function
2584 may be treated as if any non-@code{NULL} pointer it returns cannot
2585 alias any other pointer valid when the function returns.
2586 This will often improve optimization.
2587 Standard functions with this property include @code{malloc} and
2588 @code{calloc}. @code{realloc}-like functions have this property as
2589 long as the old pointer is never referred to (including comparing it
2590 to the new pointer) after the function returns a non-@code{NULL}
2593 @item mips16/nomips16
2594 @cindex @code{mips16} attribute
2595 @cindex @code{nomips16} attribute
2597 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2598 function attributes to locally select or turn off MIPS16 code generation.
2599 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2600 while MIPS16 code generation is disabled for functions with the
2601 @code{nomips16} attribute. These attributes override the
2602 @option{-mips16} and @option{-mno-mips16} options on the command line
2603 (@pxref{MIPS Options}).
2605 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2606 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2607 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2608 may interact badly with some GCC extensions such as @code{__builtin_apply}
2609 (@pxref{Constructing Calls}).
2611 @item model (@var{model-name})
2612 @cindex function addressability on the M32R/D
2613 @cindex variable addressability on the IA-64
2615 On the M32R/D, use this attribute to set the addressability of an
2616 object, and of the code generated for a function. The identifier
2617 @var{model-name} is one of @code{small}, @code{medium}, or
2618 @code{large}, representing each of the code models.
2620 Small model objects live in the lower 16MB of memory (so that their
2621 addresses can be loaded with the @code{ld24} instruction), and are
2622 callable with the @code{bl} instruction.
2624 Medium model objects may live anywhere in the 32-bit address space (the
2625 compiler will generate @code{seth/add3} instructions to load their addresses),
2626 and are callable with the @code{bl} instruction.
2628 Large model objects may live anywhere in the 32-bit address space (the
2629 compiler will generate @code{seth/add3} instructions to load their addresses),
2630 and may not be reachable with the @code{bl} instruction (the compiler will
2631 generate the much slower @code{seth/add3/jl} instruction sequence).
2633 On IA-64, use this attribute to set the addressability of an object.
2634 At present, the only supported identifier for @var{model-name} is
2635 @code{small}, indicating addressability via ``small'' (22-bit)
2636 addresses (so that their addresses can be loaded with the @code{addl}
2637 instruction). Caveat: such addressing is by definition not position
2638 independent and hence this attribute must not be used for objects
2639 defined by shared libraries.
2641 @item ms_abi/sysv_abi
2642 @cindex @code{ms_abi} attribute
2643 @cindex @code{sysv_abi} attribute
2645 On 64-bit x86_64-*-* targets, you can use an ABI attribute to indicate
2646 which calling convention should be used for a function. The @code{ms_abi}
2647 attribute tells the compiler to use the Microsoft ABI, while the
2648 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2649 GNU/Linux and other systems. The default is to use the Microsoft ABI
2650 when targeting Windows. On all other systems, the default is the AMD ABI.
2652 Note, This feature is currently sorried out for Windows targets trying to
2655 @cindex function without a prologue/epilogue code
2656 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate that
2657 the specified function does not need prologue/epilogue sequences generated by
2658 the compiler. It is up to the programmer to provide these sequences. The
2659 only statements that can be safely included in naked functions are
2660 @code{asm} statements that do not have operands. All other statements,
2661 including declarations of local variables, @code{if} statements, and so
2662 forth, should be avoided. Naked functions should be used to implement the
2663 body of an assembly function, while allowing the compiler to construct
2664 the requisite function declaration for the assembler.
2667 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2668 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2669 use the normal calling convention based on @code{jsr} and @code{rts}.
2670 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2674 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2675 Use this attribute together with @code{interrupt_handler},
2676 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2677 entry code should enable nested interrupts or exceptions.
2680 @cindex NMI handler functions on the Blackfin processor
2681 Use this attribute on the Blackfin to indicate that the specified function
2682 is an NMI handler. The compiler will generate function entry and
2683 exit sequences suitable for use in an NMI handler when this
2684 attribute is present.
2686 @item no_instrument_function
2687 @cindex @code{no_instrument_function} function attribute
2688 @opindex finstrument-functions
2689 If @option{-finstrument-functions} is given, profiling function calls will
2690 be generated at entry and exit of most user-compiled functions.
2691 Functions with this attribute will not be so instrumented.
2694 @cindex @code{noinline} function attribute
2695 This function attribute prevents a function from being considered for
2697 @c Don't enumerate the optimizations by name here; we try to be
2698 @c future-compatible with this mechanism.
2699 If the function does not have side-effects, there are optimizations
2700 other than inlining that causes function calls to be optimized away,
2701 although the function call is live. To keep such calls from being
2706 (@pxref{Extended Asm}) in the called function, to serve as a special
2709 @item nonnull (@var{arg-index}, @dots{})
2710 @cindex @code{nonnull} function attribute
2711 The @code{nonnull} attribute specifies that some function parameters should
2712 be non-null pointers. For instance, the declaration:
2716 my_memcpy (void *dest, const void *src, size_t len)
2717 __attribute__((nonnull (1, 2)));
2721 causes the compiler to check that, in calls to @code{my_memcpy},
2722 arguments @var{dest} and @var{src} are non-null. If the compiler
2723 determines that a null pointer is passed in an argument slot marked
2724 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2725 is issued. The compiler may also choose to make optimizations based
2726 on the knowledge that certain function arguments will not be null.
2728 If no argument index list is given to the @code{nonnull} attribute,
2729 all pointer arguments are marked as non-null. To illustrate, the
2730 following declaration is equivalent to the previous example:
2734 my_memcpy (void *dest, const void *src, size_t len)
2735 __attribute__((nonnull));
2739 @cindex @code{noreturn} function attribute
2740 A few standard library functions, such as @code{abort} and @code{exit},
2741 cannot return. GCC knows this automatically. Some programs define
2742 their own functions that never return. You can declare them
2743 @code{noreturn} to tell the compiler this fact. For example,
2747 void fatal () __attribute__ ((noreturn));
2750 fatal (/* @r{@dots{}} */)
2752 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2758 The @code{noreturn} keyword tells the compiler to assume that
2759 @code{fatal} cannot return. It can then optimize without regard to what
2760 would happen if @code{fatal} ever did return. This makes slightly
2761 better code. More importantly, it helps avoid spurious warnings of
2762 uninitialized variables.
2764 The @code{noreturn} keyword does not affect the exceptional path when that
2765 applies: a @code{noreturn}-marked function may still return to the caller
2766 by throwing an exception or calling @code{longjmp}.
2768 Do not assume that registers saved by the calling function are
2769 restored before calling the @code{noreturn} function.
2771 It does not make sense for a @code{noreturn} function to have a return
2772 type other than @code{void}.
2774 The attribute @code{noreturn} is not implemented in GCC versions
2775 earlier than 2.5. An alternative way to declare that a function does
2776 not return, which works in the current version and in some older
2777 versions, is as follows:
2780 typedef void voidfn ();
2782 volatile voidfn fatal;
2785 This approach does not work in GNU C++.
2788 @cindex @code{nothrow} function attribute
2789 The @code{nothrow} attribute is used to inform the compiler that a
2790 function cannot throw an exception. For example, most functions in
2791 the standard C library can be guaranteed not to throw an exception
2792 with the notable exceptions of @code{qsort} and @code{bsearch} that
2793 take function pointer arguments. The @code{nothrow} attribute is not
2794 implemented in GCC versions earlier than 3.3.
2797 @cindex @code{optimize} function attribute
2798 The @code{optimize} attribute is used to specify that a function is to
2799 be compiled with different optimization options than specified on the
2800 command line. Arguments can either be numbers or strings. Numbers
2801 are assumed to be an optimization level. Strings that begin with
2802 @code{O} are assumed to be an optimization option, while other options
2803 are assumed to be used with a @code{-f} prefix. You can also use the
2804 @samp{#pragma GCC optimize} pragma to set the optimization options
2805 that affect more than one function.
2806 @xref{Function Specific Option Pragmas}, for details about the
2807 @samp{#pragma GCC optimize} pragma.
2809 This can be used for instance to have frequently executed functions
2810 compiled with more aggressive optimization options that produce faster
2811 and larger code, while other functions can be called with less
2815 @cindex @code{pure} function attribute
2816 Many functions have no effects except the return value and their
2817 return value depends only on the parameters and/or global variables.
2818 Such a function can be subject
2819 to common subexpression elimination and loop optimization just as an
2820 arithmetic operator would be. These functions should be declared
2821 with the attribute @code{pure}. For example,
2824 int square (int) __attribute__ ((pure));
2828 says that the hypothetical function @code{square} is safe to call
2829 fewer times than the program says.
2831 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2832 Interesting non-pure functions are functions with infinite loops or those
2833 depending on volatile memory or other system resource, that may change between
2834 two consecutive calls (such as @code{feof} in a multithreading environment).
2836 The attribute @code{pure} is not implemented in GCC versions earlier
2840 @cindex @code{hot} function attribute
2841 The @code{hot} attribute is used to inform the compiler that a function is a
2842 hot spot of the compiled program. The function is optimized more aggressively
2843 and on many target it is placed into special subsection of the text section so
2844 all hot functions appears close together improving locality.
2846 When profile feedback is available, via @option{-fprofile-use}, hot functions
2847 are automatically detected and this attribute is ignored.
2849 The @code{hot} attribute is not implemented in GCC versions earlier
2853 @cindex @code{cold} function attribute
2854 The @code{cold} attribute is used to inform the compiler that a function is
2855 unlikely executed. The function is optimized for size rather than speed and on
2856 many targets it is placed into special subsection of the text section so all
2857 cold functions appears close together improving code locality of non-cold parts
2858 of program. The paths leading to call of cold functions within code are marked
2859 as unlikely by the branch prediction mechanism. It is thus useful to mark
2860 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2861 improve optimization of hot functions that do call marked functions in rare
2864 When profile feedback is available, via @option{-fprofile-use}, hot functions
2865 are automatically detected and this attribute is ignored.
2867 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
2869 @item regparm (@var{number})
2870 @cindex @code{regparm} attribute
2871 @cindex functions that are passed arguments in registers on the 386
2872 On the Intel 386, the @code{regparm} attribute causes the compiler to
2873 pass arguments number one to @var{number} if they are of integral type
2874 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2875 take a variable number of arguments will continue to be passed all of their
2876 arguments on the stack.
2878 Beware that on some ELF systems this attribute is unsuitable for
2879 global functions in shared libraries with lazy binding (which is the
2880 default). Lazy binding will send the first call via resolving code in
2881 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2882 per the standard calling conventions. Solaris 8 is affected by this.
2883 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2884 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
2885 disabled with the linker or the loader if desired, to avoid the
2889 @cindex @code{sseregparm} attribute
2890 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2891 causes the compiler to pass up to 3 floating point arguments in
2892 SSE registers instead of on the stack. Functions that take a
2893 variable number of arguments will continue to pass all of their
2894 floating point arguments on the stack.
2896 @item force_align_arg_pointer
2897 @cindex @code{force_align_arg_pointer} attribute
2898 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2899 applied to individual function definitions, generating an alternate
2900 prologue and epilogue that realigns the runtime stack if necessary.
2901 This supports mixing legacy codes that run with a 4-byte aligned stack
2902 with modern codes that keep a 16-byte stack for SSE compatibility.
2905 @cindex @code{resbank} attribute
2906 On the SH2A target, this attribute enables the high-speed register
2907 saving and restoration using a register bank for @code{interrupt_handler}
2908 routines. Saving to the bank is performed automatically after the CPU
2909 accepts an interrupt that uses a register bank.
2911 The nineteen 32-bit registers comprising general register R0 to R14,
2912 control register GBR, and system registers MACH, MACL, and PR and the
2913 vector table address offset are saved into a register bank. Register
2914 banks are stacked in first-in last-out (FILO) sequence. Restoration
2915 from the bank is executed by issuing a RESBANK instruction.
2918 @cindex @code{returns_twice} attribute
2919 The @code{returns_twice} attribute tells the compiler that a function may
2920 return more than one time. The compiler will ensure that all registers
2921 are dead before calling such a function and will emit a warning about
2922 the variables that may be clobbered after the second return from the
2923 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2924 The @code{longjmp}-like counterpart of such function, if any, might need
2925 to be marked with the @code{noreturn} attribute.
2928 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2929 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2930 all registers except the stack pointer should be saved in the prologue
2931 regardless of whether they are used or not.
2933 @item section ("@var{section-name}")
2934 @cindex @code{section} function attribute
2935 Normally, the compiler places the code it generates in the @code{text} section.
2936 Sometimes, however, you need additional sections, or you need certain
2937 particular functions to appear in special sections. The @code{section}
2938 attribute specifies that a function lives in a particular section.
2939 For example, the declaration:
2942 extern void foobar (void) __attribute__ ((section ("bar")));
2946 puts the function @code{foobar} in the @code{bar} section.
2948 Some file formats do not support arbitrary sections so the @code{section}
2949 attribute is not available on all platforms.
2950 If you need to map the entire contents of a module to a particular
2951 section, consider using the facilities of the linker instead.
2954 @cindex @code{sentinel} function attribute
2955 This function attribute ensures that a parameter in a function call is
2956 an explicit @code{NULL}. The attribute is only valid on variadic
2957 functions. By default, the sentinel is located at position zero, the
2958 last parameter of the function call. If an optional integer position
2959 argument P is supplied to the attribute, the sentinel must be located at
2960 position P counting backwards from the end of the argument list.
2963 __attribute__ ((sentinel))
2965 __attribute__ ((sentinel(0)))
2968 The attribute is automatically set with a position of 0 for the built-in
2969 functions @code{execl} and @code{execlp}. The built-in function
2970 @code{execle} has the attribute set with a position of 1.
2972 A valid @code{NULL} in this context is defined as zero with any pointer
2973 type. If your system defines the @code{NULL} macro with an integer type
2974 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2975 with a copy that redefines NULL appropriately.
2977 The warnings for missing or incorrect sentinels are enabled with
2981 See long_call/short_call.
2984 See longcall/shortcall.
2987 @cindex signal handler functions on the AVR processors
2988 Use this attribute on the AVR to indicate that the specified
2989 function is a signal handler. The compiler will generate function
2990 entry and exit sequences suitable for use in a signal handler when this
2991 attribute is present. Interrupts will be disabled inside the function.
2994 Use this attribute on the SH to indicate an @code{interrupt_handler}
2995 function should switch to an alternate stack. It expects a string
2996 argument that names a global variable holding the address of the
3001 void f () __attribute__ ((interrupt_handler,
3002 sp_switch ("alt_stack")));
3006 @cindex functions that pop the argument stack on the 386
3007 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3008 assume that the called function will pop off the stack space used to
3009 pass arguments, unless it takes a variable number of arguments.
3011 @item syscall_linkage
3012 @cindex @code{syscall_linkage} attribute
3013 This attribute is used to modify the IA64 calling convention by marking
3014 all input registers as live at all function exits. This makes it possible
3015 to restart a system call after an interrupt without having to save/restore
3016 the input registers. This also prevents kernel data from leaking into
3020 @cindex @code{target} function attribute
3021 The @code{target} attribute is used to specify that a function is to
3022 be compiled with different target options than specified on the
3023 command line. This can be used for instance to have functions
3024 compiled with a different ISA (instruction set architecture) than the
3025 default. You can also use the @samp{#pragma GCC target} pragma to set
3026 more than one function to be compiled with specific target options.
3027 @xref{Function Specific Option Pragmas}, for details about the
3028 @samp{#pragma GCC target} pragma.
3030 For instance on a 386, you could compile one function with
3031 @code{target("sse4.1,arch=core2")} and another with
3032 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
3033 compiling the first function with @option{-msse4.1} and
3034 @option{-march=core2} options, and the second function with
3035 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
3036 user to make sure that a function is only invoked on a machine that
3037 supports the particular ISA it was compiled for (for example by using
3038 @code{cpuid} on 386 to determine what feature bits and architecture
3042 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
3043 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
3046 On the 386, the following options are allowed:
3051 @cindex @code{target("abm")} attribute
3052 Enable/disable the generation of the advanced bit instructions.
3056 @cindex @code{target("aes")} attribute
3057 Enable/disable the generation of the AES instructions.
3061 @cindex @code{target("mmx")} attribute
3062 Enable/disable the generation of the MMX instructions.
3066 @cindex @code{target("pclmul")} attribute
3067 Enable/disable the generation of the PCLMUL instructions.
3071 @cindex @code{target("popcnt")} attribute
3072 Enable/disable the generation of the POPCNT instruction.
3076 @cindex @code{target("sse")} attribute
3077 Enable/disable the generation of the SSE instructions.
3081 @cindex @code{target("sse2")} attribute
3082 Enable/disable the generation of the SSE2 instructions.
3086 @cindex @code{target("sse3")} attribute
3087 Enable/disable the generation of the SSE3 instructions.
3091 @cindex @code{target("sse4")} attribute
3092 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3097 @cindex @code{target("sse4.1")} attribute
3098 Enable/disable the generation of the sse4.1 instructions.
3102 @cindex @code{target("sse4.2")} attribute
3103 Enable/disable the generation of the sse4.2 instructions.
3107 @cindex @code{target("sse4a")} attribute
3108 Enable/disable the generation of the SSE4A instructions.
3112 @cindex @code{target("sse5")} attribute
3113 Enable/disable the generation of the SSE5 instructions.
3117 @cindex @code{target("ssse3")} attribute
3118 Enable/disable the generation of the SSSE3 instructions.
3122 @cindex @code{target("cld")} attribute
3123 Enable/disable the generation of the CLD before string moves.
3125 @item fancy-math-387
3126 @itemx no-fancy-math-387
3127 @cindex @code{target("fancy-math-387")} attribute
3128 Enable/disable the generation of the @code{sin}, @code{cos}, and
3129 @code{sqrt} instructions on the 387 floating point unit.
3132 @itemx no-fused-madd
3133 @cindex @code{target("fused-madd")} attribute
3134 Enable/disable the generation of the fused multiply/add instructions.
3138 @cindex @code{target("ieee-fp")} attribute
3139 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3141 @item inline-all-stringops
3142 @itemx no-inline-all-stringops
3143 @cindex @code{target("inline-all-stringops")} attribute
3144 Enable/disable inlining of string operations.
3146 @item inline-stringops-dynamically
3147 @itemx no-inline-stringops-dynamically
3148 @cindex @code{target("inline-stringops-dynamically")} attribute
3149 Enable/disable the generation of the inline code to do small string
3150 operations and calling the library routines for large operations.
3152 @item align-stringops
3153 @itemx no-align-stringops
3154 @cindex @code{target("align-stringops")} attribute
3155 Do/do not align destination of inlined string operations.
3159 @cindex @code{target("recip")} attribute
3160 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3161 instructions followed an additional Newton-Raphson step instead of
3162 doing a floating point division.
3164 @item arch=@var{ARCH}
3165 @cindex @code{target("arch=@var{ARCH}")} attribute
3166 Specify the architecture to generate code for in compiling the function.
3168 @item tune=@var{TUNE}
3169 @cindex @code{target("tune=@var{TUNE}")} attribute
3170 Specify the architecture to tune for in compiling the function.
3172 @item fpmath=@var{FPMATH}
3173 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3174 Specify which floating point unit to use. The
3175 @code{target("fpmath=sse,387")} option must be specified as
3176 @code{target("fpmath=sse+387")} because the comma would separate
3180 On the 386, you can use either multiple strings to specify multiple
3181 options, or you can separate the option with a comma (@code{,}).
3183 On the 386, the inliner will not inline a function that has different
3184 target options than the caller, unless the callee has a subset of the
3185 target options of the caller. For example a function declared with
3186 @code{target("sse5")} can inline a function with
3187 @code{target("sse2")}, since @code{-msse5} implies @code{-msse2}.
3189 The @code{target} attribute is not implemented in GCC versions earlier
3190 than 4.4, and at present only the 386 uses it.
3193 @cindex tiny data section on the H8/300H and H8S
3194 Use this attribute on the H8/300H and H8S to indicate that the specified
3195 variable should be placed into the tiny data section.
3196 The compiler will generate more efficient code for loads and stores
3197 on data in the tiny data section. Note the tiny data area is limited to
3198 slightly under 32kbytes of data.
3201 Use this attribute on the SH for an @code{interrupt_handler} to return using
3202 @code{trapa} instead of @code{rte}. This attribute expects an integer
3203 argument specifying the trap number to be used.
3206 @cindex @code{unused} attribute.
3207 This attribute, attached to a function, means that the function is meant
3208 to be possibly unused. GCC will not produce a warning for this
3212 @cindex @code{used} attribute.
3213 This attribute, attached to a function, means that code must be emitted
3214 for the function even if it appears that the function is not referenced.
3215 This is useful, for example, when the function is referenced only in
3219 @cindex @code{version_id} attribute
3220 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3221 symbol to contain a version string, thus allowing for function level
3222 versioning. HP-UX system header files may use version level functioning
3223 for some system calls.
3226 extern int foo () __attribute__((version_id ("20040821")));
3229 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3231 @item visibility ("@var{visibility_type}")
3232 @cindex @code{visibility} attribute
3233 This attribute affects the linkage of the declaration to which it is attached.
3234 There are four supported @var{visibility_type} values: default,
3235 hidden, protected or internal visibility.
3238 void __attribute__ ((visibility ("protected")))
3239 f () @{ /* @r{Do something.} */; @}
3240 int i __attribute__ ((visibility ("hidden")));
3243 The possible values of @var{visibility_type} correspond to the
3244 visibility settings in the ELF gABI.
3247 @c keep this list of visibilities in alphabetical order.
3250 Default visibility is the normal case for the object file format.
3251 This value is available for the visibility attribute to override other
3252 options that may change the assumed visibility of entities.
3254 On ELF, default visibility means that the declaration is visible to other
3255 modules and, in shared libraries, means that the declared entity may be
3258 On Darwin, default visibility means that the declaration is visible to
3261 Default visibility corresponds to ``external linkage'' in the language.
3264 Hidden visibility indicates that the entity declared will have a new
3265 form of linkage, which we'll call ``hidden linkage''. Two
3266 declarations of an object with hidden linkage refer to the same object
3267 if they are in the same shared object.
3270 Internal visibility is like hidden visibility, but with additional
3271 processor specific semantics. Unless otherwise specified by the
3272 psABI, GCC defines internal visibility to mean that a function is
3273 @emph{never} called from another module. Compare this with hidden
3274 functions which, while they cannot be referenced directly by other
3275 modules, can be referenced indirectly via function pointers. By
3276 indicating that a function cannot be called from outside the module,
3277 GCC may for instance omit the load of a PIC register since it is known
3278 that the calling function loaded the correct value.
3281 Protected visibility is like default visibility except that it
3282 indicates that references within the defining module will bind to the
3283 definition in that module. That is, the declared entity cannot be
3284 overridden by another module.
3288 All visibilities are supported on many, but not all, ELF targets
3289 (supported when the assembler supports the @samp{.visibility}
3290 pseudo-op). Default visibility is supported everywhere. Hidden
3291 visibility is supported on Darwin targets.
3293 The visibility attribute should be applied only to declarations which
3294 would otherwise have external linkage. The attribute should be applied
3295 consistently, so that the same entity should not be declared with
3296 different settings of the attribute.
3298 In C++, the visibility attribute applies to types as well as functions
3299 and objects, because in C++ types have linkage. A class must not have
3300 greater visibility than its non-static data member types and bases,
3301 and class members default to the visibility of their class. Also, a
3302 declaration without explicit visibility is limited to the visibility
3305 In C++, you can mark member functions and static member variables of a
3306 class with the visibility attribute. This is useful if you know a
3307 particular method or static member variable should only be used from
3308 one shared object; then you can mark it hidden while the rest of the
3309 class has default visibility. Care must be taken to avoid breaking
3310 the One Definition Rule; for example, it is usually not useful to mark
3311 an inline method as hidden without marking the whole class as hidden.
3313 A C++ namespace declaration can also have the visibility attribute.
3314 This attribute applies only to the particular namespace body, not to
3315 other definitions of the same namespace; it is equivalent to using
3316 @samp{#pragma GCC visibility} before and after the namespace
3317 definition (@pxref{Visibility Pragmas}).
3319 In C++, if a template argument has limited visibility, this
3320 restriction is implicitly propagated to the template instantiation.
3321 Otherwise, template instantiations and specializations default to the
3322 visibility of their template.
3324 If both the template and enclosing class have explicit visibility, the
3325 visibility from the template is used.
3327 @item warn_unused_result
3328 @cindex @code{warn_unused_result} attribute
3329 The @code{warn_unused_result} attribute causes a warning to be emitted
3330 if a caller of the function with this attribute does not use its
3331 return value. This is useful for functions where not checking
3332 the result is either a security problem or always a bug, such as
3336 int fn () __attribute__ ((warn_unused_result));
3339 if (fn () < 0) return -1;
3345 results in warning on line 5.
3348 @cindex @code{weak} attribute
3349 The @code{weak} attribute causes the declaration to be emitted as a weak
3350 symbol rather than a global. This is primarily useful in defining
3351 library functions which can be overridden in user code, though it can
3352 also be used with non-function declarations. Weak symbols are supported
3353 for ELF targets, and also for a.out targets when using the GNU assembler
3357 @itemx weakref ("@var{target}")
3358 @cindex @code{weakref} attribute
3359 The @code{weakref} attribute marks a declaration as a weak reference.
3360 Without arguments, it should be accompanied by an @code{alias} attribute
3361 naming the target symbol. Optionally, the @var{target} may be given as
3362 an argument to @code{weakref} itself. In either case, @code{weakref}
3363 implicitly marks the declaration as @code{weak}. Without a
3364 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3365 @code{weakref} is equivalent to @code{weak}.
3368 static int x() __attribute__ ((weakref ("y")));
3369 /* is equivalent to... */
3370 static int x() __attribute__ ((weak, weakref, alias ("y")));
3372 static int x() __attribute__ ((weakref));
3373 static int x() __attribute__ ((alias ("y")));
3376 A weak reference is an alias that does not by itself require a
3377 definition to be given for the target symbol. If the target symbol is
3378 only referenced through weak references, then the becomes a @code{weak}
3379 undefined symbol. If it is directly referenced, however, then such
3380 strong references prevail, and a definition will be required for the
3381 symbol, not necessarily in the same translation unit.
3383 The effect is equivalent to moving all references to the alias to a
3384 separate translation unit, renaming the alias to the aliased symbol,
3385 declaring it as weak, compiling the two separate translation units and
3386 performing a reloadable link on them.
3388 At present, a declaration to which @code{weakref} is attached can
3389 only be @code{static}.
3393 You can specify multiple attributes in a declaration by separating them
3394 by commas within the double parentheses or by immediately following an
3395 attribute declaration with another attribute declaration.
3397 @cindex @code{#pragma}, reason for not using
3398 @cindex pragma, reason for not using
3399 Some people object to the @code{__attribute__} feature, suggesting that
3400 ISO C's @code{#pragma} should be used instead. At the time
3401 @code{__attribute__} was designed, there were two reasons for not doing
3406 It is impossible to generate @code{#pragma} commands from a macro.
3409 There is no telling what the same @code{#pragma} might mean in another
3413 These two reasons applied to almost any application that might have been
3414 proposed for @code{#pragma}. It was basically a mistake to use
3415 @code{#pragma} for @emph{anything}.
3417 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3418 to be generated from macros. In addition, a @code{#pragma GCC}
3419 namespace is now in use for GCC-specific pragmas. However, it has been
3420 found convenient to use @code{__attribute__} to achieve a natural
3421 attachment of attributes to their corresponding declarations, whereas
3422 @code{#pragma GCC} is of use for constructs that do not naturally form
3423 part of the grammar. @xref{Other Directives,,Miscellaneous
3424 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3426 @node Attribute Syntax
3427 @section Attribute Syntax
3428 @cindex attribute syntax
3430 This section describes the syntax with which @code{__attribute__} may be
3431 used, and the constructs to which attribute specifiers bind, for the C
3432 language. Some details may vary for C++ and Objective-C@. Because of
3433 infelicities in the grammar for attributes, some forms described here
3434 may not be successfully parsed in all cases.
3436 There are some problems with the semantics of attributes in C++. For
3437 example, there are no manglings for attributes, although they may affect
3438 code generation, so problems may arise when attributed types are used in
3439 conjunction with templates or overloading. Similarly, @code{typeid}
3440 does not distinguish between types with different attributes. Support
3441 for attributes in C++ may be restricted in future to attributes on
3442 declarations only, but not on nested declarators.
3444 @xref{Function Attributes}, for details of the semantics of attributes
3445 applying to functions. @xref{Variable Attributes}, for details of the
3446 semantics of attributes applying to variables. @xref{Type Attributes},
3447 for details of the semantics of attributes applying to structure, union
3448 and enumerated types.
3450 An @dfn{attribute specifier} is of the form
3451 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3452 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3453 each attribute is one of the following:
3457 Empty. Empty attributes are ignored.
3460 A word (which may be an identifier such as @code{unused}, or a reserved
3461 word such as @code{const}).
3464 A word, followed by, in parentheses, parameters for the attribute.
3465 These parameters take one of the following forms:
3469 An identifier. For example, @code{mode} attributes use this form.
3472 An identifier followed by a comma and a non-empty comma-separated list
3473 of expressions. For example, @code{format} attributes use this form.
3476 A possibly empty comma-separated list of expressions. For example,
3477 @code{format_arg} attributes use this form with the list being a single
3478 integer constant expression, and @code{alias} attributes use this form
3479 with the list being a single string constant.
3483 An @dfn{attribute specifier list} is a sequence of one or more attribute
3484 specifiers, not separated by any other tokens.
3486 In GNU C, an attribute specifier list may appear after the colon following a
3487 label, other than a @code{case} or @code{default} label. The only
3488 attribute it makes sense to use after a label is @code{unused}. This
3489 feature is intended for code generated by programs which contains labels
3490 that may be unused but which is compiled with @option{-Wall}. It would
3491 not normally be appropriate to use in it human-written code, though it
3492 could be useful in cases where the code that jumps to the label is
3493 contained within an @code{#ifdef} conditional. GNU C++ only permits
3494 attributes on labels if the attribute specifier is immediately
3495 followed by a semicolon (i.e., the label applies to an empty
3496 statement). If the semicolon is missing, C++ label attributes are
3497 ambiguous, as it is permissible for a declaration, which could begin
3498 with an attribute list, to be labelled in C++. Declarations cannot be
3499 labelled in C90 or C99, so the ambiguity does not arise there.
3501 An attribute specifier list may appear as part of a @code{struct},
3502 @code{union} or @code{enum} specifier. It may go either immediately
3503 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3504 the closing brace. The former syntax is preferred.
3505 Where attribute specifiers follow the closing brace, they are considered
3506 to relate to the structure, union or enumerated type defined, not to any
3507 enclosing declaration the type specifier appears in, and the type
3508 defined is not complete until after the attribute specifiers.
3509 @c Otherwise, there would be the following problems: a shift/reduce
3510 @c conflict between attributes binding the struct/union/enum and
3511 @c binding to the list of specifiers/qualifiers; and "aligned"
3512 @c attributes could use sizeof for the structure, but the size could be
3513 @c changed later by "packed" attributes.
3515 Otherwise, an attribute specifier appears as part of a declaration,
3516 counting declarations of unnamed parameters and type names, and relates
3517 to that declaration (which may be nested in another declaration, for
3518 example in the case of a parameter declaration), or to a particular declarator
3519 within a declaration. Where an
3520 attribute specifier is applied to a parameter declared as a function or
3521 an array, it should apply to the function or array rather than the
3522 pointer to which the parameter is implicitly converted, but this is not
3523 yet correctly implemented.
3525 Any list of specifiers and qualifiers at the start of a declaration may
3526 contain attribute specifiers, whether or not such a list may in that
3527 context contain storage class specifiers. (Some attributes, however,
3528 are essentially in the nature of storage class specifiers, and only make
3529 sense where storage class specifiers may be used; for example,
3530 @code{section}.) There is one necessary limitation to this syntax: the
3531 first old-style parameter declaration in a function definition cannot
3532 begin with an attribute specifier, because such an attribute applies to
3533 the function instead by syntax described below (which, however, is not
3534 yet implemented in this case). In some other cases, attribute
3535 specifiers are permitted by this grammar but not yet supported by the
3536 compiler. All attribute specifiers in this place relate to the
3537 declaration as a whole. In the obsolescent usage where a type of
3538 @code{int} is implied by the absence of type specifiers, such a list of
3539 specifiers and qualifiers may be an attribute specifier list with no
3540 other specifiers or qualifiers.
3542 At present, the first parameter in a function prototype must have some
3543 type specifier which is not an attribute specifier; this resolves an
3544 ambiguity in the interpretation of @code{void f(int
3545 (__attribute__((foo)) x))}, but is subject to change. At present, if
3546 the parentheses of a function declarator contain only attributes then
3547 those attributes are ignored, rather than yielding an error or warning
3548 or implying a single parameter of type int, but this is subject to
3551 An attribute specifier list may appear immediately before a declarator
3552 (other than the first) in a comma-separated list of declarators in a
3553 declaration of more than one identifier using a single list of
3554 specifiers and qualifiers. Such attribute specifiers apply
3555 only to the identifier before whose declarator they appear. For
3559 __attribute__((noreturn)) void d0 (void),
3560 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3565 the @code{noreturn} attribute applies to all the functions
3566 declared; the @code{format} attribute only applies to @code{d1}.
3568 An attribute specifier list may appear immediately before the comma,
3569 @code{=} or semicolon terminating the declaration of an identifier other
3570 than a function definition. Such attribute specifiers apply
3571 to the declared object or function. Where an
3572 assembler name for an object or function is specified (@pxref{Asm
3573 Labels}), the attribute must follow the @code{asm}
3576 An attribute specifier list may, in future, be permitted to appear after
3577 the declarator in a function definition (before any old-style parameter
3578 declarations or the function body).
3580 Attribute specifiers may be mixed with type qualifiers appearing inside
3581 the @code{[]} of a parameter array declarator, in the C99 construct by
3582 which such qualifiers are applied to the pointer to which the array is
3583 implicitly converted. Such attribute specifiers apply to the pointer,
3584 not to the array, but at present this is not implemented and they are
3587 An attribute specifier list may appear at the start of a nested
3588 declarator. At present, there are some limitations in this usage: the
3589 attributes correctly apply to the declarator, but for most individual
3590 attributes the semantics this implies are not implemented.
3591 When attribute specifiers follow the @code{*} of a pointer
3592 declarator, they may be mixed with any type qualifiers present.
3593 The following describes the formal semantics of this syntax. It will make the
3594 most sense if you are familiar with the formal specification of
3595 declarators in the ISO C standard.
3597 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3598 D1}, where @code{T} contains declaration specifiers that specify a type
3599 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3600 contains an identifier @var{ident}. The type specified for @var{ident}
3601 for derived declarators whose type does not include an attribute
3602 specifier is as in the ISO C standard.
3604 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3605 and the declaration @code{T D} specifies the type
3606 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3607 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3608 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3610 If @code{D1} has the form @code{*
3611 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3612 declaration @code{T D} specifies the type
3613 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3614 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3615 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3621 void (__attribute__((noreturn)) ****f) (void);
3625 specifies the type ``pointer to pointer to pointer to pointer to
3626 non-returning function returning @code{void}''. As another example,
3629 char *__attribute__((aligned(8))) *f;
3633 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3634 Note again that this does not work with most attributes; for example,
3635 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3636 is not yet supported.
3638 For compatibility with existing code written for compiler versions that
3639 did not implement attributes on nested declarators, some laxity is
3640 allowed in the placing of attributes. If an attribute that only applies
3641 to types is applied to a declaration, it will be treated as applying to
3642 the type of that declaration. If an attribute that only applies to
3643 declarations is applied to the type of a declaration, it will be treated
3644 as applying to that declaration; and, for compatibility with code
3645 placing the attributes immediately before the identifier declared, such
3646 an attribute applied to a function return type will be treated as
3647 applying to the function type, and such an attribute applied to an array
3648 element type will be treated as applying to the array type. If an
3649 attribute that only applies to function types is applied to a
3650 pointer-to-function type, it will be treated as applying to the pointer
3651 target type; if such an attribute is applied to a function return type
3652 that is not a pointer-to-function type, it will be treated as applying
3653 to the function type.
3655 @node Function Prototypes
3656 @section Prototypes and Old-Style Function Definitions
3657 @cindex function prototype declarations
3658 @cindex old-style function definitions
3659 @cindex promotion of formal parameters
3661 GNU C extends ISO C to allow a function prototype to override a later
3662 old-style non-prototype definition. Consider the following example:
3665 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3672 /* @r{Prototype function declaration.} */
3673 int isroot P((uid_t));
3675 /* @r{Old-style function definition.} */
3677 isroot (x) /* @r{??? lossage here ???} */
3684 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3685 not allow this example, because subword arguments in old-style
3686 non-prototype definitions are promoted. Therefore in this example the
3687 function definition's argument is really an @code{int}, which does not
3688 match the prototype argument type of @code{short}.
3690 This restriction of ISO C makes it hard to write code that is portable
3691 to traditional C compilers, because the programmer does not know
3692 whether the @code{uid_t} type is @code{short}, @code{int}, or
3693 @code{long}. Therefore, in cases like these GNU C allows a prototype
3694 to override a later old-style definition. More precisely, in GNU C, a
3695 function prototype argument type overrides the argument type specified
3696 by a later old-style definition if the former type is the same as the
3697 latter type before promotion. Thus in GNU C the above example is
3698 equivalent to the following:
3711 GNU C++ does not support old-style function definitions, so this
3712 extension is irrelevant.
3715 @section C++ Style Comments
3717 @cindex C++ comments
3718 @cindex comments, C++ style
3720 In GNU C, you may use C++ style comments, which start with @samp{//} and
3721 continue until the end of the line. Many other C implementations allow
3722 such comments, and they are included in the 1999 C standard. However,
3723 C++ style comments are not recognized if you specify an @option{-std}
3724 option specifying a version of ISO C before C99, or @option{-ansi}
3725 (equivalent to @option{-std=c89}).
3728 @section Dollar Signs in Identifier Names
3730 @cindex dollar signs in identifier names
3731 @cindex identifier names, dollar signs in
3733 In GNU C, you may normally use dollar signs in identifier names.
3734 This is because many traditional C implementations allow such identifiers.
3735 However, dollar signs in identifiers are not supported on a few target
3736 machines, typically because the target assembler does not allow them.
3738 @node Character Escapes
3739 @section The Character @key{ESC} in Constants
3741 You can use the sequence @samp{\e} in a string or character constant to
3742 stand for the ASCII character @key{ESC}.
3745 @section Inquiring on Alignment of Types or Variables
3747 @cindex type alignment
3748 @cindex variable alignment
3750 The keyword @code{__alignof__} allows you to inquire about how an object
3751 is aligned, or the minimum alignment usually required by a type. Its
3752 syntax is just like @code{sizeof}.
3754 For example, if the target machine requires a @code{double} value to be
3755 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3756 This is true on many RISC machines. On more traditional machine
3757 designs, @code{__alignof__ (double)} is 4 or even 2.
3759 Some machines never actually require alignment; they allow reference to any
3760 data type even at an odd address. For these machines, @code{__alignof__}
3761 reports the smallest alignment that GCC will give the data type, usually as
3762 mandated by the target ABI.
3764 If the operand of @code{__alignof__} is an lvalue rather than a type,
3765 its value is the required alignment for its type, taking into account
3766 any minimum alignment specified with GCC's @code{__attribute__}
3767 extension (@pxref{Variable Attributes}). For example, after this
3771 struct foo @{ int x; char y; @} foo1;
3775 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3776 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3778 It is an error to ask for the alignment of an incomplete type.
3780 @node Variable Attributes
3781 @section Specifying Attributes of Variables
3782 @cindex attribute of variables
3783 @cindex variable attributes
3785 The keyword @code{__attribute__} allows you to specify special
3786 attributes of variables or structure fields. This keyword is followed
3787 by an attribute specification inside double parentheses. Some
3788 attributes are currently defined generically for variables.
3789 Other attributes are defined for variables on particular target
3790 systems. Other attributes are available for functions
3791 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3792 Other front ends might define more attributes
3793 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3795 You may also specify attributes with @samp{__} preceding and following
3796 each keyword. This allows you to use them in header files without
3797 being concerned about a possible macro of the same name. For example,
3798 you may use @code{__aligned__} instead of @code{aligned}.
3800 @xref{Attribute Syntax}, for details of the exact syntax for using
3804 @cindex @code{aligned} attribute
3805 @item aligned (@var{alignment})
3806 This attribute specifies a minimum alignment for the variable or
3807 structure field, measured in bytes. For example, the declaration:
3810 int x __attribute__ ((aligned (16))) = 0;
3814 causes the compiler to allocate the global variable @code{x} on a
3815 16-byte boundary. On a 68040, this could be used in conjunction with
3816 an @code{asm} expression to access the @code{move16} instruction which
3817 requires 16-byte aligned operands.
3819 You can also specify the alignment of structure fields. For example, to
3820 create a double-word aligned @code{int} pair, you could write:
3823 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3827 This is an alternative to creating a union with a @code{double} member
3828 that forces the union to be double-word aligned.
3830 As in the preceding examples, you can explicitly specify the alignment
3831 (in bytes) that you wish the compiler to use for a given variable or
3832 structure field. Alternatively, you can leave out the alignment factor
3833 and just ask the compiler to align a variable or field to the
3834 default alignment for the target architecture you are compiling for.
3835 The default alignment is sufficient for all scalar types, but may not be
3836 enough for all vector types on a target which supports vector operations.
3837 The default alignment is fixed for a particular target ABI.
3839 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
3840 which is the largest alignment ever used for any data type on the
3841 target machine you are compiling for. For example, you could write:
3844 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
3847 The compiler automatically sets the alignment for the declared
3848 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
3849 often make copy operations more efficient, because the compiler can
3850 use whatever instructions copy the biggest chunks of memory when
3851 performing copies to or from the variables or fields that you have
3852 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
3853 may change depending on command line options.
3855 When used on a struct, or struct member, the @code{aligned} attribute can
3856 only increase the alignment; in order to decrease it, the @code{packed}
3857 attribute must be specified as well. When used as part of a typedef, the
3858 @code{aligned} attribute can both increase and decrease alignment, and
3859 specifying the @code{packed} attribute will generate a warning.
3861 Note that the effectiveness of @code{aligned} attributes may be limited
3862 by inherent limitations in your linker. On many systems, the linker is
3863 only able to arrange for variables to be aligned up to a certain maximum
3864 alignment. (For some linkers, the maximum supported alignment may
3865 be very very small.) If your linker is only able to align variables
3866 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3867 in an @code{__attribute__} will still only provide you with 8 byte
3868 alignment. See your linker documentation for further information.
3870 The @code{aligned} attribute can also be used for functions
3871 (@pxref{Function Attributes}.)
3873 @item cleanup (@var{cleanup_function})
3874 @cindex @code{cleanup} attribute
3875 The @code{cleanup} attribute runs a function when the variable goes
3876 out of scope. This attribute can only be applied to auto function
3877 scope variables; it may not be applied to parameters or variables
3878 with static storage duration. The function must take one parameter,
3879 a pointer to a type compatible with the variable. The return value
3880 of the function (if any) is ignored.
3882 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3883 will be run during the stack unwinding that happens during the
3884 processing of the exception. Note that the @code{cleanup} attribute
3885 does not allow the exception to be caught, only to perform an action.
3886 It is undefined what happens if @var{cleanup_function} does not
3891 @cindex @code{common} attribute
3892 @cindex @code{nocommon} attribute
3895 The @code{common} attribute requests GCC to place a variable in
3896 ``common'' storage. The @code{nocommon} attribute requests the
3897 opposite---to allocate space for it directly.
3899 These attributes override the default chosen by the
3900 @option{-fno-common} and @option{-fcommon} flags respectively.
3903 @itemx deprecated (@var{msg})
3904 @cindex @code{deprecated} attribute
3905 The @code{deprecated} attribute results in a warning if the variable
3906 is used anywhere in the source file. This is useful when identifying
3907 variables that are expected to be removed in a future version of a
3908 program. The warning also includes the location of the declaration
3909 of the deprecated variable, to enable users to easily find further
3910 information about why the variable is deprecated, or what they should
3911 do instead. Note that the warning only occurs for uses:
3914 extern int old_var __attribute__ ((deprecated));
3916 int new_fn () @{ return old_var; @}
3919 results in a warning on line 3 but not line 2. The optional msg
3920 argument, which must be a string, will be printed in the warning if
3923 The @code{deprecated} attribute can also be used for functions and
3924 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3926 @item mode (@var{mode})
3927 @cindex @code{mode} attribute
3928 This attribute specifies the data type for the declaration---whichever
3929 type corresponds to the mode @var{mode}. This in effect lets you
3930 request an integer or floating point type according to its width.
3932 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3933 indicate the mode corresponding to a one-byte integer, @samp{word} or
3934 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3935 or @samp{__pointer__} for the mode used to represent pointers.
3938 @cindex @code{packed} attribute
3939 The @code{packed} attribute specifies that a variable or structure field
3940 should have the smallest possible alignment---one byte for a variable,
3941 and one bit for a field, unless you specify a larger value with the
3942 @code{aligned} attribute.
3944 Here is a structure in which the field @code{x} is packed, so that it
3945 immediately follows @code{a}:
3951 int x[2] __attribute__ ((packed));
3955 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
3956 @code{packed} attribute on bit-fields of type @code{char}. This has
3957 been fixed in GCC 4.4 but the change can lead to differences in the
3958 structure layout. See the documentation of
3959 @option{-Wpacked-bitfield-compat} for more information.
3961 @item section ("@var{section-name}")
3962 @cindex @code{section} variable attribute
3963 Normally, the compiler places the objects it generates in sections like
3964 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3965 or you need certain particular variables to appear in special sections,
3966 for example to map to special hardware. The @code{section}
3967 attribute specifies that a variable (or function) lives in a particular
3968 section. For example, this small program uses several specific section names:
3971 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3972 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3973 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3974 int init_data __attribute__ ((section ("INITDATA")));
3978 /* @r{Initialize stack pointer} */
3979 init_sp (stack + sizeof (stack));
3981 /* @r{Initialize initialized data} */
3982 memcpy (&init_data, &data, &edata - &data);
3984 /* @r{Turn on the serial ports} */
3991 Use the @code{section} attribute with
3992 @emph{global} variables and not @emph{local} variables,
3993 as shown in the example.
3995 You may use the @code{section} attribute with initialized or
3996 uninitialized global variables but the linker requires
3997 each object be defined once, with the exception that uninitialized
3998 variables tentatively go in the @code{common} (or @code{bss}) section
3999 and can be multiply ``defined''. Using the @code{section} attribute
4000 will change what section the variable goes into and may cause the
4001 linker to issue an error if an uninitialized variable has multiple
4002 definitions. You can force a variable to be initialized with the
4003 @option{-fno-common} flag or the @code{nocommon} attribute.
4005 Some file formats do not support arbitrary sections so the @code{section}
4006 attribute is not available on all platforms.
4007 If you need to map the entire contents of a module to a particular
4008 section, consider using the facilities of the linker instead.
4011 @cindex @code{shared} variable attribute
4012 On Microsoft Windows, in addition to putting variable definitions in a named
4013 section, the section can also be shared among all running copies of an
4014 executable or DLL@. For example, this small program defines shared data
4015 by putting it in a named section @code{shared} and marking the section
4019 int foo __attribute__((section ("shared"), shared)) = 0;
4024 /* @r{Read and write foo. All running
4025 copies see the same value.} */
4031 You may only use the @code{shared} attribute along with @code{section}
4032 attribute with a fully initialized global definition because of the way
4033 linkers work. See @code{section} attribute for more information.
4035 The @code{shared} attribute is only available on Microsoft Windows@.
4037 @item tls_model ("@var{tls_model}")
4038 @cindex @code{tls_model} attribute
4039 The @code{tls_model} attribute sets thread-local storage model
4040 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
4041 overriding @option{-ftls-model=} command line switch on a per-variable
4043 The @var{tls_model} argument should be one of @code{global-dynamic},
4044 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
4046 Not all targets support this attribute.
4049 This attribute, attached to a variable, means that the variable is meant
4050 to be possibly unused. GCC will not produce a warning for this
4054 This attribute, attached to a variable, means that the variable must be
4055 emitted even if it appears that the variable is not referenced.
4057 @item vector_size (@var{bytes})
4058 This attribute specifies the vector size for the variable, measured in
4059 bytes. For example, the declaration:
4062 int foo __attribute__ ((vector_size (16)));
4066 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
4067 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
4068 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
4070 This attribute is only applicable to integral and float scalars,
4071 although arrays, pointers, and function return values are allowed in
4072 conjunction with this construct.
4074 Aggregates with this attribute are invalid, even if they are of the same
4075 size as a corresponding scalar. For example, the declaration:
4078 struct S @{ int a; @};
4079 struct S __attribute__ ((vector_size (16))) foo;
4083 is invalid even if the size of the structure is the same as the size of
4087 The @code{selectany} attribute causes an initialized global variable to
4088 have link-once semantics. When multiple definitions of the variable are
4089 encountered by the linker, the first is selected and the remainder are
4090 discarded. Following usage by the Microsoft compiler, the linker is told
4091 @emph{not} to warn about size or content differences of the multiple
4094 Although the primary usage of this attribute is for POD types, the
4095 attribute can also be applied to global C++ objects that are initialized
4096 by a constructor. In this case, the static initialization and destruction
4097 code for the object is emitted in each translation defining the object,
4098 but the calls to the constructor and destructor are protected by a
4099 link-once guard variable.
4101 The @code{selectany} attribute is only available on Microsoft Windows
4102 targets. You can use @code{__declspec (selectany)} as a synonym for
4103 @code{__attribute__ ((selectany))} for compatibility with other
4107 The @code{weak} attribute is described in @ref{Function Attributes}.
4110 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4113 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4117 @subsection Blackfin Variable Attributes
4119 Three attributes are currently defined for the Blackfin.
4125 @cindex @code{l1_data} variable attribute
4126 @cindex @code{l1_data_A} variable attribute
4127 @cindex @code{l1_data_B} variable attribute
4128 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4129 Variables with @code{l1_data} attribute will be put into the specific section
4130 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4131 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4132 attribute will be put into the specific section named @code{.l1.data.B}.
4135 @subsection M32R/D Variable Attributes
4137 One attribute is currently defined for the M32R/D@.
4140 @item model (@var{model-name})
4141 @cindex variable addressability on the M32R/D
4142 Use this attribute on the M32R/D to set the addressability of an object.
4143 The identifier @var{model-name} is one of @code{small}, @code{medium},
4144 or @code{large}, representing each of the code models.
4146 Small model objects live in the lower 16MB of memory (so that their
4147 addresses can be loaded with the @code{ld24} instruction).
4149 Medium and large model objects may live anywhere in the 32-bit address space
4150 (the compiler will generate @code{seth/add3} instructions to load their
4154 @anchor{i386 Variable Attributes}
4155 @subsection i386 Variable Attributes
4157 Two attributes are currently defined for i386 configurations:
4158 @code{ms_struct} and @code{gcc_struct}
4163 @cindex @code{ms_struct} attribute
4164 @cindex @code{gcc_struct} attribute
4166 If @code{packed} is used on a structure, or if bit-fields are used
4167 it may be that the Microsoft ABI packs them differently
4168 than GCC would normally pack them. Particularly when moving packed
4169 data between functions compiled with GCC and the native Microsoft compiler
4170 (either via function call or as data in a file), it may be necessary to access
4173 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4174 compilers to match the native Microsoft compiler.
4176 The Microsoft structure layout algorithm is fairly simple with the exception
4177 of the bitfield packing:
4179 The padding and alignment of members of structures and whether a bit field
4180 can straddle a storage-unit boundary
4183 @item Structure members are stored sequentially in the order in which they are
4184 declared: the first member has the lowest memory address and the last member
4187 @item Every data object has an alignment-requirement. The alignment-requirement
4188 for all data except structures, unions, and arrays is either the size of the
4189 object or the current packing size (specified with either the aligned attribute
4190 or the pack pragma), whichever is less. For structures, unions, and arrays,
4191 the alignment-requirement is the largest alignment-requirement of its members.
4192 Every object is allocated an offset so that:
4194 offset % alignment-requirement == 0
4196 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4197 unit if the integral types are the same size and if the next bit field fits
4198 into the current allocation unit without crossing the boundary imposed by the
4199 common alignment requirements of the bit fields.
4202 Handling of zero-length bitfields:
4204 MSVC interprets zero-length bitfields in the following ways:
4207 @item If a zero-length bitfield is inserted between two bitfields that would
4208 normally be coalesced, the bitfields will not be coalesced.
4215 unsigned long bf_1 : 12;
4217 unsigned long bf_2 : 12;
4221 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4222 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4224 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4225 alignment of the zero-length bitfield is greater than the member that follows it,
4226 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4246 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4247 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4248 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4251 Taking this into account, it is important to note the following:
4254 @item If a zero-length bitfield follows a normal bitfield, the type of the
4255 zero-length bitfield may affect the alignment of the structure as whole. For
4256 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4257 normal bitfield, and is of type short.
4259 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4260 still affect the alignment of the structure:
4270 Here, @code{t4} will take up 4 bytes.
4273 @item Zero-length bitfields following non-bitfield members are ignored:
4284 Here, @code{t5} will take up 2 bytes.
4288 @subsection PowerPC Variable Attributes
4290 Three attributes currently are defined for PowerPC configurations:
4291 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4293 For full documentation of the struct attributes please see the
4294 documentation in @ref{i386 Variable Attributes}.
4296 For documentation of @code{altivec} attribute please see the
4297 documentation in @ref{PowerPC Type Attributes}.
4299 @subsection SPU Variable Attributes
4301 The SPU supports the @code{spu_vector} attribute for variables. For
4302 documentation of this attribute please see the documentation in
4303 @ref{SPU Type Attributes}.
4305 @subsection Xstormy16 Variable Attributes
4307 One attribute is currently defined for xstormy16 configurations:
4312 @cindex @code{below100} attribute
4314 If a variable has the @code{below100} attribute (@code{BELOW100} is
4315 allowed also), GCC will place the variable in the first 0x100 bytes of
4316 memory and use special opcodes to access it. Such variables will be
4317 placed in either the @code{.bss_below100} section or the
4318 @code{.data_below100} section.
4322 @subsection AVR Variable Attributes
4326 @cindex @code{progmem} variable attribute
4327 The @code{progmem} attribute is used on the AVR to place data in the Program
4328 Memory address space. The AVR is a Harvard Architecture processor and data
4329 normally resides in the Data Memory address space.
4332 @node Type Attributes
4333 @section Specifying Attributes of Types
4334 @cindex attribute of types
4335 @cindex type attributes
4337 The keyword @code{__attribute__} allows you to specify special
4338 attributes of @code{struct} and @code{union} types when you define
4339 such types. This keyword is followed by an attribute specification
4340 inside double parentheses. Seven attributes are currently defined for
4341 types: @code{aligned}, @code{packed}, @code{transparent_union},
4342 @code{unused}, @code{deprecated}, @code{visibility}, and
4343 @code{may_alias}. Other attributes are defined for functions
4344 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4347 You may also specify any one of these attributes with @samp{__}
4348 preceding and following its keyword. This allows you to use these
4349 attributes in header files without being concerned about a possible
4350 macro of the same name. For example, you may use @code{__aligned__}
4351 instead of @code{aligned}.
4353 You may specify type attributes in an enum, struct or union type
4354 declaration or definition, or for other types in a @code{typedef}
4357 For an enum, struct or union type, you may specify attributes either
4358 between the enum, struct or union tag and the name of the type, or
4359 just past the closing curly brace of the @emph{definition}. The
4360 former syntax is preferred.
4362 @xref{Attribute Syntax}, for details of the exact syntax for using
4366 @cindex @code{aligned} attribute
4367 @item aligned (@var{alignment})
4368 This attribute specifies a minimum alignment (in bytes) for variables
4369 of the specified type. For example, the declarations:
4372 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4373 typedef int more_aligned_int __attribute__ ((aligned (8)));
4377 force the compiler to insure (as far as it can) that each variable whose
4378 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4379 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4380 variables of type @code{struct S} aligned to 8-byte boundaries allows
4381 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4382 store) instructions when copying one variable of type @code{struct S} to
4383 another, thus improving run-time efficiency.
4385 Note that the alignment of any given @code{struct} or @code{union} type
4386 is required by the ISO C standard to be at least a perfect multiple of
4387 the lowest common multiple of the alignments of all of the members of
4388 the @code{struct} or @code{union} in question. This means that you @emph{can}
4389 effectively adjust the alignment of a @code{struct} or @code{union}
4390 type by attaching an @code{aligned} attribute to any one of the members
4391 of such a type, but the notation illustrated in the example above is a
4392 more obvious, intuitive, and readable way to request the compiler to
4393 adjust the alignment of an entire @code{struct} or @code{union} type.
4395 As in the preceding example, you can explicitly specify the alignment
4396 (in bytes) that you wish the compiler to use for a given @code{struct}
4397 or @code{union} type. Alternatively, you can leave out the alignment factor
4398 and just ask the compiler to align a type to the maximum
4399 useful alignment for the target machine you are compiling for. For
4400 example, you could write:
4403 struct S @{ short f[3]; @} __attribute__ ((aligned));
4406 Whenever you leave out the alignment factor in an @code{aligned}
4407 attribute specification, the compiler automatically sets the alignment
4408 for the type to the largest alignment which is ever used for any data
4409 type on the target machine you are compiling for. Doing this can often
4410 make copy operations more efficient, because the compiler can use
4411 whatever instructions copy the biggest chunks of memory when performing
4412 copies to or from the variables which have types that you have aligned
4415 In the example above, if the size of each @code{short} is 2 bytes, then
4416 the size of the entire @code{struct S} type is 6 bytes. The smallest
4417 power of two which is greater than or equal to that is 8, so the
4418 compiler sets the alignment for the entire @code{struct S} type to 8
4421 Note that although you can ask the compiler to select a time-efficient
4422 alignment for a given type and then declare only individual stand-alone
4423 objects of that type, the compiler's ability to select a time-efficient
4424 alignment is primarily useful only when you plan to create arrays of
4425 variables having the relevant (efficiently aligned) type. If you
4426 declare or use arrays of variables of an efficiently-aligned type, then
4427 it is likely that your program will also be doing pointer arithmetic (or
4428 subscripting, which amounts to the same thing) on pointers to the
4429 relevant type, and the code that the compiler generates for these
4430 pointer arithmetic operations will often be more efficient for
4431 efficiently-aligned types than for other types.
4433 The @code{aligned} attribute can only increase the alignment; but you
4434 can decrease it by specifying @code{packed} as well. See below.
4436 Note that the effectiveness of @code{aligned} attributes may be limited
4437 by inherent limitations in your linker. On many systems, the linker is
4438 only able to arrange for variables to be aligned up to a certain maximum
4439 alignment. (For some linkers, the maximum supported alignment may
4440 be very very small.) If your linker is only able to align variables
4441 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4442 in an @code{__attribute__} will still only provide you with 8 byte
4443 alignment. See your linker documentation for further information.
4446 This attribute, attached to @code{struct} or @code{union} type
4447 definition, specifies that each member (other than zero-width bitfields)
4448 of the structure or union is placed to minimize the memory required. When
4449 attached to an @code{enum} definition, it indicates that the smallest
4450 integral type should be used.
4452 @opindex fshort-enums
4453 Specifying this attribute for @code{struct} and @code{union} types is
4454 equivalent to specifying the @code{packed} attribute on each of the
4455 structure or union members. Specifying the @option{-fshort-enums}
4456 flag on the line is equivalent to specifying the @code{packed}
4457 attribute on all @code{enum} definitions.
4459 In the following example @code{struct my_packed_struct}'s members are
4460 packed closely together, but the internal layout of its @code{s} member
4461 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4465 struct my_unpacked_struct
4471 struct __attribute__ ((__packed__)) my_packed_struct
4475 struct my_unpacked_struct s;
4479 You may only specify this attribute on the definition of an @code{enum},
4480 @code{struct} or @code{union}, not on a @code{typedef} which does not
4481 also define the enumerated type, structure or union.
4483 @item transparent_union
4484 This attribute, attached to a @code{union} type definition, indicates
4485 that any function parameter having that union type causes calls to that
4486 function to be treated in a special way.
4488 First, the argument corresponding to a transparent union type can be of
4489 any type in the union; no cast is required. Also, if the union contains
4490 a pointer type, the corresponding argument can be a null pointer
4491 constant or a void pointer expression; and if the union contains a void
4492 pointer type, the corresponding argument can be any pointer expression.
4493 If the union member type is a pointer, qualifiers like @code{const} on
4494 the referenced type must be respected, just as with normal pointer
4497 Second, the argument is passed to the function using the calling
4498 conventions of the first member of the transparent union, not the calling
4499 conventions of the union itself. All members of the union must have the
4500 same machine representation; this is necessary for this argument passing
4503 Transparent unions are designed for library functions that have multiple
4504 interfaces for compatibility reasons. For example, suppose the
4505 @code{wait} function must accept either a value of type @code{int *} to
4506 comply with Posix, or a value of type @code{union wait *} to comply with
4507 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4508 @code{wait} would accept both kinds of arguments, but it would also
4509 accept any other pointer type and this would make argument type checking
4510 less useful. Instead, @code{<sys/wait.h>} might define the interface
4514 typedef union __attribute__ ((__transparent_union__))
4518 @} wait_status_ptr_t;
4520 pid_t wait (wait_status_ptr_t);
4523 This interface allows either @code{int *} or @code{union wait *}
4524 arguments to be passed, using the @code{int *} calling convention.
4525 The program can call @code{wait} with arguments of either type:
4528 int w1 () @{ int w; return wait (&w); @}
4529 int w2 () @{ union wait w; return wait (&w); @}
4532 With this interface, @code{wait}'s implementation might look like this:
4535 pid_t wait (wait_status_ptr_t p)
4537 return waitpid (-1, p.__ip, 0);
4542 When attached to a type (including a @code{union} or a @code{struct}),
4543 this attribute means that variables of that type are meant to appear
4544 possibly unused. GCC will not produce a warning for any variables of
4545 that type, even if the variable appears to do nothing. This is often
4546 the case with lock or thread classes, which are usually defined and then
4547 not referenced, but contain constructors and destructors that have
4548 nontrivial bookkeeping functions.
4551 @itemx deprecated (@var{msg})
4552 The @code{deprecated} attribute results in a warning if the type
4553 is used anywhere in the source file. This is useful when identifying
4554 types that are expected to be removed in a future version of a program.
4555 If possible, the warning also includes the location of the declaration
4556 of the deprecated type, to enable users to easily find further
4557 information about why the type is deprecated, or what they should do
4558 instead. Note that the warnings only occur for uses and then only
4559 if the type is being applied to an identifier that itself is not being
4560 declared as deprecated.
4563 typedef int T1 __attribute__ ((deprecated));
4567 typedef T1 T3 __attribute__ ((deprecated));
4568 T3 z __attribute__ ((deprecated));
4571 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4572 warning is issued for line 4 because T2 is not explicitly
4573 deprecated. Line 5 has no warning because T3 is explicitly
4574 deprecated. Similarly for line 6. The optional msg
4575 argument, which must be a string, will be printed in the warning if
4578 The @code{deprecated} attribute can also be used for functions and
4579 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4582 Accesses through pointers to types with this attribute are not subject
4583 to type-based alias analysis, but are instead assumed to be able to alias
4584 any other type of objects. In the context of 6.5/7 an lvalue expression
4585 dereferencing such a pointer is treated like having a character type.
4586 See @option{-fstrict-aliasing} for more information on aliasing issues.
4587 This extension exists to support some vector APIs, in which pointers to
4588 one vector type are permitted to alias pointers to a different vector type.
4590 Note that an object of a type with this attribute does not have any
4596 typedef short __attribute__((__may_alias__)) short_a;
4602 short_a *b = (short_a *) &a;
4606 if (a == 0x12345678)
4613 If you replaced @code{short_a} with @code{short} in the variable
4614 declaration, the above program would abort when compiled with
4615 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4616 above in recent GCC versions.
4619 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4620 applied to class, struct, union and enum types. Unlike other type
4621 attributes, the attribute must appear between the initial keyword and
4622 the name of the type; it cannot appear after the body of the type.
4624 Note that the type visibility is applied to vague linkage entities
4625 associated with the class (vtable, typeinfo node, etc.). In
4626 particular, if a class is thrown as an exception in one shared object
4627 and caught in another, the class must have default visibility.
4628 Otherwise the two shared objects will be unable to use the same
4629 typeinfo node and exception handling will break.
4633 @subsection ARM Type Attributes
4635 On those ARM targets that support @code{dllimport} (such as Symbian
4636 OS), you can use the @code{notshared} attribute to indicate that the
4637 virtual table and other similar data for a class should not be
4638 exported from a DLL@. For example:
4641 class __declspec(notshared) C @{
4643 __declspec(dllimport) C();
4647 __declspec(dllexport)
4651 In this code, @code{C::C} is exported from the current DLL, but the
4652 virtual table for @code{C} is not exported. (You can use
4653 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4654 most Symbian OS code uses @code{__declspec}.)
4656 @anchor{i386 Type Attributes}
4657 @subsection i386 Type Attributes
4659 Two attributes are currently defined for i386 configurations:
4660 @code{ms_struct} and @code{gcc_struct}.
4666 @cindex @code{ms_struct}
4667 @cindex @code{gcc_struct}
4669 If @code{packed} is used on a structure, or if bit-fields are used
4670 it may be that the Microsoft ABI packs them differently
4671 than GCC would normally pack them. Particularly when moving packed
4672 data between functions compiled with GCC and the native Microsoft compiler
4673 (either via function call or as data in a file), it may be necessary to access
4676 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4677 compilers to match the native Microsoft compiler.
4680 To specify multiple attributes, separate them by commas within the
4681 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4684 @anchor{PowerPC Type Attributes}
4685 @subsection PowerPC Type Attributes
4687 Three attributes currently are defined for PowerPC configurations:
4688 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4690 For full documentation of the @code{ms_struct} and @code{gcc_struct}
4691 attributes please see the documentation in @ref{i386 Type Attributes}.
4693 The @code{altivec} attribute allows one to declare AltiVec vector data
4694 types supported by the AltiVec Programming Interface Manual. The
4695 attribute requires an argument to specify one of three vector types:
4696 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4697 and @code{bool__} (always followed by unsigned).
4700 __attribute__((altivec(vector__)))
4701 __attribute__((altivec(pixel__))) unsigned short
4702 __attribute__((altivec(bool__))) unsigned
4705 These attributes mainly are intended to support the @code{__vector},
4706 @code{__pixel}, and @code{__bool} AltiVec keywords.
4708 @anchor{SPU Type Attributes}
4709 @subsection SPU Type Attributes
4711 The SPU supports the @code{spu_vector} attribute for types. This attribute
4712 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
4713 Language Extensions Specification. It is intended to support the
4714 @code{__vector} keyword.
4718 @section An Inline Function is As Fast As a Macro
4719 @cindex inline functions
4720 @cindex integrating function code
4722 @cindex macros, inline alternative
4724 By declaring a function inline, you can direct GCC to make
4725 calls to that function faster. One way GCC can achieve this is to
4726 integrate that function's code into the code for its callers. This
4727 makes execution faster by eliminating the function-call overhead; in
4728 addition, if any of the actual argument values are constant, their
4729 known values may permit simplifications at compile time so that not
4730 all of the inline function's code needs to be included. The effect on
4731 code size is less predictable; object code may be larger or smaller
4732 with function inlining, depending on the particular case. You can
4733 also direct GCC to try to integrate all ``simple enough'' functions
4734 into their callers with the option @option{-finline-functions}.
4736 GCC implements three different semantics of declaring a function
4737 inline. One is available with @option{-std=gnu89} or
4738 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
4739 on all inline declarations, another when @option{-std=c99} or
4740 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
4741 is used when compiling C++.
4743 To declare a function inline, use the @code{inline} keyword in its
4744 declaration, like this:
4754 If you are writing a header file to be included in ISO C89 programs, write
4755 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4757 The three types of inlining behave similarly in two important cases:
4758 when the @code{inline} keyword is used on a @code{static} function,
4759 like the example above, and when a function is first declared without
4760 using the @code{inline} keyword and then is defined with
4761 @code{inline}, like this:
4764 extern int inc (int *a);
4772 In both of these common cases, the program behaves the same as if you
4773 had not used the @code{inline} keyword, except for its speed.
4775 @cindex inline functions, omission of
4776 @opindex fkeep-inline-functions
4777 When a function is both inline and @code{static}, if all calls to the
4778 function are integrated into the caller, and the function's address is
4779 never used, then the function's own assembler code is never referenced.
4780 In this case, GCC does not actually output assembler code for the
4781 function, unless you specify the option @option{-fkeep-inline-functions}.
4782 Some calls cannot be integrated for various reasons (in particular,
4783 calls that precede the function's definition cannot be integrated, and
4784 neither can recursive calls within the definition). If there is a
4785 nonintegrated call, then the function is compiled to assembler code as
4786 usual. The function must also be compiled as usual if the program
4787 refers to its address, because that can't be inlined.
4790 Note that certain usages in a function definition can make it unsuitable
4791 for inline substitution. Among these usages are: use of varargs, use of
4792 alloca, use of variable sized data types (@pxref{Variable Length}),
4793 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4794 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4795 will warn when a function marked @code{inline} could not be substituted,
4796 and will give the reason for the failure.
4798 @cindex automatic @code{inline} for C++ member fns
4799 @cindex @code{inline} automatic for C++ member fns
4800 @cindex member fns, automatically @code{inline}
4801 @cindex C++ member fns, automatically @code{inline}
4802 @opindex fno-default-inline
4803 As required by ISO C++, GCC considers member functions defined within
4804 the body of a class to be marked inline even if they are
4805 not explicitly declared with the @code{inline} keyword. You can
4806 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4807 Options,,Options Controlling C++ Dialect}.
4809 GCC does not inline any functions when not optimizing unless you specify
4810 the @samp{always_inline} attribute for the function, like this:
4813 /* @r{Prototype.} */
4814 inline void foo (const char) __attribute__((always_inline));
4817 The remainder of this section is specific to GNU C89 inlining.
4819 @cindex non-static inline function
4820 When an inline function is not @code{static}, then the compiler must assume
4821 that there may be calls from other source files; since a global symbol can
4822 be defined only once in any program, the function must not be defined in
4823 the other source files, so the calls therein cannot be integrated.
4824 Therefore, a non-@code{static} inline function is always compiled on its
4825 own in the usual fashion.
4827 If you specify both @code{inline} and @code{extern} in the function
4828 definition, then the definition is used only for inlining. In no case
4829 is the function compiled on its own, not even if you refer to its
4830 address explicitly. Such an address becomes an external reference, as
4831 if you had only declared the function, and had not defined it.
4833 This combination of @code{inline} and @code{extern} has almost the
4834 effect of a macro. The way to use it is to put a function definition in
4835 a header file with these keywords, and put another copy of the
4836 definition (lacking @code{inline} and @code{extern}) in a library file.
4837 The definition in the header file will cause most calls to the function
4838 to be inlined. If any uses of the function remain, they will refer to
4839 the single copy in the library.
4842 @section Assembler Instructions with C Expression Operands
4843 @cindex extended @code{asm}
4844 @cindex @code{asm} expressions
4845 @cindex assembler instructions
4848 In an assembler instruction using @code{asm}, you can specify the
4849 operands of the instruction using C expressions. This means you need not
4850 guess which registers or memory locations will contain the data you want
4853 You must specify an assembler instruction template much like what
4854 appears in a machine description, plus an operand constraint string for
4857 For example, here is how to use the 68881's @code{fsinx} instruction:
4860 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4864 Here @code{angle} is the C expression for the input operand while
4865 @code{result} is that of the output operand. Each has @samp{"f"} as its
4866 operand constraint, saying that a floating point register is required.
4867 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4868 output operands' constraints must use @samp{=}. The constraints use the
4869 same language used in the machine description (@pxref{Constraints}).
4871 Each operand is described by an operand-constraint string followed by
4872 the C expression in parentheses. A colon separates the assembler
4873 template from the first output operand and another separates the last
4874 output operand from the first input, if any. Commas separate the
4875 operands within each group. The total number of operands is currently
4876 limited to 30; this limitation may be lifted in some future version of
4879 If there are no output operands but there are input operands, you must
4880 place two consecutive colons surrounding the place where the output
4883 As of GCC version 3.1, it is also possible to specify input and output
4884 operands using symbolic names which can be referenced within the
4885 assembler code. These names are specified inside square brackets
4886 preceding the constraint string, and can be referenced inside the
4887 assembler code using @code{%[@var{name}]} instead of a percentage sign
4888 followed by the operand number. Using named operands the above example
4892 asm ("fsinx %[angle],%[output]"
4893 : [output] "=f" (result)
4894 : [angle] "f" (angle));
4898 Note that the symbolic operand names have no relation whatsoever to
4899 other C identifiers. You may use any name you like, even those of
4900 existing C symbols, but you must ensure that no two operands within the same
4901 assembler construct use the same symbolic name.
4903 Output operand expressions must be lvalues; the compiler can check this.
4904 The input operands need not be lvalues. The compiler cannot check
4905 whether the operands have data types that are reasonable for the
4906 instruction being executed. It does not parse the assembler instruction
4907 template and does not know what it means or even whether it is valid
4908 assembler input. The extended @code{asm} feature is most often used for
4909 machine instructions the compiler itself does not know exist. If
4910 the output expression cannot be directly addressed (for example, it is a
4911 bit-field), your constraint must allow a register. In that case, GCC
4912 will use the register as the output of the @code{asm}, and then store
4913 that register into the output.
4915 The ordinary output operands must be write-only; GCC will assume that
4916 the values in these operands before the instruction are dead and need
4917 not be generated. Extended asm supports input-output or read-write
4918 operands. Use the constraint character @samp{+} to indicate such an
4919 operand and list it with the output operands. You should only use
4920 read-write operands when the constraints for the operand (or the
4921 operand in which only some of the bits are to be changed) allow a
4924 You may, as an alternative, logically split its function into two
4925 separate operands, one input operand and one write-only output
4926 operand. The connection between them is expressed by constraints
4927 which say they need to be in the same location when the instruction
4928 executes. You can use the same C expression for both operands, or
4929 different expressions. For example, here we write the (fictitious)
4930 @samp{combine} instruction with @code{bar} as its read-only source
4931 operand and @code{foo} as its read-write destination:
4934 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4938 The constraint @samp{"0"} for operand 1 says that it must occupy the
4939 same location as operand 0. A number in constraint is allowed only in
4940 an input operand and it must refer to an output operand.
4942 Only a number in the constraint can guarantee that one operand will be in
4943 the same place as another. The mere fact that @code{foo} is the value
4944 of both operands is not enough to guarantee that they will be in the
4945 same place in the generated assembler code. The following would not
4949 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4952 Various optimizations or reloading could cause operands 0 and 1 to be in
4953 different registers; GCC knows no reason not to do so. For example, the
4954 compiler might find a copy of the value of @code{foo} in one register and
4955 use it for operand 1, but generate the output operand 0 in a different
4956 register (copying it afterward to @code{foo}'s own address). Of course,
4957 since the register for operand 1 is not even mentioned in the assembler
4958 code, the result will not work, but GCC can't tell that.
4960 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4961 the operand number for a matching constraint. For example:
4964 asm ("cmoveq %1,%2,%[result]"
4965 : [result] "=r"(result)
4966 : "r" (test), "r"(new), "[result]"(old));
4969 Sometimes you need to make an @code{asm} operand be a specific register,
4970 but there's no matching constraint letter for that register @emph{by
4971 itself}. To force the operand into that register, use a local variable
4972 for the operand and specify the register in the variable declaration.
4973 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4974 register constraint letter that matches the register:
4977 register int *p1 asm ("r0") = @dots{};
4978 register int *p2 asm ("r1") = @dots{};
4979 register int *result asm ("r0");
4980 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4983 @anchor{Example of asm with clobbered asm reg}
4984 In the above example, beware that a register that is call-clobbered by
4985 the target ABI will be overwritten by any function call in the
4986 assignment, including library calls for arithmetic operators.
4987 Also a register may be clobbered when generating some operations,
4988 like variable shift, memory copy or memory move on x86.
4989 Assuming it is a call-clobbered register, this may happen to @code{r0}
4990 above by the assignment to @code{p2}. If you have to use such a
4991 register, use temporary variables for expressions between the register
4996 register int *p1 asm ("r0") = @dots{};
4997 register int *p2 asm ("r1") = t1;
4998 register int *result asm ("r0");
4999 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
5002 Some instructions clobber specific hard registers. To describe this,
5003 write a third colon after the input operands, followed by the names of
5004 the clobbered hard registers (given as strings). Here is a realistic
5005 example for the VAX:
5008 asm volatile ("movc3 %0,%1,%2"
5009 : /* @r{no outputs} */
5010 : "g" (from), "g" (to), "g" (count)
5011 : "r0", "r1", "r2", "r3", "r4", "r5");
5014 You may not write a clobber description in a way that overlaps with an
5015 input or output operand. For example, you may not have an operand
5016 describing a register class with one member if you mention that register
5017 in the clobber list. Variables declared to live in specific registers
5018 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
5019 have no part mentioned in the clobber description.
5020 There is no way for you to specify that an input
5021 operand is modified without also specifying it as an output
5022 operand. Note that if all the output operands you specify are for this
5023 purpose (and hence unused), you will then also need to specify
5024 @code{volatile} for the @code{asm} construct, as described below, to
5025 prevent GCC from deleting the @code{asm} statement as unused.
5027 If you refer to a particular hardware register from the assembler code,
5028 you will probably have to list the register after the third colon to
5029 tell the compiler the register's value is modified. In some assemblers,
5030 the register names begin with @samp{%}; to produce one @samp{%} in the
5031 assembler code, you must write @samp{%%} in the input.
5033 If your assembler instruction can alter the condition code register, add
5034 @samp{cc} to the list of clobbered registers. GCC on some machines
5035 represents the condition codes as a specific hardware register;
5036 @samp{cc} serves to name this register. On other machines, the
5037 condition code is handled differently, and specifying @samp{cc} has no
5038 effect. But it is valid no matter what the machine.
5040 If your assembler instructions access memory in an unpredictable
5041 fashion, add @samp{memory} to the list of clobbered registers. This
5042 will cause GCC to not keep memory values cached in registers across the
5043 assembler instruction and not optimize stores or loads to that memory.
5044 You will also want to add the @code{volatile} keyword if the memory
5045 affected is not listed in the inputs or outputs of the @code{asm}, as
5046 the @samp{memory} clobber does not count as a side-effect of the
5047 @code{asm}. If you know how large the accessed memory is, you can add
5048 it as input or output but if this is not known, you should add
5049 @samp{memory}. As an example, if you access ten bytes of a string, you
5050 can use a memory input like:
5053 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
5056 Note that in the following example the memory input is necessary,
5057 otherwise GCC might optimize the store to @code{x} away:
5064 asm ("magic stuff accessing an 'int' pointed to by '%1'"
5065 "=&d" (r) : "a" (y), "m" (*y));
5070 You can put multiple assembler instructions together in a single
5071 @code{asm} template, separated by the characters normally used in assembly
5072 code for the system. A combination that works in most places is a newline
5073 to break the line, plus a tab character to move to the instruction field
5074 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
5075 assembler allows semicolons as a line-breaking character. Note that some
5076 assembler dialects use semicolons to start a comment.
5077 The input operands are guaranteed not to use any of the clobbered
5078 registers, and neither will the output operands' addresses, so you can
5079 read and write the clobbered registers as many times as you like. Here
5080 is an example of multiple instructions in a template; it assumes the
5081 subroutine @code{_foo} accepts arguments in registers 9 and 10:
5084 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
5086 : "g" (from), "g" (to)
5090 Unless an output operand has the @samp{&} constraint modifier, GCC
5091 may allocate it in the same register as an unrelated input operand, on
5092 the assumption the inputs are consumed before the outputs are produced.
5093 This assumption may be false if the assembler code actually consists of
5094 more than one instruction. In such a case, use @samp{&} for each output
5095 operand that may not overlap an input. @xref{Modifiers}.
5097 If you want to test the condition code produced by an assembler
5098 instruction, you must include a branch and a label in the @code{asm}
5099 construct, as follows:
5102 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5108 This assumes your assembler supports local labels, as the GNU assembler
5109 and most Unix assemblers do.
5111 Speaking of labels, jumps from one @code{asm} to another are not
5112 supported. The compiler's optimizers do not know about these jumps, and
5113 therefore they cannot take account of them when deciding how to
5116 @cindex macros containing @code{asm}
5117 Usually the most convenient way to use these @code{asm} instructions is to
5118 encapsulate them in macros that look like functions. For example,
5122 (@{ double __value, __arg = (x); \
5123 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5128 Here the variable @code{__arg} is used to make sure that the instruction
5129 operates on a proper @code{double} value, and to accept only those
5130 arguments @code{x} which can convert automatically to a @code{double}.
5132 Another way to make sure the instruction operates on the correct data
5133 type is to use a cast in the @code{asm}. This is different from using a
5134 variable @code{__arg} in that it converts more different types. For
5135 example, if the desired type were @code{int}, casting the argument to
5136 @code{int} would accept a pointer with no complaint, while assigning the
5137 argument to an @code{int} variable named @code{__arg} would warn about
5138 using a pointer unless the caller explicitly casts it.
5140 If an @code{asm} has output operands, GCC assumes for optimization
5141 purposes the instruction has no side effects except to change the output
5142 operands. This does not mean instructions with a side effect cannot be
5143 used, but you must be careful, because the compiler may eliminate them
5144 if the output operands aren't used, or move them out of loops, or
5145 replace two with one if they constitute a common subexpression. Also,
5146 if your instruction does have a side effect on a variable that otherwise
5147 appears not to change, the old value of the variable may be reused later
5148 if it happens to be found in a register.
5150 You can prevent an @code{asm} instruction from being deleted
5151 by writing the keyword @code{volatile} after
5152 the @code{asm}. For example:
5155 #define get_and_set_priority(new) \
5157 asm volatile ("get_and_set_priority %0, %1" \
5158 : "=g" (__old) : "g" (new)); \
5163 The @code{volatile} keyword indicates that the instruction has
5164 important side-effects. GCC will not delete a volatile @code{asm} if
5165 it is reachable. (The instruction can still be deleted if GCC can
5166 prove that control-flow will never reach the location of the
5167 instruction.) Note that even a volatile @code{asm} instruction
5168 can be moved relative to other code, including across jump
5169 instructions. For example, on many targets there is a system
5170 register which can be set to control the rounding mode of
5171 floating point operations. You might try
5172 setting it with a volatile @code{asm}, like this PowerPC example:
5175 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5180 This will not work reliably, as the compiler may move the addition back
5181 before the volatile @code{asm}. To make it work you need to add an
5182 artificial dependency to the @code{asm} referencing a variable in the code
5183 you don't want moved, for example:
5186 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5190 Similarly, you can't expect a
5191 sequence of volatile @code{asm} instructions to remain perfectly
5192 consecutive. If you want consecutive output, use a single @code{asm}.
5193 Also, GCC will perform some optimizations across a volatile @code{asm}
5194 instruction; GCC does not ``forget everything'' when it encounters
5195 a volatile @code{asm} instruction the way some other compilers do.
5197 An @code{asm} instruction without any output operands will be treated
5198 identically to a volatile @code{asm} instruction.
5200 It is a natural idea to look for a way to give access to the condition
5201 code left by the assembler instruction. However, when we attempted to
5202 implement this, we found no way to make it work reliably. The problem
5203 is that output operands might need reloading, which would result in
5204 additional following ``store'' instructions. On most machines, these
5205 instructions would alter the condition code before there was time to
5206 test it. This problem doesn't arise for ordinary ``test'' and
5207 ``compare'' instructions because they don't have any output operands.
5209 For reasons similar to those described above, it is not possible to give
5210 an assembler instruction access to the condition code left by previous
5213 If you are writing a header file that should be includable in ISO C
5214 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5217 @subsection Size of an @code{asm}
5219 Some targets require that GCC track the size of each instruction used in
5220 order to generate correct code. Because the final length of an
5221 @code{asm} is only known by the assembler, GCC must make an estimate as
5222 to how big it will be. The estimate is formed by counting the number of
5223 statements in the pattern of the @code{asm} and multiplying that by the
5224 length of the longest instruction on that processor. Statements in the
5225 @code{asm} are identified by newline characters and whatever statement
5226 separator characters are supported by the assembler; on most processors
5227 this is the `@code{;}' character.
5229 Normally, GCC's estimate is perfectly adequate to ensure that correct
5230 code is generated, but it is possible to confuse the compiler if you use
5231 pseudo instructions or assembler macros that expand into multiple real
5232 instructions or if you use assembler directives that expand to more
5233 space in the object file than would be needed for a single instruction.
5234 If this happens then the assembler will produce a diagnostic saying that
5235 a label is unreachable.
5237 @subsection i386 floating point asm operands
5239 There are several rules on the usage of stack-like regs in
5240 asm_operands insns. These rules apply only to the operands that are
5245 Given a set of input regs that die in an asm_operands, it is
5246 necessary to know which are implicitly popped by the asm, and
5247 which must be explicitly popped by gcc.
5249 An input reg that is implicitly popped by the asm must be
5250 explicitly clobbered, unless it is constrained to match an
5254 For any input reg that is implicitly popped by an asm, it is
5255 necessary to know how to adjust the stack to compensate for the pop.
5256 If any non-popped input is closer to the top of the reg-stack than
5257 the implicitly popped reg, it would not be possible to know what the
5258 stack looked like---it's not clear how the rest of the stack ``slides
5261 All implicitly popped input regs must be closer to the top of
5262 the reg-stack than any input that is not implicitly popped.
5264 It is possible that if an input dies in an insn, reload might
5265 use the input reg for an output reload. Consider this example:
5268 asm ("foo" : "=t" (a) : "f" (b));
5271 This asm says that input B is not popped by the asm, and that
5272 the asm pushes a result onto the reg-stack, i.e., the stack is one
5273 deeper after the asm than it was before. But, it is possible that
5274 reload will think that it can use the same reg for both the input and
5275 the output, if input B dies in this insn.
5277 If any input operand uses the @code{f} constraint, all output reg
5278 constraints must use the @code{&} earlyclobber.
5280 The asm above would be written as
5283 asm ("foo" : "=&t" (a) : "f" (b));
5287 Some operands need to be in particular places on the stack. All
5288 output operands fall in this category---there is no other way to
5289 know which regs the outputs appear in unless the user indicates
5290 this in the constraints.
5292 Output operands must specifically indicate which reg an output
5293 appears in after an asm. @code{=f} is not allowed: the operand
5294 constraints must select a class with a single reg.
5297 Output operands may not be ``inserted'' between existing stack regs.
5298 Since no 387 opcode uses a read/write operand, all output operands
5299 are dead before the asm_operands, and are pushed by the asm_operands.
5300 It makes no sense to push anywhere but the top of the reg-stack.
5302 Output operands must start at the top of the reg-stack: output
5303 operands may not ``skip'' a reg.
5306 Some asm statements may need extra stack space for internal
5307 calculations. This can be guaranteed by clobbering stack registers
5308 unrelated to the inputs and outputs.
5312 Here are a couple of reasonable asms to want to write. This asm
5313 takes one input, which is internally popped, and produces two outputs.
5316 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
5319 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
5320 and replaces them with one output. The user must code the @code{st(1)}
5321 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
5324 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
5330 @section Controlling Names Used in Assembler Code
5331 @cindex assembler names for identifiers
5332 @cindex names used in assembler code
5333 @cindex identifiers, names in assembler code
5335 You can specify the name to be used in the assembler code for a C
5336 function or variable by writing the @code{asm} (or @code{__asm__})
5337 keyword after the declarator as follows:
5340 int foo asm ("myfoo") = 2;
5344 This specifies that the name to be used for the variable @code{foo} in
5345 the assembler code should be @samp{myfoo} rather than the usual
5348 On systems where an underscore is normally prepended to the name of a C
5349 function or variable, this feature allows you to define names for the
5350 linker that do not start with an underscore.
5352 It does not make sense to use this feature with a non-static local
5353 variable since such variables do not have assembler names. If you are
5354 trying to put the variable in a particular register, see @ref{Explicit
5355 Reg Vars}. GCC presently accepts such code with a warning, but will
5356 probably be changed to issue an error, rather than a warning, in the
5359 You cannot use @code{asm} in this way in a function @emph{definition}; but
5360 you can get the same effect by writing a declaration for the function
5361 before its definition and putting @code{asm} there, like this:
5364 extern func () asm ("FUNC");
5371 It is up to you to make sure that the assembler names you choose do not
5372 conflict with any other assembler symbols. Also, you must not use a
5373 register name; that would produce completely invalid assembler code. GCC
5374 does not as yet have the ability to store static variables in registers.
5375 Perhaps that will be added.
5377 @node Explicit Reg Vars
5378 @section Variables in Specified Registers
5379 @cindex explicit register variables
5380 @cindex variables in specified registers
5381 @cindex specified registers
5382 @cindex registers, global allocation
5384 GNU C allows you to put a few global variables into specified hardware
5385 registers. You can also specify the register in which an ordinary
5386 register variable should be allocated.
5390 Global register variables reserve registers throughout the program.
5391 This may be useful in programs such as programming language
5392 interpreters which have a couple of global variables that are accessed
5396 Local register variables in specific registers do not reserve the
5397 registers, except at the point where they are used as input or output
5398 operands in an @code{asm} statement and the @code{asm} statement itself is
5399 not deleted. The compiler's data flow analysis is capable of determining
5400 where the specified registers contain live values, and where they are
5401 available for other uses. Stores into local register variables may be deleted
5402 when they appear to be dead according to dataflow analysis. References
5403 to local register variables may be deleted or moved or simplified.
5405 These local variables are sometimes convenient for use with the extended
5406 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
5407 output of the assembler instruction directly into a particular register.
5408 (This will work provided the register you specify fits the constraints
5409 specified for that operand in the @code{asm}.)
5417 @node Global Reg Vars
5418 @subsection Defining Global Register Variables
5419 @cindex global register variables
5420 @cindex registers, global variables in
5422 You can define a global register variable in GNU C like this:
5425 register int *foo asm ("a5");
5429 Here @code{a5} is the name of the register which should be used. Choose a
5430 register which is normally saved and restored by function calls on your
5431 machine, so that library routines will not clobber it.
5433 Naturally the register name is cpu-dependent, so you would need to
5434 conditionalize your program according to cpu type. The register
5435 @code{a5} would be a good choice on a 68000 for a variable of pointer
5436 type. On machines with register windows, be sure to choose a ``global''
5437 register that is not affected magically by the function call mechanism.
5439 In addition, operating systems on one type of cpu may differ in how they
5440 name the registers; then you would need additional conditionals. For
5441 example, some 68000 operating systems call this register @code{%a5}.
5443 Eventually there may be a way of asking the compiler to choose a register
5444 automatically, but first we need to figure out how it should choose and
5445 how to enable you to guide the choice. No solution is evident.
5447 Defining a global register variable in a certain register reserves that
5448 register entirely for this use, at least within the current compilation.
5449 The register will not be allocated for any other purpose in the functions
5450 in the current compilation. The register will not be saved and restored by
5451 these functions. Stores into this register are never deleted even if they
5452 would appear to be dead, but references may be deleted or moved or
5455 It is not safe to access the global register variables from signal
5456 handlers, or from more than one thread of control, because the system
5457 library routines may temporarily use the register for other things (unless
5458 you recompile them specially for the task at hand).
5460 @cindex @code{qsort}, and global register variables
5461 It is not safe for one function that uses a global register variable to
5462 call another such function @code{foo} by way of a third function
5463 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
5464 different source file in which the variable wasn't declared). This is
5465 because @code{lose} might save the register and put some other value there.
5466 For example, you can't expect a global register variable to be available in
5467 the comparison-function that you pass to @code{qsort}, since @code{qsort}
5468 might have put something else in that register. (If you are prepared to
5469 recompile @code{qsort} with the same global register variable, you can
5470 solve this problem.)
5472 If you want to recompile @code{qsort} or other source files which do not
5473 actually use your global register variable, so that they will not use that
5474 register for any other purpose, then it suffices to specify the compiler
5475 option @option{-ffixed-@var{reg}}. You need not actually add a global
5476 register declaration to their source code.
5478 A function which can alter the value of a global register variable cannot
5479 safely be called from a function compiled without this variable, because it
5480 could clobber the value the caller expects to find there on return.
5481 Therefore, the function which is the entry point into the part of the
5482 program that uses the global register variable must explicitly save and
5483 restore the value which belongs to its caller.
5485 @cindex register variable after @code{longjmp}
5486 @cindex global register after @code{longjmp}
5487 @cindex value after @code{longjmp}
5490 On most machines, @code{longjmp} will restore to each global register
5491 variable the value it had at the time of the @code{setjmp}. On some
5492 machines, however, @code{longjmp} will not change the value of global
5493 register variables. To be portable, the function that called @code{setjmp}
5494 should make other arrangements to save the values of the global register
5495 variables, and to restore them in a @code{longjmp}. This way, the same
5496 thing will happen regardless of what @code{longjmp} does.
5498 All global register variable declarations must precede all function
5499 definitions. If such a declaration could appear after function
5500 definitions, the declaration would be too late to prevent the register from
5501 being used for other purposes in the preceding functions.
5503 Global register variables may not have initial values, because an
5504 executable file has no means to supply initial contents for a register.
5506 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5507 registers, but certain library functions, such as @code{getwd}, as well
5508 as the subroutines for division and remainder, modify g3 and g4. g1 and
5509 g2 are local temporaries.
5511 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5512 Of course, it will not do to use more than a few of those.
5514 @node Local Reg Vars
5515 @subsection Specifying Registers for Local Variables
5516 @cindex local variables, specifying registers
5517 @cindex specifying registers for local variables
5518 @cindex registers for local variables
5520 You can define a local register variable with a specified register
5524 register int *foo asm ("a5");
5528 Here @code{a5} is the name of the register which should be used. Note
5529 that this is the same syntax used for defining global register
5530 variables, but for a local variable it would appear within a function.
5532 Naturally the register name is cpu-dependent, but this is not a
5533 problem, since specific registers are most often useful with explicit
5534 assembler instructions (@pxref{Extended Asm}). Both of these things
5535 generally require that you conditionalize your program according to
5538 In addition, operating systems on one type of cpu may differ in how they
5539 name the registers; then you would need additional conditionals. For
5540 example, some 68000 operating systems call this register @code{%a5}.
5542 Defining such a register variable does not reserve the register; it
5543 remains available for other uses in places where flow control determines
5544 the variable's value is not live.
5546 This option does not guarantee that GCC will generate code that has
5547 this variable in the register you specify at all times. You may not
5548 code an explicit reference to this register in the @emph{assembler
5549 instruction template} part of an @code{asm} statement and assume it will
5550 always refer to this variable. However, using the variable as an
5551 @code{asm} @emph{operand} guarantees that the specified register is used
5554 Stores into local register variables may be deleted when they appear to be dead
5555 according to dataflow analysis. References to local register variables may
5556 be deleted or moved or simplified.
5558 As for global register variables, it's recommended that you choose a
5559 register which is normally saved and restored by function calls on
5560 your machine, so that library routines will not clobber it. A common
5561 pitfall is to initialize multiple call-clobbered registers with
5562 arbitrary expressions, where a function call or library call for an
5563 arithmetic operator will overwrite a register value from a previous
5564 assignment, for example @code{r0} below:
5566 register int *p1 asm ("r0") = @dots{};
5567 register int *p2 asm ("r1") = @dots{};
5569 In those cases, a solution is to use a temporary variable for
5570 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
5572 @node Alternate Keywords
5573 @section Alternate Keywords
5574 @cindex alternate keywords
5575 @cindex keywords, alternate
5577 @option{-ansi} and the various @option{-std} options disable certain
5578 keywords. This causes trouble when you want to use GNU C extensions, or
5579 a general-purpose header file that should be usable by all programs,
5580 including ISO C programs. The keywords @code{asm}, @code{typeof} and
5581 @code{inline} are not available in programs compiled with
5582 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
5583 program compiled with @option{-std=c99}). The ISO C99 keyword
5584 @code{restrict} is only available when @option{-std=gnu99} (which will
5585 eventually be the default) or @option{-std=c99} (or the equivalent
5586 @option{-std=iso9899:1999}) is used.
5588 The way to solve these problems is to put @samp{__} at the beginning and
5589 end of each problematical keyword. For example, use @code{__asm__}
5590 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
5592 Other C compilers won't accept these alternative keywords; if you want to
5593 compile with another compiler, you can define the alternate keywords as
5594 macros to replace them with the customary keywords. It looks like this:
5602 @findex __extension__
5604 @option{-pedantic} and other options cause warnings for many GNU C extensions.
5606 prevent such warnings within one expression by writing
5607 @code{__extension__} before the expression. @code{__extension__} has no
5608 effect aside from this.
5610 @node Incomplete Enums
5611 @section Incomplete @code{enum} Types
5613 You can define an @code{enum} tag without specifying its possible values.
5614 This results in an incomplete type, much like what you get if you write
5615 @code{struct foo} without describing the elements. A later declaration
5616 which does specify the possible values completes the type.
5618 You can't allocate variables or storage using the type while it is
5619 incomplete. However, you can work with pointers to that type.
5621 This extension may not be very useful, but it makes the handling of
5622 @code{enum} more consistent with the way @code{struct} and @code{union}
5625 This extension is not supported by GNU C++.
5627 @node Function Names
5628 @section Function Names as Strings
5629 @cindex @code{__func__} identifier
5630 @cindex @code{__FUNCTION__} identifier
5631 @cindex @code{__PRETTY_FUNCTION__} identifier
5633 GCC provides three magic variables which hold the name of the current
5634 function, as a string. The first of these is @code{__func__}, which
5635 is part of the C99 standard:
5637 The identifier @code{__func__} is implicitly declared by the translator
5638 as if, immediately following the opening brace of each function
5639 definition, the declaration
5642 static const char __func__[] = "function-name";
5646 appeared, where function-name is the name of the lexically-enclosing
5647 function. This name is the unadorned name of the function.
5649 @code{__FUNCTION__} is another name for @code{__func__}. Older
5650 versions of GCC recognize only this name. However, it is not
5651 standardized. For maximum portability, we recommend you use
5652 @code{__func__}, but provide a fallback definition with the
5656 #if __STDC_VERSION__ < 199901L
5658 # define __func__ __FUNCTION__
5660 # define __func__ "<unknown>"
5665 In C, @code{__PRETTY_FUNCTION__} is yet another name for
5666 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
5667 the type signature of the function as well as its bare name. For
5668 example, this program:
5672 extern int printf (char *, ...);
5679 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
5680 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
5698 __PRETTY_FUNCTION__ = void a::sub(int)
5701 These identifiers are not preprocessor macros. In GCC 3.3 and
5702 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
5703 were treated as string literals; they could be used to initialize
5704 @code{char} arrays, and they could be concatenated with other string
5705 literals. GCC 3.4 and later treat them as variables, like
5706 @code{__func__}. In C++, @code{__FUNCTION__} and
5707 @code{__PRETTY_FUNCTION__} have always been variables.
5709 @node Return Address
5710 @section Getting the Return or Frame Address of a Function
5712 These functions may be used to get information about the callers of a
5715 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
5716 This function returns the return address of the current function, or of
5717 one of its callers. The @var{level} argument is number of frames to
5718 scan up the call stack. A value of @code{0} yields the return address
5719 of the current function, a value of @code{1} yields the return address
5720 of the caller of the current function, and so forth. When inlining
5721 the expected behavior is that the function will return the address of
5722 the function that will be returned to. To work around this behavior use
5723 the @code{noinline} function attribute.
5725 The @var{level} argument must be a constant integer.
5727 On some machines it may be impossible to determine the return address of
5728 any function other than the current one; in such cases, or when the top
5729 of the stack has been reached, this function will return @code{0} or a
5730 random value. In addition, @code{__builtin_frame_address} may be used
5731 to determine if the top of the stack has been reached.
5733 This function should only be used with a nonzero argument for debugging
5737 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
5738 This function is similar to @code{__builtin_return_address}, but it
5739 returns the address of the function frame rather than the return address
5740 of the function. Calling @code{__builtin_frame_address} with a value of
5741 @code{0} yields the frame address of the current function, a value of
5742 @code{1} yields the frame address of the caller of the current function,
5745 The frame is the area on the stack which holds local variables and saved
5746 registers. The frame address is normally the address of the first word
5747 pushed on to the stack by the function. However, the exact definition
5748 depends upon the processor and the calling convention. If the processor
5749 has a dedicated frame pointer register, and the function has a frame,
5750 then @code{__builtin_frame_address} will return the value of the frame
5753 On some machines it may be impossible to determine the frame address of
5754 any function other than the current one; in such cases, or when the top
5755 of the stack has been reached, this function will return @code{0} if
5756 the first frame pointer is properly initialized by the startup code.
5758 This function should only be used with a nonzero argument for debugging
5762 @node Vector Extensions
5763 @section Using vector instructions through built-in functions
5765 On some targets, the instruction set contains SIMD vector instructions that
5766 operate on multiple values contained in one large register at the same time.
5767 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5770 The first step in using these extensions is to provide the necessary data
5771 types. This should be done using an appropriate @code{typedef}:
5774 typedef int v4si __attribute__ ((vector_size (16)));
5777 The @code{int} type specifies the base type, while the attribute specifies
5778 the vector size for the variable, measured in bytes. For example, the
5779 declaration above causes the compiler to set the mode for the @code{v4si}
5780 type to be 16 bytes wide and divided into @code{int} sized units. For
5781 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5782 corresponding mode of @code{foo} will be @acronym{V4SI}.
5784 The @code{vector_size} attribute is only applicable to integral and
5785 float scalars, although arrays, pointers, and function return values
5786 are allowed in conjunction with this construct.
5788 All the basic integer types can be used as base types, both as signed
5789 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5790 @code{long long}. In addition, @code{float} and @code{double} can be
5791 used to build floating-point vector types.
5793 Specifying a combination that is not valid for the current architecture
5794 will cause GCC to synthesize the instructions using a narrower mode.
5795 For example, if you specify a variable of type @code{V4SI} and your
5796 architecture does not allow for this specific SIMD type, GCC will
5797 produce code that uses 4 @code{SIs}.
5799 The types defined in this manner can be used with a subset of normal C
5800 operations. Currently, GCC will allow using the following operators
5801 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.
5803 The operations behave like C++ @code{valarrays}. Addition is defined as
5804 the addition of the corresponding elements of the operands. For
5805 example, in the code below, each of the 4 elements in @var{a} will be
5806 added to the corresponding 4 elements in @var{b} and the resulting
5807 vector will be stored in @var{c}.
5810 typedef int v4si __attribute__ ((vector_size (16)));
5817 Subtraction, multiplication, division, and the logical operations
5818 operate in a similar manner. Likewise, the result of using the unary
5819 minus or complement operators on a vector type is a vector whose
5820 elements are the negative or complemented values of the corresponding
5821 elements in the operand.
5823 You can declare variables and use them in function calls and returns, as
5824 well as in assignments and some casts. You can specify a vector type as
5825 a return type for a function. Vector types can also be used as function
5826 arguments. It is possible to cast from one vector type to another,
5827 provided they are of the same size (in fact, you can also cast vectors
5828 to and from other datatypes of the same size).
5830 You cannot operate between vectors of different lengths or different
5831 signedness without a cast.
5833 A port that supports hardware vector operations, usually provides a set
5834 of built-in functions that can be used to operate on vectors. For
5835 example, a function to add two vectors and multiply the result by a
5836 third could look like this:
5839 v4si f (v4si a, v4si b, v4si c)
5841 v4si tmp = __builtin_addv4si (a, b);
5842 return __builtin_mulv4si (tmp, c);
5849 @findex __builtin_offsetof
5851 GCC implements for both C and C++ a syntactic extension to implement
5852 the @code{offsetof} macro.
5856 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5858 offsetof_member_designator:
5860 | offsetof_member_designator "." @code{identifier}
5861 | offsetof_member_designator "[" @code{expr} "]"
5864 This extension is sufficient such that
5867 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5870 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5871 may be dependent. In either case, @var{member} may consist of a single
5872 identifier, or a sequence of member accesses and array references.
5874 @node Atomic Builtins
5875 @section Built-in functions for atomic memory access
5877 The following builtins are intended to be compatible with those described
5878 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5879 section 7.4. As such, they depart from the normal GCC practice of using
5880 the ``__builtin_'' prefix, and further that they are overloaded such that
5881 they work on multiple types.
5883 The definition given in the Intel documentation allows only for the use of
5884 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5885 counterparts. GCC will allow any integral scalar or pointer type that is
5886 1, 2, 4 or 8 bytes in length.
5888 Not all operations are supported by all target processors. If a particular
5889 operation cannot be implemented on the target processor, a warning will be
5890 generated and a call an external function will be generated. The external
5891 function will carry the same name as the builtin, with an additional suffix
5892 @samp{_@var{n}} where @var{n} is the size of the data type.
5894 @c ??? Should we have a mechanism to suppress this warning? This is almost
5895 @c useful for implementing the operation under the control of an external
5898 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5899 no memory operand will be moved across the operation, either forward or
5900 backward. Further, instructions will be issued as necessary to prevent the
5901 processor from speculating loads across the operation and from queuing stores
5902 after the operation.
5904 All of the routines are described in the Intel documentation to take
5905 ``an optional list of variables protected by the memory barrier''. It's
5906 not clear what is meant by that; it could mean that @emph{only} the
5907 following variables are protected, or it could mean that these variables
5908 should in addition be protected. At present GCC ignores this list and
5909 protects all variables which are globally accessible. If in the future
5910 we make some use of this list, an empty list will continue to mean all
5911 globally accessible variables.
5914 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5915 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5916 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5917 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5918 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5919 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5920 @findex __sync_fetch_and_add
5921 @findex __sync_fetch_and_sub
5922 @findex __sync_fetch_and_or
5923 @findex __sync_fetch_and_and
5924 @findex __sync_fetch_and_xor
5925 @findex __sync_fetch_and_nand
5926 These builtins perform the operation suggested by the name, and
5927 returns the value that had previously been in memory. That is,
5930 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5931 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
5934 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
5935 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
5937 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5938 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5939 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5940 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5941 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5942 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5943 @findex __sync_add_and_fetch
5944 @findex __sync_sub_and_fetch
5945 @findex __sync_or_and_fetch
5946 @findex __sync_and_and_fetch
5947 @findex __sync_xor_and_fetch
5948 @findex __sync_nand_and_fetch
5949 These builtins perform the operation suggested by the name, and
5950 return the new value. That is,
5953 @{ *ptr @var{op}= value; return *ptr; @}
5954 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
5957 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
5958 builtin as @code{*ptr = ~(*ptr & value)} instead of
5959 @code{*ptr = ~*ptr & value}.
5961 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5962 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5963 @findex __sync_bool_compare_and_swap
5964 @findex __sync_val_compare_and_swap
5965 These builtins perform an atomic compare and swap. That is, if the current
5966 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5969 The ``bool'' version returns true if the comparison is successful and
5970 @var{newval} was written. The ``val'' version returns the contents
5971 of @code{*@var{ptr}} before the operation.
5973 @item __sync_synchronize (...)
5974 @findex __sync_synchronize
5975 This builtin issues a full memory barrier.
5977 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5978 @findex __sync_lock_test_and_set
5979 This builtin, as described by Intel, is not a traditional test-and-set
5980 operation, but rather an atomic exchange operation. It writes @var{value}
5981 into @code{*@var{ptr}}, and returns the previous contents of
5984 Many targets have only minimal support for such locks, and do not support
5985 a full exchange operation. In this case, a target may support reduced
5986 functionality here by which the @emph{only} valid value to store is the
5987 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5988 is implementation defined.
5990 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5991 This means that references after the builtin cannot move to (or be
5992 speculated to) before the builtin, but previous memory stores may not
5993 be globally visible yet, and previous memory loads may not yet be
5996 @item void __sync_lock_release (@var{type} *ptr, ...)
5997 @findex __sync_lock_release
5998 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5999 Normally this means writing the constant 0 to @code{*@var{ptr}}.
6001 This builtin is not a full barrier, but rather a @dfn{release barrier}.
6002 This means that all previous memory stores are globally visible, and all
6003 previous memory loads have been satisfied, but following memory reads
6004 are not prevented from being speculated to before the barrier.
6007 @node Object Size Checking
6008 @section Object Size Checking Builtins
6009 @findex __builtin_object_size
6010 @findex __builtin___memcpy_chk
6011 @findex __builtin___mempcpy_chk
6012 @findex __builtin___memmove_chk
6013 @findex __builtin___memset_chk
6014 @findex __builtin___strcpy_chk
6015 @findex __builtin___stpcpy_chk
6016 @findex __builtin___strncpy_chk
6017 @findex __builtin___strcat_chk
6018 @findex __builtin___strncat_chk
6019 @findex __builtin___sprintf_chk
6020 @findex __builtin___snprintf_chk
6021 @findex __builtin___vsprintf_chk
6022 @findex __builtin___vsnprintf_chk
6023 @findex __builtin___printf_chk
6024 @findex __builtin___vprintf_chk
6025 @findex __builtin___fprintf_chk
6026 @findex __builtin___vfprintf_chk
6028 GCC implements a limited buffer overflow protection mechanism
6029 that can prevent some buffer overflow attacks.
6031 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
6032 is a built-in construct that returns a constant number of bytes from
6033 @var{ptr} to the end of the object @var{ptr} pointer points to
6034 (if known at compile time). @code{__builtin_object_size} never evaluates
6035 its arguments for side-effects. If there are any side-effects in them, it
6036 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6037 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
6038 point to and all of them are known at compile time, the returned number
6039 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
6040 0 and minimum if nonzero. If it is not possible to determine which objects
6041 @var{ptr} points to at compile time, @code{__builtin_object_size} should
6042 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
6043 for @var{type} 2 or 3.
6045 @var{type} is an integer constant from 0 to 3. If the least significant
6046 bit is clear, objects are whole variables, if it is set, a closest
6047 surrounding subobject is considered the object a pointer points to.
6048 The second bit determines if maximum or minimum of remaining bytes
6052 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
6053 char *p = &var.buf1[1], *q = &var.b;
6055 /* Here the object p points to is var. */
6056 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
6057 /* The subobject p points to is var.buf1. */
6058 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
6059 /* The object q points to is var. */
6060 assert (__builtin_object_size (q, 0)
6061 == (char *) (&var + 1) - (char *) &var.b);
6062 /* The subobject q points to is var.b. */
6063 assert (__builtin_object_size (q, 1) == sizeof (var.b));
6067 There are built-in functions added for many common string operation
6068 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
6069 built-in is provided. This built-in has an additional last argument,
6070 which is the number of bytes remaining in object the @var{dest}
6071 argument points to or @code{(size_t) -1} if the size is not known.
6073 The built-in functions are optimized into the normal string functions
6074 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
6075 it is known at compile time that the destination object will not
6076 be overflown. If the compiler can determine at compile time the
6077 object will be always overflown, it issues a warning.
6079 The intended use can be e.g.
6083 #define bos0(dest) __builtin_object_size (dest, 0)
6084 #define memcpy(dest, src, n) \
6085 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
6089 /* It is unknown what object p points to, so this is optimized
6090 into plain memcpy - no checking is possible. */
6091 memcpy (p, "abcde", n);
6092 /* Destination is known and length too. It is known at compile
6093 time there will be no overflow. */
6094 memcpy (&buf[5], "abcde", 5);
6095 /* Destination is known, but the length is not known at compile time.
6096 This will result in __memcpy_chk call that can check for overflow
6098 memcpy (&buf[5], "abcde", n);
6099 /* Destination is known and it is known at compile time there will
6100 be overflow. There will be a warning and __memcpy_chk call that
6101 will abort the program at runtime. */
6102 memcpy (&buf[6], "abcde", 5);
6105 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
6106 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
6107 @code{strcat} and @code{strncat}.
6109 There are also checking built-in functions for formatted output functions.
6111 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
6112 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6113 const char *fmt, ...);
6114 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
6116 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6117 const char *fmt, va_list ap);
6120 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
6121 etc.@: functions and can contain implementation specific flags on what
6122 additional security measures the checking function might take, such as
6123 handling @code{%n} differently.
6125 The @var{os} argument is the object size @var{s} points to, like in the
6126 other built-in functions. There is a small difference in the behavior
6127 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
6128 optimized into the non-checking functions only if @var{flag} is 0, otherwise
6129 the checking function is called with @var{os} argument set to
6132 In addition to this, there are checking built-in functions
6133 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
6134 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
6135 These have just one additional argument, @var{flag}, right before
6136 format string @var{fmt}. If the compiler is able to optimize them to
6137 @code{fputc} etc.@: functions, it will, otherwise the checking function
6138 should be called and the @var{flag} argument passed to it.
6140 @node Other Builtins
6141 @section Other built-in functions provided by GCC
6142 @cindex built-in functions
6143 @findex __builtin_fpclassify
6144 @findex __builtin_isfinite
6145 @findex __builtin_isnormal
6146 @findex __builtin_isgreater
6147 @findex __builtin_isgreaterequal
6148 @findex __builtin_isinf_sign
6149 @findex __builtin_isless
6150 @findex __builtin_islessequal
6151 @findex __builtin_islessgreater
6152 @findex __builtin_isunordered
6153 @findex __builtin_powi
6154 @findex __builtin_powif
6155 @findex __builtin_powil
6313 @findex fprintf_unlocked
6315 @findex fputs_unlocked
6432 @findex printf_unlocked
6464 @findex significandf
6465 @findex significandl
6536 GCC provides a large number of built-in functions other than the ones
6537 mentioned above. Some of these are for internal use in the processing
6538 of exceptions or variable-length argument lists and will not be
6539 documented here because they may change from time to time; we do not
6540 recommend general use of these functions.
6542 The remaining functions are provided for optimization purposes.
6544 @opindex fno-builtin
6545 GCC includes built-in versions of many of the functions in the standard
6546 C library. The versions prefixed with @code{__builtin_} will always be
6547 treated as having the same meaning as the C library function even if you
6548 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
6549 Many of these functions are only optimized in certain cases; if they are
6550 not optimized in a particular case, a call to the library function will
6555 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
6556 @option{-std=c99}), the functions
6557 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
6558 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
6559 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
6560 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
6561 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
6562 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
6563 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
6564 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
6565 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
6566 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
6567 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
6568 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
6569 @code{signbitd64}, @code{signbitd128}, @code{significandf},
6570 @code{significandl}, @code{significand}, @code{sincosf},
6571 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
6572 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
6573 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
6574 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
6576 may be handled as built-in functions.
6577 All these functions have corresponding versions
6578 prefixed with @code{__builtin_}, which may be used even in strict C89
6581 The ISO C99 functions
6582 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
6583 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
6584 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
6585 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
6586 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
6587 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
6588 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
6589 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
6590 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
6591 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
6592 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
6593 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
6594 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
6595 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
6596 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
6597 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
6598 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
6599 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
6600 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
6601 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
6602 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
6603 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
6604 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
6605 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
6606 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
6607 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
6608 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
6609 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
6610 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
6611 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
6612 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
6613 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
6614 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
6615 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
6616 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
6617 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
6618 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
6619 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
6620 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
6621 are handled as built-in functions
6622 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6624 There are also built-in versions of the ISO C99 functions
6625 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
6626 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
6627 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
6628 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
6629 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
6630 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
6631 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
6632 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
6633 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
6634 that are recognized in any mode since ISO C90 reserves these names for
6635 the purpose to which ISO C99 puts them. All these functions have
6636 corresponding versions prefixed with @code{__builtin_}.
6638 The ISO C94 functions
6639 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
6640 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
6641 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
6643 are handled as built-in functions
6644 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6646 The ISO C90 functions
6647 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
6648 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
6649 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
6650 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
6651 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
6652 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
6653 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
6654 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
6655 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
6656 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
6657 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
6658 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
6659 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
6660 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
6661 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
6662 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
6663 are all recognized as built-in functions unless
6664 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
6665 is specified for an individual function). All of these functions have
6666 corresponding versions prefixed with @code{__builtin_}.
6668 GCC provides built-in versions of the ISO C99 floating point comparison
6669 macros that avoid raising exceptions for unordered operands. They have
6670 the same names as the standard macros ( @code{isgreater},
6671 @code{isgreaterequal}, @code{isless}, @code{islessequal},
6672 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
6673 prefixed. We intend for a library implementor to be able to simply
6674 @code{#define} each standard macro to its built-in equivalent.
6675 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
6676 @code{isinf_sign} and @code{isnormal} built-ins used with
6677 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
6678 builtins appear both with and without the @code{__builtin_} prefix.
6680 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
6682 You can use the built-in function @code{__builtin_types_compatible_p} to
6683 determine whether two types are the same.
6685 This built-in function returns 1 if the unqualified versions of the
6686 types @var{type1} and @var{type2} (which are types, not expressions) are
6687 compatible, 0 otherwise. The result of this built-in function can be
6688 used in integer constant expressions.
6690 This built-in function ignores top level qualifiers (e.g., @code{const},
6691 @code{volatile}). For example, @code{int} is equivalent to @code{const
6694 The type @code{int[]} and @code{int[5]} are compatible. On the other
6695 hand, @code{int} and @code{char *} are not compatible, even if the size
6696 of their types, on the particular architecture are the same. Also, the
6697 amount of pointer indirection is taken into account when determining
6698 similarity. Consequently, @code{short *} is not similar to
6699 @code{short **}. Furthermore, two types that are typedefed are
6700 considered compatible if their underlying types are compatible.
6702 An @code{enum} type is not considered to be compatible with another
6703 @code{enum} type even if both are compatible with the same integer
6704 type; this is what the C standard specifies.
6705 For example, @code{enum @{foo, bar@}} is not similar to
6706 @code{enum @{hot, dog@}}.
6708 You would typically use this function in code whose execution varies
6709 depending on the arguments' types. For example:
6714 typeof (x) tmp = (x); \
6715 if (__builtin_types_compatible_p (typeof (x), long double)) \
6716 tmp = foo_long_double (tmp); \
6717 else if (__builtin_types_compatible_p (typeof (x), double)) \
6718 tmp = foo_double (tmp); \
6719 else if (__builtin_types_compatible_p (typeof (x), float)) \
6720 tmp = foo_float (tmp); \
6727 @emph{Note:} This construct is only available for C@.
6731 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
6733 You can use the built-in function @code{__builtin_choose_expr} to
6734 evaluate code depending on the value of a constant expression. This
6735 built-in function returns @var{exp1} if @var{const_exp}, which is an
6736 integer constant expression, is nonzero. Otherwise it returns 0.
6738 This built-in function is analogous to the @samp{? :} operator in C,
6739 except that the expression returned has its type unaltered by promotion
6740 rules. Also, the built-in function does not evaluate the expression
6741 that was not chosen. For example, if @var{const_exp} evaluates to true,
6742 @var{exp2} is not evaluated even if it has side-effects.
6744 This built-in function can return an lvalue if the chosen argument is an
6747 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
6748 type. Similarly, if @var{exp2} is returned, its return type is the same
6755 __builtin_choose_expr ( \
6756 __builtin_types_compatible_p (typeof (x), double), \
6758 __builtin_choose_expr ( \
6759 __builtin_types_compatible_p (typeof (x), float), \
6761 /* @r{The void expression results in a compile-time error} \
6762 @r{when assigning the result to something.} */ \
6766 @emph{Note:} This construct is only available for C@. Furthermore, the
6767 unused expression (@var{exp1} or @var{exp2} depending on the value of
6768 @var{const_exp}) may still generate syntax errors. This may change in
6773 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6774 You can use the built-in function @code{__builtin_constant_p} to
6775 determine if a value is known to be constant at compile-time and hence
6776 that GCC can perform constant-folding on expressions involving that
6777 value. The argument of the function is the value to test. The function
6778 returns the integer 1 if the argument is known to be a compile-time
6779 constant and 0 if it is not known to be a compile-time constant. A
6780 return of 0 does not indicate that the value is @emph{not} a constant,
6781 but merely that GCC cannot prove it is a constant with the specified
6782 value of the @option{-O} option.
6784 You would typically use this function in an embedded application where
6785 memory was a critical resource. If you have some complex calculation,
6786 you may want it to be folded if it involves constants, but need to call
6787 a function if it does not. For example:
6790 #define Scale_Value(X) \
6791 (__builtin_constant_p (X) \
6792 ? ((X) * SCALE + OFFSET) : Scale (X))
6795 You may use this built-in function in either a macro or an inline
6796 function. However, if you use it in an inlined function and pass an
6797 argument of the function as the argument to the built-in, GCC will
6798 never return 1 when you call the inline function with a string constant
6799 or compound literal (@pxref{Compound Literals}) and will not return 1
6800 when you pass a constant numeric value to the inline function unless you
6801 specify the @option{-O} option.
6803 You may also use @code{__builtin_constant_p} in initializers for static
6804 data. For instance, you can write
6807 static const int table[] = @{
6808 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6814 This is an acceptable initializer even if @var{EXPRESSION} is not a
6815 constant expression, including the case where
6816 @code{__builtin_constant_p} returns 1 because @var{EXPRESSION} can be
6817 folded to a constant but @var{EXPRESSION} contains operands that would
6818 not otherwise be permitted in a static initializer (for example,
6819 @code{0 && foo ()}). GCC must be more conservative about evaluating the
6820 built-in in this case, because it has no opportunity to perform
6823 Previous versions of GCC did not accept this built-in in data
6824 initializers. The earliest version where it is completely safe is
6828 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6829 @opindex fprofile-arcs
6830 You may use @code{__builtin_expect} to provide the compiler with
6831 branch prediction information. In general, you should prefer to
6832 use actual profile feedback for this (@option{-fprofile-arcs}), as
6833 programmers are notoriously bad at predicting how their programs
6834 actually perform. However, there are applications in which this
6835 data is hard to collect.
6837 The return value is the value of @var{exp}, which should be an integral
6838 expression. The semantics of the built-in are that it is expected that
6839 @var{exp} == @var{c}. For example:
6842 if (__builtin_expect (x, 0))
6847 would indicate that we do not expect to call @code{foo}, since
6848 we expect @code{x} to be zero. Since you are limited to integral
6849 expressions for @var{exp}, you should use constructions such as
6852 if (__builtin_expect (ptr != NULL, 1))
6857 when testing pointer or floating-point values.
6860 @deftypefn {Built-in Function} void __builtin_trap (void)
6861 This function causes the program to exit abnormally. GCC implements
6862 this function by using a target-dependent mechanism (such as
6863 intentionally executing an illegal instruction) or by calling
6864 @code{abort}. The mechanism used may vary from release to release so
6865 you should not rely on any particular implementation.
6868 @deftypefn {Built-in Function} void __builtin_unreachable (void)
6869 If control flow reaches the point of the @code{__builtin_unreachable},
6870 the program is undefined. It is useful in situations where the
6871 compiler cannot deduce the unreachability of the code.
6873 One such case is immediately following an @code{asm} statement that
6874 will either never terminate, or one that transfers control elsewhere
6875 and never returns. In this example, without the
6876 @code{__builtin_unreachable}, GCC would issue a warning that control
6877 reaches the end of a non-void function. It would also generate code
6878 to return after the @code{asm}.
6881 int f (int c, int v)
6889 asm("jmp error_handler");
6890 __builtin_unreachable ();
6895 Because the @code{asm} statement unconditionally transfers control out
6896 of the function, control will never reach the end of the function
6897 body. The @code{__builtin_unreachable} is in fact unreachable and
6898 communicates this fact to the compiler.
6900 Another use for @code{__builtin_unreachable} is following a call a
6901 function that never returns but that is not declared
6902 @code{__attribute__((noreturn))}, as in this example:
6905 void function_that_never_returns (void);
6915 function_that_never_returns ();
6916 __builtin_unreachable ();
6923 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
6924 This function is used to flush the processor's instruction cache for
6925 the region of memory between @var{begin} inclusive and @var{end}
6926 exclusive. Some targets require that the instruction cache be
6927 flushed, after modifying memory containing code, in order to obtain
6928 deterministic behavior.
6930 If the target does not require instruction cache flushes,
6931 @code{__builtin___clear_cache} has no effect. Otherwise either
6932 instructions are emitted in-line to clear the instruction cache or a
6933 call to the @code{__clear_cache} function in libgcc is made.
6936 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6937 This function is used to minimize cache-miss latency by moving data into
6938 a cache before it is accessed.
6939 You can insert calls to @code{__builtin_prefetch} into code for which
6940 you know addresses of data in memory that is likely to be accessed soon.
6941 If the target supports them, data prefetch instructions will be generated.
6942 If the prefetch is done early enough before the access then the data will
6943 be in the cache by the time it is accessed.
6945 The value of @var{addr} is the address of the memory to prefetch.
6946 There are two optional arguments, @var{rw} and @var{locality}.
6947 The value of @var{rw} is a compile-time constant one or zero; one
6948 means that the prefetch is preparing for a write to the memory address
6949 and zero, the default, means that the prefetch is preparing for a read.
6950 The value @var{locality} must be a compile-time constant integer between
6951 zero and three. A value of zero means that the data has no temporal
6952 locality, so it need not be left in the cache after the access. A value
6953 of three means that the data has a high degree of temporal locality and
6954 should be left in all levels of cache possible. Values of one and two
6955 mean, respectively, a low or moderate degree of temporal locality. The
6959 for (i = 0; i < n; i++)
6962 __builtin_prefetch (&a[i+j], 1, 1);
6963 __builtin_prefetch (&b[i+j], 0, 1);
6968 Data prefetch does not generate faults if @var{addr} is invalid, but
6969 the address expression itself must be valid. For example, a prefetch
6970 of @code{p->next} will not fault if @code{p->next} is not a valid
6971 address, but evaluation will fault if @code{p} is not a valid address.
6973 If the target does not support data prefetch, the address expression
6974 is evaluated if it includes side effects but no other code is generated
6975 and GCC does not issue a warning.
6978 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6979 Returns a positive infinity, if supported by the floating-point format,
6980 else @code{DBL_MAX}. This function is suitable for implementing the
6981 ISO C macro @code{HUGE_VAL}.
6984 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6985 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6988 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6989 Similar to @code{__builtin_huge_val}, except the return
6990 type is @code{long double}.
6993 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
6994 This built-in implements the C99 fpclassify functionality. The first
6995 five int arguments should be the target library's notion of the
6996 possible FP classes and are used for return values. They must be
6997 constant values and they must appear in this order: @code{FP_NAN},
6998 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
6999 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
7000 to classify. GCC treats the last argument as type-generic, which
7001 means it does not do default promotion from float to double.
7004 @deftypefn {Built-in Function} double __builtin_inf (void)
7005 Similar to @code{__builtin_huge_val}, except a warning is generated
7006 if the target floating-point format does not support infinities.
7009 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
7010 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
7013 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
7014 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
7017 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
7018 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
7021 @deftypefn {Built-in Function} float __builtin_inff (void)
7022 Similar to @code{__builtin_inf}, except the return type is @code{float}.
7023 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
7026 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
7027 Similar to @code{__builtin_inf}, except the return
7028 type is @code{long double}.
7031 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
7032 Similar to @code{isinf}, except the return value will be negative for
7033 an argument of @code{-Inf}. Note while the parameter list is an
7034 ellipsis, this function only accepts exactly one floating point
7035 argument. GCC treats this parameter as type-generic, which means it
7036 does not do default promotion from float to double.
7039 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
7040 This is an implementation of the ISO C99 function @code{nan}.
7042 Since ISO C99 defines this function in terms of @code{strtod}, which we
7043 do not implement, a description of the parsing is in order. The string
7044 is parsed as by @code{strtol}; that is, the base is recognized by
7045 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
7046 in the significand such that the least significant bit of the number
7047 is at the least significant bit of the significand. The number is
7048 truncated to fit the significand field provided. The significand is
7049 forced to be a quiet NaN@.
7051 This function, if given a string literal all of which would have been
7052 consumed by strtol, is evaluated early enough that it is considered a
7053 compile-time constant.
7056 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
7057 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
7060 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
7061 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
7064 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
7065 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
7068 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
7069 Similar to @code{__builtin_nan}, except the return type is @code{float}.
7072 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
7073 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
7076 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
7077 Similar to @code{__builtin_nan}, except the significand is forced
7078 to be a signaling NaN@. The @code{nans} function is proposed by
7079 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
7082 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
7083 Similar to @code{__builtin_nans}, except the return type is @code{float}.
7086 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
7087 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
7090 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
7091 Returns one plus the index of the least significant 1-bit of @var{x}, or
7092 if @var{x} is zero, returns zero.
7095 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
7096 Returns the number of leading 0-bits in @var{x}, starting at the most
7097 significant bit position. If @var{x} is 0, the result is undefined.
7100 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
7101 Returns the number of trailing 0-bits in @var{x}, starting at the least
7102 significant bit position. If @var{x} is 0, the result is undefined.
7105 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
7106 Returns the number of 1-bits in @var{x}.
7109 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
7110 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
7114 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
7115 Similar to @code{__builtin_ffs}, except the argument type is
7116 @code{unsigned long}.
7119 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
7120 Similar to @code{__builtin_clz}, except the argument type is
7121 @code{unsigned long}.
7124 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
7125 Similar to @code{__builtin_ctz}, except the argument type is
7126 @code{unsigned long}.
7129 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
7130 Similar to @code{__builtin_popcount}, except the argument type is
7131 @code{unsigned long}.
7134 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
7135 Similar to @code{__builtin_parity}, except the argument type is
7136 @code{unsigned long}.
7139 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
7140 Similar to @code{__builtin_ffs}, except the argument type is
7141 @code{unsigned long long}.
7144 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
7145 Similar to @code{__builtin_clz}, except the argument type is
7146 @code{unsigned long long}.
7149 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
7150 Similar to @code{__builtin_ctz}, except the argument type is
7151 @code{unsigned long long}.
7154 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
7155 Similar to @code{__builtin_popcount}, except the argument type is
7156 @code{unsigned long long}.
7159 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
7160 Similar to @code{__builtin_parity}, except the argument type is
7161 @code{unsigned long long}.
7164 @deftypefn {Built-in Function} double __builtin_powi (double, int)
7165 Returns the first argument raised to the power of the second. Unlike the
7166 @code{pow} function no guarantees about precision and rounding are made.
7169 @deftypefn {Built-in Function} float __builtin_powif (float, int)
7170 Similar to @code{__builtin_powi}, except the argument and return types
7174 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
7175 Similar to @code{__builtin_powi}, except the argument and return types
7176 are @code{long double}.
7179 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
7180 Returns @var{x} with the order of the bytes reversed; for example,
7181 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
7185 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
7186 Similar to @code{__builtin_bswap32}, except the argument and return types
7190 @node Target Builtins
7191 @section Built-in Functions Specific to Particular Target Machines
7193 On some target machines, GCC supports many built-in functions specific
7194 to those machines. Generally these generate calls to specific machine
7195 instructions, but allow the compiler to schedule those calls.
7198 * Alpha Built-in Functions::
7199 * ARM iWMMXt Built-in Functions::
7200 * ARM NEON Intrinsics::
7201 * Blackfin Built-in Functions::
7202 * FR-V Built-in Functions::
7203 * X86 Built-in Functions::
7204 * MIPS DSP Built-in Functions::
7205 * MIPS Paired-Single Support::
7206 * MIPS Loongson Built-in Functions::
7207 * Other MIPS Built-in Functions::
7208 * picoChip Built-in Functions::
7209 * PowerPC AltiVec Built-in Functions::
7210 * SPARC VIS Built-in Functions::
7211 * SPU Built-in Functions::
7214 @node Alpha Built-in Functions
7215 @subsection Alpha Built-in Functions
7217 These built-in functions are available for the Alpha family of
7218 processors, depending on the command-line switches used.
7220 The following built-in functions are always available. They
7221 all generate the machine instruction that is part of the name.
7224 long __builtin_alpha_implver (void)
7225 long __builtin_alpha_rpcc (void)
7226 long __builtin_alpha_amask (long)
7227 long __builtin_alpha_cmpbge (long, long)
7228 long __builtin_alpha_extbl (long, long)
7229 long __builtin_alpha_extwl (long, long)
7230 long __builtin_alpha_extll (long, long)
7231 long __builtin_alpha_extql (long, long)
7232 long __builtin_alpha_extwh (long, long)
7233 long __builtin_alpha_extlh (long, long)
7234 long __builtin_alpha_extqh (long, long)
7235 long __builtin_alpha_insbl (long, long)
7236 long __builtin_alpha_inswl (long, long)
7237 long __builtin_alpha_insll (long, long)
7238 long __builtin_alpha_insql (long, long)
7239 long __builtin_alpha_inswh (long, long)
7240 long __builtin_alpha_inslh (long, long)
7241 long __builtin_alpha_insqh (long, long)
7242 long __builtin_alpha_mskbl (long, long)
7243 long __builtin_alpha_mskwl (long, long)
7244 long __builtin_alpha_mskll (long, long)
7245 long __builtin_alpha_mskql (long, long)
7246 long __builtin_alpha_mskwh (long, long)
7247 long __builtin_alpha_msklh (long, long)
7248 long __builtin_alpha_mskqh (long, long)
7249 long __builtin_alpha_umulh (long, long)
7250 long __builtin_alpha_zap (long, long)
7251 long __builtin_alpha_zapnot (long, long)
7254 The following built-in functions are always with @option{-mmax}
7255 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7256 later. They all generate the machine instruction that is part
7260 long __builtin_alpha_pklb (long)
7261 long __builtin_alpha_pkwb (long)
7262 long __builtin_alpha_unpkbl (long)
7263 long __builtin_alpha_unpkbw (long)
7264 long __builtin_alpha_minub8 (long, long)
7265 long __builtin_alpha_minsb8 (long, long)
7266 long __builtin_alpha_minuw4 (long, long)
7267 long __builtin_alpha_minsw4 (long, long)
7268 long __builtin_alpha_maxub8 (long, long)
7269 long __builtin_alpha_maxsb8 (long, long)
7270 long __builtin_alpha_maxuw4 (long, long)
7271 long __builtin_alpha_maxsw4 (long, long)
7272 long __builtin_alpha_perr (long, long)
7275 The following built-in functions are always with @option{-mcix}
7276 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
7277 later. They all generate the machine instruction that is part
7281 long __builtin_alpha_cttz (long)
7282 long __builtin_alpha_ctlz (long)
7283 long __builtin_alpha_ctpop (long)
7286 The following builtins are available on systems that use the OSF/1
7287 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
7288 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
7289 @code{rdval} and @code{wrval}.
7292 void *__builtin_thread_pointer (void)
7293 void __builtin_set_thread_pointer (void *)
7296 @node ARM iWMMXt Built-in Functions
7297 @subsection ARM iWMMXt Built-in Functions
7299 These built-in functions are available for the ARM family of
7300 processors when the @option{-mcpu=iwmmxt} switch is used:
7303 typedef int v2si __attribute__ ((vector_size (8)));
7304 typedef short v4hi __attribute__ ((vector_size (8)));
7305 typedef char v8qi __attribute__ ((vector_size (8)));
7307 int __builtin_arm_getwcx (int)
7308 void __builtin_arm_setwcx (int, int)
7309 int __builtin_arm_textrmsb (v8qi, int)
7310 int __builtin_arm_textrmsh (v4hi, int)
7311 int __builtin_arm_textrmsw (v2si, int)
7312 int __builtin_arm_textrmub (v8qi, int)
7313 int __builtin_arm_textrmuh (v4hi, int)
7314 int __builtin_arm_textrmuw (v2si, int)
7315 v8qi __builtin_arm_tinsrb (v8qi, int)
7316 v4hi __builtin_arm_tinsrh (v4hi, int)
7317 v2si __builtin_arm_tinsrw (v2si, int)
7318 long long __builtin_arm_tmia (long long, int, int)
7319 long long __builtin_arm_tmiabb (long long, int, int)
7320 long long __builtin_arm_tmiabt (long long, int, int)
7321 long long __builtin_arm_tmiaph (long long, int, int)
7322 long long __builtin_arm_tmiatb (long long, int, int)
7323 long long __builtin_arm_tmiatt (long long, int, int)
7324 int __builtin_arm_tmovmskb (v8qi)
7325 int __builtin_arm_tmovmskh (v4hi)
7326 int __builtin_arm_tmovmskw (v2si)
7327 long long __builtin_arm_waccb (v8qi)
7328 long long __builtin_arm_wacch (v4hi)
7329 long long __builtin_arm_waccw (v2si)
7330 v8qi __builtin_arm_waddb (v8qi, v8qi)
7331 v8qi __builtin_arm_waddbss (v8qi, v8qi)
7332 v8qi __builtin_arm_waddbus (v8qi, v8qi)
7333 v4hi __builtin_arm_waddh (v4hi, v4hi)
7334 v4hi __builtin_arm_waddhss (v4hi, v4hi)
7335 v4hi __builtin_arm_waddhus (v4hi, v4hi)
7336 v2si __builtin_arm_waddw (v2si, v2si)
7337 v2si __builtin_arm_waddwss (v2si, v2si)
7338 v2si __builtin_arm_waddwus (v2si, v2si)
7339 v8qi __builtin_arm_walign (v8qi, v8qi, int)
7340 long long __builtin_arm_wand(long long, long long)
7341 long long __builtin_arm_wandn (long long, long long)
7342 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
7343 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
7344 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
7345 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
7346 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
7347 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
7348 v2si __builtin_arm_wcmpeqw (v2si, v2si)
7349 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
7350 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
7351 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
7352 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
7353 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
7354 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
7355 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
7356 long long __builtin_arm_wmacsz (v4hi, v4hi)
7357 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
7358 long long __builtin_arm_wmacuz (v4hi, v4hi)
7359 v4hi __builtin_arm_wmadds (v4hi, v4hi)
7360 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
7361 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
7362 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
7363 v2si __builtin_arm_wmaxsw (v2si, v2si)
7364 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
7365 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
7366 v2si __builtin_arm_wmaxuw (v2si, v2si)
7367 v8qi __builtin_arm_wminsb (v8qi, v8qi)
7368 v4hi __builtin_arm_wminsh (v4hi, v4hi)
7369 v2si __builtin_arm_wminsw (v2si, v2si)
7370 v8qi __builtin_arm_wminub (v8qi, v8qi)
7371 v4hi __builtin_arm_wminuh (v4hi, v4hi)
7372 v2si __builtin_arm_wminuw (v2si, v2si)
7373 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
7374 v4hi __builtin_arm_wmulul (v4hi, v4hi)
7375 v4hi __builtin_arm_wmulum (v4hi, v4hi)
7376 long long __builtin_arm_wor (long long, long long)
7377 v2si __builtin_arm_wpackdss (long long, long long)
7378 v2si __builtin_arm_wpackdus (long long, long long)
7379 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
7380 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
7381 v4hi __builtin_arm_wpackwss (v2si, v2si)
7382 v4hi __builtin_arm_wpackwus (v2si, v2si)
7383 long long __builtin_arm_wrord (long long, long long)
7384 long long __builtin_arm_wrordi (long long, int)
7385 v4hi __builtin_arm_wrorh (v4hi, long long)
7386 v4hi __builtin_arm_wrorhi (v4hi, int)
7387 v2si __builtin_arm_wrorw (v2si, long long)
7388 v2si __builtin_arm_wrorwi (v2si, int)
7389 v2si __builtin_arm_wsadb (v8qi, v8qi)
7390 v2si __builtin_arm_wsadbz (v8qi, v8qi)
7391 v2si __builtin_arm_wsadh (v4hi, v4hi)
7392 v2si __builtin_arm_wsadhz (v4hi, v4hi)
7393 v4hi __builtin_arm_wshufh (v4hi, int)
7394 long long __builtin_arm_wslld (long long, long long)
7395 long long __builtin_arm_wslldi (long long, int)
7396 v4hi __builtin_arm_wsllh (v4hi, long long)
7397 v4hi __builtin_arm_wsllhi (v4hi, int)
7398 v2si __builtin_arm_wsllw (v2si, long long)
7399 v2si __builtin_arm_wsllwi (v2si, int)
7400 long long __builtin_arm_wsrad (long long, long long)
7401 long long __builtin_arm_wsradi (long long, int)
7402 v4hi __builtin_arm_wsrah (v4hi, long long)
7403 v4hi __builtin_arm_wsrahi (v4hi, int)
7404 v2si __builtin_arm_wsraw (v2si, long long)
7405 v2si __builtin_arm_wsrawi (v2si, int)
7406 long long __builtin_arm_wsrld (long long, long long)
7407 long long __builtin_arm_wsrldi (long long, int)
7408 v4hi __builtin_arm_wsrlh (v4hi, long long)
7409 v4hi __builtin_arm_wsrlhi (v4hi, int)
7410 v2si __builtin_arm_wsrlw (v2si, long long)
7411 v2si __builtin_arm_wsrlwi (v2si, int)
7412 v8qi __builtin_arm_wsubb (v8qi, v8qi)
7413 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
7414 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
7415 v4hi __builtin_arm_wsubh (v4hi, v4hi)
7416 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
7417 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
7418 v2si __builtin_arm_wsubw (v2si, v2si)
7419 v2si __builtin_arm_wsubwss (v2si, v2si)
7420 v2si __builtin_arm_wsubwus (v2si, v2si)
7421 v4hi __builtin_arm_wunpckehsb (v8qi)
7422 v2si __builtin_arm_wunpckehsh (v4hi)
7423 long long __builtin_arm_wunpckehsw (v2si)
7424 v4hi __builtin_arm_wunpckehub (v8qi)
7425 v2si __builtin_arm_wunpckehuh (v4hi)
7426 long long __builtin_arm_wunpckehuw (v2si)
7427 v4hi __builtin_arm_wunpckelsb (v8qi)
7428 v2si __builtin_arm_wunpckelsh (v4hi)
7429 long long __builtin_arm_wunpckelsw (v2si)
7430 v4hi __builtin_arm_wunpckelub (v8qi)
7431 v2si __builtin_arm_wunpckeluh (v4hi)
7432 long long __builtin_arm_wunpckeluw (v2si)
7433 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
7434 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
7435 v2si __builtin_arm_wunpckihw (v2si, v2si)
7436 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
7437 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
7438 v2si __builtin_arm_wunpckilw (v2si, v2si)
7439 long long __builtin_arm_wxor (long long, long long)
7440 long long __builtin_arm_wzero ()
7443 @node ARM NEON Intrinsics
7444 @subsection ARM NEON Intrinsics
7446 These built-in intrinsics for the ARM Advanced SIMD extension are available
7447 when the @option{-mfpu=neon} switch is used:
7449 @include arm-neon-intrinsics.texi
7451 @node Blackfin Built-in Functions
7452 @subsection Blackfin Built-in Functions
7454 Currently, there are two Blackfin-specific built-in functions. These are
7455 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
7456 using inline assembly; by using these built-in functions the compiler can
7457 automatically add workarounds for hardware errata involving these
7458 instructions. These functions are named as follows:
7461 void __builtin_bfin_csync (void)
7462 void __builtin_bfin_ssync (void)
7465 @node FR-V Built-in Functions
7466 @subsection FR-V Built-in Functions
7468 GCC provides many FR-V-specific built-in functions. In general,
7469 these functions are intended to be compatible with those described
7470 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
7471 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
7472 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
7473 pointer rather than by value.
7475 Most of the functions are named after specific FR-V instructions.
7476 Such functions are said to be ``directly mapped'' and are summarized
7477 here in tabular form.
7481 * Directly-mapped Integer Functions::
7482 * Directly-mapped Media Functions::
7483 * Raw read/write Functions::
7484 * Other Built-in Functions::
7487 @node Argument Types
7488 @subsubsection Argument Types
7490 The arguments to the built-in functions can be divided into three groups:
7491 register numbers, compile-time constants and run-time values. In order
7492 to make this classification clear at a glance, the arguments and return
7493 values are given the following pseudo types:
7495 @multitable @columnfractions .20 .30 .15 .35
7496 @item Pseudo type @tab Real C type @tab Constant? @tab Description
7497 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
7498 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
7499 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
7500 @item @code{uw2} @tab @code{unsigned long long} @tab No
7501 @tab an unsigned doubleword
7502 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
7503 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
7504 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
7505 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
7508 These pseudo types are not defined by GCC, they are simply a notational
7509 convenience used in this manual.
7511 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
7512 and @code{sw2} are evaluated at run time. They correspond to
7513 register operands in the underlying FR-V instructions.
7515 @code{const} arguments represent immediate operands in the underlying
7516 FR-V instructions. They must be compile-time constants.
7518 @code{acc} arguments are evaluated at compile time and specify the number
7519 of an accumulator register. For example, an @code{acc} argument of 2
7520 will select the ACC2 register.
7522 @code{iacc} arguments are similar to @code{acc} arguments but specify the
7523 number of an IACC register. See @pxref{Other Built-in Functions}
7526 @node Directly-mapped Integer Functions
7527 @subsubsection Directly-mapped Integer Functions
7529 The functions listed below map directly to FR-V I-type instructions.
7531 @multitable @columnfractions .45 .32 .23
7532 @item Function prototype @tab Example usage @tab Assembly output
7533 @item @code{sw1 __ADDSS (sw1, sw1)}
7534 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
7535 @tab @code{ADDSS @var{a},@var{b},@var{c}}
7536 @item @code{sw1 __SCAN (sw1, sw1)}
7537 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
7538 @tab @code{SCAN @var{a},@var{b},@var{c}}
7539 @item @code{sw1 __SCUTSS (sw1)}
7540 @tab @code{@var{b} = __SCUTSS (@var{a})}
7541 @tab @code{SCUTSS @var{a},@var{b}}
7542 @item @code{sw1 __SLASS (sw1, sw1)}
7543 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
7544 @tab @code{SLASS @var{a},@var{b},@var{c}}
7545 @item @code{void __SMASS (sw1, sw1)}
7546 @tab @code{__SMASS (@var{a}, @var{b})}
7547 @tab @code{SMASS @var{a},@var{b}}
7548 @item @code{void __SMSSS (sw1, sw1)}
7549 @tab @code{__SMSSS (@var{a}, @var{b})}
7550 @tab @code{SMSSS @var{a},@var{b}}
7551 @item @code{void __SMU (sw1, sw1)}
7552 @tab @code{__SMU (@var{a}, @var{b})}
7553 @tab @code{SMU @var{a},@var{b}}
7554 @item @code{sw2 __SMUL (sw1, sw1)}
7555 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
7556 @tab @code{SMUL @var{a},@var{b},@var{c}}
7557 @item @code{sw1 __SUBSS (sw1, sw1)}
7558 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
7559 @tab @code{SUBSS @var{a},@var{b},@var{c}}
7560 @item @code{uw2 __UMUL (uw1, uw1)}
7561 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
7562 @tab @code{UMUL @var{a},@var{b},@var{c}}
7565 @node Directly-mapped Media Functions
7566 @subsubsection Directly-mapped Media Functions
7568 The functions listed below map directly to FR-V M-type instructions.
7570 @multitable @columnfractions .45 .32 .23
7571 @item Function prototype @tab Example usage @tab Assembly output
7572 @item @code{uw1 __MABSHS (sw1)}
7573 @tab @code{@var{b} = __MABSHS (@var{a})}
7574 @tab @code{MABSHS @var{a},@var{b}}
7575 @item @code{void __MADDACCS (acc, acc)}
7576 @tab @code{__MADDACCS (@var{b}, @var{a})}
7577 @tab @code{MADDACCS @var{a},@var{b}}
7578 @item @code{sw1 __MADDHSS (sw1, sw1)}
7579 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
7580 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
7581 @item @code{uw1 __MADDHUS (uw1, uw1)}
7582 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
7583 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
7584 @item @code{uw1 __MAND (uw1, uw1)}
7585 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
7586 @tab @code{MAND @var{a},@var{b},@var{c}}
7587 @item @code{void __MASACCS (acc, acc)}
7588 @tab @code{__MASACCS (@var{b}, @var{a})}
7589 @tab @code{MASACCS @var{a},@var{b}}
7590 @item @code{uw1 __MAVEH (uw1, uw1)}
7591 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
7592 @tab @code{MAVEH @var{a},@var{b},@var{c}}
7593 @item @code{uw2 __MBTOH (uw1)}
7594 @tab @code{@var{b} = __MBTOH (@var{a})}
7595 @tab @code{MBTOH @var{a},@var{b}}
7596 @item @code{void __MBTOHE (uw1 *, uw1)}
7597 @tab @code{__MBTOHE (&@var{b}, @var{a})}
7598 @tab @code{MBTOHE @var{a},@var{b}}
7599 @item @code{void __MCLRACC (acc)}
7600 @tab @code{__MCLRACC (@var{a})}
7601 @tab @code{MCLRACC @var{a}}
7602 @item @code{void __MCLRACCA (void)}
7603 @tab @code{__MCLRACCA ()}
7604 @tab @code{MCLRACCA}
7605 @item @code{uw1 __Mcop1 (uw1, uw1)}
7606 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
7607 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
7608 @item @code{uw1 __Mcop2 (uw1, uw1)}
7609 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
7610 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
7611 @item @code{uw1 __MCPLHI (uw2, const)}
7612 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
7613 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
7614 @item @code{uw1 __MCPLI (uw2, const)}
7615 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
7616 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
7617 @item @code{void __MCPXIS (acc, sw1, sw1)}
7618 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
7619 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
7620 @item @code{void __MCPXIU (acc, uw1, uw1)}
7621 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
7622 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
7623 @item @code{void __MCPXRS (acc, sw1, sw1)}
7624 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
7625 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
7626 @item @code{void __MCPXRU (acc, uw1, uw1)}
7627 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
7628 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
7629 @item @code{uw1 __MCUT (acc, uw1)}
7630 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
7631 @tab @code{MCUT @var{a},@var{b},@var{c}}
7632 @item @code{uw1 __MCUTSS (acc, sw1)}
7633 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
7634 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
7635 @item @code{void __MDADDACCS (acc, acc)}
7636 @tab @code{__MDADDACCS (@var{b}, @var{a})}
7637 @tab @code{MDADDACCS @var{a},@var{b}}
7638 @item @code{void __MDASACCS (acc, acc)}
7639 @tab @code{__MDASACCS (@var{b}, @var{a})}
7640 @tab @code{MDASACCS @var{a},@var{b}}
7641 @item @code{uw2 __MDCUTSSI (acc, const)}
7642 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
7643 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
7644 @item @code{uw2 __MDPACKH (uw2, uw2)}
7645 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
7646 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
7647 @item @code{uw2 __MDROTLI (uw2, const)}
7648 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
7649 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
7650 @item @code{void __MDSUBACCS (acc, acc)}
7651 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
7652 @tab @code{MDSUBACCS @var{a},@var{b}}
7653 @item @code{void __MDUNPACKH (uw1 *, uw2)}
7654 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
7655 @tab @code{MDUNPACKH @var{a},@var{b}}
7656 @item @code{uw2 __MEXPDHD (uw1, const)}
7657 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
7658 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
7659 @item @code{uw1 __MEXPDHW (uw1, const)}
7660 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
7661 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
7662 @item @code{uw1 __MHDSETH (uw1, const)}
7663 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
7664 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
7665 @item @code{sw1 __MHDSETS (const)}
7666 @tab @code{@var{b} = __MHDSETS (@var{a})}
7667 @tab @code{MHDSETS #@var{a},@var{b}}
7668 @item @code{uw1 __MHSETHIH (uw1, const)}
7669 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
7670 @tab @code{MHSETHIH #@var{a},@var{b}}
7671 @item @code{sw1 __MHSETHIS (sw1, const)}
7672 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
7673 @tab @code{MHSETHIS #@var{a},@var{b}}
7674 @item @code{uw1 __MHSETLOH (uw1, const)}
7675 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
7676 @tab @code{MHSETLOH #@var{a},@var{b}}
7677 @item @code{sw1 __MHSETLOS (sw1, const)}
7678 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
7679 @tab @code{MHSETLOS #@var{a},@var{b}}
7680 @item @code{uw1 __MHTOB (uw2)}
7681 @tab @code{@var{b} = __MHTOB (@var{a})}
7682 @tab @code{MHTOB @var{a},@var{b}}
7683 @item @code{void __MMACHS (acc, sw1, sw1)}
7684 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
7685 @tab @code{MMACHS @var{a},@var{b},@var{c}}
7686 @item @code{void __MMACHU (acc, uw1, uw1)}
7687 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
7688 @tab @code{MMACHU @var{a},@var{b},@var{c}}
7689 @item @code{void __MMRDHS (acc, sw1, sw1)}
7690 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
7691 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
7692 @item @code{void __MMRDHU (acc, uw1, uw1)}
7693 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
7694 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
7695 @item @code{void __MMULHS (acc, sw1, sw1)}
7696 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
7697 @tab @code{MMULHS @var{a},@var{b},@var{c}}
7698 @item @code{void __MMULHU (acc, uw1, uw1)}
7699 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
7700 @tab @code{MMULHU @var{a},@var{b},@var{c}}
7701 @item @code{void __MMULXHS (acc, sw1, sw1)}
7702 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
7703 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
7704 @item @code{void __MMULXHU (acc, uw1, uw1)}
7705 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
7706 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
7707 @item @code{uw1 __MNOT (uw1)}
7708 @tab @code{@var{b} = __MNOT (@var{a})}
7709 @tab @code{MNOT @var{a},@var{b}}
7710 @item @code{uw1 __MOR (uw1, uw1)}
7711 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
7712 @tab @code{MOR @var{a},@var{b},@var{c}}
7713 @item @code{uw1 __MPACKH (uh, uh)}
7714 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
7715 @tab @code{MPACKH @var{a},@var{b},@var{c}}
7716 @item @code{sw2 __MQADDHSS (sw2, sw2)}
7717 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
7718 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
7719 @item @code{uw2 __MQADDHUS (uw2, uw2)}
7720 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
7721 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
7722 @item @code{void __MQCPXIS (acc, sw2, sw2)}
7723 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
7724 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
7725 @item @code{void __MQCPXIU (acc, uw2, uw2)}
7726 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
7727 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
7728 @item @code{void __MQCPXRS (acc, sw2, sw2)}
7729 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
7730 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
7731 @item @code{void __MQCPXRU (acc, uw2, uw2)}
7732 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
7733 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
7734 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
7735 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
7736 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
7737 @item @code{sw2 __MQLMTHS (sw2, sw2)}
7738 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
7739 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
7740 @item @code{void __MQMACHS (acc, sw2, sw2)}
7741 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
7742 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
7743 @item @code{void __MQMACHU (acc, uw2, uw2)}
7744 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
7745 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
7746 @item @code{void __MQMACXHS (acc, sw2, sw2)}
7747 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
7748 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
7749 @item @code{void __MQMULHS (acc, sw2, sw2)}
7750 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
7751 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
7752 @item @code{void __MQMULHU (acc, uw2, uw2)}
7753 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
7754 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
7755 @item @code{void __MQMULXHS (acc, sw2, sw2)}
7756 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
7757 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
7758 @item @code{void __MQMULXHU (acc, uw2, uw2)}
7759 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
7760 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
7761 @item @code{sw2 __MQSATHS (sw2, sw2)}
7762 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
7763 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
7764 @item @code{uw2 __MQSLLHI (uw2, int)}
7765 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
7766 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
7767 @item @code{sw2 __MQSRAHI (sw2, int)}
7768 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
7769 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
7770 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
7771 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
7772 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
7773 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
7774 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
7775 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
7776 @item @code{void __MQXMACHS (acc, sw2, sw2)}
7777 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
7778 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
7779 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
7780 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
7781 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
7782 @item @code{uw1 __MRDACC (acc)}
7783 @tab @code{@var{b} = __MRDACC (@var{a})}
7784 @tab @code{MRDACC @var{a},@var{b}}
7785 @item @code{uw1 __MRDACCG (acc)}
7786 @tab @code{@var{b} = __MRDACCG (@var{a})}
7787 @tab @code{MRDACCG @var{a},@var{b}}
7788 @item @code{uw1 __MROTLI (uw1, const)}
7789 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
7790 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
7791 @item @code{uw1 __MROTRI (uw1, const)}
7792 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
7793 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
7794 @item @code{sw1 __MSATHS (sw1, sw1)}
7795 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
7796 @tab @code{MSATHS @var{a},@var{b},@var{c}}
7797 @item @code{uw1 __MSATHU (uw1, uw1)}
7798 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
7799 @tab @code{MSATHU @var{a},@var{b},@var{c}}
7800 @item @code{uw1 __MSLLHI (uw1, const)}
7801 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
7802 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
7803 @item @code{sw1 __MSRAHI (sw1, const)}
7804 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
7805 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
7806 @item @code{uw1 __MSRLHI (uw1, const)}
7807 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
7808 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
7809 @item @code{void __MSUBACCS (acc, acc)}
7810 @tab @code{__MSUBACCS (@var{b}, @var{a})}
7811 @tab @code{MSUBACCS @var{a},@var{b}}
7812 @item @code{sw1 __MSUBHSS (sw1, sw1)}
7813 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
7814 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
7815 @item @code{uw1 __MSUBHUS (uw1, uw1)}
7816 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
7817 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
7818 @item @code{void __MTRAP (void)}
7819 @tab @code{__MTRAP ()}
7821 @item @code{uw2 __MUNPACKH (uw1)}
7822 @tab @code{@var{b} = __MUNPACKH (@var{a})}
7823 @tab @code{MUNPACKH @var{a},@var{b}}
7824 @item @code{uw1 __MWCUT (uw2, uw1)}
7825 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
7826 @tab @code{MWCUT @var{a},@var{b},@var{c}}
7827 @item @code{void __MWTACC (acc, uw1)}
7828 @tab @code{__MWTACC (@var{b}, @var{a})}
7829 @tab @code{MWTACC @var{a},@var{b}}
7830 @item @code{void __MWTACCG (acc, uw1)}
7831 @tab @code{__MWTACCG (@var{b}, @var{a})}
7832 @tab @code{MWTACCG @var{a},@var{b}}
7833 @item @code{uw1 __MXOR (uw1, uw1)}
7834 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
7835 @tab @code{MXOR @var{a},@var{b},@var{c}}
7838 @node Raw read/write Functions
7839 @subsubsection Raw read/write Functions
7841 This sections describes built-in functions related to read and write
7842 instructions to access memory. These functions generate
7843 @code{membar} instructions to flush the I/O load and stores where
7844 appropriate, as described in Fujitsu's manual described above.
7848 @item unsigned char __builtin_read8 (void *@var{data})
7849 @item unsigned short __builtin_read16 (void *@var{data})
7850 @item unsigned long __builtin_read32 (void *@var{data})
7851 @item unsigned long long __builtin_read64 (void *@var{data})
7853 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
7854 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
7855 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
7856 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
7859 @node Other Built-in Functions
7860 @subsubsection Other Built-in Functions
7862 This section describes built-in functions that are not named after
7863 a specific FR-V instruction.
7866 @item sw2 __IACCreadll (iacc @var{reg})
7867 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
7868 for future expansion and must be 0.
7870 @item sw1 __IACCreadl (iacc @var{reg})
7871 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
7872 Other values of @var{reg} are rejected as invalid.
7874 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
7875 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
7876 is reserved for future expansion and must be 0.
7878 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7879 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7880 is 1. Other values of @var{reg} are rejected as invalid.
7882 @item void __data_prefetch0 (const void *@var{x})
7883 Use the @code{dcpl} instruction to load the contents of address @var{x}
7884 into the data cache.
7886 @item void __data_prefetch (const void *@var{x})
7887 Use the @code{nldub} instruction to load the contents of address @var{x}
7888 into the data cache. The instruction will be issued in slot I1@.
7891 @node X86 Built-in Functions
7892 @subsection X86 Built-in Functions
7894 These built-in functions are available for the i386 and x86-64 family
7895 of computers, depending on the command-line switches used.
7897 Note that, if you specify command-line switches such as @option{-msse},
7898 the compiler could use the extended instruction sets even if the built-ins
7899 are not used explicitly in the program. For this reason, applications
7900 which perform runtime CPU detection must compile separate files for each
7901 supported architecture, using the appropriate flags. In particular,
7902 the file containing the CPU detection code should be compiled without
7905 The following machine modes are available for use with MMX built-in functions
7906 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7907 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7908 vector of eight 8-bit integers. Some of the built-in functions operate on
7909 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
7911 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7912 of two 32-bit floating point values.
7914 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7915 floating point values. Some instructions use a vector of four 32-bit
7916 integers, these use @code{V4SI}. Finally, some instructions operate on an
7917 entire vector register, interpreting it as a 128-bit integer, these use mode
7920 In 64-bit mode, the x86-64 family of processors uses additional built-in
7921 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7922 floating point and @code{TC} 128-bit complex floating point values.
7924 The following floating point built-in functions are available in 64-bit
7925 mode. All of them implement the function that is part of the name.
7928 __float128 __builtin_fabsq (__float128)
7929 __float128 __builtin_copysignq (__float128, __float128)
7932 The following floating point built-in functions are made available in the
7936 @item __float128 __builtin_infq (void)
7937 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7938 @findex __builtin_infq
7940 @item __float128 __builtin_huge_valq (void)
7941 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
7942 @findex __builtin_huge_valq
7945 The following built-in functions are made available by @option{-mmmx}.
7946 All of them generate the machine instruction that is part of the name.
7949 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7950 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7951 v2si __builtin_ia32_paddd (v2si, v2si)
7952 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7953 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7954 v2si __builtin_ia32_psubd (v2si, v2si)
7955 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7956 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7957 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7958 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7959 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7960 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7961 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7962 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7963 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7964 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7965 di __builtin_ia32_pand (di, di)
7966 di __builtin_ia32_pandn (di,di)
7967 di __builtin_ia32_por (di, di)
7968 di __builtin_ia32_pxor (di, di)
7969 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7970 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7971 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7972 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7973 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7974 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7975 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7976 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7977 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7978 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7979 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7980 v2si __builtin_ia32_punpckldq (v2si, v2si)
7981 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7982 v4hi __builtin_ia32_packssdw (v2si, v2si)
7983 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7985 v4hi __builtin_ia32_psllw (v4hi, v4hi)
7986 v2si __builtin_ia32_pslld (v2si, v2si)
7987 v1di __builtin_ia32_psllq (v1di, v1di)
7988 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
7989 v2si __builtin_ia32_psrld (v2si, v2si)
7990 v1di __builtin_ia32_psrlq (v1di, v1di)
7991 v4hi __builtin_ia32_psraw (v4hi, v4hi)
7992 v2si __builtin_ia32_psrad (v2si, v2si)
7993 v4hi __builtin_ia32_psllwi (v4hi, int)
7994 v2si __builtin_ia32_pslldi (v2si, int)
7995 v1di __builtin_ia32_psllqi (v1di, int)
7996 v4hi __builtin_ia32_psrlwi (v4hi, int)
7997 v2si __builtin_ia32_psrldi (v2si, int)
7998 v1di __builtin_ia32_psrlqi (v1di, int)
7999 v4hi __builtin_ia32_psrawi (v4hi, int)
8000 v2si __builtin_ia32_psradi (v2si, int)
8004 The following built-in functions are made available either with
8005 @option{-msse}, or with a combination of @option{-m3dnow} and
8006 @option{-march=athlon}. All of them generate the machine
8007 instruction that is part of the name.
8010 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
8011 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
8012 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
8013 v1di __builtin_ia32_psadbw (v8qi, v8qi)
8014 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
8015 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
8016 v8qi __builtin_ia32_pminub (v8qi, v8qi)
8017 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
8018 int __builtin_ia32_pextrw (v4hi, int)
8019 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
8020 int __builtin_ia32_pmovmskb (v8qi)
8021 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
8022 void __builtin_ia32_movntq (di *, di)
8023 void __builtin_ia32_sfence (void)
8026 The following built-in functions are available when @option{-msse} is used.
8027 All of them generate the machine instruction that is part of the name.
8030 int __builtin_ia32_comieq (v4sf, v4sf)
8031 int __builtin_ia32_comineq (v4sf, v4sf)
8032 int __builtin_ia32_comilt (v4sf, v4sf)
8033 int __builtin_ia32_comile (v4sf, v4sf)
8034 int __builtin_ia32_comigt (v4sf, v4sf)
8035 int __builtin_ia32_comige (v4sf, v4sf)
8036 int __builtin_ia32_ucomieq (v4sf, v4sf)
8037 int __builtin_ia32_ucomineq (v4sf, v4sf)
8038 int __builtin_ia32_ucomilt (v4sf, v4sf)
8039 int __builtin_ia32_ucomile (v4sf, v4sf)
8040 int __builtin_ia32_ucomigt (v4sf, v4sf)
8041 int __builtin_ia32_ucomige (v4sf, v4sf)
8042 v4sf __builtin_ia32_addps (v4sf, v4sf)
8043 v4sf __builtin_ia32_subps (v4sf, v4sf)
8044 v4sf __builtin_ia32_mulps (v4sf, v4sf)
8045 v4sf __builtin_ia32_divps (v4sf, v4sf)
8046 v4sf __builtin_ia32_addss (v4sf, v4sf)
8047 v4sf __builtin_ia32_subss (v4sf, v4sf)
8048 v4sf __builtin_ia32_mulss (v4sf, v4sf)
8049 v4sf __builtin_ia32_divss (v4sf, v4sf)
8050 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
8051 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
8052 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
8053 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
8054 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
8055 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
8056 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
8057 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
8058 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
8059 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
8060 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
8061 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
8062 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
8063 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
8064 v4si __builtin_ia32_cmpless (v4sf, v4sf)
8065 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
8066 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
8067 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
8068 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
8069 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
8070 v4sf __builtin_ia32_maxps (v4sf, v4sf)
8071 v4sf __builtin_ia32_maxss (v4sf, v4sf)
8072 v4sf __builtin_ia32_minps (v4sf, v4sf)
8073 v4sf __builtin_ia32_minss (v4sf, v4sf)
8074 v4sf __builtin_ia32_andps (v4sf, v4sf)
8075 v4sf __builtin_ia32_andnps (v4sf, v4sf)
8076 v4sf __builtin_ia32_orps (v4sf, v4sf)
8077 v4sf __builtin_ia32_xorps (v4sf, v4sf)
8078 v4sf __builtin_ia32_movss (v4sf, v4sf)
8079 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
8080 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
8081 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
8082 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
8083 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
8084 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
8085 v2si __builtin_ia32_cvtps2pi (v4sf)
8086 int __builtin_ia32_cvtss2si (v4sf)
8087 v2si __builtin_ia32_cvttps2pi (v4sf)
8088 int __builtin_ia32_cvttss2si (v4sf)
8089 v4sf __builtin_ia32_rcpps (v4sf)
8090 v4sf __builtin_ia32_rsqrtps (v4sf)
8091 v4sf __builtin_ia32_sqrtps (v4sf)
8092 v4sf __builtin_ia32_rcpss (v4sf)
8093 v4sf __builtin_ia32_rsqrtss (v4sf)
8094 v4sf __builtin_ia32_sqrtss (v4sf)
8095 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
8096 void __builtin_ia32_movntps (float *, v4sf)
8097 int __builtin_ia32_movmskps (v4sf)
8100 The following built-in functions are available when @option{-msse} is used.
8103 @item v4sf __builtin_ia32_loadaps (float *)
8104 Generates the @code{movaps} machine instruction as a load from memory.
8105 @item void __builtin_ia32_storeaps (float *, v4sf)
8106 Generates the @code{movaps} machine instruction as a store to memory.
8107 @item v4sf __builtin_ia32_loadups (float *)
8108 Generates the @code{movups} machine instruction as a load from memory.
8109 @item void __builtin_ia32_storeups (float *, v4sf)
8110 Generates the @code{movups} machine instruction as a store to memory.
8111 @item v4sf __builtin_ia32_loadsss (float *)
8112 Generates the @code{movss} machine instruction as a load from memory.
8113 @item void __builtin_ia32_storess (float *, v4sf)
8114 Generates the @code{movss} machine instruction as a store to memory.
8115 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
8116 Generates the @code{movhps} machine instruction as a load from memory.
8117 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
8118 Generates the @code{movlps} machine instruction as a load from memory
8119 @item void __builtin_ia32_storehps (v2sf *, v4sf)
8120 Generates the @code{movhps} machine instruction as a store to memory.
8121 @item void __builtin_ia32_storelps (v2sf *, v4sf)
8122 Generates the @code{movlps} machine instruction as a store to memory.
8125 The following built-in functions are available when @option{-msse2} is used.
8126 All of them generate the machine instruction that is part of the name.
8129 int __builtin_ia32_comisdeq (v2df, v2df)
8130 int __builtin_ia32_comisdlt (v2df, v2df)
8131 int __builtin_ia32_comisdle (v2df, v2df)
8132 int __builtin_ia32_comisdgt (v2df, v2df)
8133 int __builtin_ia32_comisdge (v2df, v2df)
8134 int __builtin_ia32_comisdneq (v2df, v2df)
8135 int __builtin_ia32_ucomisdeq (v2df, v2df)
8136 int __builtin_ia32_ucomisdlt (v2df, v2df)
8137 int __builtin_ia32_ucomisdle (v2df, v2df)
8138 int __builtin_ia32_ucomisdgt (v2df, v2df)
8139 int __builtin_ia32_ucomisdge (v2df, v2df)
8140 int __builtin_ia32_ucomisdneq (v2df, v2df)
8141 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
8142 v2df __builtin_ia32_cmpltpd (v2df, v2df)
8143 v2df __builtin_ia32_cmplepd (v2df, v2df)
8144 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
8145 v2df __builtin_ia32_cmpgepd (v2df, v2df)
8146 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
8147 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
8148 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
8149 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
8150 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
8151 v2df __builtin_ia32_cmpngepd (v2df, v2df)
8152 v2df __builtin_ia32_cmpordpd (v2df, v2df)
8153 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
8154 v2df __builtin_ia32_cmpltsd (v2df, v2df)
8155 v2df __builtin_ia32_cmplesd (v2df, v2df)
8156 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
8157 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
8158 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
8159 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
8160 v2df __builtin_ia32_cmpordsd (v2df, v2df)
8161 v2di __builtin_ia32_paddq (v2di, v2di)
8162 v2di __builtin_ia32_psubq (v2di, v2di)
8163 v2df __builtin_ia32_addpd (v2df, v2df)
8164 v2df __builtin_ia32_subpd (v2df, v2df)
8165 v2df __builtin_ia32_mulpd (v2df, v2df)
8166 v2df __builtin_ia32_divpd (v2df, v2df)
8167 v2df __builtin_ia32_addsd (v2df, v2df)
8168 v2df __builtin_ia32_subsd (v2df, v2df)
8169 v2df __builtin_ia32_mulsd (v2df, v2df)
8170 v2df __builtin_ia32_divsd (v2df, v2df)
8171 v2df __builtin_ia32_minpd (v2df, v2df)
8172 v2df __builtin_ia32_maxpd (v2df, v2df)
8173 v2df __builtin_ia32_minsd (v2df, v2df)
8174 v2df __builtin_ia32_maxsd (v2df, v2df)
8175 v2df __builtin_ia32_andpd (v2df, v2df)
8176 v2df __builtin_ia32_andnpd (v2df, v2df)
8177 v2df __builtin_ia32_orpd (v2df, v2df)
8178 v2df __builtin_ia32_xorpd (v2df, v2df)
8179 v2df __builtin_ia32_movsd (v2df, v2df)
8180 v2df __builtin_ia32_unpckhpd (v2df, v2df)
8181 v2df __builtin_ia32_unpcklpd (v2df, v2df)
8182 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
8183 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
8184 v4si __builtin_ia32_paddd128 (v4si, v4si)
8185 v2di __builtin_ia32_paddq128 (v2di, v2di)
8186 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
8187 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
8188 v4si __builtin_ia32_psubd128 (v4si, v4si)
8189 v2di __builtin_ia32_psubq128 (v2di, v2di)
8190 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
8191 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
8192 v2di __builtin_ia32_pand128 (v2di, v2di)
8193 v2di __builtin_ia32_pandn128 (v2di, v2di)
8194 v2di __builtin_ia32_por128 (v2di, v2di)
8195 v2di __builtin_ia32_pxor128 (v2di, v2di)
8196 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
8197 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
8198 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
8199 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
8200 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
8201 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
8202 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
8203 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
8204 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
8205 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
8206 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
8207 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8208 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8209 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8210 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8211 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8212 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8213 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8214 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8215 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8216 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8217 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8218 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8219 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8220 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8221 v2df __builtin_ia32_loadupd (double *)
8222 void __builtin_ia32_storeupd (double *, v2df)
8223 v2df __builtin_ia32_loadhpd (v2df, double const *)
8224 v2df __builtin_ia32_loadlpd (v2df, double const *)
8225 int __builtin_ia32_movmskpd (v2df)
8226 int __builtin_ia32_pmovmskb128 (v16qi)
8227 void __builtin_ia32_movnti (int *, int)
8228 void __builtin_ia32_movntpd (double *, v2df)
8229 void __builtin_ia32_movntdq (v2df *, v2df)
8230 v4si __builtin_ia32_pshufd (v4si, int)
8231 v8hi __builtin_ia32_pshuflw (v8hi, int)
8232 v8hi __builtin_ia32_pshufhw (v8hi, int)
8233 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
8234 v2df __builtin_ia32_sqrtpd (v2df)
8235 v2df __builtin_ia32_sqrtsd (v2df)
8236 v2df __builtin_ia32_shufpd (v2df, v2df, int)
8237 v2df __builtin_ia32_cvtdq2pd (v4si)
8238 v4sf __builtin_ia32_cvtdq2ps (v4si)
8239 v4si __builtin_ia32_cvtpd2dq (v2df)
8240 v2si __builtin_ia32_cvtpd2pi (v2df)
8241 v4sf __builtin_ia32_cvtpd2ps (v2df)
8242 v4si __builtin_ia32_cvttpd2dq (v2df)
8243 v2si __builtin_ia32_cvttpd2pi (v2df)
8244 v2df __builtin_ia32_cvtpi2pd (v2si)
8245 int __builtin_ia32_cvtsd2si (v2df)
8246 int __builtin_ia32_cvttsd2si (v2df)
8247 long long __builtin_ia32_cvtsd2si64 (v2df)
8248 long long __builtin_ia32_cvttsd2si64 (v2df)
8249 v4si __builtin_ia32_cvtps2dq (v4sf)
8250 v2df __builtin_ia32_cvtps2pd (v4sf)
8251 v4si __builtin_ia32_cvttps2dq (v4sf)
8252 v2df __builtin_ia32_cvtsi2sd (v2df, int)
8253 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
8254 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
8255 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
8256 void __builtin_ia32_clflush (const void *)
8257 void __builtin_ia32_lfence (void)
8258 void __builtin_ia32_mfence (void)
8259 v16qi __builtin_ia32_loaddqu (const char *)
8260 void __builtin_ia32_storedqu (char *, v16qi)
8261 v1di __builtin_ia32_pmuludq (v2si, v2si)
8262 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
8263 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
8264 v4si __builtin_ia32_pslld128 (v4si, v4si)
8265 v2di __builtin_ia32_psllq128 (v2di, v2di)
8266 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
8267 v4si __builtin_ia32_psrld128 (v4si, v4si)
8268 v2di __builtin_ia32_psrlq128 (v2di, v2di)
8269 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
8270 v4si __builtin_ia32_psrad128 (v4si, v4si)
8271 v2di __builtin_ia32_pslldqi128 (v2di, int)
8272 v8hi __builtin_ia32_psllwi128 (v8hi, int)
8273 v4si __builtin_ia32_pslldi128 (v4si, int)
8274 v2di __builtin_ia32_psllqi128 (v2di, int)
8275 v2di __builtin_ia32_psrldqi128 (v2di, int)
8276 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
8277 v4si __builtin_ia32_psrldi128 (v4si, int)
8278 v2di __builtin_ia32_psrlqi128 (v2di, int)
8279 v8hi __builtin_ia32_psrawi128 (v8hi, int)
8280 v4si __builtin_ia32_psradi128 (v4si, int)
8281 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
8282 v2di __builtin_ia32_movq128 (v2di)
8285 The following built-in functions are available when @option{-msse3} is used.
8286 All of them generate the machine instruction that is part of the name.
8289 v2df __builtin_ia32_addsubpd (v2df, v2df)
8290 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
8291 v2df __builtin_ia32_haddpd (v2df, v2df)
8292 v4sf __builtin_ia32_haddps (v4sf, v4sf)
8293 v2df __builtin_ia32_hsubpd (v2df, v2df)
8294 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
8295 v16qi __builtin_ia32_lddqu (char const *)
8296 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
8297 v2df __builtin_ia32_movddup (v2df)
8298 v4sf __builtin_ia32_movshdup (v4sf)
8299 v4sf __builtin_ia32_movsldup (v4sf)
8300 void __builtin_ia32_mwait (unsigned int, unsigned int)
8303 The following built-in functions are available when @option{-msse3} is used.
8306 @item v2df __builtin_ia32_loadddup (double const *)
8307 Generates the @code{movddup} machine instruction as a load from memory.
8310 The following built-in functions are available when @option{-mssse3} is used.
8311 All of them generate the machine instruction that is part of the name
8315 v2si __builtin_ia32_phaddd (v2si, v2si)
8316 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
8317 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
8318 v2si __builtin_ia32_phsubd (v2si, v2si)
8319 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
8320 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
8321 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
8322 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
8323 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
8324 v8qi __builtin_ia32_psignb (v8qi, v8qi)
8325 v2si __builtin_ia32_psignd (v2si, v2si)
8326 v4hi __builtin_ia32_psignw (v4hi, v4hi)
8327 v1di __builtin_ia32_palignr (v1di, v1di, int)
8328 v8qi __builtin_ia32_pabsb (v8qi)
8329 v2si __builtin_ia32_pabsd (v2si)
8330 v4hi __builtin_ia32_pabsw (v4hi)
8333 The following built-in functions are available when @option{-mssse3} is used.
8334 All of them generate the machine instruction that is part of the name
8338 v4si __builtin_ia32_phaddd128 (v4si, v4si)
8339 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
8340 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
8341 v4si __builtin_ia32_phsubd128 (v4si, v4si)
8342 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
8343 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
8344 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
8345 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
8346 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
8347 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
8348 v4si __builtin_ia32_psignd128 (v4si, v4si)
8349 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
8350 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
8351 v16qi __builtin_ia32_pabsb128 (v16qi)
8352 v4si __builtin_ia32_pabsd128 (v4si)
8353 v8hi __builtin_ia32_pabsw128 (v8hi)
8356 The following built-in functions are available when @option{-msse4.1} is
8357 used. All of them generate the machine instruction that is part of the
8361 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
8362 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
8363 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
8364 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
8365 v2df __builtin_ia32_dppd (v2df, v2df, const int)
8366 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
8367 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
8368 v2di __builtin_ia32_movntdqa (v2di *);
8369 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
8370 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
8371 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
8372 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
8373 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
8374 v8hi __builtin_ia32_phminposuw128 (v8hi)
8375 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
8376 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
8377 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
8378 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
8379 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
8380 v4si __builtin_ia32_pminsd128 (v4si, v4si)
8381 v4si __builtin_ia32_pminud128 (v4si, v4si)
8382 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
8383 v4si __builtin_ia32_pmovsxbd128 (v16qi)
8384 v2di __builtin_ia32_pmovsxbq128 (v16qi)
8385 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
8386 v2di __builtin_ia32_pmovsxdq128 (v4si)
8387 v4si __builtin_ia32_pmovsxwd128 (v8hi)
8388 v2di __builtin_ia32_pmovsxwq128 (v8hi)
8389 v4si __builtin_ia32_pmovzxbd128 (v16qi)
8390 v2di __builtin_ia32_pmovzxbq128 (v16qi)
8391 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
8392 v2di __builtin_ia32_pmovzxdq128 (v4si)
8393 v4si __builtin_ia32_pmovzxwd128 (v8hi)
8394 v2di __builtin_ia32_pmovzxwq128 (v8hi)
8395 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
8396 v4si __builtin_ia32_pmulld128 (v4si, v4si)
8397 int __builtin_ia32_ptestc128 (v2di, v2di)
8398 int __builtin_ia32_ptestnzc128 (v2di, v2di)
8399 int __builtin_ia32_ptestz128 (v2di, v2di)
8400 v2df __builtin_ia32_roundpd (v2df, const int)
8401 v4sf __builtin_ia32_roundps (v4sf, const int)
8402 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
8403 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
8406 The following built-in functions are available when @option{-msse4.1} is
8410 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
8411 Generates the @code{insertps} machine instruction.
8412 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
8413 Generates the @code{pextrb} machine instruction.
8414 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
8415 Generates the @code{pinsrb} machine instruction.
8416 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
8417 Generates the @code{pinsrd} machine instruction.
8418 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
8419 Generates the @code{pinsrq} machine instruction in 64bit mode.
8422 The following built-in functions are changed to generate new SSE4.1
8423 instructions when @option{-msse4.1} is used.
8426 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
8427 Generates the @code{extractps} machine instruction.
8428 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
8429 Generates the @code{pextrd} machine instruction.
8430 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
8431 Generates the @code{pextrq} machine instruction in 64bit mode.
8434 The following built-in functions are available when @option{-msse4.2} is
8435 used. All of them generate the machine instruction that is part of the
8439 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
8440 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
8441 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
8442 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
8443 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
8444 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
8445 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
8446 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
8447 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
8448 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
8449 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
8450 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
8451 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
8452 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
8453 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
8456 The following built-in functions are available when @option{-msse4.2} is
8460 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
8461 Generates the @code{crc32b} machine instruction.
8462 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
8463 Generates the @code{crc32w} machine instruction.
8464 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
8465 Generates the @code{crc32l} machine instruction.
8466 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
8467 Generates the @code{crc32q} machine instruction.
8470 The following built-in functions are changed to generate new SSE4.2
8471 instructions when @option{-msse4.2} is used.
8474 @item int __builtin_popcount (unsigned int)
8475 Generates the @code{popcntl} machine instruction.
8476 @item int __builtin_popcountl (unsigned long)
8477 Generates the @code{popcntl} or @code{popcntq} machine instruction,
8478 depending on the size of @code{unsigned long}.
8479 @item int __builtin_popcountll (unsigned long long)
8480 Generates the @code{popcntq} machine instruction.
8483 The following built-in functions are available when @option{-mavx} is
8484 used. All of them generate the machine instruction that is part of the
8488 v4df __builtin_ia32_addpd256 (v4df,v4df)
8489 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
8490 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
8491 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
8492 v4df __builtin_ia32_andnpd256 (v4df,v4df)
8493 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
8494 v4df __builtin_ia32_andpd256 (v4df,v4df)
8495 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
8496 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
8497 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
8498 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
8499 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
8500 v2df __builtin_ia32_cmppd (v2df,v2df,int)
8501 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
8502 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
8503 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
8504 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
8505 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
8506 v4df __builtin_ia32_cvtdq2pd256 (v4si)
8507 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
8508 v4si __builtin_ia32_cvtpd2dq256 (v4df)
8509 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
8510 v8si __builtin_ia32_cvtps2dq256 (v8sf)
8511 v4df __builtin_ia32_cvtps2pd256 (v4sf)
8512 v4si __builtin_ia32_cvttpd2dq256 (v4df)
8513 v8si __builtin_ia32_cvttps2dq256 (v8sf)
8514 v4df __builtin_ia32_divpd256 (v4df,v4df)
8515 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
8516 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
8517 v4df __builtin_ia32_haddpd256 (v4df,v4df)
8518 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
8519 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
8520 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
8521 v32qi __builtin_ia32_lddqu256 (pcchar)
8522 v32qi __builtin_ia32_loaddqu256 (pcchar)
8523 v4df __builtin_ia32_loadupd256 (pcdouble)
8524 v8sf __builtin_ia32_loadups256 (pcfloat)
8525 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
8526 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
8527 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
8528 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
8529 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
8530 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
8531 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
8532 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
8533 v4df __builtin_ia32_maxpd256 (v4df,v4df)
8534 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
8535 v4df __builtin_ia32_minpd256 (v4df,v4df)
8536 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
8537 v4df __builtin_ia32_movddup256 (v4df)
8538 int __builtin_ia32_movmskpd256 (v4df)
8539 int __builtin_ia32_movmskps256 (v8sf)
8540 v8sf __builtin_ia32_movshdup256 (v8sf)
8541 v8sf __builtin_ia32_movsldup256 (v8sf)
8542 v4df __builtin_ia32_mulpd256 (v4df,v4df)
8543 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
8544 v4df __builtin_ia32_orpd256 (v4df,v4df)
8545 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
8546 v2df __builtin_ia32_pd_pd256 (v4df)
8547 v4df __builtin_ia32_pd256_pd (v2df)
8548 v4sf __builtin_ia32_ps_ps256 (v8sf)
8549 v8sf __builtin_ia32_ps256_ps (v4sf)
8550 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
8551 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
8552 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
8553 v8sf __builtin_ia32_rcpps256 (v8sf)
8554 v4df __builtin_ia32_roundpd256 (v4df,int)
8555 v8sf __builtin_ia32_roundps256 (v8sf,int)
8556 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
8557 v8sf __builtin_ia32_rsqrtps256 (v8sf)
8558 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
8559 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
8560 v4si __builtin_ia32_si_si256 (v8si)
8561 v8si __builtin_ia32_si256_si (v4si)
8562 v4df __builtin_ia32_sqrtpd256 (v4df)
8563 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
8564 v8sf __builtin_ia32_sqrtps256 (v8sf)
8565 void __builtin_ia32_storedqu256 (pchar,v32qi)
8566 void __builtin_ia32_storeupd256 (pdouble,v4df)
8567 void __builtin_ia32_storeups256 (pfloat,v8sf)
8568 v4df __builtin_ia32_subpd256 (v4df,v4df)
8569 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
8570 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
8571 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
8572 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
8573 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
8574 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
8575 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
8576 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
8577 v4sf __builtin_ia32_vbroadcastss (pcfloat)
8578 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
8579 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
8580 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
8581 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
8582 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
8583 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
8584 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
8585 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
8586 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
8587 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
8588 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
8589 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
8590 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
8591 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
8592 v2df __builtin_ia32_vpermilpd (v2df,int)
8593 v4df __builtin_ia32_vpermilpd256 (v4df,int)
8594 v4sf __builtin_ia32_vpermilps (v4sf,int)
8595 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
8596 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
8597 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
8598 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
8599 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
8600 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
8601 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
8602 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
8603 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
8604 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
8605 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
8606 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
8607 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
8608 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
8609 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
8610 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
8611 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
8612 void __builtin_ia32_vzeroall (void)
8613 void __builtin_ia32_vzeroupper (void)
8614 v4df __builtin_ia32_xorpd256 (v4df,v4df)
8615 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
8618 The following built-in functions are available when @option{-maes} is
8619 used. All of them generate the machine instruction that is part of the
8623 v2di __builtin_ia32_aesenc128 (v2di, v2di)
8624 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
8625 v2di __builtin_ia32_aesdec128 (v2di, v2di)
8626 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
8627 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
8628 v2di __builtin_ia32_aesimc128 (v2di)
8631 The following built-in function is available when @option{-mpclmul} is
8635 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
8636 Generates the @code{pclmulqdq} machine instruction.
8639 The following built-in functions are available when @option{-msse4a} is used.
8640 All of them generate the machine instruction that is part of the name.
8643 void __builtin_ia32_movntsd (double *, v2df)
8644 void __builtin_ia32_movntss (float *, v4sf)
8645 v2di __builtin_ia32_extrq (v2di, v16qi)
8646 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
8647 v2di __builtin_ia32_insertq (v2di, v2di)
8648 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
8651 The following built-in functions are available when @option{-msse5} is used.
8652 All of them generate the machine instruction that is part of the name
8656 v2df __builtin_ia32_comeqpd (v2df, v2df)
8657 v2df __builtin_ia32_comeqps (v2df, v2df)
8658 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
8659 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
8660 v2df __builtin_ia32_comfalsepd (v2df, v2df)
8661 v2df __builtin_ia32_comfalseps (v2df, v2df)
8662 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
8663 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
8664 v2df __builtin_ia32_comgepd (v2df, v2df)
8665 v2df __builtin_ia32_comgeps (v2df, v2df)
8666 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
8667 v4sf __builtin_ia32_comgess (v4sf, v4sf)
8668 v2df __builtin_ia32_comgtpd (v2df, v2df)
8669 v2df __builtin_ia32_comgtps (v2df, v2df)
8670 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
8671 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
8672 v2df __builtin_ia32_comlepd (v2df, v2df)
8673 v2df __builtin_ia32_comleps (v2df, v2df)
8674 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
8675 v4sf __builtin_ia32_comless (v4sf, v4sf)
8676 v2df __builtin_ia32_comltpd (v2df, v2df)
8677 v2df __builtin_ia32_comltps (v2df, v2df)
8678 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
8679 v4sf __builtin_ia32_comltss (v4sf, v4sf)
8680 v2df __builtin_ia32_comnepd (v2df, v2df)
8681 v2df __builtin_ia32_comneps (v2df, v2df)
8682 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
8683 v4sf __builtin_ia32_comness (v4sf, v4sf)
8684 v2df __builtin_ia32_comordpd (v2df, v2df)
8685 v2df __builtin_ia32_comordps (v2df, v2df)
8686 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
8687 v4sf __builtin_ia32_comordss (v4sf, v4sf)
8688 v2df __builtin_ia32_comtruepd (v2df, v2df)
8689 v2df __builtin_ia32_comtrueps (v2df, v2df)
8690 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
8691 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
8692 v2df __builtin_ia32_comueqpd (v2df, v2df)
8693 v2df __builtin_ia32_comueqps (v2df, v2df)
8694 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
8695 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
8696 v2df __builtin_ia32_comugepd (v2df, v2df)
8697 v2df __builtin_ia32_comugeps (v2df, v2df)
8698 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
8699 v4sf __builtin_ia32_comugess (v4sf, v4sf)
8700 v2df __builtin_ia32_comugtpd (v2df, v2df)
8701 v2df __builtin_ia32_comugtps (v2df, v2df)
8702 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
8703 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
8704 v2df __builtin_ia32_comulepd (v2df, v2df)
8705 v2df __builtin_ia32_comuleps (v2df, v2df)
8706 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
8707 v4sf __builtin_ia32_comuless (v4sf, v4sf)
8708 v2df __builtin_ia32_comultpd (v2df, v2df)
8709 v2df __builtin_ia32_comultps (v2df, v2df)
8710 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
8711 v4sf __builtin_ia32_comultss (v4sf, v4sf)
8712 v2df __builtin_ia32_comunepd (v2df, v2df)
8713 v2df __builtin_ia32_comuneps (v2df, v2df)
8714 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
8715 v4sf __builtin_ia32_comuness (v4sf, v4sf)
8716 v2df __builtin_ia32_comunordpd (v2df, v2df)
8717 v2df __builtin_ia32_comunordps (v2df, v2df)
8718 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
8719 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
8720 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
8721 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
8722 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
8723 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
8724 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
8725 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
8726 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
8727 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
8728 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
8729 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
8730 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
8731 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
8732 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
8733 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
8734 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
8735 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
8736 v2df __builtin_ia32_frczpd (v2df)
8737 v4sf __builtin_ia32_frczps (v4sf)
8738 v2df __builtin_ia32_frczsd (v2df, v2df)
8739 v4sf __builtin_ia32_frczss (v4sf, v4sf)
8740 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
8741 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
8742 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
8743 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
8744 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
8745 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
8746 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
8747 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
8748 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8749 v4si __builtin_ia32_pcomeqd (v4si, v4si)
8750 v2di __builtin_ia32_pcomeqq (v2di, v2di)
8751 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
8752 v4si __builtin_ia32_pcomequd (v4si, v4si)
8753 v2di __builtin_ia32_pcomequq (v2di, v2di)
8754 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
8755 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8756 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
8757 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
8758 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
8759 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
8760 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
8761 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
8762 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
8763 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
8764 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
8765 v4si __builtin_ia32_pcomged (v4si, v4si)
8766 v2di __builtin_ia32_pcomgeq (v2di, v2di)
8767 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
8768 v4si __builtin_ia32_pcomgeud (v4si, v4si)
8769 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
8770 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
8771 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
8772 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
8773 v4si __builtin_ia32_pcomgtd (v4si, v4si)
8774 v2di __builtin_ia32_pcomgtq (v2di, v2di)
8775 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
8776 v4si __builtin_ia32_pcomgtud (v4si, v4si)
8777 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
8778 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
8779 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
8780 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
8781 v4si __builtin_ia32_pcomled (v4si, v4si)
8782 v2di __builtin_ia32_pcomleq (v2di, v2di)
8783 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
8784 v4si __builtin_ia32_pcomleud (v4si, v4si)
8785 v2di __builtin_ia32_pcomleuq (v2di, v2di)
8786 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
8787 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
8788 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
8789 v4si __builtin_ia32_pcomltd (v4si, v4si)
8790 v2di __builtin_ia32_pcomltq (v2di, v2di)
8791 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
8792 v4si __builtin_ia32_pcomltud (v4si, v4si)
8793 v2di __builtin_ia32_pcomltuq (v2di, v2di)
8794 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
8795 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
8796 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
8797 v4si __builtin_ia32_pcomned (v4si, v4si)
8798 v2di __builtin_ia32_pcomneq (v2di, v2di)
8799 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
8800 v4si __builtin_ia32_pcomneud (v4si, v4si)
8801 v2di __builtin_ia32_pcomneuq (v2di, v2di)
8802 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
8803 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
8804 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
8805 v4si __builtin_ia32_pcomtrued (v4si, v4si)
8806 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
8807 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
8808 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
8809 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
8810 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
8811 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
8812 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
8813 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
8814 v4si __builtin_ia32_phaddbd (v16qi)
8815 v2di __builtin_ia32_phaddbq (v16qi)
8816 v8hi __builtin_ia32_phaddbw (v16qi)
8817 v2di __builtin_ia32_phadddq (v4si)
8818 v4si __builtin_ia32_phaddubd (v16qi)
8819 v2di __builtin_ia32_phaddubq (v16qi)
8820 v8hi __builtin_ia32_phaddubw (v16qi)
8821 v2di __builtin_ia32_phaddudq (v4si)
8822 v4si __builtin_ia32_phadduwd (v8hi)
8823 v2di __builtin_ia32_phadduwq (v8hi)
8824 v4si __builtin_ia32_phaddwd (v8hi)
8825 v2di __builtin_ia32_phaddwq (v8hi)
8826 v8hi __builtin_ia32_phsubbw (v16qi)
8827 v2di __builtin_ia32_phsubdq (v4si)
8828 v4si __builtin_ia32_phsubwd (v8hi)
8829 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
8830 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
8831 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
8832 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
8833 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
8834 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
8835 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
8836 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
8837 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
8838 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
8839 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
8840 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
8841 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
8842 v16qi __builtin_ia32_protb (v16qi, v16qi)
8843 v4si __builtin_ia32_protd (v4si, v4si)
8844 v2di __builtin_ia32_protq (v2di, v2di)
8845 v8hi __builtin_ia32_protw (v8hi, v8hi)
8846 v16qi __builtin_ia32_pshab (v16qi, v16qi)
8847 v4si __builtin_ia32_pshad (v4si, v4si)
8848 v2di __builtin_ia32_pshaq (v2di, v2di)
8849 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
8850 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
8851 v4si __builtin_ia32_pshld (v4si, v4si)
8852 v2di __builtin_ia32_pshlq (v2di, v2di)
8853 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
8856 The following builtin-in functions are available when @option{-msse5}
8857 is used. The second argument must be an integer constant and generate
8858 the machine instruction that is part of the name with the @samp{_imm}
8862 v16qi __builtin_ia32_protb_imm (v16qi, int)
8863 v4si __builtin_ia32_protd_imm (v4si, int)
8864 v2di __builtin_ia32_protq_imm (v2di, int)
8865 v8hi __builtin_ia32_protw_imm (v8hi, int)
8868 The following built-in functions are available when @option{-m3dnow} is used.
8869 All of them generate the machine instruction that is part of the name.
8872 void __builtin_ia32_femms (void)
8873 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
8874 v2si __builtin_ia32_pf2id (v2sf)
8875 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
8876 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
8877 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
8878 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
8879 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
8880 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
8881 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
8882 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
8883 v2sf __builtin_ia32_pfrcp (v2sf)
8884 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
8885 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
8886 v2sf __builtin_ia32_pfrsqrt (v2sf)
8887 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
8888 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
8889 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
8890 v2sf __builtin_ia32_pi2fd (v2si)
8891 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
8894 The following built-in functions are available when both @option{-m3dnow}
8895 and @option{-march=athlon} are used. All of them generate the machine
8896 instruction that is part of the name.
8899 v2si __builtin_ia32_pf2iw (v2sf)
8900 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
8901 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
8902 v2sf __builtin_ia32_pi2fw (v2si)
8903 v2sf __builtin_ia32_pswapdsf (v2sf)
8904 v2si __builtin_ia32_pswapdsi (v2si)
8907 @node MIPS DSP Built-in Functions
8908 @subsection MIPS DSP Built-in Functions
8910 The MIPS DSP Application-Specific Extension (ASE) includes new
8911 instructions that are designed to improve the performance of DSP and
8912 media applications. It provides instructions that operate on packed
8913 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
8915 GCC supports MIPS DSP operations using both the generic
8916 vector extensions (@pxref{Vector Extensions}) and a collection of
8917 MIPS-specific built-in functions. Both kinds of support are
8918 enabled by the @option{-mdsp} command-line option.
8920 Revision 2 of the ASE was introduced in the second half of 2006.
8921 This revision adds extra instructions to the original ASE, but is
8922 otherwise backwards-compatible with it. You can select revision 2
8923 using the command-line option @option{-mdspr2}; this option implies
8926 The SCOUNT and POS bits of the DSP control register are global. The
8927 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
8928 POS bits. During optimization, the compiler will not delete these
8929 instructions and it will not delete calls to functions containing
8932 At present, GCC only provides support for operations on 32-bit
8933 vectors. The vector type associated with 8-bit integer data is
8934 usually called @code{v4i8}, the vector type associated with Q7
8935 is usually called @code{v4q7}, the vector type associated with 16-bit
8936 integer data is usually called @code{v2i16}, and the vector type
8937 associated with Q15 is usually called @code{v2q15}. They can be
8938 defined in C as follows:
8941 typedef signed char v4i8 __attribute__ ((vector_size(4)));
8942 typedef signed char v4q7 __attribute__ ((vector_size(4)));
8943 typedef short v2i16 __attribute__ ((vector_size(4)));
8944 typedef short v2q15 __attribute__ ((vector_size(4)));
8947 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
8948 initialized in the same way as aggregates. For example:
8951 v4i8 a = @{1, 2, 3, 4@};
8953 b = (v4i8) @{5, 6, 7, 8@};
8955 v2q15 c = @{0x0fcb, 0x3a75@};
8957 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
8960 @emph{Note:} The CPU's endianness determines the order in which values
8961 are packed. On little-endian targets, the first value is the least
8962 significant and the last value is the most significant. The opposite
8963 order applies to big-endian targets. For example, the code above will
8964 set the lowest byte of @code{a} to @code{1} on little-endian targets
8965 and @code{4} on big-endian targets.
8967 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
8968 representation. As shown in this example, the integer representation
8969 of a Q7 value can be obtained by multiplying the fractional value by
8970 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
8971 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
8974 The table below lists the @code{v4i8} and @code{v2q15} operations for which
8975 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
8976 and @code{c} and @code{d} are @code{v2q15} values.
8978 @multitable @columnfractions .50 .50
8979 @item C code @tab MIPS instruction
8980 @item @code{a + b} @tab @code{addu.qb}
8981 @item @code{c + d} @tab @code{addq.ph}
8982 @item @code{a - b} @tab @code{subu.qb}
8983 @item @code{c - d} @tab @code{subq.ph}
8986 The table below lists the @code{v2i16} operation for which
8987 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
8988 @code{v2i16} values.
8990 @multitable @columnfractions .50 .50
8991 @item C code @tab MIPS instruction
8992 @item @code{e * f} @tab @code{mul.ph}
8995 It is easier to describe the DSP built-in functions if we first define
8996 the following types:
9001 typedef unsigned int ui32;
9002 typedef long long a64;
9005 @code{q31} and @code{i32} are actually the same as @code{int}, but we
9006 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
9007 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
9008 @code{long long}, but we use @code{a64} to indicate values that will
9009 be placed in one of the four DSP accumulators (@code{$ac0},
9010 @code{$ac1}, @code{$ac2} or @code{$ac3}).
9012 Also, some built-in functions prefer or require immediate numbers as
9013 parameters, because the corresponding DSP instructions accept both immediate
9014 numbers and register operands, or accept immediate numbers only. The
9015 immediate parameters are listed as follows.
9024 imm_n32_31: -32 to 31.
9025 imm_n512_511: -512 to 511.
9028 The following built-in functions map directly to a particular MIPS DSP
9029 instruction. Please refer to the architecture specification
9030 for details on what each instruction does.
9033 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
9034 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
9035 q31 __builtin_mips_addq_s_w (q31, q31)
9036 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
9037 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
9038 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
9039 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
9040 q31 __builtin_mips_subq_s_w (q31, q31)
9041 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
9042 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
9043 i32 __builtin_mips_addsc (i32, i32)
9044 i32 __builtin_mips_addwc (i32, i32)
9045 i32 __builtin_mips_modsub (i32, i32)
9046 i32 __builtin_mips_raddu_w_qb (v4i8)
9047 v2q15 __builtin_mips_absq_s_ph (v2q15)
9048 q31 __builtin_mips_absq_s_w (q31)
9049 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
9050 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
9051 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
9052 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
9053 q31 __builtin_mips_preceq_w_phl (v2q15)
9054 q31 __builtin_mips_preceq_w_phr (v2q15)
9055 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
9056 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
9057 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
9058 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
9059 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
9060 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
9061 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
9062 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
9063 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
9064 v4i8 __builtin_mips_shll_qb (v4i8, i32)
9065 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
9066 v2q15 __builtin_mips_shll_ph (v2q15, i32)
9067 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
9068 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
9069 q31 __builtin_mips_shll_s_w (q31, imm0_31)
9070 q31 __builtin_mips_shll_s_w (q31, i32)
9071 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
9072 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
9073 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
9074 v2q15 __builtin_mips_shra_ph (v2q15, i32)
9075 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
9076 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
9077 q31 __builtin_mips_shra_r_w (q31, imm0_31)
9078 q31 __builtin_mips_shra_r_w (q31, i32)
9079 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
9080 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
9081 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
9082 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
9083 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
9084 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
9085 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
9086 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
9087 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
9088 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
9089 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
9090 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
9091 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
9092 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
9093 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
9094 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
9095 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
9096 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
9097 i32 __builtin_mips_bitrev (i32)
9098 i32 __builtin_mips_insv (i32, i32)
9099 v4i8 __builtin_mips_repl_qb (imm0_255)
9100 v4i8 __builtin_mips_repl_qb (i32)
9101 v2q15 __builtin_mips_repl_ph (imm_n512_511)
9102 v2q15 __builtin_mips_repl_ph (i32)
9103 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
9104 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
9105 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
9106 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
9107 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
9108 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
9109 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
9110 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
9111 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
9112 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
9113 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
9114 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
9115 i32 __builtin_mips_extr_w (a64, imm0_31)
9116 i32 __builtin_mips_extr_w (a64, i32)
9117 i32 __builtin_mips_extr_r_w (a64, imm0_31)
9118 i32 __builtin_mips_extr_s_h (a64, i32)
9119 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
9120 i32 __builtin_mips_extr_rs_w (a64, i32)
9121 i32 __builtin_mips_extr_s_h (a64, imm0_31)
9122 i32 __builtin_mips_extr_r_w (a64, i32)
9123 i32 __builtin_mips_extp (a64, imm0_31)
9124 i32 __builtin_mips_extp (a64, i32)
9125 i32 __builtin_mips_extpdp (a64, imm0_31)
9126 i32 __builtin_mips_extpdp (a64, i32)
9127 a64 __builtin_mips_shilo (a64, imm_n32_31)
9128 a64 __builtin_mips_shilo (a64, i32)
9129 a64 __builtin_mips_mthlip (a64, i32)
9130 void __builtin_mips_wrdsp (i32, imm0_63)
9131 i32 __builtin_mips_rddsp (imm0_63)
9132 i32 __builtin_mips_lbux (void *, i32)
9133 i32 __builtin_mips_lhx (void *, i32)
9134 i32 __builtin_mips_lwx (void *, i32)
9135 i32 __builtin_mips_bposge32 (void)
9138 The following built-in functions map directly to a particular MIPS DSP REV 2
9139 instruction. Please refer to the architecture specification
9140 for details on what each instruction does.
9143 v4q7 __builtin_mips_absq_s_qb (v4q7);
9144 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
9145 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
9146 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
9147 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
9148 i32 __builtin_mips_append (i32, i32, imm0_31);
9149 i32 __builtin_mips_balign (i32, i32, imm0_3);
9150 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
9151 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
9152 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
9153 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
9154 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
9155 a64 __builtin_mips_madd (a64, i32, i32);
9156 a64 __builtin_mips_maddu (a64, ui32, ui32);
9157 a64 __builtin_mips_msub (a64, i32, i32);
9158 a64 __builtin_mips_msubu (a64, ui32, ui32);
9159 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
9160 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
9161 q31 __builtin_mips_mulq_rs_w (q31, q31);
9162 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
9163 q31 __builtin_mips_mulq_s_w (q31, q31);
9164 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
9165 a64 __builtin_mips_mult (i32, i32);
9166 a64 __builtin_mips_multu (ui32, ui32);
9167 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
9168 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
9169 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
9170 i32 __builtin_mips_prepend (i32, i32, imm0_31);
9171 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
9172 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
9173 v4i8 __builtin_mips_shra_qb (v4i8, i32);
9174 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
9175 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
9176 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
9177 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
9178 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
9179 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
9180 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
9181 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
9182 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
9183 q31 __builtin_mips_addqh_w (q31, q31);
9184 q31 __builtin_mips_addqh_r_w (q31, q31);
9185 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
9186 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
9187 q31 __builtin_mips_subqh_w (q31, q31);
9188 q31 __builtin_mips_subqh_r_w (q31, q31);
9189 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
9190 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
9191 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
9192 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
9193 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
9194 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
9198 @node MIPS Paired-Single Support
9199 @subsection MIPS Paired-Single Support
9201 The MIPS64 architecture includes a number of instructions that
9202 operate on pairs of single-precision floating-point values.
9203 Each pair is packed into a 64-bit floating-point register,
9204 with one element being designated the ``upper half'' and
9205 the other being designated the ``lower half''.
9207 GCC supports paired-single operations using both the generic
9208 vector extensions (@pxref{Vector Extensions}) and a collection of
9209 MIPS-specific built-in functions. Both kinds of support are
9210 enabled by the @option{-mpaired-single} command-line option.
9212 The vector type associated with paired-single values is usually
9213 called @code{v2sf}. It can be defined in C as follows:
9216 typedef float v2sf __attribute__ ((vector_size (8)));
9219 @code{v2sf} values are initialized in the same way as aggregates.
9223 v2sf a = @{1.5, 9.1@};
9226 b = (v2sf) @{e, f@};
9229 @emph{Note:} The CPU's endianness determines which value is stored in
9230 the upper half of a register and which value is stored in the lower half.
9231 On little-endian targets, the first value is the lower one and the second
9232 value is the upper one. The opposite order applies to big-endian targets.
9233 For example, the code above will set the lower half of @code{a} to
9234 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
9236 @node MIPS Loongson Built-in Functions
9237 @subsection MIPS Loongson Built-in Functions
9239 GCC provides intrinsics to access the SIMD instructions provided by the
9240 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
9241 available after inclusion of the @code{loongson.h} header file,
9242 operate on the following 64-bit vector types:
9245 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
9246 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
9247 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
9248 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
9249 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
9250 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
9253 The intrinsics provided are listed below; each is named after the
9254 machine instruction to which it corresponds, with suffixes added as
9255 appropriate to distinguish intrinsics that expand to the same machine
9256 instruction yet have different argument types. Refer to the architecture
9257 documentation for a description of the functionality of each
9261 int16x4_t packsswh (int32x2_t s, int32x2_t t);
9262 int8x8_t packsshb (int16x4_t s, int16x4_t t);
9263 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
9264 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
9265 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
9266 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
9267 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
9268 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
9269 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
9270 uint64_t paddd_u (uint64_t s, uint64_t t);
9271 int64_t paddd_s (int64_t s, int64_t t);
9272 int16x4_t paddsh (int16x4_t s, int16x4_t t);
9273 int8x8_t paddsb (int8x8_t s, int8x8_t t);
9274 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
9275 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
9276 uint64_t pandn_ud (uint64_t s, uint64_t t);
9277 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
9278 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
9279 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
9280 int64_t pandn_sd (int64_t s, int64_t t);
9281 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
9282 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
9283 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
9284 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
9285 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
9286 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
9287 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
9288 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
9289 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
9290 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
9291 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
9292 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
9293 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
9294 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
9295 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
9296 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
9297 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
9298 uint16x4_t pextrh_u (uint16x4_t s, int field);
9299 int16x4_t pextrh_s (int16x4_t s, int field);
9300 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
9301 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
9302 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
9303 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
9304 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
9305 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
9306 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
9307 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
9308 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
9309 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
9310 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
9311 int16x4_t pminsh (int16x4_t s, int16x4_t t);
9312 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
9313 uint8x8_t pmovmskb_u (uint8x8_t s);
9314 int8x8_t pmovmskb_s (int8x8_t s);
9315 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
9316 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
9317 int16x4_t pmullh (int16x4_t s, int16x4_t t);
9318 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
9319 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
9320 uint16x4_t biadd (uint8x8_t s);
9321 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
9322 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
9323 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
9324 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
9325 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
9326 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
9327 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
9328 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
9329 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
9330 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
9331 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
9332 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
9333 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
9334 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
9335 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
9336 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
9337 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
9338 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
9339 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
9340 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
9341 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
9342 uint64_t psubd_u (uint64_t s, uint64_t t);
9343 int64_t psubd_s (int64_t s, int64_t t);
9344 int16x4_t psubsh (int16x4_t s, int16x4_t t);
9345 int8x8_t psubsb (int8x8_t s, int8x8_t t);
9346 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
9347 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
9348 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
9349 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
9350 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
9351 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
9352 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
9353 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
9354 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
9355 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
9356 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
9357 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
9358 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
9359 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
9363 * Paired-Single Arithmetic::
9364 * Paired-Single Built-in Functions::
9365 * MIPS-3D Built-in Functions::
9368 @node Paired-Single Arithmetic
9369 @subsubsection Paired-Single Arithmetic
9371 The table below lists the @code{v2sf} operations for which hardware
9372 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
9373 values and @code{x} is an integral value.
9375 @multitable @columnfractions .50 .50
9376 @item C code @tab MIPS instruction
9377 @item @code{a + b} @tab @code{add.ps}
9378 @item @code{a - b} @tab @code{sub.ps}
9379 @item @code{-a} @tab @code{neg.ps}
9380 @item @code{a * b} @tab @code{mul.ps}
9381 @item @code{a * b + c} @tab @code{madd.ps}
9382 @item @code{a * b - c} @tab @code{msub.ps}
9383 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
9384 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
9385 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
9388 Note that the multiply-accumulate instructions can be disabled
9389 using the command-line option @code{-mno-fused-madd}.
9391 @node Paired-Single Built-in Functions
9392 @subsubsection Paired-Single Built-in Functions
9394 The following paired-single functions map directly to a particular
9395 MIPS instruction. Please refer to the architecture specification
9396 for details on what each instruction does.
9399 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
9400 Pair lower lower (@code{pll.ps}).
9402 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
9403 Pair upper lower (@code{pul.ps}).
9405 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
9406 Pair lower upper (@code{plu.ps}).
9408 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
9409 Pair upper upper (@code{puu.ps}).
9411 @item v2sf __builtin_mips_cvt_ps_s (float, float)
9412 Convert pair to paired single (@code{cvt.ps.s}).
9414 @item float __builtin_mips_cvt_s_pl (v2sf)
9415 Convert pair lower to single (@code{cvt.s.pl}).
9417 @item float __builtin_mips_cvt_s_pu (v2sf)
9418 Convert pair upper to single (@code{cvt.s.pu}).
9420 @item v2sf __builtin_mips_abs_ps (v2sf)
9421 Absolute value (@code{abs.ps}).
9423 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
9424 Align variable (@code{alnv.ps}).
9426 @emph{Note:} The value of the third parameter must be 0 or 4
9427 modulo 8, otherwise the result will be unpredictable. Please read the
9428 instruction description for details.
9431 The following multi-instruction functions are also available.
9432 In each case, @var{cond} can be any of the 16 floating-point conditions:
9433 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9434 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
9435 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9438 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9439 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9440 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
9441 @code{movt.ps}/@code{movf.ps}).
9443 The @code{movt} functions return the value @var{x} computed by:
9446 c.@var{cond}.ps @var{cc},@var{a},@var{b}
9447 mov.ps @var{x},@var{c}
9448 movt.ps @var{x},@var{d},@var{cc}
9451 The @code{movf} functions are similar but use @code{movf.ps} instead
9454 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9455 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9456 Comparison of two paired-single values (@code{c.@var{cond}.ps},
9457 @code{bc1t}/@code{bc1f}).
9459 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9460 and return either the upper or lower half of the result. For example:
9464 if (__builtin_mips_upper_c_eq_ps (a, b))
9465 upper_halves_are_equal ();
9467 upper_halves_are_unequal ();
9469 if (__builtin_mips_lower_c_eq_ps (a, b))
9470 lower_halves_are_equal ();
9472 lower_halves_are_unequal ();
9476 @node MIPS-3D Built-in Functions
9477 @subsubsection MIPS-3D Built-in Functions
9479 The MIPS-3D Application-Specific Extension (ASE) includes additional
9480 paired-single instructions that are designed to improve the performance
9481 of 3D graphics operations. Support for these instructions is controlled
9482 by the @option{-mips3d} command-line option.
9484 The functions listed below map directly to a particular MIPS-3D
9485 instruction. Please refer to the architecture specification for
9486 more details on what each instruction does.
9489 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
9490 Reduction add (@code{addr.ps}).
9492 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
9493 Reduction multiply (@code{mulr.ps}).
9495 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
9496 Convert paired single to paired word (@code{cvt.pw.ps}).
9498 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
9499 Convert paired word to paired single (@code{cvt.ps.pw}).
9501 @item float __builtin_mips_recip1_s (float)
9502 @itemx double __builtin_mips_recip1_d (double)
9503 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
9504 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
9506 @item float __builtin_mips_recip2_s (float, float)
9507 @itemx double __builtin_mips_recip2_d (double, double)
9508 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
9509 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
9511 @item float __builtin_mips_rsqrt1_s (float)
9512 @itemx double __builtin_mips_rsqrt1_d (double)
9513 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
9514 Reduced precision reciprocal square root (sequence step 1)
9515 (@code{rsqrt1.@var{fmt}}).
9517 @item float __builtin_mips_rsqrt2_s (float, float)
9518 @itemx double __builtin_mips_rsqrt2_d (double, double)
9519 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
9520 Reduced precision reciprocal square root (sequence step 2)
9521 (@code{rsqrt2.@var{fmt}}).
9524 The following multi-instruction functions are also available.
9525 In each case, @var{cond} can be any of the 16 floating-point conditions:
9526 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9527 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
9528 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9531 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
9532 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
9533 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
9534 @code{bc1t}/@code{bc1f}).
9536 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
9537 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
9542 if (__builtin_mips_cabs_eq_s (a, b))
9548 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9549 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9550 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
9551 @code{bc1t}/@code{bc1f}).
9553 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
9554 and return either the upper or lower half of the result. For example:
9558 if (__builtin_mips_upper_cabs_eq_ps (a, b))
9559 upper_halves_are_equal ();
9561 upper_halves_are_unequal ();
9563 if (__builtin_mips_lower_cabs_eq_ps (a, b))
9564 lower_halves_are_equal ();
9566 lower_halves_are_unequal ();
9569 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9570 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9571 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
9572 @code{movt.ps}/@code{movf.ps}).
9574 The @code{movt} functions return the value @var{x} computed by:
9577 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
9578 mov.ps @var{x},@var{c}
9579 movt.ps @var{x},@var{d},@var{cc}
9582 The @code{movf} functions are similar but use @code{movf.ps} instead
9585 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9586 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9587 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9588 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9589 Comparison of two paired-single values
9590 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9591 @code{bc1any2t}/@code{bc1any2f}).
9593 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9594 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
9595 result is true and the @code{all} forms return true if both results are true.
9600 if (__builtin_mips_any_c_eq_ps (a, b))
9605 if (__builtin_mips_all_c_eq_ps (a, b))
9611 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9612 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9613 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9614 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9615 Comparison of four paired-single values
9616 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9617 @code{bc1any4t}/@code{bc1any4f}).
9619 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
9620 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
9621 The @code{any} forms return true if any of the four results are true
9622 and the @code{all} forms return true if all four results are true.
9627 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
9632 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
9639 @node picoChip Built-in Functions
9640 @subsection picoChip Built-in Functions
9642 GCC provides an interface to selected machine instructions from the
9643 picoChip instruction set.
9646 @item int __builtin_sbc (int @var{value})
9647 Sign bit count. Return the number of consecutive bits in @var{value}
9648 which have the same value as the sign-bit. The result is the number of
9649 leading sign bits minus one, giving the number of redundant sign bits in
9652 @item int __builtin_byteswap (int @var{value})
9653 Byte swap. Return the result of swapping the upper and lower bytes of
9656 @item int __builtin_brev (int @var{value})
9657 Bit reversal. Return the result of reversing the bits in
9658 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
9661 @item int __builtin_adds (int @var{x}, int @var{y})
9662 Saturating addition. Return the result of adding @var{x} and @var{y},
9663 storing the value 32767 if the result overflows.
9665 @item int __builtin_subs (int @var{x}, int @var{y})
9666 Saturating subtraction. Return the result of subtracting @var{y} from
9667 @var{x}, storing the value -32768 if the result overflows.
9669 @item void __builtin_halt (void)
9670 Halt. The processor will stop execution. This built-in is useful for
9671 implementing assertions.
9675 @node Other MIPS Built-in Functions
9676 @subsection Other MIPS Built-in Functions
9678 GCC provides other MIPS-specific built-in functions:
9681 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
9682 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
9683 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
9684 when this function is available.
9687 @node PowerPC AltiVec Built-in Functions
9688 @subsection PowerPC AltiVec Built-in Functions
9690 GCC provides an interface for the PowerPC family of processors to access
9691 the AltiVec operations described in Motorola's AltiVec Programming
9692 Interface Manual. The interface is made available by including
9693 @code{<altivec.h>} and using @option{-maltivec} and
9694 @option{-mabi=altivec}. The interface supports the following vector
9698 vector unsigned char
9702 vector unsigned short
9713 GCC's implementation of the high-level language interface available from
9714 C and C++ code differs from Motorola's documentation in several ways.
9719 A vector constant is a list of constant expressions within curly braces.
9722 A vector initializer requires no cast if the vector constant is of the
9723 same type as the variable it is initializing.
9726 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9727 vector type is the default signedness of the base type. The default
9728 varies depending on the operating system, so a portable program should
9729 always specify the signedness.
9732 Compiling with @option{-maltivec} adds keywords @code{__vector},
9733 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
9734 @code{bool}. When compiling ISO C, the context-sensitive substitution
9735 of the keywords @code{vector}, @code{pixel} and @code{bool} is
9736 disabled. To use them, you must include @code{<altivec.h>} instead.
9739 GCC allows using a @code{typedef} name as the type specifier for a
9743 For C, overloaded functions are implemented with macros so the following
9747 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9750 Since @code{vec_add} is a macro, the vector constant in the example
9751 is treated as four separate arguments. Wrap the entire argument in
9752 parentheses for this to work.
9755 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
9756 Internally, GCC uses built-in functions to achieve the functionality in
9757 the aforementioned header file, but they are not supported and are
9758 subject to change without notice.
9760 The following interfaces are supported for the generic and specific
9761 AltiVec operations and the AltiVec predicates. In cases where there
9762 is a direct mapping between generic and specific operations, only the
9763 generic names are shown here, although the specific operations can also
9766 Arguments that are documented as @code{const int} require literal
9767 integral values within the range required for that operation.
9770 vector signed char vec_abs (vector signed char);
9771 vector signed short vec_abs (vector signed short);
9772 vector signed int vec_abs (vector signed int);
9773 vector float vec_abs (vector float);
9775 vector signed char vec_abss (vector signed char);
9776 vector signed short vec_abss (vector signed short);
9777 vector signed int vec_abss (vector signed int);
9779 vector signed char vec_add (vector bool char, vector signed char);
9780 vector signed char vec_add (vector signed char, vector bool char);
9781 vector signed char vec_add (vector signed char, vector signed char);
9782 vector unsigned char vec_add (vector bool char, vector unsigned char);
9783 vector unsigned char vec_add (vector unsigned char, vector bool char);
9784 vector unsigned char vec_add (vector unsigned char,
9785 vector unsigned char);
9786 vector signed short vec_add (vector bool short, vector signed short);
9787 vector signed short vec_add (vector signed short, vector bool short);
9788 vector signed short vec_add (vector signed short, vector signed short);
9789 vector unsigned short vec_add (vector bool short,
9790 vector unsigned short);
9791 vector unsigned short vec_add (vector unsigned short,
9793 vector unsigned short vec_add (vector unsigned short,
9794 vector unsigned short);
9795 vector signed int vec_add (vector bool int, vector signed int);
9796 vector signed int vec_add (vector signed int, vector bool int);
9797 vector signed int vec_add (vector signed int, vector signed int);
9798 vector unsigned int vec_add (vector bool int, vector unsigned int);
9799 vector unsigned int vec_add (vector unsigned int, vector bool int);
9800 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
9801 vector float vec_add (vector float, vector float);
9803 vector float vec_vaddfp (vector float, vector float);
9805 vector signed int vec_vadduwm (vector bool int, vector signed int);
9806 vector signed int vec_vadduwm (vector signed int, vector bool int);
9807 vector signed int vec_vadduwm (vector signed int, vector signed int);
9808 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
9809 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
9810 vector unsigned int vec_vadduwm (vector unsigned int,
9811 vector unsigned int);
9813 vector signed short vec_vadduhm (vector bool short,
9814 vector signed short);
9815 vector signed short vec_vadduhm (vector signed short,
9817 vector signed short vec_vadduhm (vector signed short,
9818 vector signed short);
9819 vector unsigned short vec_vadduhm (vector bool short,
9820 vector unsigned short);
9821 vector unsigned short vec_vadduhm (vector unsigned short,
9823 vector unsigned short vec_vadduhm (vector unsigned short,
9824 vector unsigned short);
9826 vector signed char vec_vaddubm (vector bool char, vector signed char);
9827 vector signed char vec_vaddubm (vector signed char, vector bool char);
9828 vector signed char vec_vaddubm (vector signed char, vector signed char);
9829 vector unsigned char vec_vaddubm (vector bool char,
9830 vector unsigned char);
9831 vector unsigned char vec_vaddubm (vector unsigned char,
9833 vector unsigned char vec_vaddubm (vector unsigned char,
9834 vector unsigned char);
9836 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
9838 vector unsigned char vec_adds (vector bool char, vector unsigned char);
9839 vector unsigned char vec_adds (vector unsigned char, vector bool char);
9840 vector unsigned char vec_adds (vector unsigned char,
9841 vector unsigned char);
9842 vector signed char vec_adds (vector bool char, vector signed char);
9843 vector signed char vec_adds (vector signed char, vector bool char);
9844 vector signed char vec_adds (vector signed char, vector signed char);
9845 vector unsigned short vec_adds (vector bool short,
9846 vector unsigned short);
9847 vector unsigned short vec_adds (vector unsigned short,
9849 vector unsigned short vec_adds (vector unsigned short,
9850 vector unsigned short);
9851 vector signed short vec_adds (vector bool short, vector signed short);
9852 vector signed short vec_adds (vector signed short, vector bool short);
9853 vector signed short vec_adds (vector signed short, vector signed short);
9854 vector unsigned int vec_adds (vector bool int, vector unsigned int);
9855 vector unsigned int vec_adds (vector unsigned int, vector bool int);
9856 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
9857 vector signed int vec_adds (vector bool int, vector signed int);
9858 vector signed int vec_adds (vector signed int, vector bool int);
9859 vector signed int vec_adds (vector signed int, vector signed int);
9861 vector signed int vec_vaddsws (vector bool int, vector signed int);
9862 vector signed int vec_vaddsws (vector signed int, vector bool int);
9863 vector signed int vec_vaddsws (vector signed int, vector signed int);
9865 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
9866 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
9867 vector unsigned int vec_vadduws (vector unsigned int,
9868 vector unsigned int);
9870 vector signed short vec_vaddshs (vector bool short,
9871 vector signed short);
9872 vector signed short vec_vaddshs (vector signed short,
9874 vector signed short vec_vaddshs (vector signed short,
9875 vector signed short);
9877 vector unsigned short vec_vadduhs (vector bool short,
9878 vector unsigned short);
9879 vector unsigned short vec_vadduhs (vector unsigned short,
9881 vector unsigned short vec_vadduhs (vector unsigned short,
9882 vector unsigned short);
9884 vector signed char vec_vaddsbs (vector bool char, vector signed char);
9885 vector signed char vec_vaddsbs (vector signed char, vector bool char);
9886 vector signed char vec_vaddsbs (vector signed char, vector signed char);
9888 vector unsigned char vec_vaddubs (vector bool char,
9889 vector unsigned char);
9890 vector unsigned char vec_vaddubs (vector unsigned char,
9892 vector unsigned char vec_vaddubs (vector unsigned char,
9893 vector unsigned char);
9895 vector float vec_and (vector float, vector float);
9896 vector float vec_and (vector float, vector bool int);
9897 vector float vec_and (vector bool int, vector float);
9898 vector bool int vec_and (vector bool int, vector bool int);
9899 vector signed int vec_and (vector bool int, vector signed int);
9900 vector signed int vec_and (vector signed int, vector bool int);
9901 vector signed int vec_and (vector signed int, vector signed int);
9902 vector unsigned int vec_and (vector bool int, vector unsigned int);
9903 vector unsigned int vec_and (vector unsigned int, vector bool int);
9904 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
9905 vector bool short vec_and (vector bool short, vector bool short);
9906 vector signed short vec_and (vector bool short, vector signed short);
9907 vector signed short vec_and (vector signed short, vector bool short);
9908 vector signed short vec_and (vector signed short, vector signed short);
9909 vector unsigned short vec_and (vector bool short,
9910 vector unsigned short);
9911 vector unsigned short vec_and (vector unsigned short,
9913 vector unsigned short vec_and (vector unsigned short,
9914 vector unsigned short);
9915 vector signed char vec_and (vector bool char, vector signed char);
9916 vector bool char vec_and (vector bool char, vector bool char);
9917 vector signed char vec_and (vector signed char, vector bool char);
9918 vector signed char vec_and (vector signed char, vector signed char);
9919 vector unsigned char vec_and (vector bool char, vector unsigned char);
9920 vector unsigned char vec_and (vector unsigned char, vector bool char);
9921 vector unsigned char vec_and (vector unsigned char,
9922 vector unsigned char);
9924 vector float vec_andc (vector float, vector float);
9925 vector float vec_andc (vector float, vector bool int);
9926 vector float vec_andc (vector bool int, vector float);
9927 vector bool int vec_andc (vector bool int, vector bool int);
9928 vector signed int vec_andc (vector bool int, vector signed int);
9929 vector signed int vec_andc (vector signed int, vector bool int);
9930 vector signed int vec_andc (vector signed int, vector signed int);
9931 vector unsigned int vec_andc (vector bool int, vector unsigned int);
9932 vector unsigned int vec_andc (vector unsigned int, vector bool int);
9933 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
9934 vector bool short vec_andc (vector bool short, vector bool short);
9935 vector signed short vec_andc (vector bool short, vector signed short);
9936 vector signed short vec_andc (vector signed short, vector bool short);
9937 vector signed short vec_andc (vector signed short, vector signed short);
9938 vector unsigned short vec_andc (vector bool short,
9939 vector unsigned short);
9940 vector unsigned short vec_andc (vector unsigned short,
9942 vector unsigned short vec_andc (vector unsigned short,
9943 vector unsigned short);
9944 vector signed char vec_andc (vector bool char, vector signed char);
9945 vector bool char vec_andc (vector bool char, vector bool char);
9946 vector signed char vec_andc (vector signed char, vector bool char);
9947 vector signed char vec_andc (vector signed char, vector signed char);
9948 vector unsigned char vec_andc (vector bool char, vector unsigned char);
9949 vector unsigned char vec_andc (vector unsigned char, vector bool char);
9950 vector unsigned char vec_andc (vector unsigned char,
9951 vector unsigned char);
9953 vector unsigned char vec_avg (vector unsigned char,
9954 vector unsigned char);
9955 vector signed char vec_avg (vector signed char, vector signed char);
9956 vector unsigned short vec_avg (vector unsigned short,
9957 vector unsigned short);
9958 vector signed short vec_avg (vector signed short, vector signed short);
9959 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
9960 vector signed int vec_avg (vector signed int, vector signed int);
9962 vector signed int vec_vavgsw (vector signed int, vector signed int);
9964 vector unsigned int vec_vavguw (vector unsigned int,
9965 vector unsigned int);
9967 vector signed short vec_vavgsh (vector signed short,
9968 vector signed short);
9970 vector unsigned short vec_vavguh (vector unsigned short,
9971 vector unsigned short);
9973 vector signed char vec_vavgsb (vector signed char, vector signed char);
9975 vector unsigned char vec_vavgub (vector unsigned char,
9976 vector unsigned char);
9978 vector float vec_ceil (vector float);
9980 vector signed int vec_cmpb (vector float, vector float);
9982 vector bool char vec_cmpeq (vector signed char, vector signed char);
9983 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
9984 vector bool short vec_cmpeq (vector signed short, vector signed short);
9985 vector bool short vec_cmpeq (vector unsigned short,
9986 vector unsigned short);
9987 vector bool int vec_cmpeq (vector signed int, vector signed int);
9988 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
9989 vector bool int vec_cmpeq (vector float, vector float);
9991 vector bool int vec_vcmpeqfp (vector float, vector float);
9993 vector bool int vec_vcmpequw (vector signed int, vector signed int);
9994 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
9996 vector bool short vec_vcmpequh (vector signed short,
9997 vector signed short);
9998 vector bool short vec_vcmpequh (vector unsigned short,
9999 vector unsigned short);
10001 vector bool char vec_vcmpequb (vector signed char, vector signed char);
10002 vector bool char vec_vcmpequb (vector unsigned char,
10003 vector unsigned char);
10005 vector bool int vec_cmpge (vector float, vector float);
10007 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
10008 vector bool char vec_cmpgt (vector signed char, vector signed char);
10009 vector bool short vec_cmpgt (vector unsigned short,
10010 vector unsigned short);
10011 vector bool short vec_cmpgt (vector signed short, vector signed short);
10012 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
10013 vector bool int vec_cmpgt (vector signed int, vector signed int);
10014 vector bool int vec_cmpgt (vector float, vector float);
10016 vector bool int vec_vcmpgtfp (vector float, vector float);
10018 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
10020 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
10022 vector bool short vec_vcmpgtsh (vector signed short,
10023 vector signed short);
10025 vector bool short vec_vcmpgtuh (vector unsigned short,
10026 vector unsigned short);
10028 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
10030 vector bool char vec_vcmpgtub (vector unsigned char,
10031 vector unsigned char);
10033 vector bool int vec_cmple (vector float, vector float);
10035 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
10036 vector bool char vec_cmplt (vector signed char, vector signed char);
10037 vector bool short vec_cmplt (vector unsigned short,
10038 vector unsigned short);
10039 vector bool short vec_cmplt (vector signed short, vector signed short);
10040 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
10041 vector bool int vec_cmplt (vector signed int, vector signed int);
10042 vector bool int vec_cmplt (vector float, vector float);
10044 vector float vec_ctf (vector unsigned int, const int);
10045 vector float vec_ctf (vector signed int, const int);
10047 vector float vec_vcfsx (vector signed int, const int);
10049 vector float vec_vcfux (vector unsigned int, const int);
10051 vector signed int vec_cts (vector float, const int);
10053 vector unsigned int vec_ctu (vector float, const int);
10055 void vec_dss (const int);
10057 void vec_dssall (void);
10059 void vec_dst (const vector unsigned char *, int, const int);
10060 void vec_dst (const vector signed char *, int, const int);
10061 void vec_dst (const vector bool char *, int, const int);
10062 void vec_dst (const vector unsigned short *, int, const int);
10063 void vec_dst (const vector signed short *, int, const int);
10064 void vec_dst (const vector bool short *, int, const int);
10065 void vec_dst (const vector pixel *, int, const int);
10066 void vec_dst (const vector unsigned int *, int, const int);
10067 void vec_dst (const vector signed int *, int, const int);
10068 void vec_dst (const vector bool int *, int, const int);
10069 void vec_dst (const vector float *, int, const int);
10070 void vec_dst (const unsigned char *, int, const int);
10071 void vec_dst (const signed char *, int, const int);
10072 void vec_dst (const unsigned short *, int, const int);
10073 void vec_dst (const short *, int, const int);
10074 void vec_dst (const unsigned int *, int, const int);
10075 void vec_dst (const int *, int, const int);
10076 void vec_dst (const unsigned long *, int, const int);
10077 void vec_dst (const long *, int, const int);
10078 void vec_dst (const float *, int, const int);
10080 void vec_dstst (const vector unsigned char *, int, const int);
10081 void vec_dstst (const vector signed char *, int, const int);
10082 void vec_dstst (const vector bool char *, int, const int);
10083 void vec_dstst (const vector unsigned short *, int, const int);
10084 void vec_dstst (const vector signed short *, int, const int);
10085 void vec_dstst (const vector bool short *, int, const int);
10086 void vec_dstst (const vector pixel *, int, const int);
10087 void vec_dstst (const vector unsigned int *, int, const int);
10088 void vec_dstst (const vector signed int *, int, const int);
10089 void vec_dstst (const vector bool int *, int, const int);
10090 void vec_dstst (const vector float *, int, const int);
10091 void vec_dstst (const unsigned char *, int, const int);
10092 void vec_dstst (const signed char *, int, const int);
10093 void vec_dstst (const unsigned short *, int, const int);
10094 void vec_dstst (const short *, int, const int);
10095 void vec_dstst (const unsigned int *, int, const int);
10096 void vec_dstst (const int *, int, const int);
10097 void vec_dstst (const unsigned long *, int, const int);
10098 void vec_dstst (const long *, int, const int);
10099 void vec_dstst (const float *, int, const int);
10101 void vec_dststt (const vector unsigned char *, int, const int);
10102 void vec_dststt (const vector signed char *, int, const int);
10103 void vec_dststt (const vector bool char *, int, const int);
10104 void vec_dststt (const vector unsigned short *, int, const int);
10105 void vec_dststt (const vector signed short *, int, const int);
10106 void vec_dststt (const vector bool short *, int, const int);
10107 void vec_dststt (const vector pixel *, int, const int);
10108 void vec_dststt (const vector unsigned int *, int, const int);
10109 void vec_dststt (const vector signed int *, int, const int);
10110 void vec_dststt (const vector bool int *, int, const int);
10111 void vec_dststt (const vector float *, int, const int);
10112 void vec_dststt (const unsigned char *, int, const int);
10113 void vec_dststt (const signed char *, int, const int);
10114 void vec_dststt (const unsigned short *, int, const int);
10115 void vec_dststt (const short *, int, const int);
10116 void vec_dststt (const unsigned int *, int, const int);
10117 void vec_dststt (const int *, int, const int);
10118 void vec_dststt (const unsigned long *, int, const int);
10119 void vec_dststt (const long *, int, const int);
10120 void vec_dststt (const float *, int, const int);
10122 void vec_dstt (const vector unsigned char *, int, const int);
10123 void vec_dstt (const vector signed char *, int, const int);
10124 void vec_dstt (const vector bool char *, int, const int);
10125 void vec_dstt (const vector unsigned short *, int, const int);
10126 void vec_dstt (const vector signed short *, int, const int);
10127 void vec_dstt (const vector bool short *, int, const int);
10128 void vec_dstt (const vector pixel *, int, const int);
10129 void vec_dstt (const vector unsigned int *, int, const int);
10130 void vec_dstt (const vector signed int *, int, const int);
10131 void vec_dstt (const vector bool int *, int, const int);
10132 void vec_dstt (const vector float *, int, const int);
10133 void vec_dstt (const unsigned char *, int, const int);
10134 void vec_dstt (const signed char *, int, const int);
10135 void vec_dstt (const unsigned short *, int, const int);
10136 void vec_dstt (const short *, int, const int);
10137 void vec_dstt (const unsigned int *, int, const int);
10138 void vec_dstt (const int *, int, const int);
10139 void vec_dstt (const unsigned long *, int, const int);
10140 void vec_dstt (const long *, int, const int);
10141 void vec_dstt (const float *, int, const int);
10143 vector float vec_expte (vector float);
10145 vector float vec_floor (vector float);
10147 vector float vec_ld (int, const vector float *);
10148 vector float vec_ld (int, const float *);
10149 vector bool int vec_ld (int, const vector bool int *);
10150 vector signed int vec_ld (int, const vector signed int *);
10151 vector signed int vec_ld (int, const int *);
10152 vector signed int vec_ld (int, const long *);
10153 vector unsigned int vec_ld (int, const vector unsigned int *);
10154 vector unsigned int vec_ld (int, const unsigned int *);
10155 vector unsigned int vec_ld (int, const unsigned long *);
10156 vector bool short vec_ld (int, const vector bool short *);
10157 vector pixel vec_ld (int, const vector pixel *);
10158 vector signed short vec_ld (int, const vector signed short *);
10159 vector signed short vec_ld (int, const short *);
10160 vector unsigned short vec_ld (int, const vector unsigned short *);
10161 vector unsigned short vec_ld (int, const unsigned short *);
10162 vector bool char vec_ld (int, const vector bool char *);
10163 vector signed char vec_ld (int, const vector signed char *);
10164 vector signed char vec_ld (int, const signed char *);
10165 vector unsigned char vec_ld (int, const vector unsigned char *);
10166 vector unsigned char vec_ld (int, const unsigned char *);
10168 vector signed char vec_lde (int, const signed char *);
10169 vector unsigned char vec_lde (int, const unsigned char *);
10170 vector signed short vec_lde (int, const short *);
10171 vector unsigned short vec_lde (int, const unsigned short *);
10172 vector float vec_lde (int, const float *);
10173 vector signed int vec_lde (int, const int *);
10174 vector unsigned int vec_lde (int, const unsigned int *);
10175 vector signed int vec_lde (int, const long *);
10176 vector unsigned int vec_lde (int, const unsigned long *);
10178 vector float vec_lvewx (int, float *);
10179 vector signed int vec_lvewx (int, int *);
10180 vector unsigned int vec_lvewx (int, unsigned int *);
10181 vector signed int vec_lvewx (int, long *);
10182 vector unsigned int vec_lvewx (int, unsigned long *);
10184 vector signed short vec_lvehx (int, short *);
10185 vector unsigned short vec_lvehx (int, unsigned short *);
10187 vector signed char vec_lvebx (int, char *);
10188 vector unsigned char vec_lvebx (int, unsigned char *);
10190 vector float vec_ldl (int, const vector float *);
10191 vector float vec_ldl (int, const float *);
10192 vector bool int vec_ldl (int, const vector bool int *);
10193 vector signed int vec_ldl (int, const vector signed int *);
10194 vector signed int vec_ldl (int, const int *);
10195 vector signed int vec_ldl (int, const long *);
10196 vector unsigned int vec_ldl (int, const vector unsigned int *);
10197 vector unsigned int vec_ldl (int, const unsigned int *);
10198 vector unsigned int vec_ldl (int, const unsigned long *);
10199 vector bool short vec_ldl (int, const vector bool short *);
10200 vector pixel vec_ldl (int, const vector pixel *);
10201 vector signed short vec_ldl (int, const vector signed short *);
10202 vector signed short vec_ldl (int, const short *);
10203 vector unsigned short vec_ldl (int, const vector unsigned short *);
10204 vector unsigned short vec_ldl (int, const unsigned short *);
10205 vector bool char vec_ldl (int, const vector bool char *);
10206 vector signed char vec_ldl (int, const vector signed char *);
10207 vector signed char vec_ldl (int, const signed char *);
10208 vector unsigned char vec_ldl (int, const vector unsigned char *);
10209 vector unsigned char vec_ldl (int, const unsigned char *);
10211 vector float vec_loge (vector float);
10213 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
10214 vector unsigned char vec_lvsl (int, const volatile signed char *);
10215 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
10216 vector unsigned char vec_lvsl (int, const volatile short *);
10217 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
10218 vector unsigned char vec_lvsl (int, const volatile int *);
10219 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
10220 vector unsigned char vec_lvsl (int, const volatile long *);
10221 vector unsigned char vec_lvsl (int, const volatile float *);
10223 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
10224 vector unsigned char vec_lvsr (int, const volatile signed char *);
10225 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
10226 vector unsigned char vec_lvsr (int, const volatile short *);
10227 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
10228 vector unsigned char vec_lvsr (int, const volatile int *);
10229 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
10230 vector unsigned char vec_lvsr (int, const volatile long *);
10231 vector unsigned char vec_lvsr (int, const volatile float *);
10233 vector float vec_madd (vector float, vector float, vector float);
10235 vector signed short vec_madds (vector signed short,
10236 vector signed short,
10237 vector signed short);
10239 vector unsigned char vec_max (vector bool char, vector unsigned char);
10240 vector unsigned char vec_max (vector unsigned char, vector bool char);
10241 vector unsigned char vec_max (vector unsigned char,
10242 vector unsigned char);
10243 vector signed char vec_max (vector bool char, vector signed char);
10244 vector signed char vec_max (vector signed char, vector bool char);
10245 vector signed char vec_max (vector signed char, vector signed char);
10246 vector unsigned short vec_max (vector bool short,
10247 vector unsigned short);
10248 vector unsigned short vec_max (vector unsigned short,
10249 vector bool short);
10250 vector unsigned short vec_max (vector unsigned short,
10251 vector unsigned short);
10252 vector signed short vec_max (vector bool short, vector signed short);
10253 vector signed short vec_max (vector signed short, vector bool short);
10254 vector signed short vec_max (vector signed short, vector signed short);
10255 vector unsigned int vec_max (vector bool int, vector unsigned int);
10256 vector unsigned int vec_max (vector unsigned int, vector bool int);
10257 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
10258 vector signed int vec_max (vector bool int, vector signed int);
10259 vector signed int vec_max (vector signed int, vector bool int);
10260 vector signed int vec_max (vector signed int, vector signed int);
10261 vector float vec_max (vector float, vector float);
10263 vector float vec_vmaxfp (vector float, vector float);
10265 vector signed int vec_vmaxsw (vector bool int, vector signed int);
10266 vector signed int vec_vmaxsw (vector signed int, vector bool int);
10267 vector signed int vec_vmaxsw (vector signed int, vector signed int);
10269 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
10270 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
10271 vector unsigned int vec_vmaxuw (vector unsigned int,
10272 vector unsigned int);
10274 vector signed short vec_vmaxsh (vector bool short, vector signed short);
10275 vector signed short vec_vmaxsh (vector signed short, vector bool short);
10276 vector signed short vec_vmaxsh (vector signed short,
10277 vector signed short);
10279 vector unsigned short vec_vmaxuh (vector bool short,
10280 vector unsigned short);
10281 vector unsigned short vec_vmaxuh (vector unsigned short,
10282 vector bool short);
10283 vector unsigned short vec_vmaxuh (vector unsigned short,
10284 vector unsigned short);
10286 vector signed char vec_vmaxsb (vector bool char, vector signed char);
10287 vector signed char vec_vmaxsb (vector signed char, vector bool char);
10288 vector signed char vec_vmaxsb (vector signed char, vector signed char);
10290 vector unsigned char vec_vmaxub (vector bool char,
10291 vector unsigned char);
10292 vector unsigned char vec_vmaxub (vector unsigned char,
10294 vector unsigned char vec_vmaxub (vector unsigned char,
10295 vector unsigned char);
10297 vector bool char vec_mergeh (vector bool char, vector bool char);
10298 vector signed char vec_mergeh (vector signed char, vector signed char);
10299 vector unsigned char vec_mergeh (vector unsigned char,
10300 vector unsigned char);
10301 vector bool short vec_mergeh (vector bool short, vector bool short);
10302 vector pixel vec_mergeh (vector pixel, vector pixel);
10303 vector signed short vec_mergeh (vector signed short,
10304 vector signed short);
10305 vector unsigned short vec_mergeh (vector unsigned short,
10306 vector unsigned short);
10307 vector float vec_mergeh (vector float, vector float);
10308 vector bool int vec_mergeh (vector bool int, vector bool int);
10309 vector signed int vec_mergeh (vector signed int, vector signed int);
10310 vector unsigned int vec_mergeh (vector unsigned int,
10311 vector unsigned int);
10313 vector float vec_vmrghw (vector float, vector float);
10314 vector bool int vec_vmrghw (vector bool int, vector bool int);
10315 vector signed int vec_vmrghw (vector signed int, vector signed int);
10316 vector unsigned int vec_vmrghw (vector unsigned int,
10317 vector unsigned int);
10319 vector bool short vec_vmrghh (vector bool short, vector bool short);
10320 vector signed short vec_vmrghh (vector signed short,
10321 vector signed short);
10322 vector unsigned short vec_vmrghh (vector unsigned short,
10323 vector unsigned short);
10324 vector pixel vec_vmrghh (vector pixel, vector pixel);
10326 vector bool char vec_vmrghb (vector bool char, vector bool char);
10327 vector signed char vec_vmrghb (vector signed char, vector signed char);
10328 vector unsigned char vec_vmrghb (vector unsigned char,
10329 vector unsigned char);
10331 vector bool char vec_mergel (vector bool char, vector bool char);
10332 vector signed char vec_mergel (vector signed char, vector signed char);
10333 vector unsigned char vec_mergel (vector unsigned char,
10334 vector unsigned char);
10335 vector bool short vec_mergel (vector bool short, vector bool short);
10336 vector pixel vec_mergel (vector pixel, vector pixel);
10337 vector signed short vec_mergel (vector signed short,
10338 vector signed short);
10339 vector unsigned short vec_mergel (vector unsigned short,
10340 vector unsigned short);
10341 vector float vec_mergel (vector float, vector float);
10342 vector bool int vec_mergel (vector bool int, vector bool int);
10343 vector signed int vec_mergel (vector signed int, vector signed int);
10344 vector unsigned int vec_mergel (vector unsigned int,
10345 vector unsigned int);
10347 vector float vec_vmrglw (vector float, vector float);
10348 vector signed int vec_vmrglw (vector signed int, vector signed int);
10349 vector unsigned int vec_vmrglw (vector unsigned int,
10350 vector unsigned int);
10351 vector bool int vec_vmrglw (vector bool int, vector bool int);
10353 vector bool short vec_vmrglh (vector bool short, vector bool short);
10354 vector signed short vec_vmrglh (vector signed short,
10355 vector signed short);
10356 vector unsigned short vec_vmrglh (vector unsigned short,
10357 vector unsigned short);
10358 vector pixel vec_vmrglh (vector pixel, vector pixel);
10360 vector bool char vec_vmrglb (vector bool char, vector bool char);
10361 vector signed char vec_vmrglb (vector signed char, vector signed char);
10362 vector unsigned char vec_vmrglb (vector unsigned char,
10363 vector unsigned char);
10365 vector unsigned short vec_mfvscr (void);
10367 vector unsigned char vec_min (vector bool char, vector unsigned char);
10368 vector unsigned char vec_min (vector unsigned char, vector bool char);
10369 vector unsigned char vec_min (vector unsigned char,
10370 vector unsigned char);
10371 vector signed char vec_min (vector bool char, vector signed char);
10372 vector signed char vec_min (vector signed char, vector bool char);
10373 vector signed char vec_min (vector signed char, vector signed char);
10374 vector unsigned short vec_min (vector bool short,
10375 vector unsigned short);
10376 vector unsigned short vec_min (vector unsigned short,
10377 vector bool short);
10378 vector unsigned short vec_min (vector unsigned short,
10379 vector unsigned short);
10380 vector signed short vec_min (vector bool short, vector signed short);
10381 vector signed short vec_min (vector signed short, vector bool short);
10382 vector signed short vec_min (vector signed short, vector signed short);
10383 vector unsigned int vec_min (vector bool int, vector unsigned int);
10384 vector unsigned int vec_min (vector unsigned int, vector bool int);
10385 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
10386 vector signed int vec_min (vector bool int, vector signed int);
10387 vector signed int vec_min (vector signed int, vector bool int);
10388 vector signed int vec_min (vector signed int, vector signed int);
10389 vector float vec_min (vector float, vector float);
10391 vector float vec_vminfp (vector float, vector float);
10393 vector signed int vec_vminsw (vector bool int, vector signed int);
10394 vector signed int vec_vminsw (vector signed int, vector bool int);
10395 vector signed int vec_vminsw (vector signed int, vector signed int);
10397 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
10398 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
10399 vector unsigned int vec_vminuw (vector unsigned int,
10400 vector unsigned int);
10402 vector signed short vec_vminsh (vector bool short, vector signed short);
10403 vector signed short vec_vminsh (vector signed short, vector bool short);
10404 vector signed short vec_vminsh (vector signed short,
10405 vector signed short);
10407 vector unsigned short vec_vminuh (vector bool short,
10408 vector unsigned short);
10409 vector unsigned short vec_vminuh (vector unsigned short,
10410 vector bool short);
10411 vector unsigned short vec_vminuh (vector unsigned short,
10412 vector unsigned short);
10414 vector signed char vec_vminsb (vector bool char, vector signed char);
10415 vector signed char vec_vminsb (vector signed char, vector bool char);
10416 vector signed char vec_vminsb (vector signed char, vector signed char);
10418 vector unsigned char vec_vminub (vector bool char,
10419 vector unsigned char);
10420 vector unsigned char vec_vminub (vector unsigned char,
10422 vector unsigned char vec_vminub (vector unsigned char,
10423 vector unsigned char);
10425 vector signed short vec_mladd (vector signed short,
10426 vector signed short,
10427 vector signed short);
10428 vector signed short vec_mladd (vector signed short,
10429 vector unsigned short,
10430 vector unsigned short);
10431 vector signed short vec_mladd (vector unsigned short,
10432 vector signed short,
10433 vector signed short);
10434 vector unsigned short vec_mladd (vector unsigned short,
10435 vector unsigned short,
10436 vector unsigned short);
10438 vector signed short vec_mradds (vector signed short,
10439 vector signed short,
10440 vector signed short);
10442 vector unsigned int vec_msum (vector unsigned char,
10443 vector unsigned char,
10444 vector unsigned int);
10445 vector signed int vec_msum (vector signed char,
10446 vector unsigned char,
10447 vector signed int);
10448 vector unsigned int vec_msum (vector unsigned short,
10449 vector unsigned short,
10450 vector unsigned int);
10451 vector signed int vec_msum (vector signed short,
10452 vector signed short,
10453 vector signed int);
10455 vector signed int vec_vmsumshm (vector signed short,
10456 vector signed short,
10457 vector signed int);
10459 vector unsigned int vec_vmsumuhm (vector unsigned short,
10460 vector unsigned short,
10461 vector unsigned int);
10463 vector signed int vec_vmsummbm (vector signed char,
10464 vector unsigned char,
10465 vector signed int);
10467 vector unsigned int vec_vmsumubm (vector unsigned char,
10468 vector unsigned char,
10469 vector unsigned int);
10471 vector unsigned int vec_msums (vector unsigned short,
10472 vector unsigned short,
10473 vector unsigned int);
10474 vector signed int vec_msums (vector signed short,
10475 vector signed short,
10476 vector signed int);
10478 vector signed int vec_vmsumshs (vector signed short,
10479 vector signed short,
10480 vector signed int);
10482 vector unsigned int vec_vmsumuhs (vector unsigned short,
10483 vector unsigned short,
10484 vector unsigned int);
10486 void vec_mtvscr (vector signed int);
10487 void vec_mtvscr (vector unsigned int);
10488 void vec_mtvscr (vector bool int);
10489 void vec_mtvscr (vector signed short);
10490 void vec_mtvscr (vector unsigned short);
10491 void vec_mtvscr (vector bool short);
10492 void vec_mtvscr (vector pixel);
10493 void vec_mtvscr (vector signed char);
10494 void vec_mtvscr (vector unsigned char);
10495 void vec_mtvscr (vector bool char);
10497 vector unsigned short vec_mule (vector unsigned char,
10498 vector unsigned char);
10499 vector signed short vec_mule (vector signed char,
10500 vector signed char);
10501 vector unsigned int vec_mule (vector unsigned short,
10502 vector unsigned short);
10503 vector signed int vec_mule (vector signed short, vector signed short);
10505 vector signed int vec_vmulesh (vector signed short,
10506 vector signed short);
10508 vector unsigned int vec_vmuleuh (vector unsigned short,
10509 vector unsigned short);
10511 vector signed short vec_vmulesb (vector signed char,
10512 vector signed char);
10514 vector unsigned short vec_vmuleub (vector unsigned char,
10515 vector unsigned char);
10517 vector unsigned short vec_mulo (vector unsigned char,
10518 vector unsigned char);
10519 vector signed short vec_mulo (vector signed char, vector signed char);
10520 vector unsigned int vec_mulo (vector unsigned short,
10521 vector unsigned short);
10522 vector signed int vec_mulo (vector signed short, vector signed short);
10524 vector signed int vec_vmulosh (vector signed short,
10525 vector signed short);
10527 vector unsigned int vec_vmulouh (vector unsigned short,
10528 vector unsigned short);
10530 vector signed short vec_vmulosb (vector signed char,
10531 vector signed char);
10533 vector unsigned short vec_vmuloub (vector unsigned char,
10534 vector unsigned char);
10536 vector float vec_nmsub (vector float, vector float, vector float);
10538 vector float vec_nor (vector float, vector float);
10539 vector signed int vec_nor (vector signed int, vector signed int);
10540 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
10541 vector bool int vec_nor (vector bool int, vector bool int);
10542 vector signed short vec_nor (vector signed short, vector signed short);
10543 vector unsigned short vec_nor (vector unsigned short,
10544 vector unsigned short);
10545 vector bool short vec_nor (vector bool short, vector bool short);
10546 vector signed char vec_nor (vector signed char, vector signed char);
10547 vector unsigned char vec_nor (vector unsigned char,
10548 vector unsigned char);
10549 vector bool char vec_nor (vector bool char, vector bool char);
10551 vector float vec_or (vector float, vector float);
10552 vector float vec_or (vector float, vector bool int);
10553 vector float vec_or (vector bool int, vector float);
10554 vector bool int vec_or (vector bool int, vector bool int);
10555 vector signed int vec_or (vector bool int, vector signed int);
10556 vector signed int vec_or (vector signed int, vector bool int);
10557 vector signed int vec_or (vector signed int, vector signed int);
10558 vector unsigned int vec_or (vector bool int, vector unsigned int);
10559 vector unsigned int vec_or (vector unsigned int, vector bool int);
10560 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
10561 vector bool short vec_or (vector bool short, vector bool short);
10562 vector signed short vec_or (vector bool short, vector signed short);
10563 vector signed short vec_or (vector signed short, vector bool short);
10564 vector signed short vec_or (vector signed short, vector signed short);
10565 vector unsigned short vec_or (vector bool short, vector unsigned short);
10566 vector unsigned short vec_or (vector unsigned short, vector bool short);
10567 vector unsigned short vec_or (vector unsigned short,
10568 vector unsigned short);
10569 vector signed char vec_or (vector bool char, vector signed char);
10570 vector bool char vec_or (vector bool char, vector bool char);
10571 vector signed char vec_or (vector signed char, vector bool char);
10572 vector signed char vec_or (vector signed char, vector signed char);
10573 vector unsigned char vec_or (vector bool char, vector unsigned char);
10574 vector unsigned char vec_or (vector unsigned char, vector bool char);
10575 vector unsigned char vec_or (vector unsigned char,
10576 vector unsigned char);
10578 vector signed char vec_pack (vector signed short, vector signed short);
10579 vector unsigned char vec_pack (vector unsigned short,
10580 vector unsigned short);
10581 vector bool char vec_pack (vector bool short, vector bool short);
10582 vector signed short vec_pack (vector signed int, vector signed int);
10583 vector unsigned short vec_pack (vector unsigned int,
10584 vector unsigned int);
10585 vector bool short vec_pack (vector bool int, vector bool int);
10587 vector bool short vec_vpkuwum (vector bool int, vector bool int);
10588 vector signed short vec_vpkuwum (vector signed int, vector signed int);
10589 vector unsigned short vec_vpkuwum (vector unsigned int,
10590 vector unsigned int);
10592 vector bool char vec_vpkuhum (vector bool short, vector bool short);
10593 vector signed char vec_vpkuhum (vector signed short,
10594 vector signed short);
10595 vector unsigned char vec_vpkuhum (vector unsigned short,
10596 vector unsigned short);
10598 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
10600 vector unsigned char vec_packs (vector unsigned short,
10601 vector unsigned short);
10602 vector signed char vec_packs (vector signed short, vector signed short);
10603 vector unsigned short vec_packs (vector unsigned int,
10604 vector unsigned int);
10605 vector signed short vec_packs (vector signed int, vector signed int);
10607 vector signed short vec_vpkswss (vector signed int, vector signed int);
10609 vector unsigned short vec_vpkuwus (vector unsigned int,
10610 vector unsigned int);
10612 vector signed char vec_vpkshss (vector signed short,
10613 vector signed short);
10615 vector unsigned char vec_vpkuhus (vector unsigned short,
10616 vector unsigned short);
10618 vector unsigned char vec_packsu (vector unsigned short,
10619 vector unsigned short);
10620 vector unsigned char vec_packsu (vector signed short,
10621 vector signed short);
10622 vector unsigned short vec_packsu (vector unsigned int,
10623 vector unsigned int);
10624 vector unsigned short vec_packsu (vector signed int, vector signed int);
10626 vector unsigned short vec_vpkswus (vector signed int,
10627 vector signed int);
10629 vector unsigned char vec_vpkshus (vector signed short,
10630 vector signed short);
10632 vector float vec_perm (vector float,
10634 vector unsigned char);
10635 vector signed int vec_perm (vector signed int,
10637 vector unsigned char);
10638 vector unsigned int vec_perm (vector unsigned int,
10639 vector unsigned int,
10640 vector unsigned char);
10641 vector bool int vec_perm (vector bool int,
10643 vector unsigned char);
10644 vector signed short vec_perm (vector signed short,
10645 vector signed short,
10646 vector unsigned char);
10647 vector unsigned short vec_perm (vector unsigned short,
10648 vector unsigned short,
10649 vector unsigned char);
10650 vector bool short vec_perm (vector bool short,
10652 vector unsigned char);
10653 vector pixel vec_perm (vector pixel,
10655 vector unsigned char);
10656 vector signed char vec_perm (vector signed char,
10657 vector signed char,
10658 vector unsigned char);
10659 vector unsigned char vec_perm (vector unsigned char,
10660 vector unsigned char,
10661 vector unsigned char);
10662 vector bool char vec_perm (vector bool char,
10664 vector unsigned char);
10666 vector float vec_re (vector float);
10668 vector signed char vec_rl (vector signed char,
10669 vector unsigned char);
10670 vector unsigned char vec_rl (vector unsigned char,
10671 vector unsigned char);
10672 vector signed short vec_rl (vector signed short, vector unsigned short);
10673 vector unsigned short vec_rl (vector unsigned short,
10674 vector unsigned short);
10675 vector signed int vec_rl (vector signed int, vector unsigned int);
10676 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
10678 vector signed int vec_vrlw (vector signed int, vector unsigned int);
10679 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
10681 vector signed short vec_vrlh (vector signed short,
10682 vector unsigned short);
10683 vector unsigned short vec_vrlh (vector unsigned short,
10684 vector unsigned short);
10686 vector signed char vec_vrlb (vector signed char, vector unsigned char);
10687 vector unsigned char vec_vrlb (vector unsigned char,
10688 vector unsigned char);
10690 vector float vec_round (vector float);
10692 vector float vec_rsqrte (vector float);
10694 vector float vec_sel (vector float, vector float, vector bool int);
10695 vector float vec_sel (vector float, vector float, vector unsigned int);
10696 vector signed int vec_sel (vector signed int,
10699 vector signed int vec_sel (vector signed int,
10701 vector unsigned int);
10702 vector unsigned int vec_sel (vector unsigned int,
10703 vector unsigned int,
10705 vector unsigned int vec_sel (vector unsigned int,
10706 vector unsigned int,
10707 vector unsigned int);
10708 vector bool int vec_sel (vector bool int,
10711 vector bool int vec_sel (vector bool int,
10713 vector unsigned int);
10714 vector signed short vec_sel (vector signed short,
10715 vector signed short,
10716 vector bool short);
10717 vector signed short vec_sel (vector signed short,
10718 vector signed short,
10719 vector unsigned short);
10720 vector unsigned short vec_sel (vector unsigned short,
10721 vector unsigned short,
10722 vector bool short);
10723 vector unsigned short vec_sel (vector unsigned short,
10724 vector unsigned short,
10725 vector unsigned short);
10726 vector bool short vec_sel (vector bool short,
10728 vector bool short);
10729 vector bool short vec_sel (vector bool short,
10731 vector unsigned short);
10732 vector signed char vec_sel (vector signed char,
10733 vector signed char,
10735 vector signed char vec_sel (vector signed char,
10736 vector signed char,
10737 vector unsigned char);
10738 vector unsigned char vec_sel (vector unsigned char,
10739 vector unsigned char,
10741 vector unsigned char vec_sel (vector unsigned char,
10742 vector unsigned char,
10743 vector unsigned char);
10744 vector bool char vec_sel (vector bool char,
10747 vector bool char vec_sel (vector bool char,
10749 vector unsigned char);
10751 vector signed char vec_sl (vector signed char,
10752 vector unsigned char);
10753 vector unsigned char vec_sl (vector unsigned char,
10754 vector unsigned char);
10755 vector signed short vec_sl (vector signed short, vector unsigned short);
10756 vector unsigned short vec_sl (vector unsigned short,
10757 vector unsigned short);
10758 vector signed int vec_sl (vector signed int, vector unsigned int);
10759 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
10761 vector signed int vec_vslw (vector signed int, vector unsigned int);
10762 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
10764 vector signed short vec_vslh (vector signed short,
10765 vector unsigned short);
10766 vector unsigned short vec_vslh (vector unsigned short,
10767 vector unsigned short);
10769 vector signed char vec_vslb (vector signed char, vector unsigned char);
10770 vector unsigned char vec_vslb (vector unsigned char,
10771 vector unsigned char);
10773 vector float vec_sld (vector float, vector float, const int);
10774 vector signed int vec_sld (vector signed int,
10777 vector unsigned int vec_sld (vector unsigned int,
10778 vector unsigned int,
10780 vector bool int vec_sld (vector bool int,
10783 vector signed short vec_sld (vector signed short,
10784 vector signed short,
10786 vector unsigned short vec_sld (vector unsigned short,
10787 vector unsigned short,
10789 vector bool short vec_sld (vector bool short,
10792 vector pixel vec_sld (vector pixel,
10795 vector signed char vec_sld (vector signed char,
10796 vector signed char,
10798 vector unsigned char vec_sld (vector unsigned char,
10799 vector unsigned char,
10801 vector bool char vec_sld (vector bool char,
10805 vector signed int vec_sll (vector signed int,
10806 vector unsigned int);
10807 vector signed int vec_sll (vector signed int,
10808 vector unsigned short);
10809 vector signed int vec_sll (vector signed int,
10810 vector unsigned char);
10811 vector unsigned int vec_sll (vector unsigned int,
10812 vector unsigned int);
10813 vector unsigned int vec_sll (vector unsigned int,
10814 vector unsigned short);
10815 vector unsigned int vec_sll (vector unsigned int,
10816 vector unsigned char);
10817 vector bool int vec_sll (vector bool int,
10818 vector unsigned int);
10819 vector bool int vec_sll (vector bool int,
10820 vector unsigned short);
10821 vector bool int vec_sll (vector bool int,
10822 vector unsigned char);
10823 vector signed short vec_sll (vector signed short,
10824 vector unsigned int);
10825 vector signed short vec_sll (vector signed short,
10826 vector unsigned short);
10827 vector signed short vec_sll (vector signed short,
10828 vector unsigned char);
10829 vector unsigned short vec_sll (vector unsigned short,
10830 vector unsigned int);
10831 vector unsigned short vec_sll (vector unsigned short,
10832 vector unsigned short);
10833 vector unsigned short vec_sll (vector unsigned short,
10834 vector unsigned char);
10835 vector bool short vec_sll (vector bool short, vector unsigned int);
10836 vector bool short vec_sll (vector bool short, vector unsigned short);
10837 vector bool short vec_sll (vector bool short, vector unsigned char);
10838 vector pixel vec_sll (vector pixel, vector unsigned int);
10839 vector pixel vec_sll (vector pixel, vector unsigned short);
10840 vector pixel vec_sll (vector pixel, vector unsigned char);
10841 vector signed char vec_sll (vector signed char, vector unsigned int);
10842 vector signed char vec_sll (vector signed char, vector unsigned short);
10843 vector signed char vec_sll (vector signed char, vector unsigned char);
10844 vector unsigned char vec_sll (vector unsigned char,
10845 vector unsigned int);
10846 vector unsigned char vec_sll (vector unsigned char,
10847 vector unsigned short);
10848 vector unsigned char vec_sll (vector unsigned char,
10849 vector unsigned char);
10850 vector bool char vec_sll (vector bool char, vector unsigned int);
10851 vector bool char vec_sll (vector bool char, vector unsigned short);
10852 vector bool char vec_sll (vector bool char, vector unsigned char);
10854 vector float vec_slo (vector float, vector signed char);
10855 vector float vec_slo (vector float, vector unsigned char);
10856 vector signed int vec_slo (vector signed int, vector signed char);
10857 vector signed int vec_slo (vector signed int, vector unsigned char);
10858 vector unsigned int vec_slo (vector unsigned int, vector signed char);
10859 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
10860 vector signed short vec_slo (vector signed short, vector signed char);
10861 vector signed short vec_slo (vector signed short, vector unsigned char);
10862 vector unsigned short vec_slo (vector unsigned short,
10863 vector signed char);
10864 vector unsigned short vec_slo (vector unsigned short,
10865 vector unsigned char);
10866 vector pixel vec_slo (vector pixel, vector signed char);
10867 vector pixel vec_slo (vector pixel, vector unsigned char);
10868 vector signed char vec_slo (vector signed char, vector signed char);
10869 vector signed char vec_slo (vector signed char, vector unsigned char);
10870 vector unsigned char vec_slo (vector unsigned char, vector signed char);
10871 vector unsigned char vec_slo (vector unsigned char,
10872 vector unsigned char);
10874 vector signed char vec_splat (vector signed char, const int);
10875 vector unsigned char vec_splat (vector unsigned char, const int);
10876 vector bool char vec_splat (vector bool char, const int);
10877 vector signed short vec_splat (vector signed short, const int);
10878 vector unsigned short vec_splat (vector unsigned short, const int);
10879 vector bool short vec_splat (vector bool short, const int);
10880 vector pixel vec_splat (vector pixel, const int);
10881 vector float vec_splat (vector float, const int);
10882 vector signed int vec_splat (vector signed int, const int);
10883 vector unsigned int vec_splat (vector unsigned int, const int);
10884 vector bool int vec_splat (vector bool int, const int);
10886 vector float vec_vspltw (vector float, const int);
10887 vector signed int vec_vspltw (vector signed int, const int);
10888 vector unsigned int vec_vspltw (vector unsigned int, const int);
10889 vector bool int vec_vspltw (vector bool int, const int);
10891 vector bool short vec_vsplth (vector bool short, const int);
10892 vector signed short vec_vsplth (vector signed short, const int);
10893 vector unsigned short vec_vsplth (vector unsigned short, const int);
10894 vector pixel vec_vsplth (vector pixel, const int);
10896 vector signed char vec_vspltb (vector signed char, const int);
10897 vector unsigned char vec_vspltb (vector unsigned char, const int);
10898 vector bool char vec_vspltb (vector bool char, const int);
10900 vector signed char vec_splat_s8 (const int);
10902 vector signed short vec_splat_s16 (const int);
10904 vector signed int vec_splat_s32 (const int);
10906 vector unsigned char vec_splat_u8 (const int);
10908 vector unsigned short vec_splat_u16 (const int);
10910 vector unsigned int vec_splat_u32 (const int);
10912 vector signed char vec_sr (vector signed char, vector unsigned char);
10913 vector unsigned char vec_sr (vector unsigned char,
10914 vector unsigned char);
10915 vector signed short vec_sr (vector signed short,
10916 vector unsigned short);
10917 vector unsigned short vec_sr (vector unsigned short,
10918 vector unsigned short);
10919 vector signed int vec_sr (vector signed int, vector unsigned int);
10920 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
10922 vector signed int vec_vsrw (vector signed int, vector unsigned int);
10923 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
10925 vector signed short vec_vsrh (vector signed short,
10926 vector unsigned short);
10927 vector unsigned short vec_vsrh (vector unsigned short,
10928 vector unsigned short);
10930 vector signed char vec_vsrb (vector signed char, vector unsigned char);
10931 vector unsigned char vec_vsrb (vector unsigned char,
10932 vector unsigned char);
10934 vector signed char vec_sra (vector signed char, vector unsigned char);
10935 vector unsigned char vec_sra (vector unsigned char,
10936 vector unsigned char);
10937 vector signed short vec_sra (vector signed short,
10938 vector unsigned short);
10939 vector unsigned short vec_sra (vector unsigned short,
10940 vector unsigned short);
10941 vector signed int vec_sra (vector signed int, vector unsigned int);
10942 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
10944 vector signed int vec_vsraw (vector signed int, vector unsigned int);
10945 vector unsigned int vec_vsraw (vector unsigned int,
10946 vector unsigned int);
10948 vector signed short vec_vsrah (vector signed short,
10949 vector unsigned short);
10950 vector unsigned short vec_vsrah (vector unsigned short,
10951 vector unsigned short);
10953 vector signed char vec_vsrab (vector signed char, vector unsigned char);
10954 vector unsigned char vec_vsrab (vector unsigned char,
10955 vector unsigned char);
10957 vector signed int vec_srl (vector signed int, vector unsigned int);
10958 vector signed int vec_srl (vector signed int, vector unsigned short);
10959 vector signed int vec_srl (vector signed int, vector unsigned char);
10960 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
10961 vector unsigned int vec_srl (vector unsigned int,
10962 vector unsigned short);
10963 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
10964 vector bool int vec_srl (vector bool int, vector unsigned int);
10965 vector bool int vec_srl (vector bool int, vector unsigned short);
10966 vector bool int vec_srl (vector bool int, vector unsigned char);
10967 vector signed short vec_srl (vector signed short, vector unsigned int);
10968 vector signed short vec_srl (vector signed short,
10969 vector unsigned short);
10970 vector signed short vec_srl (vector signed short, vector unsigned char);
10971 vector unsigned short vec_srl (vector unsigned short,
10972 vector unsigned int);
10973 vector unsigned short vec_srl (vector unsigned short,
10974 vector unsigned short);
10975 vector unsigned short vec_srl (vector unsigned short,
10976 vector unsigned char);
10977 vector bool short vec_srl (vector bool short, vector unsigned int);
10978 vector bool short vec_srl (vector bool short, vector unsigned short);
10979 vector bool short vec_srl (vector bool short, vector unsigned char);
10980 vector pixel vec_srl (vector pixel, vector unsigned int);
10981 vector pixel vec_srl (vector pixel, vector unsigned short);
10982 vector pixel vec_srl (vector pixel, vector unsigned char);
10983 vector signed char vec_srl (vector signed char, vector unsigned int);
10984 vector signed char vec_srl (vector signed char, vector unsigned short);
10985 vector signed char vec_srl (vector signed char, vector unsigned char);
10986 vector unsigned char vec_srl (vector unsigned char,
10987 vector unsigned int);
10988 vector unsigned char vec_srl (vector unsigned char,
10989 vector unsigned short);
10990 vector unsigned char vec_srl (vector unsigned char,
10991 vector unsigned char);
10992 vector bool char vec_srl (vector bool char, vector unsigned int);
10993 vector bool char vec_srl (vector bool char, vector unsigned short);
10994 vector bool char vec_srl (vector bool char, vector unsigned char);
10996 vector float vec_sro (vector float, vector signed char);
10997 vector float vec_sro (vector float, vector unsigned char);
10998 vector signed int vec_sro (vector signed int, vector signed char);
10999 vector signed int vec_sro (vector signed int, vector unsigned char);
11000 vector unsigned int vec_sro (vector unsigned int, vector signed char);
11001 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
11002 vector signed short vec_sro (vector signed short, vector signed char);
11003 vector signed short vec_sro (vector signed short, vector unsigned char);
11004 vector unsigned short vec_sro (vector unsigned short,
11005 vector signed char);
11006 vector unsigned short vec_sro (vector unsigned short,
11007 vector unsigned char);
11008 vector pixel vec_sro (vector pixel, vector signed char);
11009 vector pixel vec_sro (vector pixel, vector unsigned char);
11010 vector signed char vec_sro (vector signed char, vector signed char);
11011 vector signed char vec_sro (vector signed char, vector unsigned char);
11012 vector unsigned char vec_sro (vector unsigned char, vector signed char);
11013 vector unsigned char vec_sro (vector unsigned char,
11014 vector unsigned char);
11016 void vec_st (vector float, int, vector float *);
11017 void vec_st (vector float, int, float *);
11018 void vec_st (vector signed int, int, vector signed int *);
11019 void vec_st (vector signed int, int, int *);
11020 void vec_st (vector unsigned int, int, vector unsigned int *);
11021 void vec_st (vector unsigned int, int, unsigned int *);
11022 void vec_st (vector bool int, int, vector bool int *);
11023 void vec_st (vector bool int, int, unsigned int *);
11024 void vec_st (vector bool int, int, int *);
11025 void vec_st (vector signed short, int, vector signed short *);
11026 void vec_st (vector signed short, int, short *);
11027 void vec_st (vector unsigned short, int, vector unsigned short *);
11028 void vec_st (vector unsigned short, int, unsigned short *);
11029 void vec_st (vector bool short, int, vector bool short *);
11030 void vec_st (vector bool short, int, unsigned short *);
11031 void vec_st (vector pixel, int, vector pixel *);
11032 void vec_st (vector pixel, int, unsigned short *);
11033 void vec_st (vector pixel, int, short *);
11034 void vec_st (vector bool short, int, short *);
11035 void vec_st (vector signed char, int, vector signed char *);
11036 void vec_st (vector signed char, int, signed char *);
11037 void vec_st (vector unsigned char, int, vector unsigned char *);
11038 void vec_st (vector unsigned char, int, unsigned char *);
11039 void vec_st (vector bool char, int, vector bool char *);
11040 void vec_st (vector bool char, int, unsigned char *);
11041 void vec_st (vector bool char, int, signed char *);
11043 void vec_ste (vector signed char, int, signed char *);
11044 void vec_ste (vector unsigned char, int, unsigned char *);
11045 void vec_ste (vector bool char, int, signed char *);
11046 void vec_ste (vector bool char, int, unsigned char *);
11047 void vec_ste (vector signed short, int, short *);
11048 void vec_ste (vector unsigned short, int, unsigned short *);
11049 void vec_ste (vector bool short, int, short *);
11050 void vec_ste (vector bool short, int, unsigned short *);
11051 void vec_ste (vector pixel, int, short *);
11052 void vec_ste (vector pixel, int, unsigned short *);
11053 void vec_ste (vector float, int, float *);
11054 void vec_ste (vector signed int, int, int *);
11055 void vec_ste (vector unsigned int, int, unsigned int *);
11056 void vec_ste (vector bool int, int, int *);
11057 void vec_ste (vector bool int, int, unsigned int *);
11059 void vec_stvewx (vector float, int, float *);
11060 void vec_stvewx (vector signed int, int, int *);
11061 void vec_stvewx (vector unsigned int, int, unsigned int *);
11062 void vec_stvewx (vector bool int, int, int *);
11063 void vec_stvewx (vector bool int, int, unsigned int *);
11065 void vec_stvehx (vector signed short, int, short *);
11066 void vec_stvehx (vector unsigned short, int, unsigned short *);
11067 void vec_stvehx (vector bool short, int, short *);
11068 void vec_stvehx (vector bool short, int, unsigned short *);
11069 void vec_stvehx (vector pixel, int, short *);
11070 void vec_stvehx (vector pixel, int, unsigned short *);
11072 void vec_stvebx (vector signed char, int, signed char *);
11073 void vec_stvebx (vector unsigned char, int, unsigned char *);
11074 void vec_stvebx (vector bool char, int, signed char *);
11075 void vec_stvebx (vector bool char, int, unsigned char *);
11077 void vec_stl (vector float, int, vector float *);
11078 void vec_stl (vector float, int, float *);
11079 void vec_stl (vector signed int, int, vector signed int *);
11080 void vec_stl (vector signed int, int, int *);
11081 void vec_stl (vector unsigned int, int, vector unsigned int *);
11082 void vec_stl (vector unsigned int, int, unsigned int *);
11083 void vec_stl (vector bool int, int, vector bool int *);
11084 void vec_stl (vector bool int, int, unsigned int *);
11085 void vec_stl (vector bool int, int, int *);
11086 void vec_stl (vector signed short, int, vector signed short *);
11087 void vec_stl (vector signed short, int, short *);
11088 void vec_stl (vector unsigned short, int, vector unsigned short *);
11089 void vec_stl (vector unsigned short, int, unsigned short *);
11090 void vec_stl (vector bool short, int, vector bool short *);
11091 void vec_stl (vector bool short, int, unsigned short *);
11092 void vec_stl (vector bool short, int, short *);
11093 void vec_stl (vector pixel, int, vector pixel *);
11094 void vec_stl (vector pixel, int, unsigned short *);
11095 void vec_stl (vector pixel, int, short *);
11096 void vec_stl (vector signed char, int, vector signed char *);
11097 void vec_stl (vector signed char, int, signed char *);
11098 void vec_stl (vector unsigned char, int, vector unsigned char *);
11099 void vec_stl (vector unsigned char, int, unsigned char *);
11100 void vec_stl (vector bool char, int, vector bool char *);
11101 void vec_stl (vector bool char, int, unsigned char *);
11102 void vec_stl (vector bool char, int, signed char *);
11104 vector signed char vec_sub (vector bool char, vector signed char);
11105 vector signed char vec_sub (vector signed char, vector bool char);
11106 vector signed char vec_sub (vector signed char, vector signed char);
11107 vector unsigned char vec_sub (vector bool char, vector unsigned char);
11108 vector unsigned char vec_sub (vector unsigned char, vector bool char);
11109 vector unsigned char vec_sub (vector unsigned char,
11110 vector unsigned char);
11111 vector signed short vec_sub (vector bool short, vector signed short);
11112 vector signed short vec_sub (vector signed short, vector bool short);
11113 vector signed short vec_sub (vector signed short, vector signed short);
11114 vector unsigned short vec_sub (vector bool short,
11115 vector unsigned short);
11116 vector unsigned short vec_sub (vector unsigned short,
11117 vector bool short);
11118 vector unsigned short vec_sub (vector unsigned short,
11119 vector unsigned short);
11120 vector signed int vec_sub (vector bool int, vector signed int);
11121 vector signed int vec_sub (vector signed int, vector bool int);
11122 vector signed int vec_sub (vector signed int, vector signed int);
11123 vector unsigned int vec_sub (vector bool int, vector unsigned int);
11124 vector unsigned int vec_sub (vector unsigned int, vector bool int);
11125 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
11126 vector float vec_sub (vector float, vector float);
11128 vector float vec_vsubfp (vector float, vector float);
11130 vector signed int vec_vsubuwm (vector bool int, vector signed int);
11131 vector signed int vec_vsubuwm (vector signed int, vector bool int);
11132 vector signed int vec_vsubuwm (vector signed int, vector signed int);
11133 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
11134 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
11135 vector unsigned int vec_vsubuwm (vector unsigned int,
11136 vector unsigned int);
11138 vector signed short vec_vsubuhm (vector bool short,
11139 vector signed short);
11140 vector signed short vec_vsubuhm (vector signed short,
11141 vector bool short);
11142 vector signed short vec_vsubuhm (vector signed short,
11143 vector signed short);
11144 vector unsigned short vec_vsubuhm (vector bool short,
11145 vector unsigned short);
11146 vector unsigned short vec_vsubuhm (vector unsigned short,
11147 vector bool short);
11148 vector unsigned short vec_vsubuhm (vector unsigned short,
11149 vector unsigned short);
11151 vector signed char vec_vsububm (vector bool char, vector signed char);
11152 vector signed char vec_vsububm (vector signed char, vector bool char);
11153 vector signed char vec_vsububm (vector signed char, vector signed char);
11154 vector unsigned char vec_vsububm (vector bool char,
11155 vector unsigned char);
11156 vector unsigned char vec_vsububm (vector unsigned char,
11158 vector unsigned char vec_vsububm (vector unsigned char,
11159 vector unsigned char);
11161 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
11163 vector unsigned char vec_subs (vector bool char, vector unsigned char);
11164 vector unsigned char vec_subs (vector unsigned char, vector bool char);
11165 vector unsigned char vec_subs (vector unsigned char,
11166 vector unsigned char);
11167 vector signed char vec_subs (vector bool char, vector signed char);
11168 vector signed char vec_subs (vector signed char, vector bool char);
11169 vector signed char vec_subs (vector signed char, vector signed char);
11170 vector unsigned short vec_subs (vector bool short,
11171 vector unsigned short);
11172 vector unsigned short vec_subs (vector unsigned short,
11173 vector bool short);
11174 vector unsigned short vec_subs (vector unsigned short,
11175 vector unsigned short);
11176 vector signed short vec_subs (vector bool short, vector signed short);
11177 vector signed short vec_subs (vector signed short, vector bool short);
11178 vector signed short vec_subs (vector signed short, vector signed short);
11179 vector unsigned int vec_subs (vector bool int, vector unsigned int);
11180 vector unsigned int vec_subs (vector unsigned int, vector bool int);
11181 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
11182 vector signed int vec_subs (vector bool int, vector signed int);
11183 vector signed int vec_subs (vector signed int, vector bool int);
11184 vector signed int vec_subs (vector signed int, vector signed int);
11186 vector signed int vec_vsubsws (vector bool int, vector signed int);
11187 vector signed int vec_vsubsws (vector signed int, vector bool int);
11188 vector signed int vec_vsubsws (vector signed int, vector signed int);
11190 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
11191 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
11192 vector unsigned int vec_vsubuws (vector unsigned int,
11193 vector unsigned int);
11195 vector signed short vec_vsubshs (vector bool short,
11196 vector signed short);
11197 vector signed short vec_vsubshs (vector signed short,
11198 vector bool short);
11199 vector signed short vec_vsubshs (vector signed short,
11200 vector signed short);
11202 vector unsigned short vec_vsubuhs (vector bool short,
11203 vector unsigned short);
11204 vector unsigned short vec_vsubuhs (vector unsigned short,
11205 vector bool short);
11206 vector unsigned short vec_vsubuhs (vector unsigned short,
11207 vector unsigned short);
11209 vector signed char vec_vsubsbs (vector bool char, vector signed char);
11210 vector signed char vec_vsubsbs (vector signed char, vector bool char);
11211 vector signed char vec_vsubsbs (vector signed char, vector signed char);
11213 vector unsigned char vec_vsububs (vector bool char,
11214 vector unsigned char);
11215 vector unsigned char vec_vsububs (vector unsigned char,
11217 vector unsigned char vec_vsububs (vector unsigned char,
11218 vector unsigned char);
11220 vector unsigned int vec_sum4s (vector unsigned char,
11221 vector unsigned int);
11222 vector signed int vec_sum4s (vector signed char, vector signed int);
11223 vector signed int vec_sum4s (vector signed short, vector signed int);
11225 vector signed int vec_vsum4shs (vector signed short, vector signed int);
11227 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
11229 vector unsigned int vec_vsum4ubs (vector unsigned char,
11230 vector unsigned int);
11232 vector signed int vec_sum2s (vector signed int, vector signed int);
11234 vector signed int vec_sums (vector signed int, vector signed int);
11236 vector float vec_trunc (vector float);
11238 vector signed short vec_unpackh (vector signed char);
11239 vector bool short vec_unpackh (vector bool char);
11240 vector signed int vec_unpackh (vector signed short);
11241 vector bool int vec_unpackh (vector bool short);
11242 vector unsigned int vec_unpackh (vector pixel);
11244 vector bool int vec_vupkhsh (vector bool short);
11245 vector signed int vec_vupkhsh (vector signed short);
11247 vector unsigned int vec_vupkhpx (vector pixel);
11249 vector bool short vec_vupkhsb (vector bool char);
11250 vector signed short vec_vupkhsb (vector signed char);
11252 vector signed short vec_unpackl (vector signed char);
11253 vector bool short vec_unpackl (vector bool char);
11254 vector unsigned int vec_unpackl (vector pixel);
11255 vector signed int vec_unpackl (vector signed short);
11256 vector bool int vec_unpackl (vector bool short);
11258 vector unsigned int vec_vupklpx (vector pixel);
11260 vector bool int vec_vupklsh (vector bool short);
11261 vector signed int vec_vupklsh (vector signed short);
11263 vector bool short vec_vupklsb (vector bool char);
11264 vector signed short vec_vupklsb (vector signed char);
11266 vector float vec_xor (vector float, vector float);
11267 vector float vec_xor (vector float, vector bool int);
11268 vector float vec_xor (vector bool int, vector float);
11269 vector bool int vec_xor (vector bool int, vector bool int);
11270 vector signed int vec_xor (vector bool int, vector signed int);
11271 vector signed int vec_xor (vector signed int, vector bool int);
11272 vector signed int vec_xor (vector signed int, vector signed int);
11273 vector unsigned int vec_xor (vector bool int, vector unsigned int);
11274 vector unsigned int vec_xor (vector unsigned int, vector bool int);
11275 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
11276 vector bool short vec_xor (vector bool short, vector bool short);
11277 vector signed short vec_xor (vector bool short, vector signed short);
11278 vector signed short vec_xor (vector signed short, vector bool short);
11279 vector signed short vec_xor (vector signed short, vector signed short);
11280 vector unsigned short vec_xor (vector bool short,
11281 vector unsigned short);
11282 vector unsigned short vec_xor (vector unsigned short,
11283 vector bool short);
11284 vector unsigned short vec_xor (vector unsigned short,
11285 vector unsigned short);
11286 vector signed char vec_xor (vector bool char, vector signed char);
11287 vector bool char vec_xor (vector bool char, vector bool char);
11288 vector signed char vec_xor (vector signed char, vector bool char);
11289 vector signed char vec_xor (vector signed char, vector signed char);
11290 vector unsigned char vec_xor (vector bool char, vector unsigned char);
11291 vector unsigned char vec_xor (vector unsigned char, vector bool char);
11292 vector unsigned char vec_xor (vector unsigned char,
11293 vector unsigned char);
11295 int vec_all_eq (vector signed char, vector bool char);
11296 int vec_all_eq (vector signed char, vector signed char);
11297 int vec_all_eq (vector unsigned char, vector bool char);
11298 int vec_all_eq (vector unsigned char, vector unsigned char);
11299 int vec_all_eq (vector bool char, vector bool char);
11300 int vec_all_eq (vector bool char, vector unsigned char);
11301 int vec_all_eq (vector bool char, vector signed char);
11302 int vec_all_eq (vector signed short, vector bool short);
11303 int vec_all_eq (vector signed short, vector signed short);
11304 int vec_all_eq (vector unsigned short, vector bool short);
11305 int vec_all_eq (vector unsigned short, vector unsigned short);
11306 int vec_all_eq (vector bool short, vector bool short);
11307 int vec_all_eq (vector bool short, vector unsigned short);
11308 int vec_all_eq (vector bool short, vector signed short);
11309 int vec_all_eq (vector pixel, vector pixel);
11310 int vec_all_eq (vector signed int, vector bool int);
11311 int vec_all_eq (vector signed int, vector signed int);
11312 int vec_all_eq (vector unsigned int, vector bool int);
11313 int vec_all_eq (vector unsigned int, vector unsigned int);
11314 int vec_all_eq (vector bool int, vector bool int);
11315 int vec_all_eq (vector bool int, vector unsigned int);
11316 int vec_all_eq (vector bool int, vector signed int);
11317 int vec_all_eq (vector float, vector float);
11319 int vec_all_ge (vector bool char, vector unsigned char);
11320 int vec_all_ge (vector unsigned char, vector bool char);
11321 int vec_all_ge (vector unsigned char, vector unsigned char);
11322 int vec_all_ge (vector bool char, vector signed char);
11323 int vec_all_ge (vector signed char, vector bool char);
11324 int vec_all_ge (vector signed char, vector signed char);
11325 int vec_all_ge (vector bool short, vector unsigned short);
11326 int vec_all_ge (vector unsigned short, vector bool short);
11327 int vec_all_ge (vector unsigned short, vector unsigned short);
11328 int vec_all_ge (vector signed short, vector signed short);
11329 int vec_all_ge (vector bool short, vector signed short);
11330 int vec_all_ge (vector signed short, vector bool short);
11331 int vec_all_ge (vector bool int, vector unsigned int);
11332 int vec_all_ge (vector unsigned int, vector bool int);
11333 int vec_all_ge (vector unsigned int, vector unsigned int);
11334 int vec_all_ge (vector bool int, vector signed int);
11335 int vec_all_ge (vector signed int, vector bool int);
11336 int vec_all_ge (vector signed int, vector signed int);
11337 int vec_all_ge (vector float, vector float);
11339 int vec_all_gt (vector bool char, vector unsigned char);
11340 int vec_all_gt (vector unsigned char, vector bool char);
11341 int vec_all_gt (vector unsigned char, vector unsigned char);
11342 int vec_all_gt (vector bool char, vector signed char);
11343 int vec_all_gt (vector signed char, vector bool char);
11344 int vec_all_gt (vector signed char, vector signed char);
11345 int vec_all_gt (vector bool short, vector unsigned short);
11346 int vec_all_gt (vector unsigned short, vector bool short);
11347 int vec_all_gt (vector unsigned short, vector unsigned short);
11348 int vec_all_gt (vector bool short, vector signed short);
11349 int vec_all_gt (vector signed short, vector bool short);
11350 int vec_all_gt (vector signed short, vector signed short);
11351 int vec_all_gt (vector bool int, vector unsigned int);
11352 int vec_all_gt (vector unsigned int, vector bool int);
11353 int vec_all_gt (vector unsigned int, vector unsigned int);
11354 int vec_all_gt (vector bool int, vector signed int);
11355 int vec_all_gt (vector signed int, vector bool int);
11356 int vec_all_gt (vector signed int, vector signed int);
11357 int vec_all_gt (vector float, vector float);
11359 int vec_all_in (vector float, vector float);
11361 int vec_all_le (vector bool char, vector unsigned char);
11362 int vec_all_le (vector unsigned char, vector bool char);
11363 int vec_all_le (vector unsigned char, vector unsigned char);
11364 int vec_all_le (vector bool char, vector signed char);
11365 int vec_all_le (vector signed char, vector bool char);
11366 int vec_all_le (vector signed char, vector signed char);
11367 int vec_all_le (vector bool short, vector unsigned short);
11368 int vec_all_le (vector unsigned short, vector bool short);
11369 int vec_all_le (vector unsigned short, vector unsigned short);
11370 int vec_all_le (vector bool short, vector signed short);
11371 int vec_all_le (vector signed short, vector bool short);
11372 int vec_all_le (vector signed short, vector signed short);
11373 int vec_all_le (vector bool int, vector unsigned int);
11374 int vec_all_le (vector unsigned int, vector bool int);
11375 int vec_all_le (vector unsigned int, vector unsigned int);
11376 int vec_all_le (vector bool int, vector signed int);
11377 int vec_all_le (vector signed int, vector bool int);
11378 int vec_all_le (vector signed int, vector signed int);
11379 int vec_all_le (vector float, vector float);
11381 int vec_all_lt (vector bool char, vector unsigned char);
11382 int vec_all_lt (vector unsigned char, vector bool char);
11383 int vec_all_lt (vector unsigned char, vector unsigned char);
11384 int vec_all_lt (vector bool char, vector signed char);
11385 int vec_all_lt (vector signed char, vector bool char);
11386 int vec_all_lt (vector signed char, vector signed char);
11387 int vec_all_lt (vector bool short, vector unsigned short);
11388 int vec_all_lt (vector unsigned short, vector bool short);
11389 int vec_all_lt (vector unsigned short, vector unsigned short);
11390 int vec_all_lt (vector bool short, vector signed short);
11391 int vec_all_lt (vector signed short, vector bool short);
11392 int vec_all_lt (vector signed short, vector signed short);
11393 int vec_all_lt (vector bool int, vector unsigned int);
11394 int vec_all_lt (vector unsigned int, vector bool int);
11395 int vec_all_lt (vector unsigned int, vector unsigned int);
11396 int vec_all_lt (vector bool int, vector signed int);
11397 int vec_all_lt (vector signed int, vector bool int);
11398 int vec_all_lt (vector signed int, vector signed int);
11399 int vec_all_lt (vector float, vector float);
11401 int vec_all_nan (vector float);
11403 int vec_all_ne (vector signed char, vector bool char);
11404 int vec_all_ne (vector signed char, vector signed char);
11405 int vec_all_ne (vector unsigned char, vector bool char);
11406 int vec_all_ne (vector unsigned char, vector unsigned char);
11407 int vec_all_ne (vector bool char, vector bool char);
11408 int vec_all_ne (vector bool char, vector unsigned char);
11409 int vec_all_ne (vector bool char, vector signed char);
11410 int vec_all_ne (vector signed short, vector bool short);
11411 int vec_all_ne (vector signed short, vector signed short);
11412 int vec_all_ne (vector unsigned short, vector bool short);
11413 int vec_all_ne (vector unsigned short, vector unsigned short);
11414 int vec_all_ne (vector bool short, vector bool short);
11415 int vec_all_ne (vector bool short, vector unsigned short);
11416 int vec_all_ne (vector bool short, vector signed short);
11417 int vec_all_ne (vector pixel, vector pixel);
11418 int vec_all_ne (vector signed int, vector bool int);
11419 int vec_all_ne (vector signed int, vector signed int);
11420 int vec_all_ne (vector unsigned int, vector bool int);
11421 int vec_all_ne (vector unsigned int, vector unsigned int);
11422 int vec_all_ne (vector bool int, vector bool int);
11423 int vec_all_ne (vector bool int, vector unsigned int);
11424 int vec_all_ne (vector bool int, vector signed int);
11425 int vec_all_ne (vector float, vector float);
11427 int vec_all_nge (vector float, vector float);
11429 int vec_all_ngt (vector float, vector float);
11431 int vec_all_nle (vector float, vector float);
11433 int vec_all_nlt (vector float, vector float);
11435 int vec_all_numeric (vector float);
11437 int vec_any_eq (vector signed char, vector bool char);
11438 int vec_any_eq (vector signed char, vector signed char);
11439 int vec_any_eq (vector unsigned char, vector bool char);
11440 int vec_any_eq (vector unsigned char, vector unsigned char);
11441 int vec_any_eq (vector bool char, vector bool char);
11442 int vec_any_eq (vector bool char, vector unsigned char);
11443 int vec_any_eq (vector bool char, vector signed char);
11444 int vec_any_eq (vector signed short, vector bool short);
11445 int vec_any_eq (vector signed short, vector signed short);
11446 int vec_any_eq (vector unsigned short, vector bool short);
11447 int vec_any_eq (vector unsigned short, vector unsigned short);
11448 int vec_any_eq (vector bool short, vector bool short);
11449 int vec_any_eq (vector bool short, vector unsigned short);
11450 int vec_any_eq (vector bool short, vector signed short);
11451 int vec_any_eq (vector pixel, vector pixel);
11452 int vec_any_eq (vector signed int, vector bool int);
11453 int vec_any_eq (vector signed int, vector signed int);
11454 int vec_any_eq (vector unsigned int, vector bool int);
11455 int vec_any_eq (vector unsigned int, vector unsigned int);
11456 int vec_any_eq (vector bool int, vector bool int);
11457 int vec_any_eq (vector bool int, vector unsigned int);
11458 int vec_any_eq (vector bool int, vector signed int);
11459 int vec_any_eq (vector float, vector float);
11461 int vec_any_ge (vector signed char, vector bool char);
11462 int vec_any_ge (vector unsigned char, vector bool char);
11463 int vec_any_ge (vector unsigned char, vector unsigned char);
11464 int vec_any_ge (vector signed char, vector signed char);
11465 int vec_any_ge (vector bool char, vector unsigned char);
11466 int vec_any_ge (vector bool char, vector signed char);
11467 int vec_any_ge (vector unsigned short, vector bool short);
11468 int vec_any_ge (vector unsigned short, vector unsigned short);
11469 int vec_any_ge (vector signed short, vector signed short);
11470 int vec_any_ge (vector signed short, vector bool short);
11471 int vec_any_ge (vector bool short, vector unsigned short);
11472 int vec_any_ge (vector bool short, vector signed short);
11473 int vec_any_ge (vector signed int, vector bool int);
11474 int vec_any_ge (vector unsigned int, vector bool int);
11475 int vec_any_ge (vector unsigned int, vector unsigned int);
11476 int vec_any_ge (vector signed int, vector signed int);
11477 int vec_any_ge (vector bool int, vector unsigned int);
11478 int vec_any_ge (vector bool int, vector signed int);
11479 int vec_any_ge (vector float, vector float);
11481 int vec_any_gt (vector bool char, vector unsigned char);
11482 int vec_any_gt (vector unsigned char, vector bool char);
11483 int vec_any_gt (vector unsigned char, vector unsigned char);
11484 int vec_any_gt (vector bool char, vector signed char);
11485 int vec_any_gt (vector signed char, vector bool char);
11486 int vec_any_gt (vector signed char, vector signed char);
11487 int vec_any_gt (vector bool short, vector unsigned short);
11488 int vec_any_gt (vector unsigned short, vector bool short);
11489 int vec_any_gt (vector unsigned short, vector unsigned short);
11490 int vec_any_gt (vector bool short, vector signed short);
11491 int vec_any_gt (vector signed short, vector bool short);
11492 int vec_any_gt (vector signed short, vector signed short);
11493 int vec_any_gt (vector bool int, vector unsigned int);
11494 int vec_any_gt (vector unsigned int, vector bool int);
11495 int vec_any_gt (vector unsigned int, vector unsigned int);
11496 int vec_any_gt (vector bool int, vector signed int);
11497 int vec_any_gt (vector signed int, vector bool int);
11498 int vec_any_gt (vector signed int, vector signed int);
11499 int vec_any_gt (vector float, vector float);
11501 int vec_any_le (vector bool char, vector unsigned char);
11502 int vec_any_le (vector unsigned char, vector bool char);
11503 int vec_any_le (vector unsigned char, vector unsigned char);
11504 int vec_any_le (vector bool char, vector signed char);
11505 int vec_any_le (vector signed char, vector bool char);
11506 int vec_any_le (vector signed char, vector signed char);
11507 int vec_any_le (vector bool short, vector unsigned short);
11508 int vec_any_le (vector unsigned short, vector bool short);
11509 int vec_any_le (vector unsigned short, vector unsigned short);
11510 int vec_any_le (vector bool short, vector signed short);
11511 int vec_any_le (vector signed short, vector bool short);
11512 int vec_any_le (vector signed short, vector signed short);
11513 int vec_any_le (vector bool int, vector unsigned int);
11514 int vec_any_le (vector unsigned int, vector bool int);
11515 int vec_any_le (vector unsigned int, vector unsigned int);
11516 int vec_any_le (vector bool int, vector signed int);
11517 int vec_any_le (vector signed int, vector bool int);
11518 int vec_any_le (vector signed int, vector signed int);
11519 int vec_any_le (vector float, vector float);
11521 int vec_any_lt (vector bool char, vector unsigned char);
11522 int vec_any_lt (vector unsigned char, vector bool char);
11523 int vec_any_lt (vector unsigned char, vector unsigned char);
11524 int vec_any_lt (vector bool char, vector signed char);
11525 int vec_any_lt (vector signed char, vector bool char);
11526 int vec_any_lt (vector signed char, vector signed char);
11527 int vec_any_lt (vector bool short, vector unsigned short);
11528 int vec_any_lt (vector unsigned short, vector bool short);
11529 int vec_any_lt (vector unsigned short, vector unsigned short);
11530 int vec_any_lt (vector bool short, vector signed short);
11531 int vec_any_lt (vector signed short, vector bool short);
11532 int vec_any_lt (vector signed short, vector signed short);
11533 int vec_any_lt (vector bool int, vector unsigned int);
11534 int vec_any_lt (vector unsigned int, vector bool int);
11535 int vec_any_lt (vector unsigned int, vector unsigned int);
11536 int vec_any_lt (vector bool int, vector signed int);
11537 int vec_any_lt (vector signed int, vector bool int);
11538 int vec_any_lt (vector signed int, vector signed int);
11539 int vec_any_lt (vector float, vector float);
11541 int vec_any_nan (vector float);
11543 int vec_any_ne (vector signed char, vector bool char);
11544 int vec_any_ne (vector signed char, vector signed char);
11545 int vec_any_ne (vector unsigned char, vector bool char);
11546 int vec_any_ne (vector unsigned char, vector unsigned char);
11547 int vec_any_ne (vector bool char, vector bool char);
11548 int vec_any_ne (vector bool char, vector unsigned char);
11549 int vec_any_ne (vector bool char, vector signed char);
11550 int vec_any_ne (vector signed short, vector bool short);
11551 int vec_any_ne (vector signed short, vector signed short);
11552 int vec_any_ne (vector unsigned short, vector bool short);
11553 int vec_any_ne (vector unsigned short, vector unsigned short);
11554 int vec_any_ne (vector bool short, vector bool short);
11555 int vec_any_ne (vector bool short, vector unsigned short);
11556 int vec_any_ne (vector bool short, vector signed short);
11557 int vec_any_ne (vector pixel, vector pixel);
11558 int vec_any_ne (vector signed int, vector bool int);
11559 int vec_any_ne (vector signed int, vector signed int);
11560 int vec_any_ne (vector unsigned int, vector bool int);
11561 int vec_any_ne (vector unsigned int, vector unsigned int);
11562 int vec_any_ne (vector bool int, vector bool int);
11563 int vec_any_ne (vector bool int, vector unsigned int);
11564 int vec_any_ne (vector bool int, vector signed int);
11565 int vec_any_ne (vector float, vector float);
11567 int vec_any_nge (vector float, vector float);
11569 int vec_any_ngt (vector float, vector float);
11571 int vec_any_nle (vector float, vector float);
11573 int vec_any_nlt (vector float, vector float);
11575 int vec_any_numeric (vector float);
11577 int vec_any_out (vector float, vector float);
11580 @node SPARC VIS Built-in Functions
11581 @subsection SPARC VIS Built-in Functions
11583 GCC supports SIMD operations on the SPARC using both the generic vector
11584 extensions (@pxref{Vector Extensions}) as well as built-in functions for
11585 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
11586 switch, the VIS extension is exposed as the following built-in functions:
11589 typedef int v2si __attribute__ ((vector_size (8)));
11590 typedef short v4hi __attribute__ ((vector_size (8)));
11591 typedef short v2hi __attribute__ ((vector_size (4)));
11592 typedef char v8qi __attribute__ ((vector_size (8)));
11593 typedef char v4qi __attribute__ ((vector_size (4)));
11595 void * __builtin_vis_alignaddr (void *, long);
11596 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
11597 v2si __builtin_vis_faligndatav2si (v2si, v2si);
11598 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
11599 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
11601 v4hi __builtin_vis_fexpand (v4qi);
11603 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
11604 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
11605 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
11606 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
11607 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
11608 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
11609 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
11611 v4qi __builtin_vis_fpack16 (v4hi);
11612 v8qi __builtin_vis_fpack32 (v2si, v2si);
11613 v2hi __builtin_vis_fpackfix (v2si);
11614 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
11616 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
11619 @node SPU Built-in Functions
11620 @subsection SPU Built-in Functions
11622 GCC provides extensions for the SPU processor as described in the
11623 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
11624 found at @uref{http://cell.scei.co.jp/} or
11625 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
11626 implementation differs in several ways.
11631 The optional extension of specifying vector constants in parentheses is
11635 A vector initializer requires no cast if the vector constant is of the
11636 same type as the variable it is initializing.
11639 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11640 vector type is the default signedness of the base type. The default
11641 varies depending on the operating system, so a portable program should
11642 always specify the signedness.
11645 By default, the keyword @code{__vector} is added. The macro
11646 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
11650 GCC allows using a @code{typedef} name as the type specifier for a
11654 For C, overloaded functions are implemented with macros so the following
11658 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11661 Since @code{spu_add} is a macro, the vector constant in the example
11662 is treated as four separate arguments. Wrap the entire argument in
11663 parentheses for this to work.
11666 The extended version of @code{__builtin_expect} is not supported.
11670 @emph{Note:} Only the interface described in the aforementioned
11671 specification is supported. Internally, GCC uses built-in functions to
11672 implement the required functionality, but these are not supported and
11673 are subject to change without notice.
11675 @node Target Format Checks
11676 @section Format Checks Specific to Particular Target Machines
11678 For some target machines, GCC supports additional options to the
11680 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
11683 * Solaris Format Checks::
11686 @node Solaris Format Checks
11687 @subsection Solaris Format Checks
11689 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
11690 check. @code{cmn_err} accepts a subset of the standard @code{printf}
11691 conversions, and the two-argument @code{%b} conversion for displaying
11692 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
11695 @section Pragmas Accepted by GCC
11699 GCC supports several types of pragmas, primarily in order to compile
11700 code originally written for other compilers. Note that in general
11701 we do not recommend the use of pragmas; @xref{Function Attributes},
11702 for further explanation.
11707 * RS/6000 and PowerPC Pragmas::
11709 * Solaris Pragmas::
11710 * Symbol-Renaming Pragmas::
11711 * Structure-Packing Pragmas::
11713 * Diagnostic Pragmas::
11714 * Visibility Pragmas::
11715 * Push/Pop Macro Pragmas::
11716 * Function Specific Option Pragmas::
11720 @subsection ARM Pragmas
11722 The ARM target defines pragmas for controlling the default addition of
11723 @code{long_call} and @code{short_call} attributes to functions.
11724 @xref{Function Attributes}, for information about the effects of these
11729 @cindex pragma, long_calls
11730 Set all subsequent functions to have the @code{long_call} attribute.
11732 @item no_long_calls
11733 @cindex pragma, no_long_calls
11734 Set all subsequent functions to have the @code{short_call} attribute.
11736 @item long_calls_off
11737 @cindex pragma, long_calls_off
11738 Do not affect the @code{long_call} or @code{short_call} attributes of
11739 subsequent functions.
11743 @subsection M32C Pragmas
11746 @item memregs @var{number}
11747 @cindex pragma, memregs
11748 Overrides the command line option @code{-memregs=} for the current
11749 file. Use with care! This pragma must be before any function in the
11750 file, and mixing different memregs values in different objects may
11751 make them incompatible. This pragma is useful when a
11752 performance-critical function uses a memreg for temporary values,
11753 as it may allow you to reduce the number of memregs used.
11757 @node RS/6000 and PowerPC Pragmas
11758 @subsection RS/6000 and PowerPC Pragmas
11760 The RS/6000 and PowerPC targets define one pragma for controlling
11761 whether or not the @code{longcall} attribute is added to function
11762 declarations by default. This pragma overrides the @option{-mlongcall}
11763 option, but not the @code{longcall} and @code{shortcall} attributes.
11764 @xref{RS/6000 and PowerPC Options}, for more information about when long
11765 calls are and are not necessary.
11769 @cindex pragma, longcall
11770 Apply the @code{longcall} attribute to all subsequent function
11774 Do not apply the @code{longcall} attribute to subsequent function
11778 @c Describe h8300 pragmas here.
11779 @c Describe sh pragmas here.
11780 @c Describe v850 pragmas here.
11782 @node Darwin Pragmas
11783 @subsection Darwin Pragmas
11785 The following pragmas are available for all architectures running the
11786 Darwin operating system. These are useful for compatibility with other
11790 @item mark @var{tokens}@dots{}
11791 @cindex pragma, mark
11792 This pragma is accepted, but has no effect.
11794 @item options align=@var{alignment}
11795 @cindex pragma, options align
11796 This pragma sets the alignment of fields in structures. The values of
11797 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
11798 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
11799 properly; to restore the previous setting, use @code{reset} for the
11802 @item segment @var{tokens}@dots{}
11803 @cindex pragma, segment
11804 This pragma is accepted, but has no effect.
11806 @item unused (@var{var} [, @var{var}]@dots{})
11807 @cindex pragma, unused
11808 This pragma declares variables to be possibly unused. GCC will not
11809 produce warnings for the listed variables. The effect is similar to
11810 that of the @code{unused} attribute, except that this pragma may appear
11811 anywhere within the variables' scopes.
11814 @node Solaris Pragmas
11815 @subsection Solaris Pragmas
11817 The Solaris target supports @code{#pragma redefine_extname}
11818 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
11819 @code{#pragma} directives for compatibility with the system compiler.
11822 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
11823 @cindex pragma, align
11825 Increase the minimum alignment of each @var{variable} to @var{alignment}.
11826 This is the same as GCC's @code{aligned} attribute @pxref{Variable
11827 Attributes}). Macro expansion occurs on the arguments to this pragma
11828 when compiling C and Objective-C@. It does not currently occur when
11829 compiling C++, but this is a bug which may be fixed in a future
11832 @item fini (@var{function} [, @var{function}]...)
11833 @cindex pragma, fini
11835 This pragma causes each listed @var{function} to be called after
11836 main, or during shared module unloading, by adding a call to the
11837 @code{.fini} section.
11839 @item init (@var{function} [, @var{function}]...)
11840 @cindex pragma, init
11842 This pragma causes each listed @var{function} to be called during
11843 initialization (before @code{main}) or during shared module loading, by
11844 adding a call to the @code{.init} section.
11848 @node Symbol-Renaming Pragmas
11849 @subsection Symbol-Renaming Pragmas
11851 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
11852 supports two @code{#pragma} directives which change the name used in
11853 assembly for a given declaration. These pragmas are only available on
11854 platforms whose system headers need them. To get this effect on all
11855 platforms supported by GCC, use the asm labels extension (@pxref{Asm
11859 @item redefine_extname @var{oldname} @var{newname}
11860 @cindex pragma, redefine_extname
11862 This pragma gives the C function @var{oldname} the assembly symbol
11863 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
11864 will be defined if this pragma is available (currently only on
11867 @item extern_prefix @var{string}
11868 @cindex pragma, extern_prefix
11870 This pragma causes all subsequent external function and variable
11871 declarations to have @var{string} prepended to their assembly symbols.
11872 This effect may be terminated with another @code{extern_prefix} pragma
11873 whose argument is an empty string. The preprocessor macro
11874 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
11875 available (currently only on Tru64 UNIX)@.
11878 These pragmas and the asm labels extension interact in a complicated
11879 manner. Here are some corner cases you may want to be aware of.
11882 @item Both pragmas silently apply only to declarations with external
11883 linkage. Asm labels do not have this restriction.
11885 @item In C++, both pragmas silently apply only to declarations with
11886 ``C'' linkage. Again, asm labels do not have this restriction.
11888 @item If any of the three ways of changing the assembly name of a
11889 declaration is applied to a declaration whose assembly name has
11890 already been determined (either by a previous use of one of these
11891 features, or because the compiler needed the assembly name in order to
11892 generate code), and the new name is different, a warning issues and
11893 the name does not change.
11895 @item The @var{oldname} used by @code{#pragma redefine_extname} is
11896 always the C-language name.
11898 @item If @code{#pragma extern_prefix} is in effect, and a declaration
11899 occurs with an asm label attached, the prefix is silently ignored for
11902 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
11903 apply to the same declaration, whichever triggered first wins, and a
11904 warning issues if they contradict each other. (We would like to have
11905 @code{#pragma redefine_extname} always win, for consistency with asm
11906 labels, but if @code{#pragma extern_prefix} triggers first we have no
11907 way of knowing that that happened.)
11910 @node Structure-Packing Pragmas
11911 @subsection Structure-Packing Pragmas
11913 For compatibility with Microsoft Windows compilers, GCC supports a
11914 set of @code{#pragma} directives which change the maximum alignment of
11915 members of structures (other than zero-width bitfields), unions, and
11916 classes subsequently defined. The @var{n} value below always is required
11917 to be a small power of two and specifies the new alignment in bytes.
11920 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
11921 @item @code{#pragma pack()} sets the alignment to the one that was in
11922 effect when compilation started (see also command line option
11923 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
11924 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
11925 setting on an internal stack and then optionally sets the new alignment.
11926 @item @code{#pragma pack(pop)} restores the alignment setting to the one
11927 saved at the top of the internal stack (and removes that stack entry).
11928 Note that @code{#pragma pack([@var{n}])} does not influence this internal
11929 stack; thus it is possible to have @code{#pragma pack(push)} followed by
11930 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
11931 @code{#pragma pack(pop)}.
11934 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
11935 @code{#pragma} which lays out a structure as the documented
11936 @code{__attribute__ ((ms_struct))}.
11938 @item @code{#pragma ms_struct on} turns on the layout for structures
11940 @item @code{#pragma ms_struct off} turns off the layout for structures
11942 @item @code{#pragma ms_struct reset} goes back to the default layout.
11946 @subsection Weak Pragmas
11948 For compatibility with SVR4, GCC supports a set of @code{#pragma}
11949 directives for declaring symbols to be weak, and defining weak
11953 @item #pragma weak @var{symbol}
11954 @cindex pragma, weak
11955 This pragma declares @var{symbol} to be weak, as if the declaration
11956 had the attribute of the same name. The pragma may appear before
11957 or after the declaration of @var{symbol}, but must appear before
11958 either its first use or its definition. It is not an error for
11959 @var{symbol} to never be defined at all.
11961 @item #pragma weak @var{symbol1} = @var{symbol2}
11962 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
11963 It is an error if @var{symbol2} is not defined in the current
11967 @node Diagnostic Pragmas
11968 @subsection Diagnostic Pragmas
11970 GCC allows the user to selectively enable or disable certain types of
11971 diagnostics, and change the kind of the diagnostic. For example, a
11972 project's policy might require that all sources compile with
11973 @option{-Werror} but certain files might have exceptions allowing
11974 specific types of warnings. Or, a project might selectively enable
11975 diagnostics and treat them as errors depending on which preprocessor
11976 macros are defined.
11979 @item #pragma GCC diagnostic @var{kind} @var{option}
11980 @cindex pragma, diagnostic
11982 Modifies the disposition of a diagnostic. Note that not all
11983 diagnostics are modifiable; at the moment only warnings (normally
11984 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
11985 Use @option{-fdiagnostics-show-option} to determine which diagnostics
11986 are controllable and which option controls them.
11988 @var{kind} is @samp{error} to treat this diagnostic as an error,
11989 @samp{warning} to treat it like a warning (even if @option{-Werror} is
11990 in effect), or @samp{ignored} if the diagnostic is to be ignored.
11991 @var{option} is a double quoted string which matches the command line
11995 #pragma GCC diagnostic warning "-Wformat"
11996 #pragma GCC diagnostic error "-Wformat"
11997 #pragma GCC diagnostic ignored "-Wformat"
12000 Note that these pragmas override any command line options. Also,
12001 while it is syntactically valid to put these pragmas anywhere in your
12002 sources, the only supported location for them is before any data or
12003 functions are defined. Doing otherwise may result in unpredictable
12004 results depending on how the optimizer manages your sources. If the
12005 same option is listed multiple times, the last one specified is the
12006 one that is in effect. This pragma is not intended to be a general
12007 purpose replacement for command line options, but for implementing
12008 strict control over project policies.
12012 GCC also offers a simple mechanism for printing messages during
12016 @item #pragma message @var{string}
12017 @cindex pragma, diagnostic
12019 Prints @var{string} as a compiler message on compilation. The message
12020 is informational only, and is neither a compilation warning nor an error.
12023 #pragma message "Compiling " __FILE__ "..."
12026 @var{string} may be parenthesized, and is printed with location
12027 information. For example,
12030 #define DO_PRAGMA(x) _Pragma (#x)
12031 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
12033 TODO(Remember to fix this)
12036 prints @samp{/tmp/file.c:4: note: #pragma message:
12037 TODO - Remember to fix this}.
12041 @node Visibility Pragmas
12042 @subsection Visibility Pragmas
12045 @item #pragma GCC visibility push(@var{visibility})
12046 @itemx #pragma GCC visibility pop
12047 @cindex pragma, visibility
12049 This pragma allows the user to set the visibility for multiple
12050 declarations without having to give each a visibility attribute
12051 @xref{Function Attributes}, for more information about visibility and
12052 the attribute syntax.
12054 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
12055 declarations. Class members and template specializations are not
12056 affected; if you want to override the visibility for a particular
12057 member or instantiation, you must use an attribute.
12062 @node Push/Pop Macro Pragmas
12063 @subsection Push/Pop Macro Pragmas
12065 For compatibility with Microsoft Windows compilers, GCC supports
12066 @samp{#pragma push_macro(@var{"macro_name"})}
12067 and @samp{#pragma pop_macro(@var{"macro_name"})}.
12070 @item #pragma push_macro(@var{"macro_name"})
12071 @cindex pragma, push_macro
12072 This pragma saves the value of the macro named as @var{macro_name} to
12073 the top of the stack for this macro.
12075 @item #pragma pop_macro(@var{"macro_name"})
12076 @cindex pragma, pop_macro
12077 This pragma sets the value of the macro named as @var{macro_name} to
12078 the value on top of the stack for this macro. If the stack for
12079 @var{macro_name} is empty, the value of the macro remains unchanged.
12086 #pragma push_macro("X")
12089 #pragma pop_macro("X")
12093 In this example, the definition of X as 1 is saved by @code{#pragma
12094 push_macro} and restored by @code{#pragma pop_macro}.
12096 @node Function Specific Option Pragmas
12097 @subsection Function Specific Option Pragmas
12100 @item #pragma GCC target (@var{"string"}...)
12101 @cindex pragma GCC target
12103 This pragma allows you to set target specific options for functions
12104 defined later in the source file. One or more strings can be
12105 specified. Each function that is defined after this point will be as
12106 if @code{attribute((target("STRING")))} was specified for that
12107 function. The parenthesis around the options is optional.
12108 @xref{Function Attributes}, for more information about the
12109 @code{target} attribute and the attribute syntax.
12111 The @samp{#pragma GCC target} pragma is not implemented in GCC
12112 versions earlier than 4.4, and is currently only implemented for the
12113 386 and x86_64 backends.
12117 @item #pragma GCC optimize (@var{"string"}...)
12118 @cindex pragma GCC optimize
12120 This pragma allows you to set global optimization options for functions
12121 defined later in the source file. One or more strings can be
12122 specified. Each function that is defined after this point will be as
12123 if @code{attribute((optimize("STRING")))} was specified for that
12124 function. The parenthesis around the options is optional.
12125 @xref{Function Attributes}, for more information about the
12126 @code{optimize} attribute and the attribute syntax.
12128 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
12129 versions earlier than 4.4.
12133 @item #pragma GCC push_options
12134 @itemx #pragma GCC pop_options
12135 @cindex pragma GCC push_options
12136 @cindex pragma GCC pop_options
12138 These pragmas maintain a stack of the current target and optimization
12139 options. It is intended for include files where you temporarily want
12140 to switch to using a different @samp{#pragma GCC target} or
12141 @samp{#pragma GCC optimize} and then to pop back to the previous
12144 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
12145 pragmas are not implemented in GCC versions earlier than 4.4.
12149 @item #pragma GCC reset_options
12150 @cindex pragma GCC reset_options
12152 This pragma clears the current @code{#pragma GCC target} and
12153 @code{#pragma GCC optimize} to use the default switches as specified
12154 on the command line.
12156 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
12157 versions earlier than 4.4.
12160 @node Unnamed Fields
12161 @section Unnamed struct/union fields within structs/unions
12165 For compatibility with other compilers, GCC allows you to define
12166 a structure or union that contains, as fields, structures and unions
12167 without names. For example:
12180 In this example, the user would be able to access members of the unnamed
12181 union with code like @samp{foo.b}. Note that only unnamed structs and
12182 unions are allowed, you may not have, for example, an unnamed
12185 You must never create such structures that cause ambiguous field definitions.
12186 For example, this structure:
12197 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
12198 Such constructs are not supported and must be avoided. In the future,
12199 such constructs may be detected and treated as compilation errors.
12201 @opindex fms-extensions
12202 Unless @option{-fms-extensions} is used, the unnamed field must be a
12203 structure or union definition without a tag (for example, @samp{struct
12204 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
12205 also be a definition with a tag such as @samp{struct foo @{ int a;
12206 @};}, a reference to a previously defined structure or union such as
12207 @samp{struct foo;}, or a reference to a @code{typedef} name for a
12208 previously defined structure or union type.
12211 @section Thread-Local Storage
12212 @cindex Thread-Local Storage
12213 @cindex @acronym{TLS}
12216 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
12217 are allocated such that there is one instance of the variable per extant
12218 thread. The run-time model GCC uses to implement this originates
12219 in the IA-64 processor-specific ABI, but has since been migrated
12220 to other processors as well. It requires significant support from
12221 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
12222 system libraries (@file{libc.so} and @file{libpthread.so}), so it
12223 is not available everywhere.
12225 At the user level, the extension is visible with a new storage
12226 class keyword: @code{__thread}. For example:
12230 extern __thread struct state s;
12231 static __thread char *p;
12234 The @code{__thread} specifier may be used alone, with the @code{extern}
12235 or @code{static} specifiers, but with no other storage class specifier.
12236 When used with @code{extern} or @code{static}, @code{__thread} must appear
12237 immediately after the other storage class specifier.
12239 The @code{__thread} specifier may be applied to any global, file-scoped
12240 static, function-scoped static, or static data member of a class. It may
12241 not be applied to block-scoped automatic or non-static data member.
12243 When the address-of operator is applied to a thread-local variable, it is
12244 evaluated at run-time and returns the address of the current thread's
12245 instance of that variable. An address so obtained may be used by any
12246 thread. When a thread terminates, any pointers to thread-local variables
12247 in that thread become invalid.
12249 No static initialization may refer to the address of a thread-local variable.
12251 In C++, if an initializer is present for a thread-local variable, it must
12252 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
12255 See @uref{http://people.redhat.com/drepper/tls.pdf,
12256 ELF Handling For Thread-Local Storage} for a detailed explanation of
12257 the four thread-local storage addressing models, and how the run-time
12258 is expected to function.
12261 * C99 Thread-Local Edits::
12262 * C++98 Thread-Local Edits::
12265 @node C99 Thread-Local Edits
12266 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
12268 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
12269 that document the exact semantics of the language extension.
12273 @cite{5.1.2 Execution environments}
12275 Add new text after paragraph 1
12278 Within either execution environment, a @dfn{thread} is a flow of
12279 control within a program. It is implementation defined whether
12280 or not there may be more than one thread associated with a program.
12281 It is implementation defined how threads beyond the first are
12282 created, the name and type of the function called at thread
12283 startup, and how threads may be terminated. However, objects
12284 with thread storage duration shall be initialized before thread
12289 @cite{6.2.4 Storage durations of objects}
12291 Add new text before paragraph 3
12294 An object whose identifier is declared with the storage-class
12295 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
12296 Its lifetime is the entire execution of the thread, and its
12297 stored value is initialized only once, prior to thread startup.
12301 @cite{6.4.1 Keywords}
12303 Add @code{__thread}.
12306 @cite{6.7.1 Storage-class specifiers}
12308 Add @code{__thread} to the list of storage class specifiers in
12311 Change paragraph 2 to
12314 With the exception of @code{__thread}, at most one storage-class
12315 specifier may be given [@dots{}]. The @code{__thread} specifier may
12316 be used alone, or immediately following @code{extern} or
12320 Add new text after paragraph 6
12323 The declaration of an identifier for a variable that has
12324 block scope that specifies @code{__thread} shall also
12325 specify either @code{extern} or @code{static}.
12327 The @code{__thread} specifier shall be used only with
12332 @node C++98 Thread-Local Edits
12333 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
12335 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
12336 that document the exact semantics of the language extension.
12340 @b{[intro.execution]}
12342 New text after paragraph 4
12345 A @dfn{thread} is a flow of control within the abstract machine.
12346 It is implementation defined whether or not there may be more than
12350 New text after paragraph 7
12353 It is unspecified whether additional action must be taken to
12354 ensure when and whether side effects are visible to other threads.
12360 Add @code{__thread}.
12363 @b{[basic.start.main]}
12365 Add after paragraph 5
12368 The thread that begins execution at the @code{main} function is called
12369 the @dfn{main thread}. It is implementation defined how functions
12370 beginning threads other than the main thread are designated or typed.
12371 A function so designated, as well as the @code{main} function, is called
12372 a @dfn{thread startup function}. It is implementation defined what
12373 happens if a thread startup function returns. It is implementation
12374 defined what happens to other threads when any thread calls @code{exit}.
12378 @b{[basic.start.init]}
12380 Add after paragraph 4
12383 The storage for an object of thread storage duration shall be
12384 statically initialized before the first statement of the thread startup
12385 function. An object of thread storage duration shall not require
12386 dynamic initialization.
12390 @b{[basic.start.term]}
12392 Add after paragraph 3
12395 The type of an object with thread storage duration shall not have a
12396 non-trivial destructor, nor shall it be an array type whose elements
12397 (directly or indirectly) have non-trivial destructors.
12403 Add ``thread storage duration'' to the list in paragraph 1.
12408 Thread, static, and automatic storage durations are associated with
12409 objects introduced by declarations [@dots{}].
12412 Add @code{__thread} to the list of specifiers in paragraph 3.
12415 @b{[basic.stc.thread]}
12417 New section before @b{[basic.stc.static]}
12420 The keyword @code{__thread} applied to a non-local object gives the
12421 object thread storage duration.
12423 A local variable or class data member declared both @code{static}
12424 and @code{__thread} gives the variable or member thread storage
12429 @b{[basic.stc.static]}
12434 All objects which have neither thread storage duration, dynamic
12435 storage duration nor are local [@dots{}].
12441 Add @code{__thread} to the list in paragraph 1.
12446 With the exception of @code{__thread}, at most one
12447 @var{storage-class-specifier} shall appear in a given
12448 @var{decl-specifier-seq}. The @code{__thread} specifier may
12449 be used alone, or immediately following the @code{extern} or
12450 @code{static} specifiers. [@dots{}]
12453 Add after paragraph 5
12456 The @code{__thread} specifier can be applied only to the names of objects
12457 and to anonymous unions.
12463 Add after paragraph 6
12466 Non-@code{static} members shall not be @code{__thread}.
12470 @node Binary constants
12471 @section Binary constants using the @samp{0b} prefix
12472 @cindex Binary constants using the @samp{0b} prefix
12474 Integer constants can be written as binary constants, consisting of a
12475 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
12476 @samp{0B}. This is particularly useful in environments that operate a
12477 lot on the bit-level (like microcontrollers).
12479 The following statements are identical:
12488 The type of these constants follows the same rules as for octal or
12489 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
12492 @node C++ Extensions
12493 @chapter Extensions to the C++ Language
12494 @cindex extensions, C++ language
12495 @cindex C++ language extensions
12497 The GNU compiler provides these extensions to the C++ language (and you
12498 can also use most of the C language extensions in your C++ programs). If you
12499 want to write code that checks whether these features are available, you can
12500 test for the GNU compiler the same way as for C programs: check for a
12501 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
12502 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
12503 Predefined Macros,cpp,The GNU C Preprocessor}).
12506 * Volatiles:: What constitutes an access to a volatile object.
12507 * Restricted Pointers:: C99 restricted pointers and references.
12508 * Vague Linkage:: Where G++ puts inlines, vtables and such.
12509 * C++ Interface:: You can use a single C++ header file for both
12510 declarations and definitions.
12511 * Template Instantiation:: Methods for ensuring that exactly one copy of
12512 each needed template instantiation is emitted.
12513 * Bound member functions:: You can extract a function pointer to the
12514 method denoted by a @samp{->*} or @samp{.*} expression.
12515 * C++ Attributes:: Variable, function, and type attributes for C++ only.
12516 * Namespace Association:: Strong using-directives for namespace association.
12517 * Type Traits:: Compiler support for type traits
12518 * Java Exceptions:: Tweaking exception handling to work with Java.
12519 * Deprecated Features:: Things will disappear from g++.
12520 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
12524 @section When is a Volatile Object Accessed?
12525 @cindex accessing volatiles
12526 @cindex volatile read
12527 @cindex volatile write
12528 @cindex volatile access
12530 Both the C and C++ standard have the concept of volatile objects. These
12531 are normally accessed by pointers and used for accessing hardware. The
12532 standards encourage compilers to refrain from optimizations concerning
12533 accesses to volatile objects. The C standard leaves it implementation
12534 defined as to what constitutes a volatile access. The C++ standard omits
12535 to specify this, except to say that C++ should behave in a similar manner
12536 to C with respect to volatiles, where possible. The minimum either
12537 standard specifies is that at a sequence point all previous accesses to
12538 volatile objects have stabilized and no subsequent accesses have
12539 occurred. Thus an implementation is free to reorder and combine
12540 volatile accesses which occur between sequence points, but cannot do so
12541 for accesses across a sequence point. The use of volatiles does not
12542 allow you to violate the restriction on updating objects multiple times
12543 within a sequence point.
12545 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
12547 The behavior differs slightly between C and C++ in the non-obvious cases:
12550 volatile int *src = @var{somevalue};
12554 With C, such expressions are rvalues, and GCC interprets this either as a
12555 read of the volatile object being pointed to or only as request to evaluate
12556 the side-effects. The C++ standard specifies that such expressions do not
12557 undergo lvalue to rvalue conversion, and that the type of the dereferenced
12558 object may be incomplete. The C++ standard does not specify explicitly
12559 that it is this lvalue to rvalue conversion which may be responsible for
12560 causing an access. However, there is reason to believe that it is,
12561 because otherwise certain simple expressions become undefined. However,
12562 because it would surprise most programmers, G++ treats dereferencing a
12563 pointer to volatile object of complete type when the value is unused as
12564 GCC would do for an equivalent type in C@. When the object has incomplete
12565 type, G++ issues a warning; if you wish to force an error, you must
12566 force a conversion to rvalue with, for instance, a static cast.
12568 When using a reference to volatile, G++ does not treat equivalent
12569 expressions as accesses to volatiles, but instead issues a warning that
12570 no volatile is accessed. The rationale for this is that otherwise it
12571 becomes difficult to determine where volatile access occur, and not
12572 possible to ignore the return value from functions returning volatile
12573 references. Again, if you wish to force a read, cast the reference to
12576 @node Restricted Pointers
12577 @section Restricting Pointer Aliasing
12578 @cindex restricted pointers
12579 @cindex restricted references
12580 @cindex restricted this pointer
12582 As with the C front end, G++ understands the C99 feature of restricted pointers,
12583 specified with the @code{__restrict__}, or @code{__restrict} type
12584 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
12585 language flag, @code{restrict} is not a keyword in C++.
12587 In addition to allowing restricted pointers, you can specify restricted
12588 references, which indicate that the reference is not aliased in the local
12592 void fn (int *__restrict__ rptr, int &__restrict__ rref)
12599 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
12600 @var{rref} refers to a (different) unaliased integer.
12602 You may also specify whether a member function's @var{this} pointer is
12603 unaliased by using @code{__restrict__} as a member function qualifier.
12606 void T::fn () __restrict__
12613 Within the body of @code{T::fn}, @var{this} will have the effective
12614 definition @code{T *__restrict__ const this}. Notice that the
12615 interpretation of a @code{__restrict__} member function qualifier is
12616 different to that of @code{const} or @code{volatile} qualifier, in that it
12617 is applied to the pointer rather than the object. This is consistent with
12618 other compilers which implement restricted pointers.
12620 As with all outermost parameter qualifiers, @code{__restrict__} is
12621 ignored in function definition matching. This means you only need to
12622 specify @code{__restrict__} in a function definition, rather than
12623 in a function prototype as well.
12625 @node Vague Linkage
12626 @section Vague Linkage
12627 @cindex vague linkage
12629 There are several constructs in C++ which require space in the object
12630 file but are not clearly tied to a single translation unit. We say that
12631 these constructs have ``vague linkage''. Typically such constructs are
12632 emitted wherever they are needed, though sometimes we can be more
12636 @item Inline Functions
12637 Inline functions are typically defined in a header file which can be
12638 included in many different compilations. Hopefully they can usually be
12639 inlined, but sometimes an out-of-line copy is necessary, if the address
12640 of the function is taken or if inlining fails. In general, we emit an
12641 out-of-line copy in all translation units where one is needed. As an
12642 exception, we only emit inline virtual functions with the vtable, since
12643 it will always require a copy.
12645 Local static variables and string constants used in an inline function
12646 are also considered to have vague linkage, since they must be shared
12647 between all inlined and out-of-line instances of the function.
12651 C++ virtual functions are implemented in most compilers using a lookup
12652 table, known as a vtable. The vtable contains pointers to the virtual
12653 functions provided by a class, and each object of the class contains a
12654 pointer to its vtable (or vtables, in some multiple-inheritance
12655 situations). If the class declares any non-inline, non-pure virtual
12656 functions, the first one is chosen as the ``key method'' for the class,
12657 and the vtable is only emitted in the translation unit where the key
12660 @emph{Note:} If the chosen key method is later defined as inline, the
12661 vtable will still be emitted in every translation unit which defines it.
12662 Make sure that any inline virtuals are declared inline in the class
12663 body, even if they are not defined there.
12665 @item type_info objects
12668 C++ requires information about types to be written out in order to
12669 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
12670 For polymorphic classes (classes with virtual functions), the type_info
12671 object is written out along with the vtable so that @samp{dynamic_cast}
12672 can determine the dynamic type of a class object at runtime. For all
12673 other types, we write out the type_info object when it is used: when
12674 applying @samp{typeid} to an expression, throwing an object, or
12675 referring to a type in a catch clause or exception specification.
12677 @item Template Instantiations
12678 Most everything in this section also applies to template instantiations,
12679 but there are other options as well.
12680 @xref{Template Instantiation,,Where's the Template?}.
12684 When used with GNU ld version 2.8 or later on an ELF system such as
12685 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
12686 these constructs will be discarded at link time. This is known as
12689 On targets that don't support COMDAT, but do support weak symbols, GCC
12690 will use them. This way one copy will override all the others, but
12691 the unused copies will still take up space in the executable.
12693 For targets which do not support either COMDAT or weak symbols,
12694 most entities with vague linkage will be emitted as local symbols to
12695 avoid duplicate definition errors from the linker. This will not happen
12696 for local statics in inlines, however, as having multiple copies will
12697 almost certainly break things.
12699 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
12700 another way to control placement of these constructs.
12702 @node C++ Interface
12703 @section #pragma interface and implementation
12705 @cindex interface and implementation headers, C++
12706 @cindex C++ interface and implementation headers
12707 @cindex pragmas, interface and implementation
12709 @code{#pragma interface} and @code{#pragma implementation} provide the
12710 user with a way of explicitly directing the compiler to emit entities
12711 with vague linkage (and debugging information) in a particular
12714 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
12715 most cases, because of COMDAT support and the ``key method'' heuristic
12716 mentioned in @ref{Vague Linkage}. Using them can actually cause your
12717 program to grow due to unnecessary out-of-line copies of inline
12718 functions. Currently (3.4) the only benefit of these
12719 @code{#pragma}s is reduced duplication of debugging information, and
12720 that should be addressed soon on DWARF 2 targets with the use of
12724 @item #pragma interface
12725 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
12726 @kindex #pragma interface
12727 Use this directive in @emph{header files} that define object classes, to save
12728 space in most of the object files that use those classes. Normally,
12729 local copies of certain information (backup copies of inline member
12730 functions, debugging information, and the internal tables that implement
12731 virtual functions) must be kept in each object file that includes class
12732 definitions. You can use this pragma to avoid such duplication. When a
12733 header file containing @samp{#pragma interface} is included in a
12734 compilation, this auxiliary information will not be generated (unless
12735 the main input source file itself uses @samp{#pragma implementation}).
12736 Instead, the object files will contain references to be resolved at link
12739 The second form of this directive is useful for the case where you have
12740 multiple headers with the same name in different directories. If you
12741 use this form, you must specify the same string to @samp{#pragma
12744 @item #pragma implementation
12745 @itemx #pragma implementation "@var{objects}.h"
12746 @kindex #pragma implementation
12747 Use this pragma in a @emph{main input file}, when you want full output from
12748 included header files to be generated (and made globally visible). The
12749 included header file, in turn, should use @samp{#pragma interface}.
12750 Backup copies of inline member functions, debugging information, and the
12751 internal tables used to implement virtual functions are all generated in
12752 implementation files.
12754 @cindex implied @code{#pragma implementation}
12755 @cindex @code{#pragma implementation}, implied
12756 @cindex naming convention, implementation headers
12757 If you use @samp{#pragma implementation} with no argument, it applies to
12758 an include file with the same basename@footnote{A file's @dfn{basename}
12759 was the name stripped of all leading path information and of trailing
12760 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
12761 file. For example, in @file{allclass.cc}, giving just
12762 @samp{#pragma implementation}
12763 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
12765 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
12766 an implementation file whenever you would include it from
12767 @file{allclass.cc} even if you never specified @samp{#pragma
12768 implementation}. This was deemed to be more trouble than it was worth,
12769 however, and disabled.
12771 Use the string argument if you want a single implementation file to
12772 include code from multiple header files. (You must also use
12773 @samp{#include} to include the header file; @samp{#pragma
12774 implementation} only specifies how to use the file---it doesn't actually
12777 There is no way to split up the contents of a single header file into
12778 multiple implementation files.
12781 @cindex inlining and C++ pragmas
12782 @cindex C++ pragmas, effect on inlining
12783 @cindex pragmas in C++, effect on inlining
12784 @samp{#pragma implementation} and @samp{#pragma interface} also have an
12785 effect on function inlining.
12787 If you define a class in a header file marked with @samp{#pragma
12788 interface}, the effect on an inline function defined in that class is
12789 similar to an explicit @code{extern} declaration---the compiler emits
12790 no code at all to define an independent version of the function. Its
12791 definition is used only for inlining with its callers.
12793 @opindex fno-implement-inlines
12794 Conversely, when you include the same header file in a main source file
12795 that declares it as @samp{#pragma implementation}, the compiler emits
12796 code for the function itself; this defines a version of the function
12797 that can be found via pointers (or by callers compiled without
12798 inlining). If all calls to the function can be inlined, you can avoid
12799 emitting the function by compiling with @option{-fno-implement-inlines}.
12800 If any calls were not inlined, you will get linker errors.
12802 @node Template Instantiation
12803 @section Where's the Template?
12804 @cindex template instantiation
12806 C++ templates are the first language feature to require more
12807 intelligence from the environment than one usually finds on a UNIX
12808 system. Somehow the compiler and linker have to make sure that each
12809 template instance occurs exactly once in the executable if it is needed,
12810 and not at all otherwise. There are two basic approaches to this
12811 problem, which are referred to as the Borland model and the Cfront model.
12814 @item Borland model
12815 Borland C++ solved the template instantiation problem by adding the code
12816 equivalent of common blocks to their linker; the compiler emits template
12817 instances in each translation unit that uses them, and the linker
12818 collapses them together. The advantage of this model is that the linker
12819 only has to consider the object files themselves; there is no external
12820 complexity to worry about. This disadvantage is that compilation time
12821 is increased because the template code is being compiled repeatedly.
12822 Code written for this model tends to include definitions of all
12823 templates in the header file, since they must be seen to be
12827 The AT&T C++ translator, Cfront, solved the template instantiation
12828 problem by creating the notion of a template repository, an
12829 automatically maintained place where template instances are stored. A
12830 more modern version of the repository works as follows: As individual
12831 object files are built, the compiler places any template definitions and
12832 instantiations encountered in the repository. At link time, the link
12833 wrapper adds in the objects in the repository and compiles any needed
12834 instances that were not previously emitted. The advantages of this
12835 model are more optimal compilation speed and the ability to use the
12836 system linker; to implement the Borland model a compiler vendor also
12837 needs to replace the linker. The disadvantages are vastly increased
12838 complexity, and thus potential for error; for some code this can be
12839 just as transparent, but in practice it can been very difficult to build
12840 multiple programs in one directory and one program in multiple
12841 directories. Code written for this model tends to separate definitions
12842 of non-inline member templates into a separate file, which should be
12843 compiled separately.
12846 When used with GNU ld version 2.8 or later on an ELF system such as
12847 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
12848 Borland model. On other systems, G++ implements neither automatic
12851 A future version of G++ will support a hybrid model whereby the compiler
12852 will emit any instantiations for which the template definition is
12853 included in the compile, and store template definitions and
12854 instantiation context information into the object file for the rest.
12855 The link wrapper will extract that information as necessary and invoke
12856 the compiler to produce the remaining instantiations. The linker will
12857 then combine duplicate instantiations.
12859 In the mean time, you have the following options for dealing with
12860 template instantiations:
12865 Compile your template-using code with @option{-frepo}. The compiler will
12866 generate files with the extension @samp{.rpo} listing all of the
12867 template instantiations used in the corresponding object files which
12868 could be instantiated there; the link wrapper, @samp{collect2}, will
12869 then update the @samp{.rpo} files to tell the compiler where to place
12870 those instantiations and rebuild any affected object files. The
12871 link-time overhead is negligible after the first pass, as the compiler
12872 will continue to place the instantiations in the same files.
12874 This is your best option for application code written for the Borland
12875 model, as it will just work. Code written for the Cfront model will
12876 need to be modified so that the template definitions are available at
12877 one or more points of instantiation; usually this is as simple as adding
12878 @code{#include <tmethods.cc>} to the end of each template header.
12880 For library code, if you want the library to provide all of the template
12881 instantiations it needs, just try to link all of its object files
12882 together; the link will fail, but cause the instantiations to be
12883 generated as a side effect. Be warned, however, that this may cause
12884 conflicts if multiple libraries try to provide the same instantiations.
12885 For greater control, use explicit instantiation as described in the next
12889 @opindex fno-implicit-templates
12890 Compile your code with @option{-fno-implicit-templates} to disable the
12891 implicit generation of template instances, and explicitly instantiate
12892 all the ones you use. This approach requires more knowledge of exactly
12893 which instances you need than do the others, but it's less
12894 mysterious and allows greater control. You can scatter the explicit
12895 instantiations throughout your program, perhaps putting them in the
12896 translation units where the instances are used or the translation units
12897 that define the templates themselves; you can put all of the explicit
12898 instantiations you need into one big file; or you can create small files
12905 template class Foo<int>;
12906 template ostream& operator <<
12907 (ostream&, const Foo<int>&);
12910 for each of the instances you need, and create a template instantiation
12911 library from those.
12913 If you are using Cfront-model code, you can probably get away with not
12914 using @option{-fno-implicit-templates} when compiling files that don't
12915 @samp{#include} the member template definitions.
12917 If you use one big file to do the instantiations, you may want to
12918 compile it without @option{-fno-implicit-templates} so you get all of the
12919 instances required by your explicit instantiations (but not by any
12920 other files) without having to specify them as well.
12922 G++ has extended the template instantiation syntax given in the ISO
12923 standard to allow forward declaration of explicit instantiations
12924 (with @code{extern}), instantiation of the compiler support data for a
12925 template class (i.e.@: the vtable) without instantiating any of its
12926 members (with @code{inline}), and instantiation of only the static data
12927 members of a template class, without the support data or member
12928 functions (with (@code{static}):
12931 extern template int max (int, int);
12932 inline template class Foo<int>;
12933 static template class Foo<int>;
12937 Do nothing. Pretend G++ does implement automatic instantiation
12938 management. Code written for the Borland model will work fine, but
12939 each translation unit will contain instances of each of the templates it
12940 uses. In a large program, this can lead to an unacceptable amount of code
12944 @node Bound member functions
12945 @section Extracting the function pointer from a bound pointer to member function
12947 @cindex pointer to member function
12948 @cindex bound pointer to member function
12950 In C++, pointer to member functions (PMFs) are implemented using a wide
12951 pointer of sorts to handle all the possible call mechanisms; the PMF
12952 needs to store information about how to adjust the @samp{this} pointer,
12953 and if the function pointed to is virtual, where to find the vtable, and
12954 where in the vtable to look for the member function. If you are using
12955 PMFs in an inner loop, you should really reconsider that decision. If
12956 that is not an option, you can extract the pointer to the function that
12957 would be called for a given object/PMF pair and call it directly inside
12958 the inner loop, to save a bit of time.
12960 Note that you will still be paying the penalty for the call through a
12961 function pointer; on most modern architectures, such a call defeats the
12962 branch prediction features of the CPU@. This is also true of normal
12963 virtual function calls.
12965 The syntax for this extension is
12969 extern int (A::*fp)();
12970 typedef int (*fptr)(A *);
12972 fptr p = (fptr)(a.*fp);
12975 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
12976 no object is needed to obtain the address of the function. They can be
12977 converted to function pointers directly:
12980 fptr p1 = (fptr)(&A::foo);
12983 @opindex Wno-pmf-conversions
12984 You must specify @option{-Wno-pmf-conversions} to use this extension.
12986 @node C++ Attributes
12987 @section C++-Specific Variable, Function, and Type Attributes
12989 Some attributes only make sense for C++ programs.
12992 @item init_priority (@var{priority})
12993 @cindex init_priority attribute
12996 In Standard C++, objects defined at namespace scope are guaranteed to be
12997 initialized in an order in strict accordance with that of their definitions
12998 @emph{in a given translation unit}. No guarantee is made for initializations
12999 across translation units. However, GNU C++ allows users to control the
13000 order of initialization of objects defined at namespace scope with the
13001 @code{init_priority} attribute by specifying a relative @var{priority},
13002 a constant integral expression currently bounded between 101 and 65535
13003 inclusive. Lower numbers indicate a higher priority.
13005 In the following example, @code{A} would normally be created before
13006 @code{B}, but the @code{init_priority} attribute has reversed that order:
13009 Some_Class A __attribute__ ((init_priority (2000)));
13010 Some_Class B __attribute__ ((init_priority (543)));
13014 Note that the particular values of @var{priority} do not matter; only their
13017 @item java_interface
13018 @cindex java_interface attribute
13020 This type attribute informs C++ that the class is a Java interface. It may
13021 only be applied to classes declared within an @code{extern "Java"} block.
13022 Calls to methods declared in this interface will be dispatched using GCJ's
13023 interface table mechanism, instead of regular virtual table dispatch.
13027 See also @ref{Namespace Association}.
13029 @node Namespace Association
13030 @section Namespace Association
13032 @strong{Caution:} The semantics of this extension are not fully
13033 defined. Users should refrain from using this extension as its
13034 semantics may change subtly over time. It is possible that this
13035 extension will be removed in future versions of G++.
13037 A using-directive with @code{__attribute ((strong))} is stronger
13038 than a normal using-directive in two ways:
13042 Templates from the used namespace can be specialized and explicitly
13043 instantiated as though they were members of the using namespace.
13046 The using namespace is considered an associated namespace of all
13047 templates in the used namespace for purposes of argument-dependent
13051 The used namespace must be nested within the using namespace so that
13052 normal unqualified lookup works properly.
13054 This is useful for composing a namespace transparently from
13055 implementation namespaces. For example:
13060 template <class T> struct A @{ @};
13062 using namespace debug __attribute ((__strong__));
13063 template <> struct A<int> @{ @}; // @r{ok to specialize}
13065 template <class T> void f (A<T>);
13070 f (std::A<float>()); // @r{lookup finds} std::f
13076 @section Type Traits
13078 The C++ front-end implements syntactic extensions that allow to
13079 determine at compile time various characteristics of a type (or of a
13083 @item __has_nothrow_assign (type)
13084 If @code{type} is const qualified or is a reference type then the trait is
13085 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
13086 is true, else if @code{type} is a cv class or union type with copy assignment
13087 operators that are known not to throw an exception then the trait is true,
13088 else it is false. Requires: @code{type} shall be a complete type, an array
13089 type of unknown bound, or is a @code{void} type.
13091 @item __has_nothrow_copy (type)
13092 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
13093 @code{type} is a cv class or union type with copy constructors that
13094 are known not to throw an exception then the trait is true, else it is false.
13095 Requires: @code{type} shall be a complete type, an array type of
13096 unknown bound, or is a @code{void} type.
13098 @item __has_nothrow_constructor (type)
13099 If @code{__has_trivial_constructor (type)} is true then the trait is
13100 true, else if @code{type} is a cv class or union type (or array
13101 thereof) with a default constructor that is known not to throw an
13102 exception then the trait is true, else it is false. Requires:
13103 @code{type} shall be a complete type, an array type of unknown bound,
13104 or is a @code{void} type.
13106 @item __has_trivial_assign (type)
13107 If @code{type} is const qualified or is a reference type then the trait is
13108 false. Otherwise if @code{__is_pod (type)} is true then the trait is
13109 true, else if @code{type} is a cv class or union type with a trivial
13110 copy assignment ([class.copy]) then the trait is true, else it is
13111 false. Requires: @code{type} shall be a complete type, an array type
13112 of unknown bound, or is a @code{void} type.
13114 @item __has_trivial_copy (type)
13115 If @code{__is_pod (type)} is true or @code{type} is a reference type
13116 then the trait is true, else if @code{type} is a cv class or union type
13117 with a trivial copy constructor ([class.copy]) then the trait
13118 is true, else it is false. Requires: @code{type} shall be a complete
13119 type, an array type of unknown bound, or is a @code{void} type.
13121 @item __has_trivial_constructor (type)
13122 If @code{__is_pod (type)} is true then the trait is true, else if
13123 @code{type} is a cv class or union type (or array thereof) with a
13124 trivial default constructor ([class.ctor]) then the trait is true,
13125 else it is false. Requires: @code{type} shall be a complete type, an
13126 array type of unknown bound, or is a @code{void} type.
13128 @item __has_trivial_destructor (type)
13129 If @code{__is_pod (type)} is true or @code{type} is a reference type then
13130 the trait is true, else if @code{type} is a cv class or union type (or
13131 array thereof) with a trivial destructor ([class.dtor]) then the trait
13132 is true, else it is false. Requires: @code{type} shall be a complete
13133 type, an array type of unknown bound, or is a @code{void} type.
13135 @item __has_virtual_destructor (type)
13136 If @code{type} is a class type with a virtual destructor
13137 ([class.dtor]) then the trait is true, else it is false. Requires:
13138 @code{type} shall be a complete type, an array type of unknown bound,
13139 or is a @code{void} type.
13141 @item __is_abstract (type)
13142 If @code{type} is an abstract class ([class.abstract]) then the trait
13143 is true, else it is false. Requires: @code{type} shall be a complete
13144 type, an array type of unknown bound, or is a @code{void} type.
13146 @item __is_base_of (base_type, derived_type)
13147 If @code{base_type} is a base class of @code{derived_type}
13148 ([class.derived]) then the trait is true, otherwise it is false.
13149 Top-level cv qualifications of @code{base_type} and
13150 @code{derived_type} are ignored. For the purposes of this trait, a
13151 class type is considered is own base. Requires: if @code{__is_class
13152 (base_type)} and @code{__is_class (derived_type)} are true and
13153 @code{base_type} and @code{derived_type} are not the same type
13154 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
13155 type. Diagnostic is produced if this requirement is not met.
13157 @item __is_class (type)
13158 If @code{type} is a cv class type, and not a union type
13159 ([basic.compound]) the trait is true, else it is false.
13161 @item __is_empty (type)
13162 If @code{__is_class (type)} is false then the trait is false.
13163 Otherwise @code{type} is considered empty if and only if: @code{type}
13164 has no non-static data members, or all non-static data members, if
13165 any, are bit-fields of length 0, and @code{type} has no virtual
13166 members, and @code{type} has no virtual base classes, and @code{type}
13167 has no base classes @code{base_type} for which
13168 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
13169 be a complete type, an array type of unknown bound, or is a
13172 @item __is_enum (type)
13173 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
13174 true, else it is false.
13176 @item __is_pod (type)
13177 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
13178 else it is false. Requires: @code{type} shall be a complete type,
13179 an array type of unknown bound, or is a @code{void} type.
13181 @item __is_polymorphic (type)
13182 If @code{type} is a polymorphic class ([class.virtual]) then the trait
13183 is true, else it is false. Requires: @code{type} shall be a complete
13184 type, an array type of unknown bound, or is a @code{void} type.
13186 @item __is_union (type)
13187 If @code{type} is a cv union type ([basic.compound]) the trait is
13188 true, else it is false.
13192 @node Java Exceptions
13193 @section Java Exceptions
13195 The Java language uses a slightly different exception handling model
13196 from C++. Normally, GNU C++ will automatically detect when you are
13197 writing C++ code that uses Java exceptions, and handle them
13198 appropriately. However, if C++ code only needs to execute destructors
13199 when Java exceptions are thrown through it, GCC will guess incorrectly.
13200 Sample problematic code is:
13203 struct S @{ ~S(); @};
13204 extern void bar(); // @r{is written in Java, and may throw exceptions}
13213 The usual effect of an incorrect guess is a link failure, complaining of
13214 a missing routine called @samp{__gxx_personality_v0}.
13216 You can inform the compiler that Java exceptions are to be used in a
13217 translation unit, irrespective of what it might think, by writing
13218 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
13219 @samp{#pragma} must appear before any functions that throw or catch
13220 exceptions, or run destructors when exceptions are thrown through them.
13222 You cannot mix Java and C++ exceptions in the same translation unit. It
13223 is believed to be safe to throw a C++ exception from one file through
13224 another file compiled for the Java exception model, or vice versa, but
13225 there may be bugs in this area.
13227 @node Deprecated Features
13228 @section Deprecated Features
13230 In the past, the GNU C++ compiler was extended to experiment with new
13231 features, at a time when the C++ language was still evolving. Now that
13232 the C++ standard is complete, some of those features are superseded by
13233 superior alternatives. Using the old features might cause a warning in
13234 some cases that the feature will be dropped in the future. In other
13235 cases, the feature might be gone already.
13237 While the list below is not exhaustive, it documents some of the options
13238 that are now deprecated:
13241 @item -fexternal-templates
13242 @itemx -falt-external-templates
13243 These are two of the many ways for G++ to implement template
13244 instantiation. @xref{Template Instantiation}. The C++ standard clearly
13245 defines how template definitions have to be organized across
13246 implementation units. G++ has an implicit instantiation mechanism that
13247 should work just fine for standard-conforming code.
13249 @item -fstrict-prototype
13250 @itemx -fno-strict-prototype
13251 Previously it was possible to use an empty prototype parameter list to
13252 indicate an unspecified number of parameters (like C), rather than no
13253 parameters, as C++ demands. This feature has been removed, except where
13254 it is required for backwards compatibility. @xref{Backwards Compatibility}.
13257 G++ allows a virtual function returning @samp{void *} to be overridden
13258 by one returning a different pointer type. This extension to the
13259 covariant return type rules is now deprecated and will be removed from a
13262 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
13263 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
13264 and are now removed from G++. Code using these operators should be
13265 modified to use @code{std::min} and @code{std::max} instead.
13267 The named return value extension has been deprecated, and is now
13270 The use of initializer lists with new expressions has been deprecated,
13271 and is now removed from G++.
13273 Floating and complex non-type template parameters have been deprecated,
13274 and are now removed from G++.
13276 The implicit typename extension has been deprecated and is now
13279 The use of default arguments in function pointers, function typedefs
13280 and other places where they are not permitted by the standard is
13281 deprecated and will be removed from a future version of G++.
13283 G++ allows floating-point literals to appear in integral constant expressions,
13284 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
13285 This extension is deprecated and will be removed from a future version.
13287 G++ allows static data members of const floating-point type to be declared
13288 with an initializer in a class definition. The standard only allows
13289 initializers for static members of const integral types and const
13290 enumeration types so this extension has been deprecated and will be removed
13291 from a future version.
13293 @node Backwards Compatibility
13294 @section Backwards Compatibility
13295 @cindex Backwards Compatibility
13296 @cindex ARM [Annotated C++ Reference Manual]
13298 Now that there is a definitive ISO standard C++, G++ has a specification
13299 to adhere to. The C++ language evolved over time, and features that
13300 used to be acceptable in previous drafts of the standard, such as the ARM
13301 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
13302 compilation of C++ written to such drafts, G++ contains some backwards
13303 compatibilities. @emph{All such backwards compatibility features are
13304 liable to disappear in future versions of G++.} They should be considered
13305 deprecated. @xref{Deprecated Features}.
13309 If a variable is declared at for scope, it used to remain in scope until
13310 the end of the scope which contained the for statement (rather than just
13311 within the for scope). G++ retains this, but issues a warning, if such a
13312 variable is accessed outside the for scope.
13314 @item Implicit C language
13315 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
13316 scope to set the language. On such systems, all header files are
13317 implicitly scoped inside a C language scope. Also, an empty prototype
13318 @code{()} will be treated as an unspecified number of arguments, rather
13319 than no arguments, as C++ demands.