1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
3 @c Free Software Foundation, Inc.
5 @c This is part of the GCC manual.
6 @c For copying conditions, see the file gcc.texi.
9 @chapter Extensions to the C Language Family
10 @cindex extensions, C language
11 @cindex C language extensions
14 GNU C provides several language features not found in ISO standard C@.
15 (The @option{-pedantic} option directs GCC to print a warning message if
16 any of these features is used.) To test for the availability of these
17 features in conditional compilation, check for a predefined macro
18 @code{__GNUC__}, which is always defined under GCC@.
20 These extensions are available in C and Objective-C@. Most of them are
21 also available in C++. @xref{C++ Extensions,,Extensions to the
22 C++ Language}, for extensions that apply @emph{only} to C++.
24 Some features that are in ISO C99 but not C89 or C++ are also, as
25 extensions, accepted by GCC in C89 mode and in C++.
28 * Statement Exprs:: Putting statements and declarations inside expressions.
29 * Local Labels:: Labels local to a block.
30 * Labels as Values:: Getting pointers to labels, and computed gotos.
31 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
32 * Constructing Calls:: Dispatching a call to another function.
33 * Typeof:: @code{typeof}: referring to the type of an expression.
34 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
35 * Long Long:: Double-word integers---@code{long long int}.
36 * Complex:: Data types for complex numbers.
37 * Floating Types:: Additional Floating Types.
38 * Decimal Float:: Decimal Floating Types.
39 * Hex Floats:: Hexadecimal floating-point constants.
40 * Fixed-Point:: Fixed-Point Types.
41 * Zero Length:: Zero-length arrays.
42 * Variable Length:: Arrays whose length is computed at run time.
43 * Empty Structures:: Structures with no members.
44 * Variadic Macros:: Macros with a variable number of arguments.
45 * Escaped Newlines:: Slightly looser rules for escaped newlines.
46 * Subscripting:: Any array can be subscripted, even if not an lvalue.
47 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
48 * Initializers:: Non-constant initializers.
49 * Compound Literals:: Compound literals give structures, unions
51 * Designated Inits:: Labeling elements of initializers.
52 * Cast to Union:: Casting to union type from any member of the union.
53 * Case Ranges:: `case 1 ... 9' and such.
54 * Mixed Declarations:: Mixing declarations and code.
55 * Function Attributes:: Declaring that functions have no side effects,
56 or that they can never return.
57 * Attribute Syntax:: Formal syntax for attributes.
58 * Function Prototypes:: Prototype declarations and old-style definitions.
59 * C++ Comments:: C++ comments are recognized.
60 * Dollar Signs:: Dollar sign is allowed in identifiers.
61 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
62 * Variable Attributes:: Specifying attributes of variables.
63 * Type Attributes:: Specifying attributes of types.
64 * Alignment:: Inquiring about the alignment of a type or variable.
65 * Inline:: Defining inline functions (as fast as macros).
66 * Extended Asm:: Assembler instructions with C expressions as operands.
67 (With them you can define ``built-in'' functions.)
68 * Constraints:: Constraints for asm operands
69 * Asm Labels:: Specifying the assembler name to use for a C symbol.
70 * Explicit Reg Vars:: Defining variables residing in specified registers.
71 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
72 * Incomplete Enums:: @code{enum foo;}, with details to follow.
73 * Function Names:: Printable strings which are the name of the current
75 * Return Address:: Getting the return or frame address of a function.
76 * Vector Extensions:: Using vector instructions through built-in functions.
77 * Offsetof:: Special syntax for implementing @code{offsetof}.
78 * Atomic Builtins:: Built-in functions for atomic memory access.
79 * Object Size Checking:: Built-in functions for limited buffer overflow
81 * Other Builtins:: Other built-in functions.
82 * Target Builtins:: Built-in functions specific to particular targets.
83 * Target Format Checks:: Format checks specific to particular targets.
84 * Pragmas:: Pragmas accepted by GCC.
85 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
86 * Thread-Local:: Per-thread variables.
87 * Binary constants:: Binary constants using the @samp{0b} prefix.
91 @section Statements and Declarations in Expressions
92 @cindex statements inside expressions
93 @cindex declarations inside expressions
94 @cindex expressions containing statements
95 @cindex macros, statements in expressions
97 @c the above section title wrapped and causes an underfull hbox.. i
98 @c changed it from "within" to "in". --mew 4feb93
99 A compound statement enclosed in parentheses may appear as an expression
100 in GNU C@. This allows you to use loops, switches, and local variables
101 within an expression.
103 Recall that a compound statement is a sequence of statements surrounded
104 by braces; in this construct, parentheses go around the braces. For
108 (@{ int y = foo (); int z;
115 is a valid (though slightly more complex than necessary) expression
116 for the absolute value of @code{foo ()}.
118 The last thing in the compound statement should be an expression
119 followed by a semicolon; the value of this subexpression serves as the
120 value of the entire construct. (If you use some other kind of statement
121 last within the braces, the construct has type @code{void}, and thus
122 effectively no value.)
124 This feature is especially useful in making macro definitions ``safe'' (so
125 that they evaluate each operand exactly once). For example, the
126 ``maximum'' function is commonly defined as a macro in standard C as
130 #define max(a,b) ((a) > (b) ? (a) : (b))
134 @cindex side effects, macro argument
135 But this definition computes either @var{a} or @var{b} twice, with bad
136 results if the operand has side effects. In GNU C, if you know the
137 type of the operands (here taken as @code{int}), you can define
138 the macro safely as follows:
141 #define maxint(a,b) \
142 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
145 Embedded statements are not allowed in constant expressions, such as
146 the value of an enumeration constant, the width of a bit-field, or
147 the initial value of a static variable.
149 If you don't know the type of the operand, you can still do this, but you
150 must use @code{typeof} (@pxref{Typeof}).
152 In G++, the result value of a statement expression undergoes array and
153 function pointer decay, and is returned by value to the enclosing
154 expression. For instance, if @code{A} is a class, then
163 will construct a temporary @code{A} object to hold the result of the
164 statement expression, and that will be used to invoke @code{Foo}.
165 Therefore the @code{this} pointer observed by @code{Foo} will not be the
168 Any temporaries created within a statement within a statement expression
169 will be destroyed at the statement's end. This makes statement
170 expressions inside macros slightly different from function calls. In
171 the latter case temporaries introduced during argument evaluation will
172 be destroyed at the end of the statement that includes the function
173 call. In the statement expression case they will be destroyed during
174 the statement expression. For instance,
177 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
178 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
188 will have different places where temporaries are destroyed. For the
189 @code{macro} case, the temporary @code{X} will be destroyed just after
190 the initialization of @code{b}. In the @code{function} case that
191 temporary will be destroyed when the function returns.
193 These considerations mean that it is probably a bad idea to use
194 statement-expressions of this form in header files that are designed to
195 work with C++. (Note that some versions of the GNU C Library contained
196 header files using statement-expression that lead to precisely this
199 Jumping into a statement expression with @code{goto} or using a
200 @code{switch} statement outside the statement expression with a
201 @code{case} or @code{default} label inside the statement expression is
202 not permitted. Jumping into a statement expression with a computed
203 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
204 Jumping out of a statement expression is permitted, but if the
205 statement expression is part of a larger expression then it is
206 unspecified which other subexpressions of that expression have been
207 evaluated except where the language definition requires certain
208 subexpressions to be evaluated before or after the statement
209 expression. In any case, as with a function call the evaluation of a
210 statement expression is not interleaved with the evaluation of other
211 parts of the containing expression. For example,
214 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
218 will call @code{foo} and @code{bar1} and will not call @code{baz} but
219 may or may not call @code{bar2}. If @code{bar2} is called, it will be
220 called after @code{foo} and before @code{bar1}
223 @section Locally Declared Labels
225 @cindex macros, local labels
227 GCC allows you to declare @dfn{local labels} in any nested block
228 scope. A local label is just like an ordinary label, but you can
229 only reference it (with a @code{goto} statement, or by taking its
230 address) within the block in which it was declared.
232 A local label declaration looks like this:
235 __label__ @var{label};
242 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
245 Local label declarations must come at the beginning of the block,
246 before any ordinary declarations or statements.
248 The label declaration defines the label @emph{name}, but does not define
249 the label itself. You must do this in the usual way, with
250 @code{@var{label}:}, within the statements of the statement expression.
252 The local label feature is useful for complex macros. If a macro
253 contains nested loops, a @code{goto} can be useful for breaking out of
254 them. However, an ordinary label whose scope is the whole function
255 cannot be used: if the macro can be expanded several times in one
256 function, the label will be multiply defined in that function. A
257 local label avoids this problem. For example:
260 #define SEARCH(value, array, target) \
263 typeof (target) _SEARCH_target = (target); \
264 typeof (*(array)) *_SEARCH_array = (array); \
267 for (i = 0; i < max; i++) \
268 for (j = 0; j < max; j++) \
269 if (_SEARCH_array[i][j] == _SEARCH_target) \
270 @{ (value) = i; goto found; @} \
276 This could also be written using a statement-expression:
279 #define SEARCH(array, target) \
282 typeof (target) _SEARCH_target = (target); \
283 typeof (*(array)) *_SEARCH_array = (array); \
286 for (i = 0; i < max; i++) \
287 for (j = 0; j < max; j++) \
288 if (_SEARCH_array[i][j] == _SEARCH_target) \
289 @{ value = i; goto found; @} \
296 Local label declarations also make the labels they declare visible to
297 nested functions, if there are any. @xref{Nested Functions}, for details.
299 @node Labels as Values
300 @section Labels as Values
301 @cindex labels as values
302 @cindex computed gotos
303 @cindex goto with computed label
304 @cindex address of a label
306 You can get the address of a label defined in the current function
307 (or a containing function) with the unary operator @samp{&&}. The
308 value has type @code{void *}. This value is a constant and can be used
309 wherever a constant of that type is valid. For example:
317 To use these values, you need to be able to jump to one. This is done
318 with the computed goto statement@footnote{The analogous feature in
319 Fortran is called an assigned goto, but that name seems inappropriate in
320 C, where one can do more than simply store label addresses in label
321 variables.}, @code{goto *@var{exp};}. For example,
328 Any expression of type @code{void *} is allowed.
330 One way of using these constants is in initializing a static array that
331 will serve as a jump table:
334 static void *array[] = @{ &&foo, &&bar, &&hack @};
337 Then you can select a label with indexing, like this:
344 Note that this does not check whether the subscript is in bounds---array
345 indexing in C never does that.
347 Such an array of label values serves a purpose much like that of the
348 @code{switch} statement. The @code{switch} statement is cleaner, so
349 use that rather than an array unless the problem does not fit a
350 @code{switch} statement very well.
352 Another use of label values is in an interpreter for threaded code.
353 The labels within the interpreter function can be stored in the
354 threaded code for super-fast dispatching.
356 You may not use this mechanism to jump to code in a different function.
357 If you do that, totally unpredictable things will happen. The best way to
358 avoid this is to store the label address only in automatic variables and
359 never pass it as an argument.
361 An alternate way to write the above example is
364 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
366 goto *(&&foo + array[i]);
370 This is more friendly to code living in shared libraries, as it reduces
371 the number of dynamic relocations that are needed, and by consequence,
372 allows the data to be read-only.
374 The @code{&&foo} expressions for the same label might have different values
375 if the containing function is inlined or cloned. If a program relies on
376 them being always the same, @code{__attribute__((__noinline__))} should
377 be used to prevent inlining. If @code{&&foo} is used
378 in a static variable initializer, inlining is forbidden.
380 @node Nested Functions
381 @section Nested Functions
382 @cindex nested functions
383 @cindex downward funargs
386 A @dfn{nested function} is a function defined inside another function.
387 (Nested functions are not supported for GNU C++.) The nested function's
388 name is local to the block where it is defined. For example, here we
389 define a nested function named @code{square}, and call it twice:
393 foo (double a, double b)
395 double square (double z) @{ return z * z; @}
397 return square (a) + square (b);
402 The nested function can access all the variables of the containing
403 function that are visible at the point of its definition. This is
404 called @dfn{lexical scoping}. For example, here we show a nested
405 function which uses an inherited variable named @code{offset}:
409 bar (int *array, int offset, int size)
411 int access (int *array, int index)
412 @{ return array[index + offset]; @}
415 for (i = 0; i < size; i++)
416 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
421 Nested function definitions are permitted within functions in the places
422 where variable definitions are allowed; that is, in any block, mixed
423 with the other declarations and statements in the block.
425 It is possible to call the nested function from outside the scope of its
426 name by storing its address or passing the address to another function:
429 hack (int *array, int size)
431 void store (int index, int value)
432 @{ array[index] = value; @}
434 intermediate (store, size);
438 Here, the function @code{intermediate} receives the address of
439 @code{store} as an argument. If @code{intermediate} calls @code{store},
440 the arguments given to @code{store} are used to store into @code{array}.
441 But this technique works only so long as the containing function
442 (@code{hack}, in this example) does not exit.
444 If you try to call the nested function through its address after the
445 containing function has exited, all hell will break loose. If you try
446 to call it after a containing scope level has exited, and if it refers
447 to some of the variables that are no longer in scope, you may be lucky,
448 but it's not wise to take the risk. If, however, the nested function
449 does not refer to anything that has gone out of scope, you should be
452 GCC implements taking the address of a nested function using a technique
453 called @dfn{trampolines}. A paper describing them is available as
456 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
458 A nested function can jump to a label inherited from a containing
459 function, provided the label was explicitly declared in the containing
460 function (@pxref{Local Labels}). Such a jump returns instantly to the
461 containing function, exiting the nested function which did the
462 @code{goto} and any intermediate functions as well. Here is an example:
466 bar (int *array, int offset, int size)
469 int access (int *array, int index)
473 return array[index + offset];
477 for (i = 0; i < size; i++)
478 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
482 /* @r{Control comes here from @code{access}
483 if it detects an error.} */
490 A nested function always has no linkage. Declaring one with
491 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
492 before its definition, use @code{auto} (which is otherwise meaningless
493 for function declarations).
496 bar (int *array, int offset, int size)
499 auto int access (int *, int);
501 int access (int *array, int index)
505 return array[index + offset];
511 @node Constructing Calls
512 @section Constructing Function Calls
513 @cindex constructing calls
514 @cindex forwarding calls
516 Using the built-in functions described below, you can record
517 the arguments a function received, and call another function
518 with the same arguments, without knowing the number or types
521 You can also record the return value of that function call,
522 and later return that value, without knowing what data type
523 the function tried to return (as long as your caller expects
526 However, these built-in functions may interact badly with some
527 sophisticated features or other extensions of the language. It
528 is, therefore, not recommended to use them outside very simple
529 functions acting as mere forwarders for their arguments.
531 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
532 This built-in function returns a pointer to data
533 describing how to perform a call with the same arguments as were passed
534 to the current function.
536 The function saves the arg pointer register, structure value address,
537 and all registers that might be used to pass arguments to a function
538 into a block of memory allocated on the stack. Then it returns the
539 address of that block.
542 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
543 This built-in function invokes @var{function}
544 with a copy of the parameters described by @var{arguments}
547 The value of @var{arguments} should be the value returned by
548 @code{__builtin_apply_args}. The argument @var{size} specifies the size
549 of the stack argument data, in bytes.
551 This function returns a pointer to data describing
552 how to return whatever value was returned by @var{function}. The data
553 is saved in a block of memory allocated on the stack.
555 It is not always simple to compute the proper value for @var{size}. The
556 value is used by @code{__builtin_apply} to compute the amount of data
557 that should be pushed on the stack and copied from the incoming argument
561 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
562 This built-in function returns the value described by @var{result} from
563 the containing function. You should specify, for @var{result}, a value
564 returned by @code{__builtin_apply}.
567 @deftypefn {Built-in Function} __builtin_va_arg_pack ()
568 This built-in function represents all anonymous arguments of an inline
569 function. It can be used only in inline functions which will be always
570 inlined, never compiled as a separate function, such as those using
571 @code{__attribute__ ((__always_inline__))} or
572 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
573 It must be only passed as last argument to some other function
574 with variable arguments. This is useful for writing small wrapper
575 inlines for variable argument functions, when using preprocessor
576 macros is undesirable. For example:
578 extern int myprintf (FILE *f, const char *format, ...);
579 extern inline __attribute__ ((__gnu_inline__)) int
580 myprintf (FILE *f, const char *format, ...)
582 int r = fprintf (f, "myprintf: ");
585 int s = fprintf (f, format, __builtin_va_arg_pack ());
593 @deftypefn {Built-in Function} __builtin_va_arg_pack_len ()
594 This built-in function returns the number of anonymous arguments of
595 an inline function. It can be used only in inline functions which
596 will be always inlined, never compiled as a separate function, such
597 as those using @code{__attribute__ ((__always_inline__))} or
598 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
599 For example following will do link or runtime checking of open
600 arguments for optimized code:
603 extern inline __attribute__((__gnu_inline__)) int
604 myopen (const char *path, int oflag, ...)
606 if (__builtin_va_arg_pack_len () > 1)
607 warn_open_too_many_arguments ();
609 if (__builtin_constant_p (oflag))
611 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
613 warn_open_missing_mode ();
614 return __open_2 (path, oflag);
616 return open (path, oflag, __builtin_va_arg_pack ());
619 if (__builtin_va_arg_pack_len () < 1)
620 return __open_2 (path, oflag);
622 return open (path, oflag, __builtin_va_arg_pack ());
629 @section Referring to a Type with @code{typeof}
632 @cindex macros, types of arguments
634 Another way to refer to the type of an expression is with @code{typeof}.
635 The syntax of using of this keyword looks like @code{sizeof}, but the
636 construct acts semantically like a type name defined with @code{typedef}.
638 There are two ways of writing the argument to @code{typeof}: with an
639 expression or with a type. Here is an example with an expression:
646 This assumes that @code{x} is an array of pointers to functions;
647 the type described is that of the values of the functions.
649 Here is an example with a typename as the argument:
656 Here the type described is that of pointers to @code{int}.
658 If you are writing a header file that must work when included in ISO C
659 programs, write @code{__typeof__} instead of @code{typeof}.
660 @xref{Alternate Keywords}.
662 A @code{typeof}-construct can be used anywhere a typedef name could be
663 used. For example, you can use it in a declaration, in a cast, or inside
664 of @code{sizeof} or @code{typeof}.
666 @code{typeof} is often useful in conjunction with the
667 statements-within-expressions feature. Here is how the two together can
668 be used to define a safe ``maximum'' macro that operates on any
669 arithmetic type and evaluates each of its arguments exactly once:
673 (@{ typeof (a) _a = (a); \
674 typeof (b) _b = (b); \
675 _a > _b ? _a : _b; @})
678 @cindex underscores in variables in macros
679 @cindex @samp{_} in variables in macros
680 @cindex local variables in macros
681 @cindex variables, local, in macros
682 @cindex macros, local variables in
684 The reason for using names that start with underscores for the local
685 variables is to avoid conflicts with variable names that occur within the
686 expressions that are substituted for @code{a} and @code{b}. Eventually we
687 hope to design a new form of declaration syntax that allows you to declare
688 variables whose scopes start only after their initializers; this will be a
689 more reliable way to prevent such conflicts.
692 Some more examples of the use of @code{typeof}:
696 This declares @code{y} with the type of what @code{x} points to.
703 This declares @code{y} as an array of such values.
710 This declares @code{y} as an array of pointers to characters:
713 typeof (typeof (char *)[4]) y;
717 It is equivalent to the following traditional C declaration:
723 To see the meaning of the declaration using @code{typeof}, and why it
724 might be a useful way to write, rewrite it with these macros:
727 #define pointer(T) typeof(T *)
728 #define array(T, N) typeof(T [N])
732 Now the declaration can be rewritten this way:
735 array (pointer (char), 4) y;
739 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
740 pointers to @code{char}.
743 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
744 a more limited extension which permitted one to write
747 typedef @var{T} = @var{expr};
751 with the effect of declaring @var{T} to have the type of the expression
752 @var{expr}. This extension does not work with GCC 3 (versions between
753 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
754 relies on it should be rewritten to use @code{typeof}:
757 typedef typeof(@var{expr}) @var{T};
761 This will work with all versions of GCC@.
764 @section Conditionals with Omitted Operands
765 @cindex conditional expressions, extensions
766 @cindex omitted middle-operands
767 @cindex middle-operands, omitted
768 @cindex extensions, @code{?:}
769 @cindex @code{?:} extensions
771 The middle operand in a conditional expression may be omitted. Then
772 if the first operand is nonzero, its value is the value of the conditional
775 Therefore, the expression
782 has the value of @code{x} if that is nonzero; otherwise, the value of
785 This example is perfectly equivalent to
791 @cindex side effect in ?:
792 @cindex ?: side effect
794 In this simple case, the ability to omit the middle operand is not
795 especially useful. When it becomes useful is when the first operand does,
796 or may (if it is a macro argument), contain a side effect. Then repeating
797 the operand in the middle would perform the side effect twice. Omitting
798 the middle operand uses the value already computed without the undesirable
799 effects of recomputing it.
802 @section Double-Word Integers
803 @cindex @code{long long} data types
804 @cindex double-word arithmetic
805 @cindex multiprecision arithmetic
806 @cindex @code{LL} integer suffix
807 @cindex @code{ULL} integer suffix
809 ISO C99 supports data types for integers that are at least 64 bits wide,
810 and as an extension GCC supports them in C89 mode and in C++.
811 Simply write @code{long long int} for a signed integer, or
812 @code{unsigned long long int} for an unsigned integer. To make an
813 integer constant of type @code{long long int}, add the suffix @samp{LL}
814 to the integer. To make an integer constant of type @code{unsigned long
815 long int}, add the suffix @samp{ULL} to the integer.
817 You can use these types in arithmetic like any other integer types.
818 Addition, subtraction, and bitwise boolean operations on these types
819 are open-coded on all types of machines. Multiplication is open-coded
820 if the machine supports fullword-to-doubleword a widening multiply
821 instruction. Division and shifts are open-coded only on machines that
822 provide special support. The operations that are not open-coded use
823 special library routines that come with GCC@.
825 There may be pitfalls when you use @code{long long} types for function
826 arguments, unless you declare function prototypes. If a function
827 expects type @code{int} for its argument, and you pass a value of type
828 @code{long long int}, confusion will result because the caller and the
829 subroutine will disagree about the number of bytes for the argument.
830 Likewise, if the function expects @code{long long int} and you pass
831 @code{int}. The best way to avoid such problems is to use prototypes.
834 @section Complex Numbers
835 @cindex complex numbers
836 @cindex @code{_Complex} keyword
837 @cindex @code{__complex__} keyword
839 ISO C99 supports complex floating data types, and as an extension GCC
840 supports them in C89 mode and in C++, and supports complex integer data
841 types which are not part of ISO C99. You can declare complex types
842 using the keyword @code{_Complex}. As an extension, the older GNU
843 keyword @code{__complex__} is also supported.
845 For example, @samp{_Complex double x;} declares @code{x} as a
846 variable whose real part and imaginary part are both of type
847 @code{double}. @samp{_Complex short int y;} declares @code{y} to
848 have real and imaginary parts of type @code{short int}; this is not
849 likely to be useful, but it shows that the set of complex types is
852 To write a constant with a complex data type, use the suffix @samp{i} or
853 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
854 has type @code{_Complex float} and @code{3i} has type
855 @code{_Complex int}. Such a constant always has a pure imaginary
856 value, but you can form any complex value you like by adding one to a
857 real constant. This is a GNU extension; if you have an ISO C99
858 conforming C library (such as GNU libc), and want to construct complex
859 constants of floating type, you should include @code{<complex.h>} and
860 use the macros @code{I} or @code{_Complex_I} instead.
862 @cindex @code{__real__} keyword
863 @cindex @code{__imag__} keyword
864 To extract the real part of a complex-valued expression @var{exp}, write
865 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
866 extract the imaginary part. This is a GNU extension; for values of
867 floating type, you should use the ISO C99 functions @code{crealf},
868 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
869 @code{cimagl}, declared in @code{<complex.h>} and also provided as
870 built-in functions by GCC@.
872 @cindex complex conjugation
873 The operator @samp{~} performs complex conjugation when used on a value
874 with a complex type. This is a GNU extension; for values of
875 floating type, you should use the ISO C99 functions @code{conjf},
876 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
877 provided as built-in functions by GCC@.
879 GCC can allocate complex automatic variables in a noncontiguous
880 fashion; it's even possible for the real part to be in a register while
881 the imaginary part is on the stack (or vice-versa). Only the DWARF2
882 debug info format can represent this, so use of DWARF2 is recommended.
883 If you are using the stabs debug info format, GCC describes a noncontiguous
884 complex variable as if it were two separate variables of noncomplex type.
885 If the variable's actual name is @code{foo}, the two fictitious
886 variables are named @code{foo$real} and @code{foo$imag}. You can
887 examine and set these two fictitious variables with your debugger.
890 @section Additional Floating Types
891 @cindex additional floating types
892 @cindex @code{__float80} data type
893 @cindex @code{__float128} data type
894 @cindex @code{w} floating point suffix
895 @cindex @code{q} floating point suffix
896 @cindex @code{W} floating point suffix
897 @cindex @code{Q} floating point suffix
899 As an extension, the GNU C compiler supports additional floating
900 types, @code{__float80} and @code{__float128} to support 80bit
901 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
902 Support for additional types includes the arithmetic operators:
903 add, subtract, multiply, divide; unary arithmetic operators;
904 relational operators; equality operators; and conversions to and from
905 integer and other floating types. Use a suffix @samp{w} or @samp{W}
906 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
907 for @code{_float128}. You can declare complex types using the
908 corresponding internal complex type, @code{XCmode} for @code{__float80}
909 type and @code{TCmode} for @code{__float128} type:
912 typedef _Complex float __attribute__((mode(TC))) _Complex128;
913 typedef _Complex float __attribute__((mode(XC))) _Complex80;
916 Not all targets support additional floating point types. @code{__float80}
917 is supported on i386, x86_64 and ia64 targets and target @code{__float128}
918 is supported on x86_64 and ia64 targets.
921 @section Decimal Floating Types
922 @cindex decimal floating types
923 @cindex @code{_Decimal32} data type
924 @cindex @code{_Decimal64} data type
925 @cindex @code{_Decimal128} data type
926 @cindex @code{df} integer suffix
927 @cindex @code{dd} integer suffix
928 @cindex @code{dl} integer suffix
929 @cindex @code{DF} integer suffix
930 @cindex @code{DD} integer suffix
931 @cindex @code{DL} integer suffix
933 As an extension, the GNU C compiler supports decimal floating types as
934 defined in the N1312 draft of ISO/IEC WDTR24732. Support for decimal
935 floating types in GCC will evolve as the draft technical report changes.
936 Calling conventions for any target might also change. Not all targets
937 support decimal floating types.
939 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
940 @code{_Decimal128}. They use a radix of ten, unlike the floating types
941 @code{float}, @code{double}, and @code{long double} whose radix is not
942 specified by the C standard but is usually two.
944 Support for decimal floating types includes the arithmetic operators
945 add, subtract, multiply, divide; unary arithmetic operators;
946 relational operators; equality operators; and conversions to and from
947 integer and other floating types. Use a suffix @samp{df} or
948 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
949 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
952 GCC support of decimal float as specified by the draft technical report
957 Pragma @code{FLOAT_CONST_DECIMAL64} is not supported, nor is the @samp{d}
958 suffix for literal constants of type @code{double}.
961 When the value of a decimal floating type cannot be represented in the
962 integer type to which it is being converted, the result is undefined
963 rather than the result value specified by the draft technical report.
966 GCC does not provide the C library functionality associated with
967 @file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
968 @file{wchar.h}, which must come from a separate C library implementation.
969 Because of this the GNU C compiler does not define macro
970 @code{__STDC_DEC_FP__} to indicate that the implementation conforms to
971 the technical report.
974 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
975 are supported by the DWARF2 debug information format.
981 ISO C99 supports floating-point numbers written not only in the usual
982 decimal notation, such as @code{1.55e1}, but also numbers such as
983 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
984 supports this in C89 mode (except in some cases when strictly
985 conforming) and in C++. In that format the
986 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
987 mandatory. The exponent is a decimal number that indicates the power of
988 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
995 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
996 is the same as @code{1.55e1}.
998 Unlike for floating-point numbers in the decimal notation the exponent
999 is always required in the hexadecimal notation. Otherwise the compiler
1000 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
1001 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1002 extension for floating-point constants of type @code{float}.
1005 @section Fixed-Point Types
1006 @cindex fixed-point types
1007 @cindex @code{_Fract} data type
1008 @cindex @code{_Accum} data type
1009 @cindex @code{_Sat} data type
1010 @cindex @code{hr} fixed-suffix
1011 @cindex @code{r} fixed-suffix
1012 @cindex @code{lr} fixed-suffix
1013 @cindex @code{llr} fixed-suffix
1014 @cindex @code{uhr} fixed-suffix
1015 @cindex @code{ur} fixed-suffix
1016 @cindex @code{ulr} fixed-suffix
1017 @cindex @code{ullr} fixed-suffix
1018 @cindex @code{hk} fixed-suffix
1019 @cindex @code{k} fixed-suffix
1020 @cindex @code{lk} fixed-suffix
1021 @cindex @code{llk} fixed-suffix
1022 @cindex @code{uhk} fixed-suffix
1023 @cindex @code{uk} fixed-suffix
1024 @cindex @code{ulk} fixed-suffix
1025 @cindex @code{ullk} fixed-suffix
1026 @cindex @code{HR} fixed-suffix
1027 @cindex @code{R} fixed-suffix
1028 @cindex @code{LR} fixed-suffix
1029 @cindex @code{LLR} fixed-suffix
1030 @cindex @code{UHR} fixed-suffix
1031 @cindex @code{UR} fixed-suffix
1032 @cindex @code{ULR} fixed-suffix
1033 @cindex @code{ULLR} fixed-suffix
1034 @cindex @code{HK} fixed-suffix
1035 @cindex @code{K} fixed-suffix
1036 @cindex @code{LK} fixed-suffix
1037 @cindex @code{LLK} fixed-suffix
1038 @cindex @code{UHK} fixed-suffix
1039 @cindex @code{UK} fixed-suffix
1040 @cindex @code{ULK} fixed-suffix
1041 @cindex @code{ULLK} fixed-suffix
1043 As an extension, the GNU C compiler supports fixed-point types as
1044 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1045 types in GCC will evolve as the draft technical report changes.
1046 Calling conventions for any target might also change. Not all targets
1047 support fixed-point types.
1049 The fixed-point types are
1050 @code{short _Fract},
1053 @code{long long _Fract},
1054 @code{unsigned short _Fract},
1055 @code{unsigned _Fract},
1056 @code{unsigned long _Fract},
1057 @code{unsigned long long _Fract},
1058 @code{_Sat short _Fract},
1060 @code{_Sat long _Fract},
1061 @code{_Sat long long _Fract},
1062 @code{_Sat unsigned short _Fract},
1063 @code{_Sat unsigned _Fract},
1064 @code{_Sat unsigned long _Fract},
1065 @code{_Sat unsigned long long _Fract},
1066 @code{short _Accum},
1069 @code{long long _Accum},
1070 @code{unsigned short _Accum},
1071 @code{unsigned _Accum},
1072 @code{unsigned long _Accum},
1073 @code{unsigned long long _Accum},
1074 @code{_Sat short _Accum},
1076 @code{_Sat long _Accum},
1077 @code{_Sat long long _Accum},
1078 @code{_Sat unsigned short _Accum},
1079 @code{_Sat unsigned _Accum},
1080 @code{_Sat unsigned long _Accum},
1081 @code{_Sat unsigned long long _Accum}.
1083 Fixed-point data values contain fractional and optional integral parts.
1084 The format of fixed-point data varies and depends on the target machine.
1086 Support for fixed-point types includes:
1089 prefix and postfix increment and decrement operators (@code{++}, @code{--})
1091 unary arithmetic operators (@code{+}, @code{-}, @code{!})
1093 binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
1095 binary shift operators (@code{<<}, @code{>>})
1097 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
1099 equality operators (@code{==}, @code{!=})
1101 assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
1102 @code{<<=}, @code{>>=})
1104 conversions to and from integer, floating-point, or fixed-point types
1107 Use a suffix in a fixed-point literal constant:
1109 @item @samp{hr} or @samp{HR} for @code{short _Fract} and
1110 @code{_Sat short _Fract}
1111 @item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
1112 @item @samp{lr} or @samp{LR} for @code{long _Fract} and
1113 @code{_Sat long _Fract}
1114 @item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1115 @code{_Sat long long _Fract}
1116 @item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1117 @code{_Sat unsigned short _Fract}
1118 @item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1119 @code{_Sat unsigned _Fract}
1120 @item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1121 @code{_Sat unsigned long _Fract}
1122 @item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1123 and @code{_Sat unsigned long long _Fract}
1124 @item @samp{hk} or @samp{HK} for @code{short _Accum} and
1125 @code{_Sat short _Accum}
1126 @item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
1127 @item @samp{lk} or @samp{LK} for @code{long _Accum} and
1128 @code{_Sat long _Accum}
1129 @item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1130 @code{_Sat long long _Accum}
1131 @item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1132 @code{_Sat unsigned short _Accum}
1133 @item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1134 @code{_Sat unsigned _Accum}
1135 @item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1136 @code{_Sat unsigned long _Accum}
1137 @item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1138 and @code{_Sat unsigned long long _Accum}
1141 GCC support of fixed-point types as specified by the draft technical report
1146 Pragmas to control overflow and rounding behaviors are not implemented.
1149 Fixed-point types are supported by the DWARF2 debug information format.
1152 @section Arrays of Length Zero
1153 @cindex arrays of length zero
1154 @cindex zero-length arrays
1155 @cindex length-zero arrays
1156 @cindex flexible array members
1158 Zero-length arrays are allowed in GNU C@. They are very useful as the
1159 last element of a structure which is really a header for a variable-length
1168 struct line *thisline = (struct line *)
1169 malloc (sizeof (struct line) + this_length);
1170 thisline->length = this_length;
1173 In ISO C90, you would have to give @code{contents} a length of 1, which
1174 means either you waste space or complicate the argument to @code{malloc}.
1176 In ISO C99, you would use a @dfn{flexible array member}, which is
1177 slightly different in syntax and semantics:
1181 Flexible array members are written as @code{contents[]} without
1185 Flexible array members have incomplete type, and so the @code{sizeof}
1186 operator may not be applied. As a quirk of the original implementation
1187 of zero-length arrays, @code{sizeof} evaluates to zero.
1190 Flexible array members may only appear as the last member of a
1191 @code{struct} that is otherwise non-empty.
1194 A structure containing a flexible array member, or a union containing
1195 such a structure (possibly recursively), may not be a member of a
1196 structure or an element of an array. (However, these uses are
1197 permitted by GCC as extensions.)
1200 GCC versions before 3.0 allowed zero-length arrays to be statically
1201 initialized, as if they were flexible arrays. In addition to those
1202 cases that were useful, it also allowed initializations in situations
1203 that would corrupt later data. Non-empty initialization of zero-length
1204 arrays is now treated like any case where there are more initializer
1205 elements than the array holds, in that a suitable warning about "excess
1206 elements in array" is given, and the excess elements (all of them, in
1207 this case) are ignored.
1209 Instead GCC allows static initialization of flexible array members.
1210 This is equivalent to defining a new structure containing the original
1211 structure followed by an array of sufficient size to contain the data.
1212 I.e.@: in the following, @code{f1} is constructed as if it were declared
1218 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1221 struct f1 f1; int data[3];
1222 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1226 The convenience of this extension is that @code{f1} has the desired
1227 type, eliminating the need to consistently refer to @code{f2.f1}.
1229 This has symmetry with normal static arrays, in that an array of
1230 unknown size is also written with @code{[]}.
1232 Of course, this extension only makes sense if the extra data comes at
1233 the end of a top-level object, as otherwise we would be overwriting
1234 data at subsequent offsets. To avoid undue complication and confusion
1235 with initialization of deeply nested arrays, we simply disallow any
1236 non-empty initialization except when the structure is the top-level
1237 object. For example:
1240 struct foo @{ int x; int y[]; @};
1241 struct bar @{ struct foo z; @};
1243 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1244 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1245 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1246 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1249 @node Empty Structures
1250 @section Structures With No Members
1251 @cindex empty structures
1252 @cindex zero-size structures
1254 GCC permits a C structure to have no members:
1261 The structure will have size zero. In C++, empty structures are part
1262 of the language. G++ treats empty structures as if they had a single
1263 member of type @code{char}.
1265 @node Variable Length
1266 @section Arrays of Variable Length
1267 @cindex variable-length arrays
1268 @cindex arrays of variable length
1271 Variable-length automatic arrays are allowed in ISO C99, and as an
1272 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1273 implementation of variable-length arrays does not yet conform in detail
1274 to the ISO C99 standard.) These arrays are
1275 declared like any other automatic arrays, but with a length that is not
1276 a constant expression. The storage is allocated at the point of
1277 declaration and deallocated when the brace-level is exited. For
1282 concat_fopen (char *s1, char *s2, char *mode)
1284 char str[strlen (s1) + strlen (s2) + 1];
1287 return fopen (str, mode);
1291 @cindex scope of a variable length array
1292 @cindex variable-length array scope
1293 @cindex deallocating variable length arrays
1294 Jumping or breaking out of the scope of the array name deallocates the
1295 storage. Jumping into the scope is not allowed; you get an error
1298 @cindex @code{alloca} vs variable-length arrays
1299 You can use the function @code{alloca} to get an effect much like
1300 variable-length arrays. The function @code{alloca} is available in
1301 many other C implementations (but not in all). On the other hand,
1302 variable-length arrays are more elegant.
1304 There are other differences between these two methods. Space allocated
1305 with @code{alloca} exists until the containing @emph{function} returns.
1306 The space for a variable-length array is deallocated as soon as the array
1307 name's scope ends. (If you use both variable-length arrays and
1308 @code{alloca} in the same function, deallocation of a variable-length array
1309 will also deallocate anything more recently allocated with @code{alloca}.)
1311 You can also use variable-length arrays as arguments to functions:
1315 tester (int len, char data[len][len])
1321 The length of an array is computed once when the storage is allocated
1322 and is remembered for the scope of the array in case you access it with
1325 If you want to pass the array first and the length afterward, you can
1326 use a forward declaration in the parameter list---another GNU extension.
1330 tester (int len; char data[len][len], int len)
1336 @cindex parameter forward declaration
1337 The @samp{int len} before the semicolon is a @dfn{parameter forward
1338 declaration}, and it serves the purpose of making the name @code{len}
1339 known when the declaration of @code{data} is parsed.
1341 You can write any number of such parameter forward declarations in the
1342 parameter list. They can be separated by commas or semicolons, but the
1343 last one must end with a semicolon, which is followed by the ``real''
1344 parameter declarations. Each forward declaration must match a ``real''
1345 declaration in parameter name and data type. ISO C99 does not support
1346 parameter forward declarations.
1348 @node Variadic Macros
1349 @section Macros with a Variable Number of Arguments.
1350 @cindex variable number of arguments
1351 @cindex macro with variable arguments
1352 @cindex rest argument (in macro)
1353 @cindex variadic macros
1355 In the ISO C standard of 1999, a macro can be declared to accept a
1356 variable number of arguments much as a function can. The syntax for
1357 defining the macro is similar to that of a function. Here is an
1361 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1364 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1365 such a macro, it represents the zero or more tokens until the closing
1366 parenthesis that ends the invocation, including any commas. This set of
1367 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1368 wherever it appears. See the CPP manual for more information.
1370 GCC has long supported variadic macros, and used a different syntax that
1371 allowed you to give a name to the variable arguments just like any other
1372 argument. Here is an example:
1375 #define debug(format, args...) fprintf (stderr, format, args)
1378 This is in all ways equivalent to the ISO C example above, but arguably
1379 more readable and descriptive.
1381 GNU CPP has two further variadic macro extensions, and permits them to
1382 be used with either of the above forms of macro definition.
1384 In standard C, you are not allowed to leave the variable argument out
1385 entirely; but you are allowed to pass an empty argument. For example,
1386 this invocation is invalid in ISO C, because there is no comma after
1393 GNU CPP permits you to completely omit the variable arguments in this
1394 way. In the above examples, the compiler would complain, though since
1395 the expansion of the macro still has the extra comma after the format
1398 To help solve this problem, CPP behaves specially for variable arguments
1399 used with the token paste operator, @samp{##}. If instead you write
1402 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1405 and if the variable arguments are omitted or empty, the @samp{##}
1406 operator causes the preprocessor to remove the comma before it. If you
1407 do provide some variable arguments in your macro invocation, GNU CPP
1408 does not complain about the paste operation and instead places the
1409 variable arguments after the comma. Just like any other pasted macro
1410 argument, these arguments are not macro expanded.
1412 @node Escaped Newlines
1413 @section Slightly Looser Rules for Escaped Newlines
1414 @cindex escaped newlines
1415 @cindex newlines (escaped)
1417 Recently, the preprocessor has relaxed its treatment of escaped
1418 newlines. Previously, the newline had to immediately follow a
1419 backslash. The current implementation allows whitespace in the form
1420 of spaces, horizontal and vertical tabs, and form feeds between the
1421 backslash and the subsequent newline. The preprocessor issues a
1422 warning, but treats it as a valid escaped newline and combines the two
1423 lines to form a single logical line. This works within comments and
1424 tokens, as well as between tokens. Comments are @emph{not} treated as
1425 whitespace for the purposes of this relaxation, since they have not
1426 yet been replaced with spaces.
1429 @section Non-Lvalue Arrays May Have Subscripts
1430 @cindex subscripting
1431 @cindex arrays, non-lvalue
1433 @cindex subscripting and function values
1434 In ISO C99, arrays that are not lvalues still decay to pointers, and
1435 may be subscripted, although they may not be modified or used after
1436 the next sequence point and the unary @samp{&} operator may not be
1437 applied to them. As an extension, GCC allows such arrays to be
1438 subscripted in C89 mode, though otherwise they do not decay to
1439 pointers outside C99 mode. For example,
1440 this is valid in GNU C though not valid in C89:
1444 struct foo @{int a[4];@};
1450 return f().a[index];
1456 @section Arithmetic on @code{void}- and Function-Pointers
1457 @cindex void pointers, arithmetic
1458 @cindex void, size of pointer to
1459 @cindex function pointers, arithmetic
1460 @cindex function, size of pointer to
1462 In GNU C, addition and subtraction operations are supported on pointers to
1463 @code{void} and on pointers to functions. This is done by treating the
1464 size of a @code{void} or of a function as 1.
1466 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1467 and on function types, and returns 1.
1469 @opindex Wpointer-arith
1470 The option @option{-Wpointer-arith} requests a warning if these extensions
1474 @section Non-Constant Initializers
1475 @cindex initializers, non-constant
1476 @cindex non-constant initializers
1478 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1479 automatic variable are not required to be constant expressions in GNU C@.
1480 Here is an example of an initializer with run-time varying elements:
1483 foo (float f, float g)
1485 float beat_freqs[2] = @{ f-g, f+g @};
1490 @node Compound Literals
1491 @section Compound Literals
1492 @cindex constructor expressions
1493 @cindex initializations in expressions
1494 @cindex structures, constructor expression
1495 @cindex expressions, constructor
1496 @cindex compound literals
1497 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1499 ISO C99 supports compound literals. A compound literal looks like
1500 a cast containing an initializer. Its value is an object of the
1501 type specified in the cast, containing the elements specified in
1502 the initializer; it is an lvalue. As an extension, GCC supports
1503 compound literals in C89 mode and in C++.
1505 Usually, the specified type is a structure. Assume that
1506 @code{struct foo} and @code{structure} are declared as shown:
1509 struct foo @{int a; char b[2];@} structure;
1513 Here is an example of constructing a @code{struct foo} with a compound literal:
1516 structure = ((struct foo) @{x + y, 'a', 0@});
1520 This is equivalent to writing the following:
1524 struct foo temp = @{x + y, 'a', 0@};
1529 You can also construct an array. If all the elements of the compound literal
1530 are (made up of) simple constant expressions, suitable for use in
1531 initializers of objects of static storage duration, then the compound
1532 literal can be coerced to a pointer to its first element and used in
1533 such an initializer, as shown here:
1536 char **foo = (char *[]) @{ "x", "y", "z" @};
1539 Compound literals for scalar types and union types are is
1540 also allowed, but then the compound literal is equivalent
1543 As a GNU extension, GCC allows initialization of objects with static storage
1544 duration by compound literals (which is not possible in ISO C99, because
1545 the initializer is not a constant).
1546 It is handled as if the object was initialized only with the bracket
1547 enclosed list if the types of the compound literal and the object match.
1548 The initializer list of the compound literal must be constant.
1549 If the object being initialized has array type of unknown size, the size is
1550 determined by compound literal size.
1553 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1554 static int y[] = (int []) @{1, 2, 3@};
1555 static int z[] = (int [3]) @{1@};
1559 The above lines are equivalent to the following:
1561 static struct foo x = @{1, 'a', 'b'@};
1562 static int y[] = @{1, 2, 3@};
1563 static int z[] = @{1, 0, 0@};
1566 @node Designated Inits
1567 @section Designated Initializers
1568 @cindex initializers with labeled elements
1569 @cindex labeled elements in initializers
1570 @cindex case labels in initializers
1571 @cindex designated initializers
1573 Standard C89 requires the elements of an initializer to appear in a fixed
1574 order, the same as the order of the elements in the array or structure
1577 In ISO C99 you can give the elements in any order, specifying the array
1578 indices or structure field names they apply to, and GNU C allows this as
1579 an extension in C89 mode as well. This extension is not
1580 implemented in GNU C++.
1582 To specify an array index, write
1583 @samp{[@var{index}] =} before the element value. For example,
1586 int a[6] = @{ [4] = 29, [2] = 15 @};
1593 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1597 The index values must be constant expressions, even if the array being
1598 initialized is automatic.
1600 An alternative syntax for this which has been obsolete since GCC 2.5 but
1601 GCC still accepts is to write @samp{[@var{index}]} before the element
1602 value, with no @samp{=}.
1604 To initialize a range of elements to the same value, write
1605 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1606 extension. For example,
1609 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1613 If the value in it has side-effects, the side-effects will happen only once,
1614 not for each initialized field by the range initializer.
1617 Note that the length of the array is the highest value specified
1620 In a structure initializer, specify the name of a field to initialize
1621 with @samp{.@var{fieldname} =} before the element value. For example,
1622 given the following structure,
1625 struct point @{ int x, y; @};
1629 the following initialization
1632 struct point p = @{ .y = yvalue, .x = xvalue @};
1639 struct point p = @{ xvalue, yvalue @};
1642 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1643 @samp{@var{fieldname}:}, as shown here:
1646 struct point p = @{ y: yvalue, x: xvalue @};
1650 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1651 @dfn{designator}. You can also use a designator (or the obsolete colon
1652 syntax) when initializing a union, to specify which element of the union
1653 should be used. For example,
1656 union foo @{ int i; double d; @};
1658 union foo f = @{ .d = 4 @};
1662 will convert 4 to a @code{double} to store it in the union using
1663 the second element. By contrast, casting 4 to type @code{union foo}
1664 would store it into the union as the integer @code{i}, since it is
1665 an integer. (@xref{Cast to Union}.)
1667 You can combine this technique of naming elements with ordinary C
1668 initialization of successive elements. Each initializer element that
1669 does not have a designator applies to the next consecutive element of the
1670 array or structure. For example,
1673 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1680 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1683 Labeling the elements of an array initializer is especially useful
1684 when the indices are characters or belong to an @code{enum} type.
1689 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1690 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1693 @cindex designator lists
1694 You can also write a series of @samp{.@var{fieldname}} and
1695 @samp{[@var{index}]} designators before an @samp{=} to specify a
1696 nested subobject to initialize; the list is taken relative to the
1697 subobject corresponding to the closest surrounding brace pair. For
1698 example, with the @samp{struct point} declaration above:
1701 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1705 If the same field is initialized multiple times, it will have value from
1706 the last initialization. If any such overridden initialization has
1707 side-effect, it is unspecified whether the side-effect happens or not.
1708 Currently, GCC will discard them and issue a warning.
1711 @section Case Ranges
1713 @cindex ranges in case statements
1715 You can specify a range of consecutive values in a single @code{case} label,
1719 case @var{low} ... @var{high}:
1723 This has the same effect as the proper number of individual @code{case}
1724 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1726 This feature is especially useful for ranges of ASCII character codes:
1732 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1733 it may be parsed wrong when you use it with integer values. For example,
1748 @section Cast to a Union Type
1749 @cindex cast to a union
1750 @cindex union, casting to a
1752 A cast to union type is similar to other casts, except that the type
1753 specified is a union type. You can specify the type either with
1754 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1755 a constructor though, not a cast, and hence does not yield an lvalue like
1756 normal casts. (@xref{Compound Literals}.)
1758 The types that may be cast to the union type are those of the members
1759 of the union. Thus, given the following union and variables:
1762 union foo @{ int i; double d; @};
1768 both @code{x} and @code{y} can be cast to type @code{union foo}.
1770 Using the cast as the right-hand side of an assignment to a variable of
1771 union type is equivalent to storing in a member of the union:
1776 u = (union foo) x @equiv{} u.i = x
1777 u = (union foo) y @equiv{} u.d = y
1780 You can also use the union cast as a function argument:
1783 void hack (union foo);
1785 hack ((union foo) x);
1788 @node Mixed Declarations
1789 @section Mixed Declarations and Code
1790 @cindex mixed declarations and code
1791 @cindex declarations, mixed with code
1792 @cindex code, mixed with declarations
1794 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1795 within compound statements. As an extension, GCC also allows this in
1796 C89 mode. For example, you could do:
1805 Each identifier is visible from where it is declared until the end of
1806 the enclosing block.
1808 @node Function Attributes
1809 @section Declaring Attributes of Functions
1810 @cindex function attributes
1811 @cindex declaring attributes of functions
1812 @cindex functions that never return
1813 @cindex functions that return more than once
1814 @cindex functions that have no side effects
1815 @cindex functions in arbitrary sections
1816 @cindex functions that behave like malloc
1817 @cindex @code{volatile} applied to function
1818 @cindex @code{const} applied to function
1819 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1820 @cindex functions with non-null pointer arguments
1821 @cindex functions that are passed arguments in registers on the 386
1822 @cindex functions that pop the argument stack on the 386
1823 @cindex functions that do not pop the argument stack on the 386
1824 @cindex functions that have different compilation options on the 386
1825 @cindex functions that have different optimization options
1827 In GNU C, you declare certain things about functions called in your program
1828 which help the compiler optimize function calls and check your code more
1831 The keyword @code{__attribute__} allows you to specify special
1832 attributes when making a declaration. This keyword is followed by an
1833 attribute specification inside double parentheses. The following
1834 attributes are currently defined for functions on all targets:
1835 @code{aligned}, @code{alloc_size}, @code{noreturn},
1836 @code{returns_twice}, @code{noinline}, @code{always_inline},
1837 @code{flatten}, @code{pure}, @code{const}, @code{nothrow},
1838 @code{sentinel}, @code{format}, @code{format_arg},
1839 @code{no_instrument_function}, @code{section}, @code{constructor},
1840 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1841 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1842 @code{nonnull}, @code{gnu_inline}, @code{externally_visible},
1843 @code{hot}, @code{cold}, @code{artificial}, @code{error}
1845 Several other attributes are defined for functions on particular
1846 target systems. Other attributes, including @code{section} are
1847 supported for variables declarations (@pxref{Variable Attributes}) and
1848 for types (@pxref{Type Attributes}).
1850 You may also specify attributes with @samp{__} preceding and following
1851 each keyword. This allows you to use them in header files without
1852 being concerned about a possible macro of the same name. For example,
1853 you may use @code{__noreturn__} instead of @code{noreturn}.
1855 @xref{Attribute Syntax}, for details of the exact syntax for using
1859 @c Keep this table alphabetized by attribute name. Treat _ as space.
1861 @item alias ("@var{target}")
1862 @cindex @code{alias} attribute
1863 The @code{alias} attribute causes the declaration to be emitted as an
1864 alias for another symbol, which must be specified. For instance,
1867 void __f () @{ /* @r{Do something.} */; @}
1868 void f () __attribute__ ((weak, alias ("__f")));
1871 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1872 mangled name for the target must be used. It is an error if @samp{__f}
1873 is not defined in the same translation unit.
1875 Not all target machines support this attribute.
1877 @item aligned (@var{alignment})
1878 @cindex @code{aligned} attribute
1879 This attribute specifies a minimum alignment for the function,
1882 You cannot use this attribute to decrease the alignment of a function,
1883 only to increase it. However, when you explicitly specify a function
1884 alignment this will override the effect of the
1885 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1888 Note that the effectiveness of @code{aligned} attributes may be
1889 limited by inherent limitations in your linker. On many systems, the
1890 linker is only able to arrange for functions to be aligned up to a
1891 certain maximum alignment. (For some linkers, the maximum supported
1892 alignment may be very very small.) See your linker documentation for
1893 further information.
1895 The @code{aligned} attribute can also be used for variables and fields
1896 (@pxref{Variable Attributes}.)
1899 @cindex @code{alloc_size} attribute
1900 The @code{alloc_size} attribute is used to tell the compiler that the
1901 function return value points to memory, where the size is given by
1902 one or two of the functions parameters. GCC uses this
1903 information to improve the correctness of @code{__builtin_object_size}.
1905 The function parameter(s) denoting the allocated size are specified by
1906 one or two integer arguments supplied to the attribute. The allocated size
1907 is either the value of the single function argument specified or the product
1908 of the two function arguments specified. Argument numbering starts at
1914 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1915 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
1918 declares that my_calloc will return memory of the size given by
1919 the product of parameter 1 and 2 and that my_realloc will return memory
1920 of the size given by parameter 2.
1923 @cindex @code{always_inline} function attribute
1924 Generally, functions are not inlined unless optimization is specified.
1925 For functions declared inline, this attribute inlines the function even
1926 if no optimization level was specified.
1929 @cindex @code{gnu_inline} function attribute
1930 This attribute should be used with a function which is also declared
1931 with the @code{inline} keyword. It directs GCC to treat the function
1932 as if it were defined in gnu89 mode even when compiling in C99 or
1935 If the function is declared @code{extern}, then this definition of the
1936 function is used only for inlining. In no case is the function
1937 compiled as a standalone function, not even if you take its address
1938 explicitly. Such an address becomes an external reference, as if you
1939 had only declared the function, and had not defined it. This has
1940 almost the effect of a macro. The way to use this is to put a
1941 function definition in a header file with this attribute, and put
1942 another copy of the function, without @code{extern}, in a library
1943 file. The definition in the header file will cause most calls to the
1944 function to be inlined. If any uses of the function remain, they will
1945 refer to the single copy in the library. Note that the two
1946 definitions of the functions need not be precisely the same, although
1947 if they do not have the same effect your program may behave oddly.
1949 In C, if the function is neither @code{extern} nor @code{static}, then
1950 the function is compiled as a standalone function, as well as being
1951 inlined where possible.
1953 This is how GCC traditionally handled functions declared
1954 @code{inline}. Since ISO C99 specifies a different semantics for
1955 @code{inline}, this function attribute is provided as a transition
1956 measure and as a useful feature in its own right. This attribute is
1957 available in GCC 4.1.3 and later. It is available if either of the
1958 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1959 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1960 Function is As Fast As a Macro}.
1962 In C++, this attribute does not depend on @code{extern} in any way,
1963 but it still requires the @code{inline} keyword to enable its special
1967 @cindex @code{artificial} function attribute
1968 This attribute is useful for small inline wrappers which if possible
1969 should appear during debugging as a unit, depending on the debug
1970 info format it will either mean marking the function as artificial
1971 or using the caller location for all instructions within the inlined
1975 @cindex @code{flatten} function attribute
1976 Generally, inlining into a function is limited. For a function marked with
1977 this attribute, every call inside this function will be inlined, if possible.
1978 Whether the function itself is considered for inlining depends on its size and
1979 the current inlining parameters.
1981 @item error ("@var{message}")
1982 @cindex @code{error} function attribute
1983 If this attribute is used on a function declaration and a call to such a function
1984 is not eliminated through dead code elimination or other optimizations, an error
1985 which will include @var{message} will be diagnosed. This is useful
1986 for compile time checking, especially together with @code{__builtin_constant_p}
1987 and inline functions where checking the inline function arguments is not
1988 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
1989 While it is possible to leave the function undefined and thus invoke
1990 a link failure, when using this attribute the problem will be diagnosed
1991 earlier and with exact location of the call even in presence of inline
1992 functions or when not emitting debugging information.
1994 @item warning ("@var{message}")
1995 @cindex @code{warning} function attribute
1996 If this attribute is used on a function declaration and a call to such a function
1997 is not eliminated through dead code elimination or other optimizations, a warning
1998 which will include @var{message} will be diagnosed. This is useful
1999 for compile time checking, especially together with @code{__builtin_constant_p}
2000 and inline functions. While it is possible to define the function with
2001 a message in @code{.gnu.warning*} section, when using this attribute the problem
2002 will be diagnosed earlier and with exact location of the call even in presence
2003 of inline functions or when not emitting debugging information.
2006 @cindex functions that do pop the argument stack on the 386
2008 On the Intel 386, the @code{cdecl} attribute causes the compiler to
2009 assume that the calling function will pop off the stack space used to
2010 pass arguments. This is
2011 useful to override the effects of the @option{-mrtd} switch.
2014 @cindex @code{const} function attribute
2015 Many functions do not examine any values except their arguments, and
2016 have no effects except the return value. Basically this is just slightly
2017 more strict class than the @code{pure} attribute below, since function is not
2018 allowed to read global memory.
2020 @cindex pointer arguments
2021 Note that a function that has pointer arguments and examines the data
2022 pointed to must @emph{not} be declared @code{const}. Likewise, a
2023 function that calls a non-@code{const} function usually must not be
2024 @code{const}. It does not make sense for a @code{const} function to
2027 The attribute @code{const} is not implemented in GCC versions earlier
2028 than 2.5. An alternative way to declare that a function has no side
2029 effects, which works in the current version and in some older versions,
2033 typedef int intfn ();
2035 extern const intfn square;
2038 This approach does not work in GNU C++ from 2.6.0 on, since the language
2039 specifies that the @samp{const} must be attached to the return value.
2043 @itemx constructor (@var{priority})
2044 @itemx destructor (@var{priority})
2045 @cindex @code{constructor} function attribute
2046 @cindex @code{destructor} function attribute
2047 The @code{constructor} attribute causes the function to be called
2048 automatically before execution enters @code{main ()}. Similarly, the
2049 @code{destructor} attribute causes the function to be called
2050 automatically after @code{main ()} has completed or @code{exit ()} has
2051 been called. Functions with these attributes are useful for
2052 initializing data that will be used implicitly during the execution of
2055 You may provide an optional integer priority to control the order in
2056 which constructor and destructor functions are run. A constructor
2057 with a smaller priority number runs before a constructor with a larger
2058 priority number; the opposite relationship holds for destructors. So,
2059 if you have a constructor that allocates a resource and a destructor
2060 that deallocates the same resource, both functions typically have the
2061 same priority. The priorities for constructor and destructor
2062 functions are the same as those specified for namespace-scope C++
2063 objects (@pxref{C++ Attributes}).
2065 These attributes are not currently implemented for Objective-C@.
2068 @cindex @code{deprecated} attribute.
2069 The @code{deprecated} attribute results in a warning if the function
2070 is used anywhere in the source file. This is useful when identifying
2071 functions that are expected to be removed in a future version of a
2072 program. The warning also includes the location of the declaration
2073 of the deprecated function, to enable users to easily find further
2074 information about why the function is deprecated, or what they should
2075 do instead. Note that the warnings only occurs for uses:
2078 int old_fn () __attribute__ ((deprecated));
2080 int (*fn_ptr)() = old_fn;
2083 results in a warning on line 3 but not line 2.
2085 The @code{deprecated} attribute can also be used for variables and
2086 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2089 @cindex @code{__declspec(dllexport)}
2090 On Microsoft Windows targets and Symbian OS targets the
2091 @code{dllexport} attribute causes the compiler to provide a global
2092 pointer to a pointer in a DLL, so that it can be referenced with the
2093 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2094 name is formed by combining @code{_imp__} and the function or variable
2097 You can use @code{__declspec(dllexport)} as a synonym for
2098 @code{__attribute__ ((dllexport))} for compatibility with other
2101 On systems that support the @code{visibility} attribute, this
2102 attribute also implies ``default'' visibility. It is an error to
2103 explicitly specify any other visibility.
2105 Currently, the @code{dllexport} attribute is ignored for inlined
2106 functions, unless the @option{-fkeep-inline-functions} flag has been
2107 used. The attribute is also ignored for undefined symbols.
2109 When applied to C++ classes, the attribute marks defined non-inlined
2110 member functions and static data members as exports. Static consts
2111 initialized in-class are not marked unless they are also defined
2114 For Microsoft Windows targets there are alternative methods for
2115 including the symbol in the DLL's export table such as using a
2116 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2117 the @option{--export-all} linker flag.
2120 @cindex @code{__declspec(dllimport)}
2121 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2122 attribute causes the compiler to reference a function or variable via
2123 a global pointer to a pointer that is set up by the DLL exporting the
2124 symbol. The attribute implies @code{extern}. On Microsoft Windows
2125 targets, the pointer name is formed by combining @code{_imp__} and the
2126 function or variable name.
2128 You can use @code{__declspec(dllimport)} as a synonym for
2129 @code{__attribute__ ((dllimport))} for compatibility with other
2132 On systems that support the @code{visibility} attribute, this
2133 attribute also implies ``default'' visibility. It is an error to
2134 explicitly specify any other visibility.
2136 Currently, the attribute is ignored for inlined functions. If the
2137 attribute is applied to a symbol @emph{definition}, an error is reported.
2138 If a symbol previously declared @code{dllimport} is later defined, the
2139 attribute is ignored in subsequent references, and a warning is emitted.
2140 The attribute is also overridden by a subsequent declaration as
2143 When applied to C++ classes, the attribute marks non-inlined
2144 member functions and static data members as imports. However, the
2145 attribute is ignored for virtual methods to allow creation of vtables
2148 On the SH Symbian OS target the @code{dllimport} attribute also has
2149 another affect---it can cause the vtable and run-time type information
2150 for a class to be exported. This happens when the class has a
2151 dllimport'ed constructor or a non-inline, non-pure virtual function
2152 and, for either of those two conditions, the class also has a inline
2153 constructor or destructor and has a key function that is defined in
2154 the current translation unit.
2156 For Microsoft Windows based targets the use of the @code{dllimport}
2157 attribute on functions is not necessary, but provides a small
2158 performance benefit by eliminating a thunk in the DLL@. The use of the
2159 @code{dllimport} attribute on imported variables was required on older
2160 versions of the GNU linker, but can now be avoided by passing the
2161 @option{--enable-auto-import} switch to the GNU linker. As with
2162 functions, using the attribute for a variable eliminates a thunk in
2165 One drawback to using this attribute is that a pointer to a
2166 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2167 address. However, a pointer to a @emph{function} with the
2168 @code{dllimport} attribute can be used as a constant initializer; in
2169 this case, the address of a stub function in the import lib is
2170 referenced. On Microsoft Windows targets, the attribute can be disabled
2171 for functions by setting the @option{-mnop-fun-dllimport} flag.
2174 @cindex eight bit data on the H8/300, H8/300H, and H8S
2175 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2176 variable should be placed into the eight bit data section.
2177 The compiler will generate more efficient code for certain operations
2178 on data in the eight bit data area. Note the eight bit data area is limited to
2181 You must use GAS and GLD from GNU binutils version 2.7 or later for
2182 this attribute to work correctly.
2184 @item exception_handler
2185 @cindex exception handler functions on the Blackfin processor
2186 Use this attribute on the Blackfin to indicate that the specified function
2187 is an exception handler. The compiler will generate function entry and
2188 exit sequences suitable for use in an exception handler when this
2189 attribute is present.
2191 @item externally_visible
2192 @cindex @code{externally_visible} attribute.
2193 This attribute, attached to a global variable or function, nullifies
2194 the effect of the @option{-fwhole-program} command-line option, so the
2195 object remains visible outside the current compilation unit.
2198 @cindex functions which handle memory bank switching
2199 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2200 use a calling convention that takes care of switching memory banks when
2201 entering and leaving a function. This calling convention is also the
2202 default when using the @option{-mlong-calls} option.
2204 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2205 to call and return from a function.
2207 On 68HC11 the compiler will generate a sequence of instructions
2208 to invoke a board-specific routine to switch the memory bank and call the
2209 real function. The board-specific routine simulates a @code{call}.
2210 At the end of a function, it will jump to a board-specific routine
2211 instead of using @code{rts}. The board-specific return routine simulates
2215 @cindex functions that pop the argument stack on the 386
2216 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2217 pass the first argument (if of integral type) in the register ECX and
2218 the second argument (if of integral type) in the register EDX@. Subsequent
2219 and other typed arguments are passed on the stack. The called function will
2220 pop the arguments off the stack. If the number of arguments is variable all
2221 arguments are pushed on the stack.
2223 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2224 @cindex @code{format} function attribute
2226 The @code{format} attribute specifies that a function takes @code{printf},
2227 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2228 should be type-checked against a format string. For example, the
2233 my_printf (void *my_object, const char *my_format, ...)
2234 __attribute__ ((format (printf, 2, 3)));
2238 causes the compiler to check the arguments in calls to @code{my_printf}
2239 for consistency with the @code{printf} style format string argument
2242 The parameter @var{archetype} determines how the format string is
2243 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2244 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2245 @code{strfmon}. (You can also use @code{__printf__},
2246 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2247 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2248 @code{ms_strftime} are also present.
2249 @var{archtype} values such as @code{printf} refer to the formats accepted
2250 by the system's C run-time library, while @code{gnu_} values always refer
2251 to the formats accepted by the GNU C Library. On Microsoft Windows
2252 targets, @code{ms_} values refer to the formats accepted by the
2253 @file{msvcrt.dll} library.
2254 The parameter @var{string-index}
2255 specifies which argument is the format string argument (starting
2256 from 1), while @var{first-to-check} is the number of the first
2257 argument to check against the format string. For functions
2258 where the arguments are not available to be checked (such as
2259 @code{vprintf}), specify the third parameter as zero. In this case the
2260 compiler only checks the format string for consistency. For
2261 @code{strftime} formats, the third parameter is required to be zero.
2262 Since non-static C++ methods have an implicit @code{this} argument, the
2263 arguments of such methods should be counted from two, not one, when
2264 giving values for @var{string-index} and @var{first-to-check}.
2266 In the example above, the format string (@code{my_format}) is the second
2267 argument of the function @code{my_print}, and the arguments to check
2268 start with the third argument, so the correct parameters for the format
2269 attribute are 2 and 3.
2271 @opindex ffreestanding
2272 @opindex fno-builtin
2273 The @code{format} attribute allows you to identify your own functions
2274 which take format strings as arguments, so that GCC can check the
2275 calls to these functions for errors. The compiler always (unless
2276 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2277 for the standard library functions @code{printf}, @code{fprintf},
2278 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2279 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2280 warnings are requested (using @option{-Wformat}), so there is no need to
2281 modify the header file @file{stdio.h}. In C99 mode, the functions
2282 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2283 @code{vsscanf} are also checked. Except in strictly conforming C
2284 standard modes, the X/Open function @code{strfmon} is also checked as
2285 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2286 @xref{C Dialect Options,,Options Controlling C Dialect}.
2288 The target may provide additional types of format checks.
2289 @xref{Target Format Checks,,Format Checks Specific to Particular
2292 @item format_arg (@var{string-index})
2293 @cindex @code{format_arg} function attribute
2294 @opindex Wformat-nonliteral
2295 The @code{format_arg} attribute specifies that a function takes a format
2296 string for a @code{printf}, @code{scanf}, @code{strftime} or
2297 @code{strfmon} style function and modifies it (for example, to translate
2298 it into another language), so the result can be passed to a
2299 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2300 function (with the remaining arguments to the format function the same
2301 as they would have been for the unmodified string). For example, the
2306 my_dgettext (char *my_domain, const char *my_format)
2307 __attribute__ ((format_arg (2)));
2311 causes the compiler to check the arguments in calls to a @code{printf},
2312 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2313 format string argument is a call to the @code{my_dgettext} function, for
2314 consistency with the format string argument @code{my_format}. If the
2315 @code{format_arg} attribute had not been specified, all the compiler
2316 could tell in such calls to format functions would be that the format
2317 string argument is not constant; this would generate a warning when
2318 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2319 without the attribute.
2321 The parameter @var{string-index} specifies which argument is the format
2322 string argument (starting from one). Since non-static C++ methods have
2323 an implicit @code{this} argument, the arguments of such methods should
2324 be counted from two.
2326 The @code{format-arg} attribute allows you to identify your own
2327 functions which modify format strings, so that GCC can check the
2328 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2329 type function whose operands are a call to one of your own function.
2330 The compiler always treats @code{gettext}, @code{dgettext}, and
2331 @code{dcgettext} in this manner except when strict ISO C support is
2332 requested by @option{-ansi} or an appropriate @option{-std} option, or
2333 @option{-ffreestanding} or @option{-fno-builtin}
2334 is used. @xref{C Dialect Options,,Options
2335 Controlling C Dialect}.
2337 @item function_vector
2338 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2339 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2340 function should be called through the function vector. Calling a
2341 function through the function vector will reduce code size, however;
2342 the function vector has a limited size (maximum 128 entries on the H8/300
2343 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2345 In SH2A target, this attribute declares a function to be called using the
2346 TBR relative addressing mode. The argument to this attribute is the entry
2347 number of the same function in a vector table containing all the TBR
2348 relative addressable functions. For the successful jump, register TBR
2349 should contain the start address of this TBR relative vector table.
2350 In the startup routine of the user application, user needs to care of this
2351 TBR register initialization. The TBR relative vector table can have at
2352 max 256 function entries. The jumps to these functions will be generated
2353 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2354 You must use GAS and GLD from GNU binutils version 2.7 or later for
2355 this attribute to work correctly.
2357 Please refer the example of M16C target, to see the use of this
2358 attribute while declaring a function,
2360 In an application, for a function being called once, this attribute will
2361 save at least 8 bytes of code; and if other successive calls are being
2362 made to the same function, it will save 2 bytes of code per each of these
2365 On M16C/M32C targets, the @code{function_vector} attribute declares a
2366 special page subroutine call function. Use of this attribute reduces
2367 the code size by 2 bytes for each call generated to the
2368 subroutine. The argument to the attribute is the vector number entry
2369 from the special page vector table which contains the 16 low-order
2370 bits of the subroutine's entry address. Each vector table has special
2371 page number (18 to 255) which are used in @code{jsrs} instruction.
2372 Jump addresses of the routines are generated by adding 0x0F0000 (in
2373 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2374 byte addresses set in the vector table. Therefore you need to ensure
2375 that all the special page vector routines should get mapped within the
2376 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2379 In the following example 2 bytes will be saved for each call to
2380 function @code{foo}.
2383 void foo (void) __attribute__((function_vector(0x18)));
2394 If functions are defined in one file and are called in another file,
2395 then be sure to write this declaration in both files.
2397 This attribute is ignored for R8C target.
2400 @cindex interrupt handler functions
2401 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k,
2402 and Xstormy16 ports to indicate that the specified function is an
2403 interrupt handler. The compiler will generate function entry and exit
2404 sequences suitable for use in an interrupt handler when this attribute
2407 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2408 SH processors can be specified via the @code{interrupt_handler} attribute.
2410 Note, on the AVR, interrupts will be enabled inside the function.
2412 Note, for the ARM, you can specify the kind of interrupt to be handled by
2413 adding an optional parameter to the interrupt attribute like this:
2416 void f () __attribute__ ((interrupt ("IRQ")));
2419 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2421 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2422 may be called with a word aligned stack pointer.
2424 @item interrupt_handler
2425 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2426 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2427 indicate that the specified function is an interrupt handler. The compiler
2428 will generate function entry and exit sequences suitable for use in an
2429 interrupt handler when this attribute is present.
2431 @item interrupt_thread
2432 @cindex interrupt thread functions on fido
2433 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2434 that the specified function is an interrupt handler that is designed
2435 to run as a thread. The compiler omits generate prologue/epilogue
2436 sequences and replaces the return instruction with a @code{sleep}
2437 instruction. This attribute is available only on fido.
2440 @cindex interrupt service routines on ARM
2441 Use this attribute on ARM to write Interrupt Service Routines. This is an
2442 alias to the @code{interrupt} attribute above.
2445 @cindex User stack pointer in interrupts on the Blackfin
2446 When used together with @code{interrupt_handler}, @code{exception_handler}
2447 or @code{nmi_handler}, code will be generated to load the stack pointer
2448 from the USP register in the function prologue.
2451 @cindex @code{l1_text} function attribute
2452 This attribute specifies a function to be placed into L1 Instruction
2453 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2454 With @option{-mfdpic}, function calls with a such function as the callee
2455 or caller will use inlined PLT.
2457 @item long_call/short_call
2458 @cindex indirect calls on ARM
2459 This attribute specifies how a particular function is called on
2460 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2461 command line switch and @code{#pragma long_calls} settings. The
2462 @code{long_call} attribute indicates that the function might be far
2463 away from the call site and require a different (more expensive)
2464 calling sequence. The @code{short_call} attribute always places
2465 the offset to the function from the call site into the @samp{BL}
2466 instruction directly.
2468 @item longcall/shortcall
2469 @cindex functions called via pointer on the RS/6000 and PowerPC
2470 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2471 indicates that the function might be far away from the call site and
2472 require a different (more expensive) calling sequence. The
2473 @code{shortcall} attribute indicates that the function is always close
2474 enough for the shorter calling sequence to be used. These attributes
2475 override both the @option{-mlongcall} switch and, on the RS/6000 and
2476 PowerPC, the @code{#pragma longcall} setting.
2478 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2479 calls are necessary.
2481 @item long_call/near/far
2482 @cindex indirect calls on MIPS
2483 These attributes specify how a particular function is called on MIPS@.
2484 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2485 command-line switch. The @code{long_call} and @code{far} attributes are
2486 synonyms, and cause the compiler to always call
2487 the function by first loading its address into a register, and then using
2488 the contents of that register. The @code{near} attribute has the opposite
2489 effect; it specifies that non-PIC calls should be made using the more
2490 efficient @code{jal} instruction.
2493 @cindex @code{malloc} attribute
2494 The @code{malloc} attribute is used to tell the compiler that a function
2495 may be treated as if any non-@code{NULL} pointer it returns cannot
2496 alias any other pointer valid when the function returns.
2497 This will often improve optimization.
2498 Standard functions with this property include @code{malloc} and
2499 @code{calloc}. @code{realloc}-like functions have this property as
2500 long as the old pointer is never referred to (including comparing it
2501 to the new pointer) after the function returns a non-@code{NULL}
2504 @item mips16/nomips16
2505 @cindex @code{mips16} attribute
2506 @cindex @code{nomips16} attribute
2508 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2509 function attributes to locally select or turn off MIPS16 code generation.
2510 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2511 while MIPS16 code generation is disabled for functions with the
2512 @code{nomips16} attribute. These attributes override the
2513 @option{-mips16} and @option{-mno-mips16} options on the command line
2514 (@pxref{MIPS Options}).
2516 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2517 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2518 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2519 may interact badly with some GCC extensions such as @code{__builtin_apply}
2520 (@pxref{Constructing Calls}).
2522 @item model (@var{model-name})
2523 @cindex function addressability on the M32R/D
2524 @cindex variable addressability on the IA-64
2526 On the M32R/D, use this attribute to set the addressability of an
2527 object, and of the code generated for a function. The identifier
2528 @var{model-name} is one of @code{small}, @code{medium}, or
2529 @code{large}, representing each of the code models.
2531 Small model objects live in the lower 16MB of memory (so that their
2532 addresses can be loaded with the @code{ld24} instruction), and are
2533 callable with the @code{bl} instruction.
2535 Medium model objects may live anywhere in the 32-bit address space (the
2536 compiler will generate @code{seth/add3} instructions to load their addresses),
2537 and are callable with the @code{bl} instruction.
2539 Large model objects may live anywhere in the 32-bit address space (the
2540 compiler will generate @code{seth/add3} instructions to load their addresses),
2541 and may not be reachable with the @code{bl} instruction (the compiler will
2542 generate the much slower @code{seth/add3/jl} instruction sequence).
2544 On IA-64, use this attribute to set the addressability of an object.
2545 At present, the only supported identifier for @var{model-name} is
2546 @code{small}, indicating addressability via ``small'' (22-bit)
2547 addresses (so that their addresses can be loaded with the @code{addl}
2548 instruction). Caveat: such addressing is by definition not position
2549 independent and hence this attribute must not be used for objects
2550 defined by shared libraries.
2552 @item ms_abi/sysv_abi
2553 @cindex @code{ms_abi} attribute
2554 @cindex @code{sysv_abi} attribute
2556 On 64-bit x86_64-*-* targets, you can use an ABI attribute to indicate
2557 which calling convention should be used for a function. The @code{ms_abi}
2558 attribute tells the compiler to use the Microsoft ABI, while the
2559 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2560 GNU/Linux and other systems. The default is to use the Microsoft ABI
2561 when targeting Windows. On all other systems, the default is the AMD ABI.
2563 Note, This feature is currently sorried out for Windows targets trying to
2566 @cindex function without a prologue/epilogue code
2567 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate that
2568 the specified function does not need prologue/epilogue sequences generated by
2569 the compiler. It is up to the programmer to provide these sequences. The
2570 only statements that can be safely included in naked functions are
2571 @code{asm} statements that do not have operands. All other statements,
2572 including declarations of local variables, @code{if} statements, and so
2573 forth, should be avoided. Naked functions should be used to implement the
2574 body of an assembly function, while allowing the compiler to construct
2575 the requisite function declaration for the assembler.
2578 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2579 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2580 use the normal calling convention based on @code{jsr} and @code{rts}.
2581 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2585 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2586 Use this attribute together with @code{interrupt_handler},
2587 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2588 entry code should enable nested interrupts or exceptions.
2591 @cindex NMI handler functions on the Blackfin processor
2592 Use this attribute on the Blackfin to indicate that the specified function
2593 is an NMI handler. The compiler will generate function entry and
2594 exit sequences suitable for use in an NMI handler when this
2595 attribute is present.
2597 @item no_instrument_function
2598 @cindex @code{no_instrument_function} function attribute
2599 @opindex finstrument-functions
2600 If @option{-finstrument-functions} is given, profiling function calls will
2601 be generated at entry and exit of most user-compiled functions.
2602 Functions with this attribute will not be so instrumented.
2605 @cindex @code{noinline} function attribute
2606 This function attribute prevents a function from being considered for
2608 @c Don't enumerate the optimizations by name here; we try to be
2609 @c future-compatible with this mechanism.
2610 If the function does not have side-effects, there are optimizations
2611 other than inlining that causes function calls to be optimized away,
2612 although the function call is live. To keep such calls from being
2617 (@pxref{Extended Asm}) in the called function, to serve as a special
2620 @item nonnull (@var{arg-index}, @dots{})
2621 @cindex @code{nonnull} function attribute
2622 The @code{nonnull} attribute specifies that some function parameters should
2623 be non-null pointers. For instance, the declaration:
2627 my_memcpy (void *dest, const void *src, size_t len)
2628 __attribute__((nonnull (1, 2)));
2632 causes the compiler to check that, in calls to @code{my_memcpy},
2633 arguments @var{dest} and @var{src} are non-null. If the compiler
2634 determines that a null pointer is passed in an argument slot marked
2635 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2636 is issued. The compiler may also choose to make optimizations based
2637 on the knowledge that certain function arguments will not be null.
2639 If no argument index list is given to the @code{nonnull} attribute,
2640 all pointer arguments are marked as non-null. To illustrate, the
2641 following declaration is equivalent to the previous example:
2645 my_memcpy (void *dest, const void *src, size_t len)
2646 __attribute__((nonnull));
2650 @cindex @code{noreturn} function attribute
2651 A few standard library functions, such as @code{abort} and @code{exit},
2652 cannot return. GCC knows this automatically. Some programs define
2653 their own functions that never return. You can declare them
2654 @code{noreturn} to tell the compiler this fact. For example,
2658 void fatal () __attribute__ ((noreturn));
2661 fatal (/* @r{@dots{}} */)
2663 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2669 The @code{noreturn} keyword tells the compiler to assume that
2670 @code{fatal} cannot return. It can then optimize without regard to what
2671 would happen if @code{fatal} ever did return. This makes slightly
2672 better code. More importantly, it helps avoid spurious warnings of
2673 uninitialized variables.
2675 The @code{noreturn} keyword does not affect the exceptional path when that
2676 applies: a @code{noreturn}-marked function may still return to the caller
2677 by throwing an exception or calling @code{longjmp}.
2679 Do not assume that registers saved by the calling function are
2680 restored before calling the @code{noreturn} function.
2682 It does not make sense for a @code{noreturn} function to have a return
2683 type other than @code{void}.
2685 The attribute @code{noreturn} is not implemented in GCC versions
2686 earlier than 2.5. An alternative way to declare that a function does
2687 not return, which works in the current version and in some older
2688 versions, is as follows:
2691 typedef void voidfn ();
2693 volatile voidfn fatal;
2696 This approach does not work in GNU C++.
2699 @cindex @code{nothrow} function attribute
2700 The @code{nothrow} attribute is used to inform the compiler that a
2701 function cannot throw an exception. For example, most functions in
2702 the standard C library can be guaranteed not to throw an exception
2703 with the notable exceptions of @code{qsort} and @code{bsearch} that
2704 take function pointer arguments. The @code{nothrow} attribute is not
2705 implemented in GCC versions earlier than 3.3.
2708 @cindex @code{optimize} function attribute
2709 The @code{optimize} attribute is used to specify that a function is to
2710 be compiled with different optimization options than specified on the
2711 command line. Arguments can either be numbers or strings. Numbers
2712 are assumed to be an optimization level. Strings that begin with
2713 @code{O} are assumed to be an optimization option, while other options
2714 are assumed to be used with a @code{-f} prefix. You can also use the
2715 @samp{#pragma GCC optimize} pragma to set the optimization options
2716 that affect more than one function.
2717 @xref{Function Specific Option Pragmas}, for details about the
2718 @samp{#pragma GCC optimize} pragma.
2720 This can be used for instance to have frequently executed functions
2721 compiled with more aggressive optimization options that produce faster
2722 and larger code, while other functions can be called with less
2726 @cindex @code{pure} function attribute
2727 Many functions have no effects except the return value and their
2728 return value depends only on the parameters and/or global variables.
2729 Such a function can be subject
2730 to common subexpression elimination and loop optimization just as an
2731 arithmetic operator would be. These functions should be declared
2732 with the attribute @code{pure}. For example,
2735 int square (int) __attribute__ ((pure));
2739 says that the hypothetical function @code{square} is safe to call
2740 fewer times than the program says.
2742 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2743 Interesting non-pure functions are functions with infinite loops or those
2744 depending on volatile memory or other system resource, that may change between
2745 two consecutive calls (such as @code{feof} in a multithreading environment).
2747 The attribute @code{pure} is not implemented in GCC versions earlier
2751 @cindex @code{hot} function attribute
2752 The @code{hot} attribute is used to inform the compiler that a function is a
2753 hot spot of the compiled program. The function is optimized more aggressively
2754 and on many target it is placed into special subsection of the text section so
2755 all hot functions appears close together improving locality.
2757 When profile feedback is available, via @option{-fprofile-use}, hot functions
2758 are automatically detected and this attribute is ignored.
2760 The @code{hot} attribute is not implemented in GCC versions earlier
2764 @cindex @code{cold} function attribute
2765 The @code{cold} attribute is used to inform the compiler that a function is
2766 unlikely executed. The function is optimized for size rather than speed and on
2767 many targets it is placed into special subsection of the text section so all
2768 cold functions appears close together improving code locality of non-cold parts
2769 of program. The paths leading to call of cold functions within code are marked
2770 as unlikely by the branch prediction mechanism. It is thus useful to mark
2771 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2772 improve optimization of hot functions that do call marked functions in rare
2775 When profile feedback is available, via @option{-fprofile-use}, hot functions
2776 are automatically detected and this attribute is ignored.
2778 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
2780 @item regparm (@var{number})
2781 @cindex @code{regparm} attribute
2782 @cindex functions that are passed arguments in registers on the 386
2783 On the Intel 386, the @code{regparm} attribute causes the compiler to
2784 pass arguments number one to @var{number} if they are of integral type
2785 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2786 take a variable number of arguments will continue to be passed all of their
2787 arguments on the stack.
2789 Beware that on some ELF systems this attribute is unsuitable for
2790 global functions in shared libraries with lazy binding (which is the
2791 default). Lazy binding will send the first call via resolving code in
2792 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2793 per the standard calling conventions. Solaris 8 is affected by this.
2794 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2795 safe since the loaders there save EAX, EDX and ECX. (Lazy binding can be
2796 disabled with the linker or the loader if desired, to avoid the
2800 @cindex @code{sseregparm} attribute
2801 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2802 causes the compiler to pass up to 3 floating point arguments in
2803 SSE registers instead of on the stack. Functions that take a
2804 variable number of arguments will continue to pass all of their
2805 floating point arguments on the stack.
2807 @item force_align_arg_pointer
2808 @cindex @code{force_align_arg_pointer} attribute
2809 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2810 applied to individual function definitions, generating an alternate
2811 prologue and epilogue that realigns the runtime stack if necessary.
2812 This supports mixing legacy codes that run with a 4-byte aligned stack
2813 with modern codes that keep a 16-byte stack for SSE compatibility.
2816 @cindex @code{resbank} attribute
2817 On the SH2A target, this attribute enables the high-speed register
2818 saving and restoration using a register bank for @code{interrupt_handler}
2819 routines. Saving to the bank is performed automatically after the CPU
2820 accepts an interrupt that uses a register bank.
2822 The nineteen 32-bit registers comprising general register R0 to R14,
2823 control register GBR, and system registers MACH, MACL, and PR and the
2824 vector table address offset are saved into a register bank. Register
2825 banks are stacked in first-in last-out (FILO) sequence. Restoration
2826 from the bank is executed by issuing a RESBANK instruction.
2829 @cindex @code{returns_twice} attribute
2830 The @code{returns_twice} attribute tells the compiler that a function may
2831 return more than one time. The compiler will ensure that all registers
2832 are dead before calling such a function and will emit a warning about
2833 the variables that may be clobbered after the second return from the
2834 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2835 The @code{longjmp}-like counterpart of such function, if any, might need
2836 to be marked with the @code{noreturn} attribute.
2839 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2840 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2841 all registers except the stack pointer should be saved in the prologue
2842 regardless of whether they are used or not.
2844 @item section ("@var{section-name}")
2845 @cindex @code{section} function attribute
2846 Normally, the compiler places the code it generates in the @code{text} section.
2847 Sometimes, however, you need additional sections, or you need certain
2848 particular functions to appear in special sections. The @code{section}
2849 attribute specifies that a function lives in a particular section.
2850 For example, the declaration:
2853 extern void foobar (void) __attribute__ ((section ("bar")));
2857 puts the function @code{foobar} in the @code{bar} section.
2859 Some file formats do not support arbitrary sections so the @code{section}
2860 attribute is not available on all platforms.
2861 If you need to map the entire contents of a module to a particular
2862 section, consider using the facilities of the linker instead.
2865 @cindex @code{sentinel} function attribute
2866 This function attribute ensures that a parameter in a function call is
2867 an explicit @code{NULL}. The attribute is only valid on variadic
2868 functions. By default, the sentinel is located at position zero, the
2869 last parameter of the function call. If an optional integer position
2870 argument P is supplied to the attribute, the sentinel must be located at
2871 position P counting backwards from the end of the argument list.
2874 __attribute__ ((sentinel))
2876 __attribute__ ((sentinel(0)))
2879 The attribute is automatically set with a position of 0 for the built-in
2880 functions @code{execl} and @code{execlp}. The built-in function
2881 @code{execle} has the attribute set with a position of 1.
2883 A valid @code{NULL} in this context is defined as zero with any pointer
2884 type. If your system defines the @code{NULL} macro with an integer type
2885 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2886 with a copy that redefines NULL appropriately.
2888 The warnings for missing or incorrect sentinels are enabled with
2892 See long_call/short_call.
2895 See longcall/shortcall.
2898 @cindex signal handler functions on the AVR processors
2899 Use this attribute on the AVR to indicate that the specified
2900 function is a signal handler. The compiler will generate function
2901 entry and exit sequences suitable for use in a signal handler when this
2902 attribute is present. Interrupts will be disabled inside the function.
2905 Use this attribute on the SH to indicate an @code{interrupt_handler}
2906 function should switch to an alternate stack. It expects a string
2907 argument that names a global variable holding the address of the
2912 void f () __attribute__ ((interrupt_handler,
2913 sp_switch ("alt_stack")));
2917 @cindex functions that pop the argument stack on the 386
2918 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2919 assume that the called function will pop off the stack space used to
2920 pass arguments, unless it takes a variable number of arguments.
2922 @item syscall_linkage
2923 @cindex @code{syscall_linkage} attribute
2924 This attribute is used to modify the IA64 calling convention by marking
2925 all input registers as live at all function exits. This makes it possible
2926 to restart a system call after an interrupt without having to save/restore
2927 the input registers. This also prevents kernel data from leaking into
2931 @cindex @code{target} function attribute
2932 The @code{target} attribute is used to specify that a function is to
2933 be compiled with different target options than specified on the
2934 command line. This can be used for instance to have functions
2935 compiled with a different ISA (instruction set architecture) than the
2936 default. You can also use the @samp{#pragma GCC target} pragma to set
2937 more than one function to be compiled with specific target options.
2938 @xref{Function Specific Option Pragmas}, for details about the
2939 @samp{#pragma GCC target} pragma.
2941 For instance on a 386, you could compile one function with
2942 @code{target("sse4.1,arch=core2")} and another with
2943 @code{target("sse4a,arch=amdfam10")} that would be equivalent to
2944 compiling the first function with @option{-msse4.1} and
2945 @option{-march=core2} options, and the second function with
2946 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
2947 user to make sure that a function is only invoked on a machine that
2948 supports the particular ISA it was compiled for (for example by using
2949 @code{cpuid} on 386 to determine what feature bits and architecture
2953 int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
2954 int sse3_func (void) __attribute__ ((__target__ ("sse3")));
2957 On the 386, the following options are allowed:
2962 @cindex @code{target("abm")} attribute
2963 Enable/disable the generation of the advanced bit instructions.
2967 @cindex @code{target("aes")} attribute
2968 Enable/disable the generation of the AES instructions.
2972 @cindex @code{target("mmx")} attribute
2973 Enable/disable the generation of the MMX instructions.
2977 @cindex @code{target("pclmul")} attribute
2978 Enable/disable the generation of the PCLMUL instructions.
2982 @cindex @code{target("popcnt")} attribute
2983 Enable/disable the generation of the POPCNT instruction.
2987 @cindex @code{target("sse")} attribute
2988 Enable/disable the generation of the SSE instructions.
2992 @cindex @code{target("sse2")} attribute
2993 Enable/disable the generation of the SSE2 instructions.
2997 @cindex @code{target("sse3")} attribute
2998 Enable/disable the generation of the SSE3 instructions.
3002 @cindex @code{target("sse4")} attribute
3003 Enable/disable the generation of the SSE4 instructions (both SSE4.1
3008 @cindex @code{target("sse4.1")} attribute
3009 Enable/disable the generation of the sse4.1 instructions.
3013 @cindex @code{target("sse4.2")} attribute
3014 Enable/disable the generation of the sse4.2 instructions.
3018 @cindex @code{target("sse4a")} attribute
3019 Enable/disable the generation of the SSE4A instructions.
3023 @cindex @code{target("sse5")} attribute
3024 Enable/disable the generation of the SSE5 instructions.
3028 @cindex @code{target("ssse3")} attribute
3029 Enable/disable the generation of the SSSE3 instructions.
3033 @cindex @code{target("cld")} attribute
3034 Enable/disable the generation of the CLD before string moves.
3036 @item fancy-math-387
3037 @itemx no-fancy-math-387
3038 @cindex @code{target("fancy-math-387")} attribute
3039 Enable/disable the generation of the @code{sin}, @code{cos}, and
3040 @code{sqrt} instructions on the 387 floating point unit.
3043 @itemx no-fused-madd
3044 @cindex @code{target("fused-madd")} attribute
3045 Enable/disable the generation of the fused multiply/add instructions.
3049 @cindex @code{target("ieee-fp")} attribute
3050 Enable/disable the generation of floating point that depends on IEEE arithmetic.
3052 @item inline-all-stringops
3053 @itemx no-inline-all-stringops
3054 @cindex @code{target("inline-all-stringops")} attribute
3055 Enable/disable inlining of string operations.
3057 @item inline-stringops-dynamically
3058 @itemx no-inline-stringops-dynamically
3059 @cindex @code{target("inline-stringops-dynamically")} attribute
3060 Enable/disable the generation of the inline code to do small string
3061 operations and calling the library routines for large operations.
3063 @item align-stringops
3064 @itemx no-align-stringops
3065 @cindex @code{target("align-stringops")} attribute
3066 Do/do not align destination of inlined string operations.
3070 @cindex @code{target("recip")} attribute
3071 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
3072 instructions followed an additional Newton-Raphson step instead of
3073 doing a floating point division.
3075 @item arch=@var{ARCH}
3076 @cindex @code{target("arch=@var{ARCH}")} attribute
3077 Specify the architecture to generate code for in compiling the function.
3079 @item tune=@var{TUNE}
3080 @cindex @code{target("tune=@var{TUNE}")} attribute
3081 Specify the architecture to tune for in compiling the function.
3083 @item fpmath=@var{FPMATH}
3084 @cindex @code{target("fpmath=@var{FPMATH}")} attribute
3085 Specify which floating point unit to use. The
3086 @code{target("fpmath=sse,387")} option must be specified as
3087 @code{target("fpmath=sse+387")} because the comma would separate
3091 On the 386, you can use either multiple strings to specify multiple
3092 options, or you can separate the option with a comma (@code{,}).
3094 On the 386, the inliner will not inline a function that has different
3095 target options than the caller, unless the callee has a subset of the
3096 target options of the caller. For example a function declared with
3097 @code{target("sse5")} can inline a function with
3098 @code{target("sse2")}, since @code{-msse5} implies @code{-msse2}.
3100 The @code{target} attribute is not implemented in GCC versions earlier
3101 than 4.4, and at present only the 386 uses it.
3104 @cindex tiny data section on the H8/300H and H8S
3105 Use this attribute on the H8/300H and H8S to indicate that the specified
3106 variable should be placed into the tiny data section.
3107 The compiler will generate more efficient code for loads and stores
3108 on data in the tiny data section. Note the tiny data area is limited to
3109 slightly under 32kbytes of data.
3112 Use this attribute on the SH for an @code{interrupt_handler} to return using
3113 @code{trapa} instead of @code{rte}. This attribute expects an integer
3114 argument specifying the trap number to be used.
3117 @cindex @code{unused} attribute.
3118 This attribute, attached to a function, means that the function is meant
3119 to be possibly unused. GCC will not produce a warning for this
3123 @cindex @code{used} attribute.
3124 This attribute, attached to a function, means that code must be emitted
3125 for the function even if it appears that the function is not referenced.
3126 This is useful, for example, when the function is referenced only in
3130 @cindex @code{version_id} attribute
3131 This IA64 HP-UX attribute, attached to a global variable or function, renames a
3132 symbol to contain a version string, thus allowing for function level
3133 versioning. HP-UX system header files may use version level functioning
3134 for some system calls.
3137 extern int foo () __attribute__((version_id ("20040821")));
3140 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3142 @item visibility ("@var{visibility_type}")
3143 @cindex @code{visibility} attribute
3144 This attribute affects the linkage of the declaration to which it is attached.
3145 There are four supported @var{visibility_type} values: default,
3146 hidden, protected or internal visibility.
3149 void __attribute__ ((visibility ("protected")))
3150 f () @{ /* @r{Do something.} */; @}
3151 int i __attribute__ ((visibility ("hidden")));
3154 The possible values of @var{visibility_type} correspond to the
3155 visibility settings in the ELF gABI.
3158 @c keep this list of visibilities in alphabetical order.
3161 Default visibility is the normal case for the object file format.
3162 This value is available for the visibility attribute to override other
3163 options that may change the assumed visibility of entities.
3165 On ELF, default visibility means that the declaration is visible to other
3166 modules and, in shared libraries, means that the declared entity may be
3169 On Darwin, default visibility means that the declaration is visible to
3172 Default visibility corresponds to ``external linkage'' in the language.
3175 Hidden visibility indicates that the entity declared will have a new
3176 form of linkage, which we'll call ``hidden linkage''. Two
3177 declarations of an object with hidden linkage refer to the same object
3178 if they are in the same shared object.
3181 Internal visibility is like hidden visibility, but with additional
3182 processor specific semantics. Unless otherwise specified by the
3183 psABI, GCC defines internal visibility to mean that a function is
3184 @emph{never} called from another module. Compare this with hidden
3185 functions which, while they cannot be referenced directly by other
3186 modules, can be referenced indirectly via function pointers. By
3187 indicating that a function cannot be called from outside the module,
3188 GCC may for instance omit the load of a PIC register since it is known
3189 that the calling function loaded the correct value.
3192 Protected visibility is like default visibility except that it
3193 indicates that references within the defining module will bind to the
3194 definition in that module. That is, the declared entity cannot be
3195 overridden by another module.
3199 All visibilities are supported on many, but not all, ELF targets
3200 (supported when the assembler supports the @samp{.visibility}
3201 pseudo-op). Default visibility is supported everywhere. Hidden
3202 visibility is supported on Darwin targets.
3204 The visibility attribute should be applied only to declarations which
3205 would otherwise have external linkage. The attribute should be applied
3206 consistently, so that the same entity should not be declared with
3207 different settings of the attribute.
3209 In C++, the visibility attribute applies to types as well as functions
3210 and objects, because in C++ types have linkage. A class must not have
3211 greater visibility than its non-static data member types and bases,
3212 and class members default to the visibility of their class. Also, a
3213 declaration without explicit visibility is limited to the visibility
3216 In C++, you can mark member functions and static member variables of a
3217 class with the visibility attribute. This is useful if you know a
3218 particular method or static member variable should only be used from
3219 one shared object; then you can mark it hidden while the rest of the
3220 class has default visibility. Care must be taken to avoid breaking
3221 the One Definition Rule; for example, it is usually not useful to mark
3222 an inline method as hidden without marking the whole class as hidden.
3224 A C++ namespace declaration can also have the visibility attribute.
3225 This attribute applies only to the particular namespace body, not to
3226 other definitions of the same namespace; it is equivalent to using
3227 @samp{#pragma GCC visibility} before and after the namespace
3228 definition (@pxref{Visibility Pragmas}).
3230 In C++, if a template argument has limited visibility, this
3231 restriction is implicitly propagated to the template instantiation.
3232 Otherwise, template instantiations and specializations default to the
3233 visibility of their template.
3235 If both the template and enclosing class have explicit visibility, the
3236 visibility from the template is used.
3238 @item warn_unused_result
3239 @cindex @code{warn_unused_result} attribute
3240 The @code{warn_unused_result} attribute causes a warning to be emitted
3241 if a caller of the function with this attribute does not use its
3242 return value. This is useful for functions where not checking
3243 the result is either a security problem or always a bug, such as
3247 int fn () __attribute__ ((warn_unused_result));
3250 if (fn () < 0) return -1;
3256 results in warning on line 5.
3259 @cindex @code{weak} attribute
3260 The @code{weak} attribute causes the declaration to be emitted as a weak
3261 symbol rather than a global. This is primarily useful in defining
3262 library functions which can be overridden in user code, though it can
3263 also be used with non-function declarations. Weak symbols are supported
3264 for ELF targets, and also for a.out targets when using the GNU assembler
3268 @itemx weakref ("@var{target}")
3269 @cindex @code{weakref} attribute
3270 The @code{weakref} attribute marks a declaration as a weak reference.
3271 Without arguments, it should be accompanied by an @code{alias} attribute
3272 naming the target symbol. Optionally, the @var{target} may be given as
3273 an argument to @code{weakref} itself. In either case, @code{weakref}
3274 implicitly marks the declaration as @code{weak}. Without a
3275 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3276 @code{weakref} is equivalent to @code{weak}.
3279 static int x() __attribute__ ((weakref ("y")));
3280 /* is equivalent to... */
3281 static int x() __attribute__ ((weak, weakref, alias ("y")));
3283 static int x() __attribute__ ((weakref));
3284 static int x() __attribute__ ((alias ("y")));
3287 A weak reference is an alias that does not by itself require a
3288 definition to be given for the target symbol. If the target symbol is
3289 only referenced through weak references, then the becomes a @code{weak}
3290 undefined symbol. If it is directly referenced, however, then such
3291 strong references prevail, and a definition will be required for the
3292 symbol, not necessarily in the same translation unit.
3294 The effect is equivalent to moving all references to the alias to a
3295 separate translation unit, renaming the alias to the aliased symbol,
3296 declaring it as weak, compiling the two separate translation units and
3297 performing a reloadable link on them.
3299 At present, a declaration to which @code{weakref} is attached can
3300 only be @code{static}.
3304 You can specify multiple attributes in a declaration by separating them
3305 by commas within the double parentheses or by immediately following an
3306 attribute declaration with another attribute declaration.
3308 @cindex @code{#pragma}, reason for not using
3309 @cindex pragma, reason for not using
3310 Some people object to the @code{__attribute__} feature, suggesting that
3311 ISO C's @code{#pragma} should be used instead. At the time
3312 @code{__attribute__} was designed, there were two reasons for not doing
3317 It is impossible to generate @code{#pragma} commands from a macro.
3320 There is no telling what the same @code{#pragma} might mean in another
3324 These two reasons applied to almost any application that might have been
3325 proposed for @code{#pragma}. It was basically a mistake to use
3326 @code{#pragma} for @emph{anything}.
3328 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3329 to be generated from macros. In addition, a @code{#pragma GCC}
3330 namespace is now in use for GCC-specific pragmas. However, it has been
3331 found convenient to use @code{__attribute__} to achieve a natural
3332 attachment of attributes to their corresponding declarations, whereas
3333 @code{#pragma GCC} is of use for constructs that do not naturally form
3334 part of the grammar. @xref{Other Directives,,Miscellaneous
3335 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3337 @node Attribute Syntax
3338 @section Attribute Syntax
3339 @cindex attribute syntax
3341 This section describes the syntax with which @code{__attribute__} may be
3342 used, and the constructs to which attribute specifiers bind, for the C
3343 language. Some details may vary for C++ and Objective-C@. Because of
3344 infelicities in the grammar for attributes, some forms described here
3345 may not be successfully parsed in all cases.
3347 There are some problems with the semantics of attributes in C++. For
3348 example, there are no manglings for attributes, although they may affect
3349 code generation, so problems may arise when attributed types are used in
3350 conjunction with templates or overloading. Similarly, @code{typeid}
3351 does not distinguish between types with different attributes. Support
3352 for attributes in C++ may be restricted in future to attributes on
3353 declarations only, but not on nested declarators.
3355 @xref{Function Attributes}, for details of the semantics of attributes
3356 applying to functions. @xref{Variable Attributes}, for details of the
3357 semantics of attributes applying to variables. @xref{Type Attributes},
3358 for details of the semantics of attributes applying to structure, union
3359 and enumerated types.
3361 An @dfn{attribute specifier} is of the form
3362 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3363 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3364 each attribute is one of the following:
3368 Empty. Empty attributes are ignored.
3371 A word (which may be an identifier such as @code{unused}, or a reserved
3372 word such as @code{const}).
3375 A word, followed by, in parentheses, parameters for the attribute.
3376 These parameters take one of the following forms:
3380 An identifier. For example, @code{mode} attributes use this form.
3383 An identifier followed by a comma and a non-empty comma-separated list
3384 of expressions. For example, @code{format} attributes use this form.
3387 A possibly empty comma-separated list of expressions. For example,
3388 @code{format_arg} attributes use this form with the list being a single
3389 integer constant expression, and @code{alias} attributes use this form
3390 with the list being a single string constant.
3394 An @dfn{attribute specifier list} is a sequence of one or more attribute
3395 specifiers, not separated by any other tokens.
3397 In GNU C, an attribute specifier list may appear after the colon following a
3398 label, other than a @code{case} or @code{default} label. The only
3399 attribute it makes sense to use after a label is @code{unused}. This
3400 feature is intended for code generated by programs which contains labels
3401 that may be unused but which is compiled with @option{-Wall}. It would
3402 not normally be appropriate to use in it human-written code, though it
3403 could be useful in cases where the code that jumps to the label is
3404 contained within an @code{#ifdef} conditional. GNU C++ does not permit
3405 such placement of attribute lists, as it is permissible for a
3406 declaration, which could begin with an attribute list, to be labelled in
3407 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
3408 does not arise there.
3410 An attribute specifier list may appear as part of a @code{struct},
3411 @code{union} or @code{enum} specifier. It may go either immediately
3412 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3413 the closing brace. The former syntax is preferred.
3414 Where attribute specifiers follow the closing brace, they are considered
3415 to relate to the structure, union or enumerated type defined, not to any
3416 enclosing declaration the type specifier appears in, and the type
3417 defined is not complete until after the attribute specifiers.
3418 @c Otherwise, there would be the following problems: a shift/reduce
3419 @c conflict between attributes binding the struct/union/enum and
3420 @c binding to the list of specifiers/qualifiers; and "aligned"
3421 @c attributes could use sizeof for the structure, but the size could be
3422 @c changed later by "packed" attributes.
3424 Otherwise, an attribute specifier appears as part of a declaration,
3425 counting declarations of unnamed parameters and type names, and relates
3426 to that declaration (which may be nested in another declaration, for
3427 example in the case of a parameter declaration), or to a particular declarator
3428 within a declaration. Where an
3429 attribute specifier is applied to a parameter declared as a function or
3430 an array, it should apply to the function or array rather than the
3431 pointer to which the parameter is implicitly converted, but this is not
3432 yet correctly implemented.
3434 Any list of specifiers and qualifiers at the start of a declaration may
3435 contain attribute specifiers, whether or not such a list may in that
3436 context contain storage class specifiers. (Some attributes, however,
3437 are essentially in the nature of storage class specifiers, and only make
3438 sense where storage class specifiers may be used; for example,
3439 @code{section}.) There is one necessary limitation to this syntax: the
3440 first old-style parameter declaration in a function definition cannot
3441 begin with an attribute specifier, because such an attribute applies to
3442 the function instead by syntax described below (which, however, is not
3443 yet implemented in this case). In some other cases, attribute
3444 specifiers are permitted by this grammar but not yet supported by the
3445 compiler. All attribute specifiers in this place relate to the
3446 declaration as a whole. In the obsolescent usage where a type of
3447 @code{int} is implied by the absence of type specifiers, such a list of
3448 specifiers and qualifiers may be an attribute specifier list with no
3449 other specifiers or qualifiers.
3451 At present, the first parameter in a function prototype must have some
3452 type specifier which is not an attribute specifier; this resolves an
3453 ambiguity in the interpretation of @code{void f(int
3454 (__attribute__((foo)) x))}, but is subject to change. At present, if
3455 the parentheses of a function declarator contain only attributes then
3456 those attributes are ignored, rather than yielding an error or warning
3457 or implying a single parameter of type int, but this is subject to
3460 An attribute specifier list may appear immediately before a declarator
3461 (other than the first) in a comma-separated list of declarators in a
3462 declaration of more than one identifier using a single list of
3463 specifiers and qualifiers. Such attribute specifiers apply
3464 only to the identifier before whose declarator they appear. For
3468 __attribute__((noreturn)) void d0 (void),
3469 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3474 the @code{noreturn} attribute applies to all the functions
3475 declared; the @code{format} attribute only applies to @code{d1}.
3477 An attribute specifier list may appear immediately before the comma,
3478 @code{=} or semicolon terminating the declaration of an identifier other
3479 than a function definition. Such attribute specifiers apply
3480 to the declared object or function. Where an
3481 assembler name for an object or function is specified (@pxref{Asm
3482 Labels}), the attribute must follow the @code{asm}
3485 An attribute specifier list may, in future, be permitted to appear after
3486 the declarator in a function definition (before any old-style parameter
3487 declarations or the function body).
3489 Attribute specifiers may be mixed with type qualifiers appearing inside
3490 the @code{[]} of a parameter array declarator, in the C99 construct by
3491 which such qualifiers are applied to the pointer to which the array is
3492 implicitly converted. Such attribute specifiers apply to the pointer,
3493 not to the array, but at present this is not implemented and they are
3496 An attribute specifier list may appear at the start of a nested
3497 declarator. At present, there are some limitations in this usage: the
3498 attributes correctly apply to the declarator, but for most individual
3499 attributes the semantics this implies are not implemented.
3500 When attribute specifiers follow the @code{*} of a pointer
3501 declarator, they may be mixed with any type qualifiers present.
3502 The following describes the formal semantics of this syntax. It will make the
3503 most sense if you are familiar with the formal specification of
3504 declarators in the ISO C standard.
3506 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3507 D1}, where @code{T} contains declaration specifiers that specify a type
3508 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3509 contains an identifier @var{ident}. The type specified for @var{ident}
3510 for derived declarators whose type does not include an attribute
3511 specifier is as in the ISO C standard.
3513 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3514 and the declaration @code{T D} specifies the type
3515 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3516 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3517 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3519 If @code{D1} has the form @code{*
3520 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3521 declaration @code{T D} specifies the type
3522 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3523 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3524 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3530 void (__attribute__((noreturn)) ****f) (void);
3534 specifies the type ``pointer to pointer to pointer to pointer to
3535 non-returning function returning @code{void}''. As another example,
3538 char *__attribute__((aligned(8))) *f;
3542 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3543 Note again that this does not work with most attributes; for example,
3544 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3545 is not yet supported.
3547 For compatibility with existing code written for compiler versions that
3548 did not implement attributes on nested declarators, some laxity is
3549 allowed in the placing of attributes. If an attribute that only applies
3550 to types is applied to a declaration, it will be treated as applying to
3551 the type of that declaration. If an attribute that only applies to
3552 declarations is applied to the type of a declaration, it will be treated
3553 as applying to that declaration; and, for compatibility with code
3554 placing the attributes immediately before the identifier declared, such
3555 an attribute applied to a function return type will be treated as
3556 applying to the function type, and such an attribute applied to an array
3557 element type will be treated as applying to the array type. If an
3558 attribute that only applies to function types is applied to a
3559 pointer-to-function type, it will be treated as applying to the pointer
3560 target type; if such an attribute is applied to a function return type
3561 that is not a pointer-to-function type, it will be treated as applying
3562 to the function type.
3564 @node Function Prototypes
3565 @section Prototypes and Old-Style Function Definitions
3566 @cindex function prototype declarations
3567 @cindex old-style function definitions
3568 @cindex promotion of formal parameters
3570 GNU C extends ISO C to allow a function prototype to override a later
3571 old-style non-prototype definition. Consider the following example:
3574 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3581 /* @r{Prototype function declaration.} */
3582 int isroot P((uid_t));
3584 /* @r{Old-style function definition.} */
3586 isroot (x) /* @r{??? lossage here ???} */
3593 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3594 not allow this example, because subword arguments in old-style
3595 non-prototype definitions are promoted. Therefore in this example the
3596 function definition's argument is really an @code{int}, which does not
3597 match the prototype argument type of @code{short}.
3599 This restriction of ISO C makes it hard to write code that is portable
3600 to traditional C compilers, because the programmer does not know
3601 whether the @code{uid_t} type is @code{short}, @code{int}, or
3602 @code{long}. Therefore, in cases like these GNU C allows a prototype
3603 to override a later old-style definition. More precisely, in GNU C, a
3604 function prototype argument type overrides the argument type specified
3605 by a later old-style definition if the former type is the same as the
3606 latter type before promotion. Thus in GNU C the above example is
3607 equivalent to the following:
3620 GNU C++ does not support old-style function definitions, so this
3621 extension is irrelevant.
3624 @section C++ Style Comments
3626 @cindex C++ comments
3627 @cindex comments, C++ style
3629 In GNU C, you may use C++ style comments, which start with @samp{//} and
3630 continue until the end of the line. Many other C implementations allow
3631 such comments, and they are included in the 1999 C standard. However,
3632 C++ style comments are not recognized if you specify an @option{-std}
3633 option specifying a version of ISO C before C99, or @option{-ansi}
3634 (equivalent to @option{-std=c89}).
3637 @section Dollar Signs in Identifier Names
3639 @cindex dollar signs in identifier names
3640 @cindex identifier names, dollar signs in
3642 In GNU C, you may normally use dollar signs in identifier names.
3643 This is because many traditional C implementations allow such identifiers.
3644 However, dollar signs in identifiers are not supported on a few target
3645 machines, typically because the target assembler does not allow them.
3647 @node Character Escapes
3648 @section The Character @key{ESC} in Constants
3650 You can use the sequence @samp{\e} in a string or character constant to
3651 stand for the ASCII character @key{ESC}.
3654 @section Inquiring on Alignment of Types or Variables
3656 @cindex type alignment
3657 @cindex variable alignment
3659 The keyword @code{__alignof__} allows you to inquire about how an object
3660 is aligned, or the minimum alignment usually required by a type. Its
3661 syntax is just like @code{sizeof}.
3663 For example, if the target machine requires a @code{double} value to be
3664 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3665 This is true on many RISC machines. On more traditional machine
3666 designs, @code{__alignof__ (double)} is 4 or even 2.
3668 Some machines never actually require alignment; they allow reference to any
3669 data type even at an odd address. For these machines, @code{__alignof__}
3670 reports the smallest alignment that GCC will give the data type, usually as
3671 mandated by the target ABI.
3673 If the operand of @code{__alignof__} is an lvalue rather than a type,
3674 its value is the required alignment for its type, taking into account
3675 any minimum alignment specified with GCC's @code{__attribute__}
3676 extension (@pxref{Variable Attributes}). For example, after this
3680 struct foo @{ int x; char y; @} foo1;
3684 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3685 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3687 It is an error to ask for the alignment of an incomplete type.
3689 @node Variable Attributes
3690 @section Specifying Attributes of Variables
3691 @cindex attribute of variables
3692 @cindex variable attributes
3694 The keyword @code{__attribute__} allows you to specify special
3695 attributes of variables or structure fields. This keyword is followed
3696 by an attribute specification inside double parentheses. Some
3697 attributes are currently defined generically for variables.
3698 Other attributes are defined for variables on particular target
3699 systems. Other attributes are available for functions
3700 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3701 Other front ends might define more attributes
3702 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3704 You may also specify attributes with @samp{__} preceding and following
3705 each keyword. This allows you to use them in header files without
3706 being concerned about a possible macro of the same name. For example,
3707 you may use @code{__aligned__} instead of @code{aligned}.
3709 @xref{Attribute Syntax}, for details of the exact syntax for using
3713 @cindex @code{aligned} attribute
3714 @item aligned (@var{alignment})
3715 This attribute specifies a minimum alignment for the variable or
3716 structure field, measured in bytes. For example, the declaration:
3719 int x __attribute__ ((aligned (16))) = 0;
3723 causes the compiler to allocate the global variable @code{x} on a
3724 16-byte boundary. On a 68040, this could be used in conjunction with
3725 an @code{asm} expression to access the @code{move16} instruction which
3726 requires 16-byte aligned operands.
3728 You can also specify the alignment of structure fields. For example, to
3729 create a double-word aligned @code{int} pair, you could write:
3732 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3736 This is an alternative to creating a union with a @code{double} member
3737 that forces the union to be double-word aligned.
3739 As in the preceding examples, you can explicitly specify the alignment
3740 (in bytes) that you wish the compiler to use for a given variable or
3741 structure field. Alternatively, you can leave out the alignment factor
3742 and just ask the compiler to align a variable or field to the
3743 default alignment for the target architecture you are compiling for.
3744 The default alignment is sufficient for all scalar types, but may not be
3745 enough for all vector types on a target which supports vector operations.
3746 The default alignment is fixed for a particular target ABI.
3748 Gcc also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
3749 which is the largest alignment ever used for any data type on the
3750 target machine you are compiling for. For example, you could write:
3753 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
3756 The compiler automatically sets the alignment for the declared
3757 variable or field to @code{__BIGGEST_ALIGNMENT__}. Doing this can
3758 often make copy operations more efficient, because the compiler can
3759 use whatever instructions copy the biggest chunks of memory when
3760 performing copies to or from the variables or fields that you have
3761 aligned this way. Note that the value of @code{__BIGGEST_ALIGNMENT__}
3762 may change depending on command line options.
3764 When used on a struct, or struct member, the @code{aligned} attribute can
3765 only increase the alignment; in order to decrease it, the @code{packed}
3766 attribute must be specified as well. When used as part of a typedef, the
3767 @code{aligned} attribute can both increase and decrease alignment, and
3768 specifying the @code{packed} attribute will generate a warning.
3770 Note that the effectiveness of @code{aligned} attributes may be limited
3771 by inherent limitations in your linker. On many systems, the linker is
3772 only able to arrange for variables to be aligned up to a certain maximum
3773 alignment. (For some linkers, the maximum supported alignment may
3774 be very very small.) If your linker is only able to align variables
3775 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3776 in an @code{__attribute__} will still only provide you with 8 byte
3777 alignment. See your linker documentation for further information.
3779 The @code{aligned} attribute can also be used for functions
3780 (@pxref{Function Attributes}.)
3782 @item cleanup (@var{cleanup_function})
3783 @cindex @code{cleanup} attribute
3784 The @code{cleanup} attribute runs a function when the variable goes
3785 out of scope. This attribute can only be applied to auto function
3786 scope variables; it may not be applied to parameters or variables
3787 with static storage duration. The function must take one parameter,
3788 a pointer to a type compatible with the variable. The return value
3789 of the function (if any) is ignored.
3791 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3792 will be run during the stack unwinding that happens during the
3793 processing of the exception. Note that the @code{cleanup} attribute
3794 does not allow the exception to be caught, only to perform an action.
3795 It is undefined what happens if @var{cleanup_function} does not
3800 @cindex @code{common} attribute
3801 @cindex @code{nocommon} attribute
3804 The @code{common} attribute requests GCC to place a variable in
3805 ``common'' storage. The @code{nocommon} attribute requests the
3806 opposite---to allocate space for it directly.
3808 These attributes override the default chosen by the
3809 @option{-fno-common} and @option{-fcommon} flags respectively.
3812 @cindex @code{deprecated} attribute
3813 The @code{deprecated} attribute results in a warning if the variable
3814 is used anywhere in the source file. This is useful when identifying
3815 variables that are expected to be removed in a future version of a
3816 program. The warning also includes the location of the declaration
3817 of the deprecated variable, to enable users to easily find further
3818 information about why the variable is deprecated, or what they should
3819 do instead. Note that the warning only occurs for uses:
3822 extern int old_var __attribute__ ((deprecated));
3824 int new_fn () @{ return old_var; @}
3827 results in a warning on line 3 but not line 2.
3829 The @code{deprecated} attribute can also be used for functions and
3830 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3832 @item mode (@var{mode})
3833 @cindex @code{mode} attribute
3834 This attribute specifies the data type for the declaration---whichever
3835 type corresponds to the mode @var{mode}. This in effect lets you
3836 request an integer or floating point type according to its width.
3838 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3839 indicate the mode corresponding to a one-byte integer, @samp{word} or
3840 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3841 or @samp{__pointer__} for the mode used to represent pointers.
3844 @cindex @code{packed} attribute
3845 The @code{packed} attribute specifies that a variable or structure field
3846 should have the smallest possible alignment---one byte for a variable,
3847 and one bit for a field, unless you specify a larger value with the
3848 @code{aligned} attribute.
3850 Here is a structure in which the field @code{x} is packed, so that it
3851 immediately follows @code{a}:
3857 int x[2] __attribute__ ((packed));
3861 @emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
3862 @code{packed} attribute on bit-fields of type @code{char}. This has
3863 been fixed in GCC 4.4 but the change can lead to differences in the
3864 structure layout. See the documentation of
3865 @option{-Wpacked-bitfield-compat} for more information.
3867 @item section ("@var{section-name}")
3868 @cindex @code{section} variable attribute
3869 Normally, the compiler places the objects it generates in sections like
3870 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3871 or you need certain particular variables to appear in special sections,
3872 for example to map to special hardware. The @code{section}
3873 attribute specifies that a variable (or function) lives in a particular
3874 section. For example, this small program uses several specific section names:
3877 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3878 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3879 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3880 int init_data __attribute__ ((section ("INITDATA")));
3884 /* @r{Initialize stack pointer} */
3885 init_sp (stack + sizeof (stack));
3887 /* @r{Initialize initialized data} */
3888 memcpy (&init_data, &data, &edata - &data);
3890 /* @r{Turn on the serial ports} */
3897 Use the @code{section} attribute with
3898 @emph{global} variables and not @emph{local} variables,
3899 as shown in the example.
3901 You may use the @code{section} attribute with initialized or
3902 uninitialized global variables but the linker requires
3903 each object be defined once, with the exception that uninitialized
3904 variables tentatively go in the @code{common} (or @code{bss}) section
3905 and can be multiply ``defined''. Using the @code{section} attribute
3906 will change what section the variable goes into and may cause the
3907 linker to issue an error if an uninitialized variable has multiple
3908 definitions. You can force a variable to be initialized with the
3909 @option{-fno-common} flag or the @code{nocommon} attribute.
3911 Some file formats do not support arbitrary sections so the @code{section}
3912 attribute is not available on all platforms.
3913 If you need to map the entire contents of a module to a particular
3914 section, consider using the facilities of the linker instead.
3917 @cindex @code{shared} variable attribute
3918 On Microsoft Windows, in addition to putting variable definitions in a named
3919 section, the section can also be shared among all running copies of an
3920 executable or DLL@. For example, this small program defines shared data
3921 by putting it in a named section @code{shared} and marking the section
3925 int foo __attribute__((section ("shared"), shared)) = 0;
3930 /* @r{Read and write foo. All running
3931 copies see the same value.} */
3937 You may only use the @code{shared} attribute along with @code{section}
3938 attribute with a fully initialized global definition because of the way
3939 linkers work. See @code{section} attribute for more information.
3941 The @code{shared} attribute is only available on Microsoft Windows@.
3943 @item tls_model ("@var{tls_model}")
3944 @cindex @code{tls_model} attribute
3945 The @code{tls_model} attribute sets thread-local storage model
3946 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3947 overriding @option{-ftls-model=} command line switch on a per-variable
3949 The @var{tls_model} argument should be one of @code{global-dynamic},
3950 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3952 Not all targets support this attribute.
3955 This attribute, attached to a variable, means that the variable is meant
3956 to be possibly unused. GCC will not produce a warning for this
3960 This attribute, attached to a variable, means that the variable must be
3961 emitted even if it appears that the variable is not referenced.
3963 @item vector_size (@var{bytes})
3964 This attribute specifies the vector size for the variable, measured in
3965 bytes. For example, the declaration:
3968 int foo __attribute__ ((vector_size (16)));
3972 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3973 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3974 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3976 This attribute is only applicable to integral and float scalars,
3977 although arrays, pointers, and function return values are allowed in
3978 conjunction with this construct.
3980 Aggregates with this attribute are invalid, even if they are of the same
3981 size as a corresponding scalar. For example, the declaration:
3984 struct S @{ int a; @};
3985 struct S __attribute__ ((vector_size (16))) foo;
3989 is invalid even if the size of the structure is the same as the size of
3993 The @code{selectany} attribute causes an initialized global variable to
3994 have link-once semantics. When multiple definitions of the variable are
3995 encountered by the linker, the first is selected and the remainder are
3996 discarded. Following usage by the Microsoft compiler, the linker is told
3997 @emph{not} to warn about size or content differences of the multiple
4000 Although the primary usage of this attribute is for POD types, the
4001 attribute can also be applied to global C++ objects that are initialized
4002 by a constructor. In this case, the static initialization and destruction
4003 code for the object is emitted in each translation defining the object,
4004 but the calls to the constructor and destructor are protected by a
4005 link-once guard variable.
4007 The @code{selectany} attribute is only available on Microsoft Windows
4008 targets. You can use @code{__declspec (selectany)} as a synonym for
4009 @code{__attribute__ ((selectany))} for compatibility with other
4013 The @code{weak} attribute is described in @ref{Function Attributes}.
4016 The @code{dllimport} attribute is described in @ref{Function Attributes}.
4019 The @code{dllexport} attribute is described in @ref{Function Attributes}.
4023 @subsection Blackfin Variable Attributes
4025 Three attributes are currently defined for the Blackfin.
4031 @cindex @code{l1_data} variable attribute
4032 @cindex @code{l1_data_A} variable attribute
4033 @cindex @code{l1_data_B} variable attribute
4034 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
4035 Variables with @code{l1_data} attribute will be put into the specific section
4036 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4037 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4038 attribute will be put into the specific section named @code{.l1.data.B}.
4041 @subsection M32R/D Variable Attributes
4043 One attribute is currently defined for the M32R/D@.
4046 @item model (@var{model-name})
4047 @cindex variable addressability on the M32R/D
4048 Use this attribute on the M32R/D to set the addressability of an object.
4049 The identifier @var{model-name} is one of @code{small}, @code{medium},
4050 or @code{large}, representing each of the code models.
4052 Small model objects live in the lower 16MB of memory (so that their
4053 addresses can be loaded with the @code{ld24} instruction).
4055 Medium and large model objects may live anywhere in the 32-bit address space
4056 (the compiler will generate @code{seth/add3} instructions to load their
4060 @anchor{i386 Variable Attributes}
4061 @subsection i386 Variable Attributes
4063 Two attributes are currently defined for i386 configurations:
4064 @code{ms_struct} and @code{gcc_struct}
4069 @cindex @code{ms_struct} attribute
4070 @cindex @code{gcc_struct} attribute
4072 If @code{packed} is used on a structure, or if bit-fields are used
4073 it may be that the Microsoft ABI packs them differently
4074 than GCC would normally pack them. Particularly when moving packed
4075 data between functions compiled with GCC and the native Microsoft compiler
4076 (either via function call or as data in a file), it may be necessary to access
4079 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4080 compilers to match the native Microsoft compiler.
4082 The Microsoft structure layout algorithm is fairly simple with the exception
4083 of the bitfield packing:
4085 The padding and alignment of members of structures and whether a bit field
4086 can straddle a storage-unit boundary
4089 @item Structure members are stored sequentially in the order in which they are
4090 declared: the first member has the lowest memory address and the last member
4093 @item Every data object has an alignment-requirement. The alignment-requirement
4094 for all data except structures, unions, and arrays is either the size of the
4095 object or the current packing size (specified with either the aligned attribute
4096 or the pack pragma), whichever is less. For structures, unions, and arrays,
4097 the alignment-requirement is the largest alignment-requirement of its members.
4098 Every object is allocated an offset so that:
4100 offset % alignment-requirement == 0
4102 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4103 unit if the integral types are the same size and if the next bit field fits
4104 into the current allocation unit without crossing the boundary imposed by the
4105 common alignment requirements of the bit fields.
4108 Handling of zero-length bitfields:
4110 MSVC interprets zero-length bitfields in the following ways:
4113 @item If a zero-length bitfield is inserted between two bitfields that would
4114 normally be coalesced, the bitfields will not be coalesced.
4121 unsigned long bf_1 : 12;
4123 unsigned long bf_2 : 12;
4127 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4128 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4130 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4131 alignment of the zero-length bitfield is greater than the member that follows it,
4132 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4152 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4153 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4154 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4157 Taking this into account, it is important to note the following:
4160 @item If a zero-length bitfield follows a normal bitfield, the type of the
4161 zero-length bitfield may affect the alignment of the structure as whole. For
4162 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4163 normal bitfield, and is of type short.
4165 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4166 still affect the alignment of the structure:
4176 Here, @code{t4} will take up 4 bytes.
4179 @item Zero-length bitfields following non-bitfield members are ignored:
4190 Here, @code{t5} will take up 2 bytes.
4194 @subsection PowerPC Variable Attributes
4196 Three attributes currently are defined for PowerPC configurations:
4197 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4199 For full documentation of the struct attributes please see the
4200 documentation in @ref{i386 Variable Attributes}.
4202 For documentation of @code{altivec} attribute please see the
4203 documentation in @ref{PowerPC Type Attributes}.
4205 @subsection SPU Variable Attributes
4207 The SPU supports the @code{spu_vector} attribute for variables. For
4208 documentation of this attribute please see the documentation in
4209 @ref{SPU Type Attributes}.
4211 @subsection Xstormy16 Variable Attributes
4213 One attribute is currently defined for xstormy16 configurations:
4218 @cindex @code{below100} attribute
4220 If a variable has the @code{below100} attribute (@code{BELOW100} is
4221 allowed also), GCC will place the variable in the first 0x100 bytes of
4222 memory and use special opcodes to access it. Such variables will be
4223 placed in either the @code{.bss_below100} section or the
4224 @code{.data_below100} section.
4228 @subsection AVR Variable Attributes
4232 @cindex @code{progmem} variable attribute
4233 The @code{progmem} attribute is used on the AVR to place data in the Program
4234 Memory address space. The AVR is a Harvard Architecture processor and data
4235 normally resides in the Data Memory address space.
4238 @node Type Attributes
4239 @section Specifying Attributes of Types
4240 @cindex attribute of types
4241 @cindex type attributes
4243 The keyword @code{__attribute__} allows you to specify special
4244 attributes of @code{struct} and @code{union} types when you define
4245 such types. This keyword is followed by an attribute specification
4246 inside double parentheses. Seven attributes are currently defined for
4247 types: @code{aligned}, @code{packed}, @code{transparent_union},
4248 @code{unused}, @code{deprecated}, @code{visibility}, and
4249 @code{may_alias}. Other attributes are defined for functions
4250 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4253 You may also specify any one of these attributes with @samp{__}
4254 preceding and following its keyword. This allows you to use these
4255 attributes in header files without being concerned about a possible
4256 macro of the same name. For example, you may use @code{__aligned__}
4257 instead of @code{aligned}.
4259 You may specify type attributes in an enum, struct or union type
4260 declaration or definition, or for other types in a @code{typedef}
4263 For an enum, struct or union type, you may specify attributes either
4264 between the enum, struct or union tag and the name of the type, or
4265 just past the closing curly brace of the @emph{definition}. The
4266 former syntax is preferred.
4268 @xref{Attribute Syntax}, for details of the exact syntax for using
4272 @cindex @code{aligned} attribute
4273 @item aligned (@var{alignment})
4274 This attribute specifies a minimum alignment (in bytes) for variables
4275 of the specified type. For example, the declarations:
4278 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4279 typedef int more_aligned_int __attribute__ ((aligned (8)));
4283 force the compiler to insure (as far as it can) that each variable whose
4284 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4285 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4286 variables of type @code{struct S} aligned to 8-byte boundaries allows
4287 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4288 store) instructions when copying one variable of type @code{struct S} to
4289 another, thus improving run-time efficiency.
4291 Note that the alignment of any given @code{struct} or @code{union} type
4292 is required by the ISO C standard to be at least a perfect multiple of
4293 the lowest common multiple of the alignments of all of the members of
4294 the @code{struct} or @code{union} in question. This means that you @emph{can}
4295 effectively adjust the alignment of a @code{struct} or @code{union}
4296 type by attaching an @code{aligned} attribute to any one of the members
4297 of such a type, but the notation illustrated in the example above is a
4298 more obvious, intuitive, and readable way to request the compiler to
4299 adjust the alignment of an entire @code{struct} or @code{union} type.
4301 As in the preceding example, you can explicitly specify the alignment
4302 (in bytes) that you wish the compiler to use for a given @code{struct}
4303 or @code{union} type. Alternatively, you can leave out the alignment factor
4304 and just ask the compiler to align a type to the maximum
4305 useful alignment for the target machine you are compiling for. For
4306 example, you could write:
4309 struct S @{ short f[3]; @} __attribute__ ((aligned));
4312 Whenever you leave out the alignment factor in an @code{aligned}
4313 attribute specification, the compiler automatically sets the alignment
4314 for the type to the largest alignment which is ever used for any data
4315 type on the target machine you are compiling for. Doing this can often
4316 make copy operations more efficient, because the compiler can use
4317 whatever instructions copy the biggest chunks of memory when performing
4318 copies to or from the variables which have types that you have aligned
4321 In the example above, if the size of each @code{short} is 2 bytes, then
4322 the size of the entire @code{struct S} type is 6 bytes. The smallest
4323 power of two which is greater than or equal to that is 8, so the
4324 compiler sets the alignment for the entire @code{struct S} type to 8
4327 Note that although you can ask the compiler to select a time-efficient
4328 alignment for a given type and then declare only individual stand-alone
4329 objects of that type, the compiler's ability to select a time-efficient
4330 alignment is primarily useful only when you plan to create arrays of
4331 variables having the relevant (efficiently aligned) type. If you
4332 declare or use arrays of variables of an efficiently-aligned type, then
4333 it is likely that your program will also be doing pointer arithmetic (or
4334 subscripting, which amounts to the same thing) on pointers to the
4335 relevant type, and the code that the compiler generates for these
4336 pointer arithmetic operations will often be more efficient for
4337 efficiently-aligned types than for other types.
4339 The @code{aligned} attribute can only increase the alignment; but you
4340 can decrease it by specifying @code{packed} as well. See below.
4342 Note that the effectiveness of @code{aligned} attributes may be limited
4343 by inherent limitations in your linker. On many systems, the linker is
4344 only able to arrange for variables to be aligned up to a certain maximum
4345 alignment. (For some linkers, the maximum supported alignment may
4346 be very very small.) If your linker is only able to align variables
4347 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4348 in an @code{__attribute__} will still only provide you with 8 byte
4349 alignment. See your linker documentation for further information.
4352 This attribute, attached to @code{struct} or @code{union} type
4353 definition, specifies that each member (other than zero-width bitfields)
4354 of the structure or union is placed to minimize the memory required. When
4355 attached to an @code{enum} definition, it indicates that the smallest
4356 integral type should be used.
4358 @opindex fshort-enums
4359 Specifying this attribute for @code{struct} and @code{union} types is
4360 equivalent to specifying the @code{packed} attribute on each of the
4361 structure or union members. Specifying the @option{-fshort-enums}
4362 flag on the line is equivalent to specifying the @code{packed}
4363 attribute on all @code{enum} definitions.
4365 In the following example @code{struct my_packed_struct}'s members are
4366 packed closely together, but the internal layout of its @code{s} member
4367 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4371 struct my_unpacked_struct
4377 struct __attribute__ ((__packed__)) my_packed_struct
4381 struct my_unpacked_struct s;
4385 You may only specify this attribute on the definition of a @code{enum},
4386 @code{struct} or @code{union}, not on a @code{typedef} which does not
4387 also define the enumerated type, structure or union.
4389 @item transparent_union
4390 This attribute, attached to a @code{union} type definition, indicates
4391 that any function parameter having that union type causes calls to that
4392 function to be treated in a special way.
4394 First, the argument corresponding to a transparent union type can be of
4395 any type in the union; no cast is required. Also, if the union contains
4396 a pointer type, the corresponding argument can be a null pointer
4397 constant or a void pointer expression; and if the union contains a void
4398 pointer type, the corresponding argument can be any pointer expression.
4399 If the union member type is a pointer, qualifiers like @code{const} on
4400 the referenced type must be respected, just as with normal pointer
4403 Second, the argument is passed to the function using the calling
4404 conventions of the first member of the transparent union, not the calling
4405 conventions of the union itself. All members of the union must have the
4406 same machine representation; this is necessary for this argument passing
4409 Transparent unions are designed for library functions that have multiple
4410 interfaces for compatibility reasons. For example, suppose the
4411 @code{wait} function must accept either a value of type @code{int *} to
4412 comply with Posix, or a value of type @code{union wait *} to comply with
4413 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4414 @code{wait} would accept both kinds of arguments, but it would also
4415 accept any other pointer type and this would make argument type checking
4416 less useful. Instead, @code{<sys/wait.h>} might define the interface
4420 typedef union __attribute__ ((__transparent_union__))
4424 @} wait_status_ptr_t;
4426 pid_t wait (wait_status_ptr_t);
4429 This interface allows either @code{int *} or @code{union wait *}
4430 arguments to be passed, using the @code{int *} calling convention.
4431 The program can call @code{wait} with arguments of either type:
4434 int w1 () @{ int w; return wait (&w); @}
4435 int w2 () @{ union wait w; return wait (&w); @}
4438 With this interface, @code{wait}'s implementation might look like this:
4441 pid_t wait (wait_status_ptr_t p)
4443 return waitpid (-1, p.__ip, 0);
4448 When attached to a type (including a @code{union} or a @code{struct}),
4449 this attribute means that variables of that type are meant to appear
4450 possibly unused. GCC will not produce a warning for any variables of
4451 that type, even if the variable appears to do nothing. This is often
4452 the case with lock or thread classes, which are usually defined and then
4453 not referenced, but contain constructors and destructors that have
4454 nontrivial bookkeeping functions.
4457 The @code{deprecated} attribute results in a warning if the type
4458 is used anywhere in the source file. This is useful when identifying
4459 types that are expected to be removed in a future version of a program.
4460 If possible, the warning also includes the location of the declaration
4461 of the deprecated type, to enable users to easily find further
4462 information about why the type is deprecated, or what they should do
4463 instead. Note that the warnings only occur for uses and then only
4464 if the type is being applied to an identifier that itself is not being
4465 declared as deprecated.
4468 typedef int T1 __attribute__ ((deprecated));
4472 typedef T1 T3 __attribute__ ((deprecated));
4473 T3 z __attribute__ ((deprecated));
4476 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4477 warning is issued for line 4 because T2 is not explicitly
4478 deprecated. Line 5 has no warning because T3 is explicitly
4479 deprecated. Similarly for line 6.
4481 The @code{deprecated} attribute can also be used for functions and
4482 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4485 Accesses through pointers to types with this attribute are not subject
4486 to type-based alias analysis, but are instead assumed to be able to alias
4487 any other type of objects. In the context of 6.5/7 an lvalue expression
4488 dereferencing such a pointer is treated like having a character type.
4489 See @option{-fstrict-aliasing} for more information on aliasing issues.
4490 This extension exists to support some vector APIs, in which pointers to
4491 one vector type are permitted to alias pointers to a different vector type.
4493 Note that an object of a type with this attribute does not have any
4499 typedef short __attribute__((__may_alias__)) short_a;
4505 short_a *b = (short_a *) &a;
4509 if (a == 0x12345678)
4516 If you replaced @code{short_a} with @code{short} in the variable
4517 declaration, the above program would abort when compiled with
4518 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4519 above in recent GCC versions.
4522 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4523 applied to class, struct, union and enum types. Unlike other type
4524 attributes, the attribute must appear between the initial keyword and
4525 the name of the type; it cannot appear after the body of the type.
4527 Note that the type visibility is applied to vague linkage entities
4528 associated with the class (vtable, typeinfo node, etc.). In
4529 particular, if a class is thrown as an exception in one shared object
4530 and caught in another, the class must have default visibility.
4531 Otherwise the two shared objects will be unable to use the same
4532 typeinfo node and exception handling will break.
4536 @subsection ARM Type Attributes
4538 On those ARM targets that support @code{dllimport} (such as Symbian
4539 OS), you can use the @code{notshared} attribute to indicate that the
4540 virtual table and other similar data for a class should not be
4541 exported from a DLL@. For example:
4544 class __declspec(notshared) C @{
4546 __declspec(dllimport) C();
4550 __declspec(dllexport)
4554 In this code, @code{C::C} is exported from the current DLL, but the
4555 virtual table for @code{C} is not exported. (You can use
4556 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4557 most Symbian OS code uses @code{__declspec}.)
4559 @anchor{i386 Type Attributes}
4560 @subsection i386 Type Attributes
4562 Two attributes are currently defined for i386 configurations:
4563 @code{ms_struct} and @code{gcc_struct}.
4569 @cindex @code{ms_struct}
4570 @cindex @code{gcc_struct}
4572 If @code{packed} is used on a structure, or if bit-fields are used
4573 it may be that the Microsoft ABI packs them differently
4574 than GCC would normally pack them. Particularly when moving packed
4575 data between functions compiled with GCC and the native Microsoft compiler
4576 (either via function call or as data in a file), it may be necessary to access
4579 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4580 compilers to match the native Microsoft compiler.
4583 To specify multiple attributes, separate them by commas within the
4584 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4587 @anchor{PowerPC Type Attributes}
4588 @subsection PowerPC Type Attributes
4590 Three attributes currently are defined for PowerPC configurations:
4591 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4593 For full documentation of the @code{ms_struct} and @code{gcc_struct}
4594 attributes please see the documentation in @ref{i386 Type Attributes}.
4596 The @code{altivec} attribute allows one to declare AltiVec vector data
4597 types supported by the AltiVec Programming Interface Manual. The
4598 attribute requires an argument to specify one of three vector types:
4599 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4600 and @code{bool__} (always followed by unsigned).
4603 __attribute__((altivec(vector__)))
4604 __attribute__((altivec(pixel__))) unsigned short
4605 __attribute__((altivec(bool__))) unsigned
4608 These attributes mainly are intended to support the @code{__vector},
4609 @code{__pixel}, and @code{__bool} AltiVec keywords.
4611 @anchor{SPU Type Attributes}
4612 @subsection SPU Type Attributes
4614 The SPU supports the @code{spu_vector} attribute for types. This attribute
4615 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
4616 Language Extensions Specification. It is intended to support the
4617 @code{__vector} keyword.
4621 @section An Inline Function is As Fast As a Macro
4622 @cindex inline functions
4623 @cindex integrating function code
4625 @cindex macros, inline alternative
4627 By declaring a function inline, you can direct GCC to make
4628 calls to that function faster. One way GCC can achieve this is to
4629 integrate that function's code into the code for its callers. This
4630 makes execution faster by eliminating the function-call overhead; in
4631 addition, if any of the actual argument values are constant, their
4632 known values may permit simplifications at compile time so that not
4633 all of the inline function's code needs to be included. The effect on
4634 code size is less predictable; object code may be larger or smaller
4635 with function inlining, depending on the particular case. You can
4636 also direct GCC to try to integrate all ``simple enough'' functions
4637 into their callers with the option @option{-finline-functions}.
4639 GCC implements three different semantics of declaring a function
4640 inline. One is available with @option{-std=gnu89} or
4641 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
4642 on all inline declarations, another when @option{-std=c99} or
4643 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
4644 is used when compiling C++.
4646 To declare a function inline, use the @code{inline} keyword in its
4647 declaration, like this:
4657 If you are writing a header file to be included in ISO C89 programs, write
4658 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4660 The three types of inlining behave similarly in two important cases:
4661 when the @code{inline} keyword is used on a @code{static} function,
4662 like the example above, and when a function is first declared without
4663 using the @code{inline} keyword and then is defined with
4664 @code{inline}, like this:
4667 extern int inc (int *a);
4675 In both of these common cases, the program behaves the same as if you
4676 had not used the @code{inline} keyword, except for its speed.
4678 @cindex inline functions, omission of
4679 @opindex fkeep-inline-functions
4680 When a function is both inline and @code{static}, if all calls to the
4681 function are integrated into the caller, and the function's address is
4682 never used, then the function's own assembler code is never referenced.
4683 In this case, GCC does not actually output assembler code for the
4684 function, unless you specify the option @option{-fkeep-inline-functions}.
4685 Some calls cannot be integrated for various reasons (in particular,
4686 calls that precede the function's definition cannot be integrated, and
4687 neither can recursive calls within the definition). If there is a
4688 nonintegrated call, then the function is compiled to assembler code as
4689 usual. The function must also be compiled as usual if the program
4690 refers to its address, because that can't be inlined.
4693 Note that certain usages in a function definition can make it unsuitable
4694 for inline substitution. Among these usages are: use of varargs, use of
4695 alloca, use of variable sized data types (@pxref{Variable Length}),
4696 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4697 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4698 will warn when a function marked @code{inline} could not be substituted,
4699 and will give the reason for the failure.
4701 @cindex automatic @code{inline} for C++ member fns
4702 @cindex @code{inline} automatic for C++ member fns
4703 @cindex member fns, automatically @code{inline}
4704 @cindex C++ member fns, automatically @code{inline}
4705 @opindex fno-default-inline
4706 As required by ISO C++, GCC considers member functions defined within
4707 the body of a class to be marked inline even if they are
4708 not explicitly declared with the @code{inline} keyword. You can
4709 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4710 Options,,Options Controlling C++ Dialect}.
4712 GCC does not inline any functions when not optimizing unless you specify
4713 the @samp{always_inline} attribute for the function, like this:
4716 /* @r{Prototype.} */
4717 inline void foo (const char) __attribute__((always_inline));
4720 The remainder of this section is specific to GNU C89 inlining.
4722 @cindex non-static inline function
4723 When an inline function is not @code{static}, then the compiler must assume
4724 that there may be calls from other source files; since a global symbol can
4725 be defined only once in any program, the function must not be defined in
4726 the other source files, so the calls therein cannot be integrated.
4727 Therefore, a non-@code{static} inline function is always compiled on its
4728 own in the usual fashion.
4730 If you specify both @code{inline} and @code{extern} in the function
4731 definition, then the definition is used only for inlining. In no case
4732 is the function compiled on its own, not even if you refer to its
4733 address explicitly. Such an address becomes an external reference, as
4734 if you had only declared the function, and had not defined it.
4736 This combination of @code{inline} and @code{extern} has almost the
4737 effect of a macro. The way to use it is to put a function definition in
4738 a header file with these keywords, and put another copy of the
4739 definition (lacking @code{inline} and @code{extern}) in a library file.
4740 The definition in the header file will cause most calls to the function
4741 to be inlined. If any uses of the function remain, they will refer to
4742 the single copy in the library.
4745 @section Assembler Instructions with C Expression Operands
4746 @cindex extended @code{asm}
4747 @cindex @code{asm} expressions
4748 @cindex assembler instructions
4751 In an assembler instruction using @code{asm}, you can specify the
4752 operands of the instruction using C expressions. This means you need not
4753 guess which registers or memory locations will contain the data you want
4756 You must specify an assembler instruction template much like what
4757 appears in a machine description, plus an operand constraint string for
4760 For example, here is how to use the 68881's @code{fsinx} instruction:
4763 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4767 Here @code{angle} is the C expression for the input operand while
4768 @code{result} is that of the output operand. Each has @samp{"f"} as its
4769 operand constraint, saying that a floating point register is required.
4770 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4771 output operands' constraints must use @samp{=}. The constraints use the
4772 same language used in the machine description (@pxref{Constraints}).
4774 Each operand is described by an operand-constraint string followed by
4775 the C expression in parentheses. A colon separates the assembler
4776 template from the first output operand and another separates the last
4777 output operand from the first input, if any. Commas separate the
4778 operands within each group. The total number of operands is currently
4779 limited to 30; this limitation may be lifted in some future version of
4782 If there are no output operands but there are input operands, you must
4783 place two consecutive colons surrounding the place where the output
4786 As of GCC version 3.1, it is also possible to specify input and output
4787 operands using symbolic names which can be referenced within the
4788 assembler code. These names are specified inside square brackets
4789 preceding the constraint string, and can be referenced inside the
4790 assembler code using @code{%[@var{name}]} instead of a percentage sign
4791 followed by the operand number. Using named operands the above example
4795 asm ("fsinx %[angle],%[output]"
4796 : [output] "=f" (result)
4797 : [angle] "f" (angle));
4801 Note that the symbolic operand names have no relation whatsoever to
4802 other C identifiers. You may use any name you like, even those of
4803 existing C symbols, but you must ensure that no two operands within the same
4804 assembler construct use the same symbolic name.
4806 Output operand expressions must be lvalues; the compiler can check this.
4807 The input operands need not be lvalues. The compiler cannot check
4808 whether the operands have data types that are reasonable for the
4809 instruction being executed. It does not parse the assembler instruction
4810 template and does not know what it means or even whether it is valid
4811 assembler input. The extended @code{asm} feature is most often used for
4812 machine instructions the compiler itself does not know exist. If
4813 the output expression cannot be directly addressed (for example, it is a
4814 bit-field), your constraint must allow a register. In that case, GCC
4815 will use the register as the output of the @code{asm}, and then store
4816 that register into the output.
4818 The ordinary output operands must be write-only; GCC will assume that
4819 the values in these operands before the instruction are dead and need
4820 not be generated. Extended asm supports input-output or read-write
4821 operands. Use the constraint character @samp{+} to indicate such an
4822 operand and list it with the output operands. You should only use
4823 read-write operands when the constraints for the operand (or the
4824 operand in which only some of the bits are to be changed) allow a
4827 You may, as an alternative, logically split its function into two
4828 separate operands, one input operand and one write-only output
4829 operand. The connection between them is expressed by constraints
4830 which say they need to be in the same location when the instruction
4831 executes. You can use the same C expression for both operands, or
4832 different expressions. For example, here we write the (fictitious)
4833 @samp{combine} instruction with @code{bar} as its read-only source
4834 operand and @code{foo} as its read-write destination:
4837 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4841 The constraint @samp{"0"} for operand 1 says that it must occupy the
4842 same location as operand 0. A number in constraint is allowed only in
4843 an input operand and it must refer to an output operand.
4845 Only a number in the constraint can guarantee that one operand will be in
4846 the same place as another. The mere fact that @code{foo} is the value
4847 of both operands is not enough to guarantee that they will be in the
4848 same place in the generated assembler code. The following would not
4852 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4855 Various optimizations or reloading could cause operands 0 and 1 to be in
4856 different registers; GCC knows no reason not to do so. For example, the
4857 compiler might find a copy of the value of @code{foo} in one register and
4858 use it for operand 1, but generate the output operand 0 in a different
4859 register (copying it afterward to @code{foo}'s own address). Of course,
4860 since the register for operand 1 is not even mentioned in the assembler
4861 code, the result will not work, but GCC can't tell that.
4863 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4864 the operand number for a matching constraint. For example:
4867 asm ("cmoveq %1,%2,%[result]"
4868 : [result] "=r"(result)
4869 : "r" (test), "r"(new), "[result]"(old));
4872 Sometimes you need to make an @code{asm} operand be a specific register,
4873 but there's no matching constraint letter for that register @emph{by
4874 itself}. To force the operand into that register, use a local variable
4875 for the operand and specify the register in the variable declaration.
4876 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4877 register constraint letter that matches the register:
4880 register int *p1 asm ("r0") = @dots{};
4881 register int *p2 asm ("r1") = @dots{};
4882 register int *result asm ("r0");
4883 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4886 @anchor{Example of asm with clobbered asm reg}
4887 In the above example, beware that a register that is call-clobbered by
4888 the target ABI will be overwritten by any function call in the
4889 assignment, including library calls for arithmetic operators.
4890 Also a register may be clobbered when generating some operations,
4891 like variable shift, memory copy or memory move on x86.
4892 Assuming it is a call-clobbered register, this may happen to @code{r0}
4893 above by the assignment to @code{p2}. If you have to use such a
4894 register, use temporary variables for expressions between the register
4899 register int *p1 asm ("r0") = @dots{};
4900 register int *p2 asm ("r1") = t1;
4901 register int *result asm ("r0");
4902 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4905 Some instructions clobber specific hard registers. To describe this,
4906 write a third colon after the input operands, followed by the names of
4907 the clobbered hard registers (given as strings). Here is a realistic
4908 example for the VAX:
4911 asm volatile ("movc3 %0,%1,%2"
4912 : /* @r{no outputs} */
4913 : "g" (from), "g" (to), "g" (count)
4914 : "r0", "r1", "r2", "r3", "r4", "r5");
4917 You may not write a clobber description in a way that overlaps with an
4918 input or output operand. For example, you may not have an operand
4919 describing a register class with one member if you mention that register
4920 in the clobber list. Variables declared to live in specific registers
4921 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4922 have no part mentioned in the clobber description.
4923 There is no way for you to specify that an input
4924 operand is modified without also specifying it as an output
4925 operand. Note that if all the output operands you specify are for this
4926 purpose (and hence unused), you will then also need to specify
4927 @code{volatile} for the @code{asm} construct, as described below, to
4928 prevent GCC from deleting the @code{asm} statement as unused.
4930 If you refer to a particular hardware register from the assembler code,
4931 you will probably have to list the register after the third colon to
4932 tell the compiler the register's value is modified. In some assemblers,
4933 the register names begin with @samp{%}; to produce one @samp{%} in the
4934 assembler code, you must write @samp{%%} in the input.
4936 If your assembler instruction can alter the condition code register, add
4937 @samp{cc} to the list of clobbered registers. GCC on some machines
4938 represents the condition codes as a specific hardware register;
4939 @samp{cc} serves to name this register. On other machines, the
4940 condition code is handled differently, and specifying @samp{cc} has no
4941 effect. But it is valid no matter what the machine.
4943 If your assembler instructions access memory in an unpredictable
4944 fashion, add @samp{memory} to the list of clobbered registers. This
4945 will cause GCC to not keep memory values cached in registers across the
4946 assembler instruction and not optimize stores or loads to that memory.
4947 You will also want to add the @code{volatile} keyword if the memory
4948 affected is not listed in the inputs or outputs of the @code{asm}, as
4949 the @samp{memory} clobber does not count as a side-effect of the
4950 @code{asm}. If you know how large the accessed memory is, you can add
4951 it as input or output but if this is not known, you should add
4952 @samp{memory}. As an example, if you access ten bytes of a string, you
4953 can use a memory input like:
4956 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4959 Note that in the following example the memory input is necessary,
4960 otherwise GCC might optimize the store to @code{x} away:
4967 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4968 "=&d" (r) : "a" (y), "m" (*y));
4973 You can put multiple assembler instructions together in a single
4974 @code{asm} template, separated by the characters normally used in assembly
4975 code for the system. A combination that works in most places is a newline
4976 to break the line, plus a tab character to move to the instruction field
4977 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4978 assembler allows semicolons as a line-breaking character. Note that some
4979 assembler dialects use semicolons to start a comment.
4980 The input operands are guaranteed not to use any of the clobbered
4981 registers, and neither will the output operands' addresses, so you can
4982 read and write the clobbered registers as many times as you like. Here
4983 is an example of multiple instructions in a template; it assumes the
4984 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4987 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4989 : "g" (from), "g" (to)
4993 Unless an output operand has the @samp{&} constraint modifier, GCC
4994 may allocate it in the same register as an unrelated input operand, on
4995 the assumption the inputs are consumed before the outputs are produced.
4996 This assumption may be false if the assembler code actually consists of
4997 more than one instruction. In such a case, use @samp{&} for each output
4998 operand that may not overlap an input. @xref{Modifiers}.
5000 If you want to test the condition code produced by an assembler
5001 instruction, you must include a branch and a label in the @code{asm}
5002 construct, as follows:
5005 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
5011 This assumes your assembler supports local labels, as the GNU assembler
5012 and most Unix assemblers do.
5014 Speaking of labels, jumps from one @code{asm} to another are not
5015 supported. The compiler's optimizers do not know about these jumps, and
5016 therefore they cannot take account of them when deciding how to
5019 @cindex macros containing @code{asm}
5020 Usually the most convenient way to use these @code{asm} instructions is to
5021 encapsulate them in macros that look like functions. For example,
5025 (@{ double __value, __arg = (x); \
5026 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
5031 Here the variable @code{__arg} is used to make sure that the instruction
5032 operates on a proper @code{double} value, and to accept only those
5033 arguments @code{x} which can convert automatically to a @code{double}.
5035 Another way to make sure the instruction operates on the correct data
5036 type is to use a cast in the @code{asm}. This is different from using a
5037 variable @code{__arg} in that it converts more different types. For
5038 example, if the desired type were @code{int}, casting the argument to
5039 @code{int} would accept a pointer with no complaint, while assigning the
5040 argument to an @code{int} variable named @code{__arg} would warn about
5041 using a pointer unless the caller explicitly casts it.
5043 If an @code{asm} has output operands, GCC assumes for optimization
5044 purposes the instruction has no side effects except to change the output
5045 operands. This does not mean instructions with a side effect cannot be
5046 used, but you must be careful, because the compiler may eliminate them
5047 if the output operands aren't used, or move them out of loops, or
5048 replace two with one if they constitute a common subexpression. Also,
5049 if your instruction does have a side effect on a variable that otherwise
5050 appears not to change, the old value of the variable may be reused later
5051 if it happens to be found in a register.
5053 You can prevent an @code{asm} instruction from being deleted
5054 by writing the keyword @code{volatile} after
5055 the @code{asm}. For example:
5058 #define get_and_set_priority(new) \
5060 asm volatile ("get_and_set_priority %0, %1" \
5061 : "=g" (__old) : "g" (new)); \
5066 The @code{volatile} keyword indicates that the instruction has
5067 important side-effects. GCC will not delete a volatile @code{asm} if
5068 it is reachable. (The instruction can still be deleted if GCC can
5069 prove that control-flow will never reach the location of the
5070 instruction.) Note that even a volatile @code{asm} instruction
5071 can be moved relative to other code, including across jump
5072 instructions. For example, on many targets there is a system
5073 register which can be set to control the rounding mode of
5074 floating point operations. You might try
5075 setting it with a volatile @code{asm}, like this PowerPC example:
5078 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5083 This will not work reliably, as the compiler may move the addition back
5084 before the volatile @code{asm}. To make it work you need to add an
5085 artificial dependency to the @code{asm} referencing a variable in the code
5086 you don't want moved, for example:
5089 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5093 Similarly, you can't expect a
5094 sequence of volatile @code{asm} instructions to remain perfectly
5095 consecutive. If you want consecutive output, use a single @code{asm}.
5096 Also, GCC will perform some optimizations across a volatile @code{asm}
5097 instruction; GCC does not ``forget everything'' when it encounters
5098 a volatile @code{asm} instruction the way some other compilers do.
5100 An @code{asm} instruction without any output operands will be treated
5101 identically to a volatile @code{asm} instruction.
5103 It is a natural idea to look for a way to give access to the condition
5104 code left by the assembler instruction. However, when we attempted to
5105 implement this, we found no way to make it work reliably. The problem
5106 is that output operands might need reloading, which would result in
5107 additional following ``store'' instructions. On most machines, these
5108 instructions would alter the condition code before there was time to
5109 test it. This problem doesn't arise for ordinary ``test'' and
5110 ``compare'' instructions because they don't have any output operands.
5112 For reasons similar to those described above, it is not possible to give
5113 an assembler instruction access to the condition code left by previous
5116 If you are writing a header file that should be includable in ISO C
5117 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5120 @subsection Size of an @code{asm}
5122 Some targets require that GCC track the size of each instruction used in
5123 order to generate correct code. Because the final length of an
5124 @code{asm} is only known by the assembler, GCC must make an estimate as
5125 to how big it will be. The estimate is formed by counting the number of
5126 statements in the pattern of the @code{asm} and multiplying that by the
5127 length of the longest instruction on that processor. Statements in the
5128 @code{asm} are identified by newline characters and whatever statement
5129 separator characters are supported by the assembler; on most processors
5130 this is the `@code{;}' character.
5132 Normally, GCC's estimate is perfectly adequate to ensure that correct
5133 code is generated, but it is possible to confuse the compiler if you use
5134 pseudo instructions or assembler macros that expand into multiple real
5135 instructions or if you use assembler directives that expand to more
5136 space in the object file than would be needed for a single instruction.
5137 If this happens then the assembler will produce a diagnostic saying that
5138 a label is unreachable.
5140 @subsection i386 floating point asm operands
5142 There are several rules on the usage of stack-like regs in
5143 asm_operands insns. These rules apply only to the operands that are
5148 Given a set of input regs that die in an asm_operands, it is
5149 necessary to know which are implicitly popped by the asm, and
5150 which must be explicitly popped by gcc.
5152 An input reg that is implicitly popped by the asm must be
5153 explicitly clobbered, unless it is constrained to match an
5157 For any input reg that is implicitly popped by an asm, it is
5158 necessary to know how to adjust the stack to compensate for the pop.
5159 If any non-popped input is closer to the top of the reg-stack than
5160 the implicitly popped reg, it would not be possible to know what the
5161 stack looked like---it's not clear how the rest of the stack ``slides
5164 All implicitly popped input regs must be closer to the top of
5165 the reg-stack than any input that is not implicitly popped.
5167 It is possible that if an input dies in an insn, reload might
5168 use the input reg for an output reload. Consider this example:
5171 asm ("foo" : "=t" (a) : "f" (b));
5174 This asm says that input B is not popped by the asm, and that
5175 the asm pushes a result onto the reg-stack, i.e., the stack is one
5176 deeper after the asm than it was before. But, it is possible that
5177 reload will think that it can use the same reg for both the input and
5178 the output, if input B dies in this insn.
5180 If any input operand uses the @code{f} constraint, all output reg
5181 constraints must use the @code{&} earlyclobber.
5183 The asm above would be written as
5186 asm ("foo" : "=&t" (a) : "f" (b));
5190 Some operands need to be in particular places on the stack. All
5191 output operands fall in this category---there is no other way to
5192 know which regs the outputs appear in unless the user indicates
5193 this in the constraints.
5195 Output operands must specifically indicate which reg an output
5196 appears in after an asm. @code{=f} is not allowed: the operand
5197 constraints must select a class with a single reg.
5200 Output operands may not be ``inserted'' between existing stack regs.
5201 Since no 387 opcode uses a read/write operand, all output operands
5202 are dead before the asm_operands, and are pushed by the asm_operands.
5203 It makes no sense to push anywhere but the top of the reg-stack.
5205 Output operands must start at the top of the reg-stack: output
5206 operands may not ``skip'' a reg.
5209 Some asm statements may need extra stack space for internal
5210 calculations. This can be guaranteed by clobbering stack registers
5211 unrelated to the inputs and outputs.
5215 Here are a couple of reasonable asms to want to write. This asm
5216 takes one input, which is internally popped, and produces two outputs.
5219 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
5222 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
5223 and replaces them with one output. The user must code the @code{st(1)}
5224 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
5227 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
5233 @section Controlling Names Used in Assembler Code
5234 @cindex assembler names for identifiers
5235 @cindex names used in assembler code
5236 @cindex identifiers, names in assembler code
5238 You can specify the name to be used in the assembler code for a C
5239 function or variable by writing the @code{asm} (or @code{__asm__})
5240 keyword after the declarator as follows:
5243 int foo asm ("myfoo") = 2;
5247 This specifies that the name to be used for the variable @code{foo} in
5248 the assembler code should be @samp{myfoo} rather than the usual
5251 On systems where an underscore is normally prepended to the name of a C
5252 function or variable, this feature allows you to define names for the
5253 linker that do not start with an underscore.
5255 It does not make sense to use this feature with a non-static local
5256 variable since such variables do not have assembler names. If you are
5257 trying to put the variable in a particular register, see @ref{Explicit
5258 Reg Vars}. GCC presently accepts such code with a warning, but will
5259 probably be changed to issue an error, rather than a warning, in the
5262 You cannot use @code{asm} in this way in a function @emph{definition}; but
5263 you can get the same effect by writing a declaration for the function
5264 before its definition and putting @code{asm} there, like this:
5267 extern func () asm ("FUNC");
5274 It is up to you to make sure that the assembler names you choose do not
5275 conflict with any other assembler symbols. Also, you must not use a
5276 register name; that would produce completely invalid assembler code. GCC
5277 does not as yet have the ability to store static variables in registers.
5278 Perhaps that will be added.
5280 @node Explicit Reg Vars
5281 @section Variables in Specified Registers
5282 @cindex explicit register variables
5283 @cindex variables in specified registers
5284 @cindex specified registers
5285 @cindex registers, global allocation
5287 GNU C allows you to put a few global variables into specified hardware
5288 registers. You can also specify the register in which an ordinary
5289 register variable should be allocated.
5293 Global register variables reserve registers throughout the program.
5294 This may be useful in programs such as programming language
5295 interpreters which have a couple of global variables that are accessed
5299 Local register variables in specific registers do not reserve the
5300 registers, except at the point where they are used as input or output
5301 operands in an @code{asm} statement and the @code{asm} statement itself is
5302 not deleted. The compiler's data flow analysis is capable of determining
5303 where the specified registers contain live values, and where they are
5304 available for other uses. Stores into local register variables may be deleted
5305 when they appear to be dead according to dataflow analysis. References
5306 to local register variables may be deleted or moved or simplified.
5308 These local variables are sometimes convenient for use with the extended
5309 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
5310 output of the assembler instruction directly into a particular register.
5311 (This will work provided the register you specify fits the constraints
5312 specified for that operand in the @code{asm}.)
5320 @node Global Reg Vars
5321 @subsection Defining Global Register Variables
5322 @cindex global register variables
5323 @cindex registers, global variables in
5325 You can define a global register variable in GNU C like this:
5328 register int *foo asm ("a5");
5332 Here @code{a5} is the name of the register which should be used. Choose a
5333 register which is normally saved and restored by function calls on your
5334 machine, so that library routines will not clobber it.
5336 Naturally the register name is cpu-dependent, so you would need to
5337 conditionalize your program according to cpu type. The register
5338 @code{a5} would be a good choice on a 68000 for a variable of pointer
5339 type. On machines with register windows, be sure to choose a ``global''
5340 register that is not affected magically by the function call mechanism.
5342 In addition, operating systems on one type of cpu may differ in how they
5343 name the registers; then you would need additional conditionals. For
5344 example, some 68000 operating systems call this register @code{%a5}.
5346 Eventually there may be a way of asking the compiler to choose a register
5347 automatically, but first we need to figure out how it should choose and
5348 how to enable you to guide the choice. No solution is evident.
5350 Defining a global register variable in a certain register reserves that
5351 register entirely for this use, at least within the current compilation.
5352 The register will not be allocated for any other purpose in the functions
5353 in the current compilation. The register will not be saved and restored by
5354 these functions. Stores into this register are never deleted even if they
5355 would appear to be dead, but references may be deleted or moved or
5358 It is not safe to access the global register variables from signal
5359 handlers, or from more than one thread of control, because the system
5360 library routines may temporarily use the register for other things (unless
5361 you recompile them specially for the task at hand).
5363 @cindex @code{qsort}, and global register variables
5364 It is not safe for one function that uses a global register variable to
5365 call another such function @code{foo} by way of a third function
5366 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
5367 different source file in which the variable wasn't declared). This is
5368 because @code{lose} might save the register and put some other value there.
5369 For example, you can't expect a global register variable to be available in
5370 the comparison-function that you pass to @code{qsort}, since @code{qsort}
5371 might have put something else in that register. (If you are prepared to
5372 recompile @code{qsort} with the same global register variable, you can
5373 solve this problem.)
5375 If you want to recompile @code{qsort} or other source files which do not
5376 actually use your global register variable, so that they will not use that
5377 register for any other purpose, then it suffices to specify the compiler
5378 option @option{-ffixed-@var{reg}}. You need not actually add a global
5379 register declaration to their source code.
5381 A function which can alter the value of a global register variable cannot
5382 safely be called from a function compiled without this variable, because it
5383 could clobber the value the caller expects to find there on return.
5384 Therefore, the function which is the entry point into the part of the
5385 program that uses the global register variable must explicitly save and
5386 restore the value which belongs to its caller.
5388 @cindex register variable after @code{longjmp}
5389 @cindex global register after @code{longjmp}
5390 @cindex value after @code{longjmp}
5393 On most machines, @code{longjmp} will restore to each global register
5394 variable the value it had at the time of the @code{setjmp}. On some
5395 machines, however, @code{longjmp} will not change the value of global
5396 register variables. To be portable, the function that called @code{setjmp}
5397 should make other arrangements to save the values of the global register
5398 variables, and to restore them in a @code{longjmp}. This way, the same
5399 thing will happen regardless of what @code{longjmp} does.
5401 All global register variable declarations must precede all function
5402 definitions. If such a declaration could appear after function
5403 definitions, the declaration would be too late to prevent the register from
5404 being used for other purposes in the preceding functions.
5406 Global register variables may not have initial values, because an
5407 executable file has no means to supply initial contents for a register.
5409 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5410 registers, but certain library functions, such as @code{getwd}, as well
5411 as the subroutines for division and remainder, modify g3 and g4. g1 and
5412 g2 are local temporaries.
5414 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5415 Of course, it will not do to use more than a few of those.
5417 @node Local Reg Vars
5418 @subsection Specifying Registers for Local Variables
5419 @cindex local variables, specifying registers
5420 @cindex specifying registers for local variables
5421 @cindex registers for local variables
5423 You can define a local register variable with a specified register
5427 register int *foo asm ("a5");
5431 Here @code{a5} is the name of the register which should be used. Note
5432 that this is the same syntax used for defining global register
5433 variables, but for a local variable it would appear within a function.
5435 Naturally the register name is cpu-dependent, but this is not a
5436 problem, since specific registers are most often useful with explicit
5437 assembler instructions (@pxref{Extended Asm}). Both of these things
5438 generally require that you conditionalize your program according to
5441 In addition, operating systems on one type of cpu may differ in how they
5442 name the registers; then you would need additional conditionals. For
5443 example, some 68000 operating systems call this register @code{%a5}.
5445 Defining such a register variable does not reserve the register; it
5446 remains available for other uses in places where flow control determines
5447 the variable's value is not live.
5449 This option does not guarantee that GCC will generate code that has
5450 this variable in the register you specify at all times. You may not
5451 code an explicit reference to this register in the @emph{assembler
5452 instruction template} part of an @code{asm} statement and assume it will
5453 always refer to this variable. However, using the variable as an
5454 @code{asm} @emph{operand} guarantees that the specified register is used
5457 Stores into local register variables may be deleted when they appear to be dead
5458 according to dataflow analysis. References to local register variables may
5459 be deleted or moved or simplified.
5461 As for global register variables, it's recommended that you choose a
5462 register which is normally saved and restored by function calls on
5463 your machine, so that library routines will not clobber it. A common
5464 pitfall is to initialize multiple call-clobbered registers with
5465 arbitrary expressions, where a function call or library call for an
5466 arithmetic operator will overwrite a register value from a previous
5467 assignment, for example @code{r0} below:
5469 register int *p1 asm ("r0") = @dots{};
5470 register int *p2 asm ("r1") = @dots{};
5472 In those cases, a solution is to use a temporary variable for
5473 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
5475 @node Alternate Keywords
5476 @section Alternate Keywords
5477 @cindex alternate keywords
5478 @cindex keywords, alternate
5480 @option{-ansi} and the various @option{-std} options disable certain
5481 keywords. This causes trouble when you want to use GNU C extensions, or
5482 a general-purpose header file that should be usable by all programs,
5483 including ISO C programs. The keywords @code{asm}, @code{typeof} and
5484 @code{inline} are not available in programs compiled with
5485 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
5486 program compiled with @option{-std=c99}). The ISO C99 keyword
5487 @code{restrict} is only available when @option{-std=gnu99} (which will
5488 eventually be the default) or @option{-std=c99} (or the equivalent
5489 @option{-std=iso9899:1999}) is used.
5491 The way to solve these problems is to put @samp{__} at the beginning and
5492 end of each problematical keyword. For example, use @code{__asm__}
5493 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
5495 Other C compilers won't accept these alternative keywords; if you want to
5496 compile with another compiler, you can define the alternate keywords as
5497 macros to replace them with the customary keywords. It looks like this:
5505 @findex __extension__
5507 @option{-pedantic} and other options cause warnings for many GNU C extensions.
5509 prevent such warnings within one expression by writing
5510 @code{__extension__} before the expression. @code{__extension__} has no
5511 effect aside from this.
5513 @node Incomplete Enums
5514 @section Incomplete @code{enum} Types
5516 You can define an @code{enum} tag without specifying its possible values.
5517 This results in an incomplete type, much like what you get if you write
5518 @code{struct foo} without describing the elements. A later declaration
5519 which does specify the possible values completes the type.
5521 You can't allocate variables or storage using the type while it is
5522 incomplete. However, you can work with pointers to that type.
5524 This extension may not be very useful, but it makes the handling of
5525 @code{enum} more consistent with the way @code{struct} and @code{union}
5528 This extension is not supported by GNU C++.
5530 @node Function Names
5531 @section Function Names as Strings
5532 @cindex @code{__func__} identifier
5533 @cindex @code{__FUNCTION__} identifier
5534 @cindex @code{__PRETTY_FUNCTION__} identifier
5536 GCC provides three magic variables which hold the name of the current
5537 function, as a string. The first of these is @code{__func__}, which
5538 is part of the C99 standard:
5540 The identifier @code{__func__} is implicitly declared by the translator
5541 as if, immediately following the opening brace of each function
5542 definition, the declaration
5545 static const char __func__[] = "function-name";
5549 appeared, where function-name is the name of the lexically-enclosing
5550 function. This name is the unadorned name of the function.
5552 @code{__FUNCTION__} is another name for @code{__func__}. Older
5553 versions of GCC recognize only this name. However, it is not
5554 standardized. For maximum portability, we recommend you use
5555 @code{__func__}, but provide a fallback definition with the
5559 #if __STDC_VERSION__ < 199901L
5561 # define __func__ __FUNCTION__
5563 # define __func__ "<unknown>"
5568 In C, @code{__PRETTY_FUNCTION__} is yet another name for
5569 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
5570 the type signature of the function as well as its bare name. For
5571 example, this program:
5575 extern int printf (char *, ...);
5582 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
5583 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
5601 __PRETTY_FUNCTION__ = void a::sub(int)
5604 These identifiers are not preprocessor macros. In GCC 3.3 and
5605 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
5606 were treated as string literals; they could be used to initialize
5607 @code{char} arrays, and they could be concatenated with other string
5608 literals. GCC 3.4 and later treat them as variables, like
5609 @code{__func__}. In C++, @code{__FUNCTION__} and
5610 @code{__PRETTY_FUNCTION__} have always been variables.
5612 @node Return Address
5613 @section Getting the Return or Frame Address of a Function
5615 These functions may be used to get information about the callers of a
5618 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
5619 This function returns the return address of the current function, or of
5620 one of its callers. The @var{level} argument is number of frames to
5621 scan up the call stack. A value of @code{0} yields the return address
5622 of the current function, a value of @code{1} yields the return address
5623 of the caller of the current function, and so forth. When inlining
5624 the expected behavior is that the function will return the address of
5625 the function that will be returned to. To work around this behavior use
5626 the @code{noinline} function attribute.
5628 The @var{level} argument must be a constant integer.
5630 On some machines it may be impossible to determine the return address of
5631 any function other than the current one; in such cases, or when the top
5632 of the stack has been reached, this function will return @code{0} or a
5633 random value. In addition, @code{__builtin_frame_address} may be used
5634 to determine if the top of the stack has been reached.
5636 This function should only be used with a nonzero argument for debugging
5640 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
5641 This function is similar to @code{__builtin_return_address}, but it
5642 returns the address of the function frame rather than the return address
5643 of the function. Calling @code{__builtin_frame_address} with a value of
5644 @code{0} yields the frame address of the current function, a value of
5645 @code{1} yields the frame address of the caller of the current function,
5648 The frame is the area on the stack which holds local variables and saved
5649 registers. The frame address is normally the address of the first word
5650 pushed on to the stack by the function. However, the exact definition
5651 depends upon the processor and the calling convention. If the processor
5652 has a dedicated frame pointer register, and the function has a frame,
5653 then @code{__builtin_frame_address} will return the value of the frame
5656 On some machines it may be impossible to determine the frame address of
5657 any function other than the current one; in such cases, or when the top
5658 of the stack has been reached, this function will return @code{0} if
5659 the first frame pointer is properly initialized by the startup code.
5661 This function should only be used with a nonzero argument for debugging
5665 @node Vector Extensions
5666 @section Using vector instructions through built-in functions
5668 On some targets, the instruction set contains SIMD vector instructions that
5669 operate on multiple values contained in one large register at the same time.
5670 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5673 The first step in using these extensions is to provide the necessary data
5674 types. This should be done using an appropriate @code{typedef}:
5677 typedef int v4si __attribute__ ((vector_size (16)));
5680 The @code{int} type specifies the base type, while the attribute specifies
5681 the vector size for the variable, measured in bytes. For example, the
5682 declaration above causes the compiler to set the mode for the @code{v4si}
5683 type to be 16 bytes wide and divided into @code{int} sized units. For
5684 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5685 corresponding mode of @code{foo} will be @acronym{V4SI}.
5687 The @code{vector_size} attribute is only applicable to integral and
5688 float scalars, although arrays, pointers, and function return values
5689 are allowed in conjunction with this construct.
5691 All the basic integer types can be used as base types, both as signed
5692 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5693 @code{long long}. In addition, @code{float} and @code{double} can be
5694 used to build floating-point vector types.
5696 Specifying a combination that is not valid for the current architecture
5697 will cause GCC to synthesize the instructions using a narrower mode.
5698 For example, if you specify a variable of type @code{V4SI} and your
5699 architecture does not allow for this specific SIMD type, GCC will
5700 produce code that uses 4 @code{SIs}.
5702 The types defined in this manner can be used with a subset of normal C
5703 operations. Currently, GCC will allow using the following operators
5704 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5706 The operations behave like C++ @code{valarrays}. Addition is defined as
5707 the addition of the corresponding elements of the operands. For
5708 example, in the code below, each of the 4 elements in @var{a} will be
5709 added to the corresponding 4 elements in @var{b} and the resulting
5710 vector will be stored in @var{c}.
5713 typedef int v4si __attribute__ ((vector_size (16)));
5720 Subtraction, multiplication, division, and the logical operations
5721 operate in a similar manner. Likewise, the result of using the unary
5722 minus or complement operators on a vector type is a vector whose
5723 elements are the negative or complemented values of the corresponding
5724 elements in the operand.
5726 You can declare variables and use them in function calls and returns, as
5727 well as in assignments and some casts. You can specify a vector type as
5728 a return type for a function. Vector types can also be used as function
5729 arguments. It is possible to cast from one vector type to another,
5730 provided they are of the same size (in fact, you can also cast vectors
5731 to and from other datatypes of the same size).
5733 You cannot operate between vectors of different lengths or different
5734 signedness without a cast.
5736 A port that supports hardware vector operations, usually provides a set
5737 of built-in functions that can be used to operate on vectors. For
5738 example, a function to add two vectors and multiply the result by a
5739 third could look like this:
5742 v4si f (v4si a, v4si b, v4si c)
5744 v4si tmp = __builtin_addv4si (a, b);
5745 return __builtin_mulv4si (tmp, c);
5752 @findex __builtin_offsetof
5754 GCC implements for both C and C++ a syntactic extension to implement
5755 the @code{offsetof} macro.
5759 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5761 offsetof_member_designator:
5763 | offsetof_member_designator "." @code{identifier}
5764 | offsetof_member_designator "[" @code{expr} "]"
5767 This extension is sufficient such that
5770 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5773 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5774 may be dependent. In either case, @var{member} may consist of a single
5775 identifier, or a sequence of member accesses and array references.
5777 @node Atomic Builtins
5778 @section Built-in functions for atomic memory access
5780 The following builtins are intended to be compatible with those described
5781 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5782 section 7.4. As such, they depart from the normal GCC practice of using
5783 the ``__builtin_'' prefix, and further that they are overloaded such that
5784 they work on multiple types.
5786 The definition given in the Intel documentation allows only for the use of
5787 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5788 counterparts. GCC will allow any integral scalar or pointer type that is
5789 1, 2, 4 or 8 bytes in length.
5791 Not all operations are supported by all target processors. If a particular
5792 operation cannot be implemented on the target processor, a warning will be
5793 generated and a call an external function will be generated. The external
5794 function will carry the same name as the builtin, with an additional suffix
5795 @samp{_@var{n}} where @var{n} is the size of the data type.
5797 @c ??? Should we have a mechanism to suppress this warning? This is almost
5798 @c useful for implementing the operation under the control of an external
5801 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5802 no memory operand will be moved across the operation, either forward or
5803 backward. Further, instructions will be issued as necessary to prevent the
5804 processor from speculating loads across the operation and from queuing stores
5805 after the operation.
5807 All of the routines are described in the Intel documentation to take
5808 ``an optional list of variables protected by the memory barrier''. It's
5809 not clear what is meant by that; it could mean that @emph{only} the
5810 following variables are protected, or it could mean that these variables
5811 should in addition be protected. At present GCC ignores this list and
5812 protects all variables which are globally accessible. If in the future
5813 we make some use of this list, an empty list will continue to mean all
5814 globally accessible variables.
5817 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5818 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5819 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5820 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5821 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5822 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5823 @findex __sync_fetch_and_add
5824 @findex __sync_fetch_and_sub
5825 @findex __sync_fetch_and_or
5826 @findex __sync_fetch_and_and
5827 @findex __sync_fetch_and_xor
5828 @findex __sync_fetch_and_nand
5829 These builtins perform the operation suggested by the name, and
5830 returns the value that had previously been in memory. That is,
5833 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5834 @{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @} // nand
5837 @emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
5838 builtin as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.
5840 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5841 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5842 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5843 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5844 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5845 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5846 @findex __sync_add_and_fetch
5847 @findex __sync_sub_and_fetch
5848 @findex __sync_or_and_fetch
5849 @findex __sync_and_and_fetch
5850 @findex __sync_xor_and_fetch
5851 @findex __sync_nand_and_fetch
5852 These builtins perform the operation suggested by the name, and
5853 return the new value. That is,
5856 @{ *ptr @var{op}= value; return *ptr; @}
5857 @{ *ptr = ~(*ptr & value); return *ptr; @} // nand
5860 @emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
5861 builtin as @code{*ptr = ~(*ptr & value)} instead of
5862 @code{*ptr = ~*ptr & value}.
5864 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5865 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5866 @findex __sync_bool_compare_and_swap
5867 @findex __sync_val_compare_and_swap
5868 These builtins perform an atomic compare and swap. That is, if the current
5869 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5872 The ``bool'' version returns true if the comparison is successful and
5873 @var{newval} was written. The ``val'' version returns the contents
5874 of @code{*@var{ptr}} before the operation.
5876 @item __sync_synchronize (...)
5877 @findex __sync_synchronize
5878 This builtin issues a full memory barrier.
5880 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5881 @findex __sync_lock_test_and_set
5882 This builtin, as described by Intel, is not a traditional test-and-set
5883 operation, but rather an atomic exchange operation. It writes @var{value}
5884 into @code{*@var{ptr}}, and returns the previous contents of
5887 Many targets have only minimal support for such locks, and do not support
5888 a full exchange operation. In this case, a target may support reduced
5889 functionality here by which the @emph{only} valid value to store is the
5890 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5891 is implementation defined.
5893 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5894 This means that references after the builtin cannot move to (or be
5895 speculated to) before the builtin, but previous memory stores may not
5896 be globally visible yet, and previous memory loads may not yet be
5899 @item void __sync_lock_release (@var{type} *ptr, ...)
5900 @findex __sync_lock_release
5901 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5902 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5904 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5905 This means that all previous memory stores are globally visible, and all
5906 previous memory loads have been satisfied, but following memory reads
5907 are not prevented from being speculated to before the barrier.
5910 @node Object Size Checking
5911 @section Object Size Checking Builtins
5912 @findex __builtin_object_size
5913 @findex __builtin___memcpy_chk
5914 @findex __builtin___mempcpy_chk
5915 @findex __builtin___memmove_chk
5916 @findex __builtin___memset_chk
5917 @findex __builtin___strcpy_chk
5918 @findex __builtin___stpcpy_chk
5919 @findex __builtin___strncpy_chk
5920 @findex __builtin___strcat_chk
5921 @findex __builtin___strncat_chk
5922 @findex __builtin___sprintf_chk
5923 @findex __builtin___snprintf_chk
5924 @findex __builtin___vsprintf_chk
5925 @findex __builtin___vsnprintf_chk
5926 @findex __builtin___printf_chk
5927 @findex __builtin___vprintf_chk
5928 @findex __builtin___fprintf_chk
5929 @findex __builtin___vfprintf_chk
5931 GCC implements a limited buffer overflow protection mechanism
5932 that can prevent some buffer overflow attacks.
5934 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5935 is a built-in construct that returns a constant number of bytes from
5936 @var{ptr} to the end of the object @var{ptr} pointer points to
5937 (if known at compile time). @code{__builtin_object_size} never evaluates
5938 its arguments for side-effects. If there are any side-effects in them, it
5939 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5940 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5941 point to and all of them are known at compile time, the returned number
5942 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5943 0 and minimum if nonzero. If it is not possible to determine which objects
5944 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5945 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5946 for @var{type} 2 or 3.
5948 @var{type} is an integer constant from 0 to 3. If the least significant
5949 bit is clear, objects are whole variables, if it is set, a closest
5950 surrounding subobject is considered the object a pointer points to.
5951 The second bit determines if maximum or minimum of remaining bytes
5955 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5956 char *p = &var.buf1[1], *q = &var.b;
5958 /* Here the object p points to is var. */
5959 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5960 /* The subobject p points to is var.buf1. */
5961 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5962 /* The object q points to is var. */
5963 assert (__builtin_object_size (q, 0)
5964 == (char *) (&var + 1) - (char *) &var.b);
5965 /* The subobject q points to is var.b. */
5966 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5970 There are built-in functions added for many common string operation
5971 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
5972 built-in is provided. This built-in has an additional last argument,
5973 which is the number of bytes remaining in object the @var{dest}
5974 argument points to or @code{(size_t) -1} if the size is not known.
5976 The built-in functions are optimized into the normal string functions
5977 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5978 it is known at compile time that the destination object will not
5979 be overflown. If the compiler can determine at compile time the
5980 object will be always overflown, it issues a warning.
5982 The intended use can be e.g.
5986 #define bos0(dest) __builtin_object_size (dest, 0)
5987 #define memcpy(dest, src, n) \
5988 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5992 /* It is unknown what object p points to, so this is optimized
5993 into plain memcpy - no checking is possible. */
5994 memcpy (p, "abcde", n);
5995 /* Destination is known and length too. It is known at compile
5996 time there will be no overflow. */
5997 memcpy (&buf[5], "abcde", 5);
5998 /* Destination is known, but the length is not known at compile time.
5999 This will result in __memcpy_chk call that can check for overflow
6001 memcpy (&buf[5], "abcde", n);
6002 /* Destination is known and it is known at compile time there will
6003 be overflow. There will be a warning and __memcpy_chk call that
6004 will abort the program at runtime. */
6005 memcpy (&buf[6], "abcde", 5);
6008 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
6009 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
6010 @code{strcat} and @code{strncat}.
6012 There are also checking built-in functions for formatted output functions.
6014 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
6015 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6016 const char *fmt, ...);
6017 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
6019 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
6020 const char *fmt, va_list ap);
6023 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
6024 etc.@: functions and can contain implementation specific flags on what
6025 additional security measures the checking function might take, such as
6026 handling @code{%n} differently.
6028 The @var{os} argument is the object size @var{s} points to, like in the
6029 other built-in functions. There is a small difference in the behavior
6030 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
6031 optimized into the non-checking functions only if @var{flag} is 0, otherwise
6032 the checking function is called with @var{os} argument set to
6035 In addition to this, there are checking built-in functions
6036 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
6037 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
6038 These have just one additional argument, @var{flag}, right before
6039 format string @var{fmt}. If the compiler is able to optimize them to
6040 @code{fputc} etc.@: functions, it will, otherwise the checking function
6041 should be called and the @var{flag} argument passed to it.
6043 @node Other Builtins
6044 @section Other built-in functions provided by GCC
6045 @cindex built-in functions
6046 @findex __builtin_fpclassify
6047 @findex __builtin_isfinite
6048 @findex __builtin_isnormal
6049 @findex __builtin_isgreater
6050 @findex __builtin_isgreaterequal
6051 @findex __builtin_isinf_sign
6052 @findex __builtin_isless
6053 @findex __builtin_islessequal
6054 @findex __builtin_islessgreater
6055 @findex __builtin_isunordered
6056 @findex __builtin_powi
6057 @findex __builtin_powif
6058 @findex __builtin_powil
6216 @findex fprintf_unlocked
6218 @findex fputs_unlocked
6335 @findex printf_unlocked
6367 @findex significandf
6368 @findex significandl
6439 GCC provides a large number of built-in functions other than the ones
6440 mentioned above. Some of these are for internal use in the processing
6441 of exceptions or variable-length argument lists and will not be
6442 documented here because they may change from time to time; we do not
6443 recommend general use of these functions.
6445 The remaining functions are provided for optimization purposes.
6447 @opindex fno-builtin
6448 GCC includes built-in versions of many of the functions in the standard
6449 C library. The versions prefixed with @code{__builtin_} will always be
6450 treated as having the same meaning as the C library function even if you
6451 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
6452 Many of these functions are only optimized in certain cases; if they are
6453 not optimized in a particular case, a call to the library function will
6458 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
6459 @option{-std=c99}), the functions
6460 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
6461 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
6462 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
6463 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
6464 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
6465 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
6466 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
6467 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
6468 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
6469 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
6470 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
6471 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
6472 @code{signbitd64}, @code{signbitd128}, @code{significandf},
6473 @code{significandl}, @code{significand}, @code{sincosf},
6474 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
6475 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
6476 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
6477 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
6479 may be handled as built-in functions.
6480 All these functions have corresponding versions
6481 prefixed with @code{__builtin_}, which may be used even in strict C89
6484 The ISO C99 functions
6485 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
6486 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
6487 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
6488 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
6489 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
6490 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
6491 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
6492 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
6493 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
6494 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
6495 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
6496 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
6497 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
6498 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
6499 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
6500 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
6501 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
6502 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
6503 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
6504 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
6505 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
6506 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
6507 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
6508 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
6509 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
6510 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
6511 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
6512 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
6513 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
6514 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
6515 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
6516 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
6517 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
6518 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
6519 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
6520 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
6521 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
6522 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
6523 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
6524 are handled as built-in functions
6525 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6527 There are also built-in versions of the ISO C99 functions
6528 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
6529 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
6530 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
6531 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
6532 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
6533 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
6534 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
6535 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
6536 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
6537 that are recognized in any mode since ISO C90 reserves these names for
6538 the purpose to which ISO C99 puts them. All these functions have
6539 corresponding versions prefixed with @code{__builtin_}.
6541 The ISO C94 functions
6542 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
6543 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
6544 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
6546 are handled as built-in functions
6547 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6549 The ISO C90 functions
6550 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
6551 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
6552 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
6553 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
6554 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
6555 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
6556 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
6557 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
6558 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
6559 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
6560 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
6561 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
6562 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
6563 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
6564 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
6565 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
6566 are all recognized as built-in functions unless
6567 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
6568 is specified for an individual function). All of these functions have
6569 corresponding versions prefixed with @code{__builtin_}.
6571 GCC provides built-in versions of the ISO C99 floating point comparison
6572 macros that avoid raising exceptions for unordered operands. They have
6573 the same names as the standard macros ( @code{isgreater},
6574 @code{isgreaterequal}, @code{isless}, @code{islessequal},
6575 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
6576 prefixed. We intend for a library implementor to be able to simply
6577 @code{#define} each standard macro to its built-in equivalent.
6578 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
6579 @code{isinf_sign} and @code{isnormal} built-ins used with
6580 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
6581 builtins appear both with and without the @code{__builtin_} prefix.
6583 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
6585 You can use the built-in function @code{__builtin_types_compatible_p} to
6586 determine whether two types are the same.
6588 This built-in function returns 1 if the unqualified versions of the
6589 types @var{type1} and @var{type2} (which are types, not expressions) are
6590 compatible, 0 otherwise. The result of this built-in function can be
6591 used in integer constant expressions.
6593 This built-in function ignores top level qualifiers (e.g., @code{const},
6594 @code{volatile}). For example, @code{int} is equivalent to @code{const
6597 The type @code{int[]} and @code{int[5]} are compatible. On the other
6598 hand, @code{int} and @code{char *} are not compatible, even if the size
6599 of their types, on the particular architecture are the same. Also, the
6600 amount of pointer indirection is taken into account when determining
6601 similarity. Consequently, @code{short *} is not similar to
6602 @code{short **}. Furthermore, two types that are typedefed are
6603 considered compatible if their underlying types are compatible.
6605 An @code{enum} type is not considered to be compatible with another
6606 @code{enum} type even if both are compatible with the same integer
6607 type; this is what the C standard specifies.
6608 For example, @code{enum @{foo, bar@}} is not similar to
6609 @code{enum @{hot, dog@}}.
6611 You would typically use this function in code whose execution varies
6612 depending on the arguments' types. For example:
6617 typeof (x) tmp = (x); \
6618 if (__builtin_types_compatible_p (typeof (x), long double)) \
6619 tmp = foo_long_double (tmp); \
6620 else if (__builtin_types_compatible_p (typeof (x), double)) \
6621 tmp = foo_double (tmp); \
6622 else if (__builtin_types_compatible_p (typeof (x), float)) \
6623 tmp = foo_float (tmp); \
6630 @emph{Note:} This construct is only available for C@.
6634 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
6636 You can use the built-in function @code{__builtin_choose_expr} to
6637 evaluate code depending on the value of a constant expression. This
6638 built-in function returns @var{exp1} if @var{const_exp}, which is a
6639 constant expression that must be able to be determined at compile time,
6640 is nonzero. Otherwise it returns 0.
6642 This built-in function is analogous to the @samp{? :} operator in C,
6643 except that the expression returned has its type unaltered by promotion
6644 rules. Also, the built-in function does not evaluate the expression
6645 that was not chosen. For example, if @var{const_exp} evaluates to true,
6646 @var{exp2} is not evaluated even if it has side-effects.
6648 This built-in function can return an lvalue if the chosen argument is an
6651 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
6652 type. Similarly, if @var{exp2} is returned, its return type is the same
6659 __builtin_choose_expr ( \
6660 __builtin_types_compatible_p (typeof (x), double), \
6662 __builtin_choose_expr ( \
6663 __builtin_types_compatible_p (typeof (x), float), \
6665 /* @r{The void expression results in a compile-time error} \
6666 @r{when assigning the result to something.} */ \
6670 @emph{Note:} This construct is only available for C@. Furthermore, the
6671 unused expression (@var{exp1} or @var{exp2} depending on the value of
6672 @var{const_exp}) may still generate syntax errors. This may change in
6677 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6678 You can use the built-in function @code{__builtin_constant_p} to
6679 determine if a value is known to be constant at compile-time and hence
6680 that GCC can perform constant-folding on expressions involving that
6681 value. The argument of the function is the value to test. The function
6682 returns the integer 1 if the argument is known to be a compile-time
6683 constant and 0 if it is not known to be a compile-time constant. A
6684 return of 0 does not indicate that the value is @emph{not} a constant,
6685 but merely that GCC cannot prove it is a constant with the specified
6686 value of the @option{-O} option.
6688 You would typically use this function in an embedded application where
6689 memory was a critical resource. If you have some complex calculation,
6690 you may want it to be folded if it involves constants, but need to call
6691 a function if it does not. For example:
6694 #define Scale_Value(X) \
6695 (__builtin_constant_p (X) \
6696 ? ((X) * SCALE + OFFSET) : Scale (X))
6699 You may use this built-in function in either a macro or an inline
6700 function. However, if you use it in an inlined function and pass an
6701 argument of the function as the argument to the built-in, GCC will
6702 never return 1 when you call the inline function with a string constant
6703 or compound literal (@pxref{Compound Literals}) and will not return 1
6704 when you pass a constant numeric value to the inline function unless you
6705 specify the @option{-O} option.
6707 You may also use @code{__builtin_constant_p} in initializers for static
6708 data. For instance, you can write
6711 static const int table[] = @{
6712 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6718 This is an acceptable initializer even if @var{EXPRESSION} is not a
6719 constant expression. GCC must be more conservative about evaluating the
6720 built-in in this case, because it has no opportunity to perform
6723 Previous versions of GCC did not accept this built-in in data
6724 initializers. The earliest version where it is completely safe is
6728 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6729 @opindex fprofile-arcs
6730 You may use @code{__builtin_expect} to provide the compiler with
6731 branch prediction information. In general, you should prefer to
6732 use actual profile feedback for this (@option{-fprofile-arcs}), as
6733 programmers are notoriously bad at predicting how their programs
6734 actually perform. However, there are applications in which this
6735 data is hard to collect.
6737 The return value is the value of @var{exp}, which should be an integral
6738 expression. The semantics of the built-in are that it is expected that
6739 @var{exp} == @var{c}. For example:
6742 if (__builtin_expect (x, 0))
6747 would indicate that we do not expect to call @code{foo}, since
6748 we expect @code{x} to be zero. Since you are limited to integral
6749 expressions for @var{exp}, you should use constructions such as
6752 if (__builtin_expect (ptr != NULL, 1))
6757 when testing pointer or floating-point values.
6760 @deftypefn {Built-in Function} void __builtin_trap (void)
6761 This function causes the program to exit abnormally. GCC implements
6762 this function by using a target-dependent mechanism (such as
6763 intentionally executing an illegal instruction) or by calling
6764 @code{abort}. The mechanism used may vary from release to release so
6765 you should not rely on any particular implementation.
6768 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
6769 This function is used to flush the processor's instruction cache for
6770 the region of memory between @var{begin} inclusive and @var{end}
6771 exclusive. Some targets require that the instruction cache be
6772 flushed, after modifying memory containing code, in order to obtain
6773 deterministic behavior.
6775 If the target does not require instruction cache flushes,
6776 @code{__builtin___clear_cache} has no effect. Otherwise either
6777 instructions are emitted in-line to clear the instruction cache or a
6778 call to the @code{__clear_cache} function in libgcc is made.
6781 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6782 This function is used to minimize cache-miss latency by moving data into
6783 a cache before it is accessed.
6784 You can insert calls to @code{__builtin_prefetch} into code for which
6785 you know addresses of data in memory that is likely to be accessed soon.
6786 If the target supports them, data prefetch instructions will be generated.
6787 If the prefetch is done early enough before the access then the data will
6788 be in the cache by the time it is accessed.
6790 The value of @var{addr} is the address of the memory to prefetch.
6791 There are two optional arguments, @var{rw} and @var{locality}.
6792 The value of @var{rw} is a compile-time constant one or zero; one
6793 means that the prefetch is preparing for a write to the memory address
6794 and zero, the default, means that the prefetch is preparing for a read.
6795 The value @var{locality} must be a compile-time constant integer between
6796 zero and three. A value of zero means that the data has no temporal
6797 locality, so it need not be left in the cache after the access. A value
6798 of three means that the data has a high degree of temporal locality and
6799 should be left in all levels of cache possible. Values of one and two
6800 mean, respectively, a low or moderate degree of temporal locality. The
6804 for (i = 0; i < n; i++)
6807 __builtin_prefetch (&a[i+j], 1, 1);
6808 __builtin_prefetch (&b[i+j], 0, 1);
6813 Data prefetch does not generate faults if @var{addr} is invalid, but
6814 the address expression itself must be valid. For example, a prefetch
6815 of @code{p->next} will not fault if @code{p->next} is not a valid
6816 address, but evaluation will fault if @code{p} is not a valid address.
6818 If the target does not support data prefetch, the address expression
6819 is evaluated if it includes side effects but no other code is generated
6820 and GCC does not issue a warning.
6823 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6824 Returns a positive infinity, if supported by the floating-point format,
6825 else @code{DBL_MAX}. This function is suitable for implementing the
6826 ISO C macro @code{HUGE_VAL}.
6829 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6830 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6833 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6834 Similar to @code{__builtin_huge_val}, except the return
6835 type is @code{long double}.
6838 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
6839 This built-in implements the C99 fpclassify functionality. The first
6840 five int arguments should be the target library's notion of the
6841 possible FP classes and are used for return values. They must be
6842 constant values and they must appear in this order: @code{FP_NAN},
6843 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
6844 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
6845 to classify. GCC treats the last argument as type-generic, which
6846 means it does not do default promotion from float to double.
6849 @deftypefn {Built-in Function} double __builtin_inf (void)
6850 Similar to @code{__builtin_huge_val}, except a warning is generated
6851 if the target floating-point format does not support infinities.
6854 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6855 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6858 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6859 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6862 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6863 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6866 @deftypefn {Built-in Function} float __builtin_inff (void)
6867 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6868 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6871 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6872 Similar to @code{__builtin_inf}, except the return
6873 type is @code{long double}.
6876 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
6877 Similar to @code{isinf}, except the return value will be negative for
6878 an argument of @code{-Inf}. Note while the parameter list is an
6879 ellipsis, this function only accepts exactly one floating point
6880 argument. GCC treats this parameter as type-generic, which means it
6881 does not do default promotion from float to double.
6884 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6885 This is an implementation of the ISO C99 function @code{nan}.
6887 Since ISO C99 defines this function in terms of @code{strtod}, which we
6888 do not implement, a description of the parsing is in order. The string
6889 is parsed as by @code{strtol}; that is, the base is recognized by
6890 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6891 in the significand such that the least significant bit of the number
6892 is at the least significant bit of the significand. The number is
6893 truncated to fit the significand field provided. The significand is
6894 forced to be a quiet NaN@.
6896 This function, if given a string literal all of which would have been
6897 consumed by strtol, is evaluated early enough that it is considered a
6898 compile-time constant.
6901 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6902 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6905 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6906 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6909 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6910 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6913 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6914 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6917 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6918 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6921 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6922 Similar to @code{__builtin_nan}, except the significand is forced
6923 to be a signaling NaN@. The @code{nans} function is proposed by
6924 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6927 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6928 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6931 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6932 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6935 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6936 Returns one plus the index of the least significant 1-bit of @var{x}, or
6937 if @var{x} is zero, returns zero.
6940 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6941 Returns the number of leading 0-bits in @var{x}, starting at the most
6942 significant bit position. If @var{x} is 0, the result is undefined.
6945 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6946 Returns the number of trailing 0-bits in @var{x}, starting at the least
6947 significant bit position. If @var{x} is 0, the result is undefined.
6950 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6951 Returns the number of 1-bits in @var{x}.
6954 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6955 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6959 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6960 Similar to @code{__builtin_ffs}, except the argument type is
6961 @code{unsigned long}.
6964 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6965 Similar to @code{__builtin_clz}, except the argument type is
6966 @code{unsigned long}.
6969 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6970 Similar to @code{__builtin_ctz}, except the argument type is
6971 @code{unsigned long}.
6974 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6975 Similar to @code{__builtin_popcount}, except the argument type is
6976 @code{unsigned long}.
6979 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6980 Similar to @code{__builtin_parity}, except the argument type is
6981 @code{unsigned long}.
6984 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6985 Similar to @code{__builtin_ffs}, except the argument type is
6986 @code{unsigned long long}.
6989 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6990 Similar to @code{__builtin_clz}, except the argument type is
6991 @code{unsigned long long}.
6994 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6995 Similar to @code{__builtin_ctz}, except the argument type is
6996 @code{unsigned long long}.
6999 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
7000 Similar to @code{__builtin_popcount}, except the argument type is
7001 @code{unsigned long long}.
7004 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
7005 Similar to @code{__builtin_parity}, except the argument type is
7006 @code{unsigned long long}.
7009 @deftypefn {Built-in Function} double __builtin_powi (double, int)
7010 Returns the first argument raised to the power of the second. Unlike the
7011 @code{pow} function no guarantees about precision and rounding are made.
7014 @deftypefn {Built-in Function} float __builtin_powif (float, int)
7015 Similar to @code{__builtin_powi}, except the argument and return types
7019 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
7020 Similar to @code{__builtin_powi}, except the argument and return types
7021 are @code{long double}.
7024 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
7025 Returns @var{x} with the order of the bytes reversed; for example,
7026 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
7030 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
7031 Similar to @code{__builtin_bswap32}, except the argument and return types
7035 @node Target Builtins
7036 @section Built-in Functions Specific to Particular Target Machines
7038 On some target machines, GCC supports many built-in functions specific
7039 to those machines. Generally these generate calls to specific machine
7040 instructions, but allow the compiler to schedule those calls.
7043 * Alpha Built-in Functions::
7044 * ARM iWMMXt Built-in Functions::
7045 * ARM NEON Intrinsics::
7046 * Blackfin Built-in Functions::
7047 * FR-V Built-in Functions::
7048 * X86 Built-in Functions::
7049 * MIPS DSP Built-in Functions::
7050 * MIPS Paired-Single Support::
7051 * MIPS Loongson Built-in Functions::
7052 * Other MIPS Built-in Functions::
7053 * picoChip Built-in Functions::
7054 * PowerPC AltiVec Built-in Functions::
7055 * SPARC VIS Built-in Functions::
7056 * SPU Built-in Functions::
7059 @node Alpha Built-in Functions
7060 @subsection Alpha Built-in Functions
7062 These built-in functions are available for the Alpha family of
7063 processors, depending on the command-line switches used.
7065 The following built-in functions are always available. They
7066 all generate the machine instruction that is part of the name.
7069 long __builtin_alpha_implver (void)
7070 long __builtin_alpha_rpcc (void)
7071 long __builtin_alpha_amask (long)
7072 long __builtin_alpha_cmpbge (long, long)
7073 long __builtin_alpha_extbl (long, long)
7074 long __builtin_alpha_extwl (long, long)
7075 long __builtin_alpha_extll (long, long)
7076 long __builtin_alpha_extql (long, long)
7077 long __builtin_alpha_extwh (long, long)
7078 long __builtin_alpha_extlh (long, long)
7079 long __builtin_alpha_extqh (long, long)
7080 long __builtin_alpha_insbl (long, long)
7081 long __builtin_alpha_inswl (long, long)
7082 long __builtin_alpha_insll (long, long)
7083 long __builtin_alpha_insql (long, long)
7084 long __builtin_alpha_inswh (long, long)
7085 long __builtin_alpha_inslh (long, long)
7086 long __builtin_alpha_insqh (long, long)
7087 long __builtin_alpha_mskbl (long, long)
7088 long __builtin_alpha_mskwl (long, long)
7089 long __builtin_alpha_mskll (long, long)
7090 long __builtin_alpha_mskql (long, long)
7091 long __builtin_alpha_mskwh (long, long)
7092 long __builtin_alpha_msklh (long, long)
7093 long __builtin_alpha_mskqh (long, long)
7094 long __builtin_alpha_umulh (long, long)
7095 long __builtin_alpha_zap (long, long)
7096 long __builtin_alpha_zapnot (long, long)
7099 The following built-in functions are always with @option{-mmax}
7100 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7101 later. They all generate the machine instruction that is part
7105 long __builtin_alpha_pklb (long)
7106 long __builtin_alpha_pkwb (long)
7107 long __builtin_alpha_unpkbl (long)
7108 long __builtin_alpha_unpkbw (long)
7109 long __builtin_alpha_minub8 (long, long)
7110 long __builtin_alpha_minsb8 (long, long)
7111 long __builtin_alpha_minuw4 (long, long)
7112 long __builtin_alpha_minsw4 (long, long)
7113 long __builtin_alpha_maxub8 (long, long)
7114 long __builtin_alpha_maxsb8 (long, long)
7115 long __builtin_alpha_maxuw4 (long, long)
7116 long __builtin_alpha_maxsw4 (long, long)
7117 long __builtin_alpha_perr (long, long)
7120 The following built-in functions are always with @option{-mcix}
7121 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
7122 later. They all generate the machine instruction that is part
7126 long __builtin_alpha_cttz (long)
7127 long __builtin_alpha_ctlz (long)
7128 long __builtin_alpha_ctpop (long)
7131 The following builtins are available on systems that use the OSF/1
7132 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
7133 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
7134 @code{rdval} and @code{wrval}.
7137 void *__builtin_thread_pointer (void)
7138 void __builtin_set_thread_pointer (void *)
7141 @node ARM iWMMXt Built-in Functions
7142 @subsection ARM iWMMXt Built-in Functions
7144 These built-in functions are available for the ARM family of
7145 processors when the @option{-mcpu=iwmmxt} switch is used:
7148 typedef int v2si __attribute__ ((vector_size (8)));
7149 typedef short v4hi __attribute__ ((vector_size (8)));
7150 typedef char v8qi __attribute__ ((vector_size (8)));
7152 int __builtin_arm_getwcx (int)
7153 void __builtin_arm_setwcx (int, int)
7154 int __builtin_arm_textrmsb (v8qi, int)
7155 int __builtin_arm_textrmsh (v4hi, int)
7156 int __builtin_arm_textrmsw (v2si, int)
7157 int __builtin_arm_textrmub (v8qi, int)
7158 int __builtin_arm_textrmuh (v4hi, int)
7159 int __builtin_arm_textrmuw (v2si, int)
7160 v8qi __builtin_arm_tinsrb (v8qi, int)
7161 v4hi __builtin_arm_tinsrh (v4hi, int)
7162 v2si __builtin_arm_tinsrw (v2si, int)
7163 long long __builtin_arm_tmia (long long, int, int)
7164 long long __builtin_arm_tmiabb (long long, int, int)
7165 long long __builtin_arm_tmiabt (long long, int, int)
7166 long long __builtin_arm_tmiaph (long long, int, int)
7167 long long __builtin_arm_tmiatb (long long, int, int)
7168 long long __builtin_arm_tmiatt (long long, int, int)
7169 int __builtin_arm_tmovmskb (v8qi)
7170 int __builtin_arm_tmovmskh (v4hi)
7171 int __builtin_arm_tmovmskw (v2si)
7172 long long __builtin_arm_waccb (v8qi)
7173 long long __builtin_arm_wacch (v4hi)
7174 long long __builtin_arm_waccw (v2si)
7175 v8qi __builtin_arm_waddb (v8qi, v8qi)
7176 v8qi __builtin_arm_waddbss (v8qi, v8qi)
7177 v8qi __builtin_arm_waddbus (v8qi, v8qi)
7178 v4hi __builtin_arm_waddh (v4hi, v4hi)
7179 v4hi __builtin_arm_waddhss (v4hi, v4hi)
7180 v4hi __builtin_arm_waddhus (v4hi, v4hi)
7181 v2si __builtin_arm_waddw (v2si, v2si)
7182 v2si __builtin_arm_waddwss (v2si, v2si)
7183 v2si __builtin_arm_waddwus (v2si, v2si)
7184 v8qi __builtin_arm_walign (v8qi, v8qi, int)
7185 long long __builtin_arm_wand(long long, long long)
7186 long long __builtin_arm_wandn (long long, long long)
7187 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
7188 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
7189 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
7190 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
7191 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
7192 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
7193 v2si __builtin_arm_wcmpeqw (v2si, v2si)
7194 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
7195 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
7196 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
7197 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
7198 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
7199 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
7200 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
7201 long long __builtin_arm_wmacsz (v4hi, v4hi)
7202 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
7203 long long __builtin_arm_wmacuz (v4hi, v4hi)
7204 v4hi __builtin_arm_wmadds (v4hi, v4hi)
7205 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
7206 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
7207 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
7208 v2si __builtin_arm_wmaxsw (v2si, v2si)
7209 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
7210 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
7211 v2si __builtin_arm_wmaxuw (v2si, v2si)
7212 v8qi __builtin_arm_wminsb (v8qi, v8qi)
7213 v4hi __builtin_arm_wminsh (v4hi, v4hi)
7214 v2si __builtin_arm_wminsw (v2si, v2si)
7215 v8qi __builtin_arm_wminub (v8qi, v8qi)
7216 v4hi __builtin_arm_wminuh (v4hi, v4hi)
7217 v2si __builtin_arm_wminuw (v2si, v2si)
7218 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
7219 v4hi __builtin_arm_wmulul (v4hi, v4hi)
7220 v4hi __builtin_arm_wmulum (v4hi, v4hi)
7221 long long __builtin_arm_wor (long long, long long)
7222 v2si __builtin_arm_wpackdss (long long, long long)
7223 v2si __builtin_arm_wpackdus (long long, long long)
7224 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
7225 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
7226 v4hi __builtin_arm_wpackwss (v2si, v2si)
7227 v4hi __builtin_arm_wpackwus (v2si, v2si)
7228 long long __builtin_arm_wrord (long long, long long)
7229 long long __builtin_arm_wrordi (long long, int)
7230 v4hi __builtin_arm_wrorh (v4hi, long long)
7231 v4hi __builtin_arm_wrorhi (v4hi, int)
7232 v2si __builtin_arm_wrorw (v2si, long long)
7233 v2si __builtin_arm_wrorwi (v2si, int)
7234 v2si __builtin_arm_wsadb (v8qi, v8qi)
7235 v2si __builtin_arm_wsadbz (v8qi, v8qi)
7236 v2si __builtin_arm_wsadh (v4hi, v4hi)
7237 v2si __builtin_arm_wsadhz (v4hi, v4hi)
7238 v4hi __builtin_arm_wshufh (v4hi, int)
7239 long long __builtin_arm_wslld (long long, long long)
7240 long long __builtin_arm_wslldi (long long, int)
7241 v4hi __builtin_arm_wsllh (v4hi, long long)
7242 v4hi __builtin_arm_wsllhi (v4hi, int)
7243 v2si __builtin_arm_wsllw (v2si, long long)
7244 v2si __builtin_arm_wsllwi (v2si, int)
7245 long long __builtin_arm_wsrad (long long, long long)
7246 long long __builtin_arm_wsradi (long long, int)
7247 v4hi __builtin_arm_wsrah (v4hi, long long)
7248 v4hi __builtin_arm_wsrahi (v4hi, int)
7249 v2si __builtin_arm_wsraw (v2si, long long)
7250 v2si __builtin_arm_wsrawi (v2si, int)
7251 long long __builtin_arm_wsrld (long long, long long)
7252 long long __builtin_arm_wsrldi (long long, int)
7253 v4hi __builtin_arm_wsrlh (v4hi, long long)
7254 v4hi __builtin_arm_wsrlhi (v4hi, int)
7255 v2si __builtin_arm_wsrlw (v2si, long long)
7256 v2si __builtin_arm_wsrlwi (v2si, int)
7257 v8qi __builtin_arm_wsubb (v8qi, v8qi)
7258 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
7259 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
7260 v4hi __builtin_arm_wsubh (v4hi, v4hi)
7261 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
7262 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
7263 v2si __builtin_arm_wsubw (v2si, v2si)
7264 v2si __builtin_arm_wsubwss (v2si, v2si)
7265 v2si __builtin_arm_wsubwus (v2si, v2si)
7266 v4hi __builtin_arm_wunpckehsb (v8qi)
7267 v2si __builtin_arm_wunpckehsh (v4hi)
7268 long long __builtin_arm_wunpckehsw (v2si)
7269 v4hi __builtin_arm_wunpckehub (v8qi)
7270 v2si __builtin_arm_wunpckehuh (v4hi)
7271 long long __builtin_arm_wunpckehuw (v2si)
7272 v4hi __builtin_arm_wunpckelsb (v8qi)
7273 v2si __builtin_arm_wunpckelsh (v4hi)
7274 long long __builtin_arm_wunpckelsw (v2si)
7275 v4hi __builtin_arm_wunpckelub (v8qi)
7276 v2si __builtin_arm_wunpckeluh (v4hi)
7277 long long __builtin_arm_wunpckeluw (v2si)
7278 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
7279 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
7280 v2si __builtin_arm_wunpckihw (v2si, v2si)
7281 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
7282 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
7283 v2si __builtin_arm_wunpckilw (v2si, v2si)
7284 long long __builtin_arm_wxor (long long, long long)
7285 long long __builtin_arm_wzero ()
7288 @node ARM NEON Intrinsics
7289 @subsection ARM NEON Intrinsics
7291 These built-in intrinsics for the ARM Advanced SIMD extension are available
7292 when the @option{-mfpu=neon} switch is used:
7294 @include arm-neon-intrinsics.texi
7296 @node Blackfin Built-in Functions
7297 @subsection Blackfin Built-in Functions
7299 Currently, there are two Blackfin-specific built-in functions. These are
7300 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
7301 using inline assembly; by using these built-in functions the compiler can
7302 automatically add workarounds for hardware errata involving these
7303 instructions. These functions are named as follows:
7306 void __builtin_bfin_csync (void)
7307 void __builtin_bfin_ssync (void)
7310 @node FR-V Built-in Functions
7311 @subsection FR-V Built-in Functions
7313 GCC provides many FR-V-specific built-in functions. In general,
7314 these functions are intended to be compatible with those described
7315 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
7316 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
7317 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
7318 pointer rather than by value.
7320 Most of the functions are named after specific FR-V instructions.
7321 Such functions are said to be ``directly mapped'' and are summarized
7322 here in tabular form.
7326 * Directly-mapped Integer Functions::
7327 * Directly-mapped Media Functions::
7328 * Raw read/write Functions::
7329 * Other Built-in Functions::
7332 @node Argument Types
7333 @subsubsection Argument Types
7335 The arguments to the built-in functions can be divided into three groups:
7336 register numbers, compile-time constants and run-time values. In order
7337 to make this classification clear at a glance, the arguments and return
7338 values are given the following pseudo types:
7340 @multitable @columnfractions .20 .30 .15 .35
7341 @item Pseudo type @tab Real C type @tab Constant? @tab Description
7342 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
7343 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
7344 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
7345 @item @code{uw2} @tab @code{unsigned long long} @tab No
7346 @tab an unsigned doubleword
7347 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
7348 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
7349 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
7350 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
7353 These pseudo types are not defined by GCC, they are simply a notational
7354 convenience used in this manual.
7356 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
7357 and @code{sw2} are evaluated at run time. They correspond to
7358 register operands in the underlying FR-V instructions.
7360 @code{const} arguments represent immediate operands in the underlying
7361 FR-V instructions. They must be compile-time constants.
7363 @code{acc} arguments are evaluated at compile time and specify the number
7364 of an accumulator register. For example, an @code{acc} argument of 2
7365 will select the ACC2 register.
7367 @code{iacc} arguments are similar to @code{acc} arguments but specify the
7368 number of an IACC register. See @pxref{Other Built-in Functions}
7371 @node Directly-mapped Integer Functions
7372 @subsubsection Directly-mapped Integer Functions
7374 The functions listed below map directly to FR-V I-type instructions.
7376 @multitable @columnfractions .45 .32 .23
7377 @item Function prototype @tab Example usage @tab Assembly output
7378 @item @code{sw1 __ADDSS (sw1, sw1)}
7379 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
7380 @tab @code{ADDSS @var{a},@var{b},@var{c}}
7381 @item @code{sw1 __SCAN (sw1, sw1)}
7382 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
7383 @tab @code{SCAN @var{a},@var{b},@var{c}}
7384 @item @code{sw1 __SCUTSS (sw1)}
7385 @tab @code{@var{b} = __SCUTSS (@var{a})}
7386 @tab @code{SCUTSS @var{a},@var{b}}
7387 @item @code{sw1 __SLASS (sw1, sw1)}
7388 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
7389 @tab @code{SLASS @var{a},@var{b},@var{c}}
7390 @item @code{void __SMASS (sw1, sw1)}
7391 @tab @code{__SMASS (@var{a}, @var{b})}
7392 @tab @code{SMASS @var{a},@var{b}}
7393 @item @code{void __SMSSS (sw1, sw1)}
7394 @tab @code{__SMSSS (@var{a}, @var{b})}
7395 @tab @code{SMSSS @var{a},@var{b}}
7396 @item @code{void __SMU (sw1, sw1)}
7397 @tab @code{__SMU (@var{a}, @var{b})}
7398 @tab @code{SMU @var{a},@var{b}}
7399 @item @code{sw2 __SMUL (sw1, sw1)}
7400 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
7401 @tab @code{SMUL @var{a},@var{b},@var{c}}
7402 @item @code{sw1 __SUBSS (sw1, sw1)}
7403 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
7404 @tab @code{SUBSS @var{a},@var{b},@var{c}}
7405 @item @code{uw2 __UMUL (uw1, uw1)}
7406 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
7407 @tab @code{UMUL @var{a},@var{b},@var{c}}
7410 @node Directly-mapped Media Functions
7411 @subsubsection Directly-mapped Media Functions
7413 The functions listed below map directly to FR-V M-type instructions.
7415 @multitable @columnfractions .45 .32 .23
7416 @item Function prototype @tab Example usage @tab Assembly output
7417 @item @code{uw1 __MABSHS (sw1)}
7418 @tab @code{@var{b} = __MABSHS (@var{a})}
7419 @tab @code{MABSHS @var{a},@var{b}}
7420 @item @code{void __MADDACCS (acc, acc)}
7421 @tab @code{__MADDACCS (@var{b}, @var{a})}
7422 @tab @code{MADDACCS @var{a},@var{b}}
7423 @item @code{sw1 __MADDHSS (sw1, sw1)}
7424 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
7425 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
7426 @item @code{uw1 __MADDHUS (uw1, uw1)}
7427 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
7428 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
7429 @item @code{uw1 __MAND (uw1, uw1)}
7430 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
7431 @tab @code{MAND @var{a},@var{b},@var{c}}
7432 @item @code{void __MASACCS (acc, acc)}
7433 @tab @code{__MASACCS (@var{b}, @var{a})}
7434 @tab @code{MASACCS @var{a},@var{b}}
7435 @item @code{uw1 __MAVEH (uw1, uw1)}
7436 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
7437 @tab @code{MAVEH @var{a},@var{b},@var{c}}
7438 @item @code{uw2 __MBTOH (uw1)}
7439 @tab @code{@var{b} = __MBTOH (@var{a})}
7440 @tab @code{MBTOH @var{a},@var{b}}
7441 @item @code{void __MBTOHE (uw1 *, uw1)}
7442 @tab @code{__MBTOHE (&@var{b}, @var{a})}
7443 @tab @code{MBTOHE @var{a},@var{b}}
7444 @item @code{void __MCLRACC (acc)}
7445 @tab @code{__MCLRACC (@var{a})}
7446 @tab @code{MCLRACC @var{a}}
7447 @item @code{void __MCLRACCA (void)}
7448 @tab @code{__MCLRACCA ()}
7449 @tab @code{MCLRACCA}
7450 @item @code{uw1 __Mcop1 (uw1, uw1)}
7451 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
7452 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
7453 @item @code{uw1 __Mcop2 (uw1, uw1)}
7454 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
7455 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
7456 @item @code{uw1 __MCPLHI (uw2, const)}
7457 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
7458 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
7459 @item @code{uw1 __MCPLI (uw2, const)}
7460 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
7461 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
7462 @item @code{void __MCPXIS (acc, sw1, sw1)}
7463 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
7464 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
7465 @item @code{void __MCPXIU (acc, uw1, uw1)}
7466 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
7467 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
7468 @item @code{void __MCPXRS (acc, sw1, sw1)}
7469 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
7470 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
7471 @item @code{void __MCPXRU (acc, uw1, uw1)}
7472 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
7473 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
7474 @item @code{uw1 __MCUT (acc, uw1)}
7475 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
7476 @tab @code{MCUT @var{a},@var{b},@var{c}}
7477 @item @code{uw1 __MCUTSS (acc, sw1)}
7478 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
7479 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
7480 @item @code{void __MDADDACCS (acc, acc)}
7481 @tab @code{__MDADDACCS (@var{b}, @var{a})}
7482 @tab @code{MDADDACCS @var{a},@var{b}}
7483 @item @code{void __MDASACCS (acc, acc)}
7484 @tab @code{__MDASACCS (@var{b}, @var{a})}
7485 @tab @code{MDASACCS @var{a},@var{b}}
7486 @item @code{uw2 __MDCUTSSI (acc, const)}
7487 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
7488 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
7489 @item @code{uw2 __MDPACKH (uw2, uw2)}
7490 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
7491 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
7492 @item @code{uw2 __MDROTLI (uw2, const)}
7493 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
7494 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
7495 @item @code{void __MDSUBACCS (acc, acc)}
7496 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
7497 @tab @code{MDSUBACCS @var{a},@var{b}}
7498 @item @code{void __MDUNPACKH (uw1 *, uw2)}
7499 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
7500 @tab @code{MDUNPACKH @var{a},@var{b}}
7501 @item @code{uw2 __MEXPDHD (uw1, const)}
7502 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
7503 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
7504 @item @code{uw1 __MEXPDHW (uw1, const)}
7505 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
7506 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
7507 @item @code{uw1 __MHDSETH (uw1, const)}
7508 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
7509 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
7510 @item @code{sw1 __MHDSETS (const)}
7511 @tab @code{@var{b} = __MHDSETS (@var{a})}
7512 @tab @code{MHDSETS #@var{a},@var{b}}
7513 @item @code{uw1 __MHSETHIH (uw1, const)}
7514 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
7515 @tab @code{MHSETHIH #@var{a},@var{b}}
7516 @item @code{sw1 __MHSETHIS (sw1, const)}
7517 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
7518 @tab @code{MHSETHIS #@var{a},@var{b}}
7519 @item @code{uw1 __MHSETLOH (uw1, const)}
7520 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
7521 @tab @code{MHSETLOH #@var{a},@var{b}}
7522 @item @code{sw1 __MHSETLOS (sw1, const)}
7523 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
7524 @tab @code{MHSETLOS #@var{a},@var{b}}
7525 @item @code{uw1 __MHTOB (uw2)}
7526 @tab @code{@var{b} = __MHTOB (@var{a})}
7527 @tab @code{MHTOB @var{a},@var{b}}
7528 @item @code{void __MMACHS (acc, sw1, sw1)}
7529 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
7530 @tab @code{MMACHS @var{a},@var{b},@var{c}}
7531 @item @code{void __MMACHU (acc, uw1, uw1)}
7532 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
7533 @tab @code{MMACHU @var{a},@var{b},@var{c}}
7534 @item @code{void __MMRDHS (acc, sw1, sw1)}
7535 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
7536 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
7537 @item @code{void __MMRDHU (acc, uw1, uw1)}
7538 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
7539 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
7540 @item @code{void __MMULHS (acc, sw1, sw1)}
7541 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
7542 @tab @code{MMULHS @var{a},@var{b},@var{c}}
7543 @item @code{void __MMULHU (acc, uw1, uw1)}
7544 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
7545 @tab @code{MMULHU @var{a},@var{b},@var{c}}
7546 @item @code{void __MMULXHS (acc, sw1, sw1)}
7547 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
7548 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
7549 @item @code{void __MMULXHU (acc, uw1, uw1)}
7550 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
7551 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
7552 @item @code{uw1 __MNOT (uw1)}
7553 @tab @code{@var{b} = __MNOT (@var{a})}
7554 @tab @code{MNOT @var{a},@var{b}}
7555 @item @code{uw1 __MOR (uw1, uw1)}
7556 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
7557 @tab @code{MOR @var{a},@var{b},@var{c}}
7558 @item @code{uw1 __MPACKH (uh, uh)}
7559 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
7560 @tab @code{MPACKH @var{a},@var{b},@var{c}}
7561 @item @code{sw2 __MQADDHSS (sw2, sw2)}
7562 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
7563 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
7564 @item @code{uw2 __MQADDHUS (uw2, uw2)}
7565 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
7566 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
7567 @item @code{void __MQCPXIS (acc, sw2, sw2)}
7568 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
7569 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
7570 @item @code{void __MQCPXIU (acc, uw2, uw2)}
7571 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
7572 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
7573 @item @code{void __MQCPXRS (acc, sw2, sw2)}
7574 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
7575 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
7576 @item @code{void __MQCPXRU (acc, uw2, uw2)}
7577 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
7578 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
7579 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
7580 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
7581 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
7582 @item @code{sw2 __MQLMTHS (sw2, sw2)}
7583 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
7584 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
7585 @item @code{void __MQMACHS (acc, sw2, sw2)}
7586 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
7587 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
7588 @item @code{void __MQMACHU (acc, uw2, uw2)}
7589 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
7590 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
7591 @item @code{void __MQMACXHS (acc, sw2, sw2)}
7592 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
7593 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
7594 @item @code{void __MQMULHS (acc, sw2, sw2)}
7595 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
7596 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
7597 @item @code{void __MQMULHU (acc, uw2, uw2)}
7598 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
7599 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
7600 @item @code{void __MQMULXHS (acc, sw2, sw2)}
7601 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
7602 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
7603 @item @code{void __MQMULXHU (acc, uw2, uw2)}
7604 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
7605 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
7606 @item @code{sw2 __MQSATHS (sw2, sw2)}
7607 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
7608 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
7609 @item @code{uw2 __MQSLLHI (uw2, int)}
7610 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
7611 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
7612 @item @code{sw2 __MQSRAHI (sw2, int)}
7613 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
7614 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
7615 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
7616 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
7617 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
7618 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
7619 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
7620 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
7621 @item @code{void __MQXMACHS (acc, sw2, sw2)}
7622 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
7623 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
7624 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
7625 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
7626 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
7627 @item @code{uw1 __MRDACC (acc)}
7628 @tab @code{@var{b} = __MRDACC (@var{a})}
7629 @tab @code{MRDACC @var{a},@var{b}}
7630 @item @code{uw1 __MRDACCG (acc)}
7631 @tab @code{@var{b} = __MRDACCG (@var{a})}
7632 @tab @code{MRDACCG @var{a},@var{b}}
7633 @item @code{uw1 __MROTLI (uw1, const)}
7634 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
7635 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
7636 @item @code{uw1 __MROTRI (uw1, const)}
7637 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
7638 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
7639 @item @code{sw1 __MSATHS (sw1, sw1)}
7640 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
7641 @tab @code{MSATHS @var{a},@var{b},@var{c}}
7642 @item @code{uw1 __MSATHU (uw1, uw1)}
7643 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
7644 @tab @code{MSATHU @var{a},@var{b},@var{c}}
7645 @item @code{uw1 __MSLLHI (uw1, const)}
7646 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
7647 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
7648 @item @code{sw1 __MSRAHI (sw1, const)}
7649 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
7650 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
7651 @item @code{uw1 __MSRLHI (uw1, const)}
7652 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
7653 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
7654 @item @code{void __MSUBACCS (acc, acc)}
7655 @tab @code{__MSUBACCS (@var{b}, @var{a})}
7656 @tab @code{MSUBACCS @var{a},@var{b}}
7657 @item @code{sw1 __MSUBHSS (sw1, sw1)}
7658 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
7659 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
7660 @item @code{uw1 __MSUBHUS (uw1, uw1)}
7661 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
7662 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
7663 @item @code{void __MTRAP (void)}
7664 @tab @code{__MTRAP ()}
7666 @item @code{uw2 __MUNPACKH (uw1)}
7667 @tab @code{@var{b} = __MUNPACKH (@var{a})}
7668 @tab @code{MUNPACKH @var{a},@var{b}}
7669 @item @code{uw1 __MWCUT (uw2, uw1)}
7670 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
7671 @tab @code{MWCUT @var{a},@var{b},@var{c}}
7672 @item @code{void __MWTACC (acc, uw1)}
7673 @tab @code{__MWTACC (@var{b}, @var{a})}
7674 @tab @code{MWTACC @var{a},@var{b}}
7675 @item @code{void __MWTACCG (acc, uw1)}
7676 @tab @code{__MWTACCG (@var{b}, @var{a})}
7677 @tab @code{MWTACCG @var{a},@var{b}}
7678 @item @code{uw1 __MXOR (uw1, uw1)}
7679 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
7680 @tab @code{MXOR @var{a},@var{b},@var{c}}
7683 @node Raw read/write Functions
7684 @subsubsection Raw read/write Functions
7686 This sections describes built-in functions related to read and write
7687 instructions to access memory. These functions generate
7688 @code{membar} instructions to flush the I/O load and stores where
7689 appropriate, as described in Fujitsu's manual described above.
7693 @item unsigned char __builtin_read8 (void *@var{data})
7694 @item unsigned short __builtin_read16 (void *@var{data})
7695 @item unsigned long __builtin_read32 (void *@var{data})
7696 @item unsigned long long __builtin_read64 (void *@var{data})
7698 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
7699 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
7700 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
7701 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
7704 @node Other Built-in Functions
7705 @subsubsection Other Built-in Functions
7707 This section describes built-in functions that are not named after
7708 a specific FR-V instruction.
7711 @item sw2 __IACCreadll (iacc @var{reg})
7712 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
7713 for future expansion and must be 0.
7715 @item sw1 __IACCreadl (iacc @var{reg})
7716 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
7717 Other values of @var{reg} are rejected as invalid.
7719 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
7720 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
7721 is reserved for future expansion and must be 0.
7723 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7724 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7725 is 1. Other values of @var{reg} are rejected as invalid.
7727 @item void __data_prefetch0 (const void *@var{x})
7728 Use the @code{dcpl} instruction to load the contents of address @var{x}
7729 into the data cache.
7731 @item void __data_prefetch (const void *@var{x})
7732 Use the @code{nldub} instruction to load the contents of address @var{x}
7733 into the data cache. The instruction will be issued in slot I1@.
7736 @node X86 Built-in Functions
7737 @subsection X86 Built-in Functions
7739 These built-in functions are available for the i386 and x86-64 family
7740 of computers, depending on the command-line switches used.
7742 Note that, if you specify command-line switches such as @option{-msse},
7743 the compiler could use the extended instruction sets even if the built-ins
7744 are not used explicitly in the program. For this reason, applications
7745 which perform runtime CPU detection must compile separate files for each
7746 supported architecture, using the appropriate flags. In particular,
7747 the file containing the CPU detection code should be compiled without
7750 The following machine modes are available for use with MMX built-in functions
7751 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7752 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7753 vector of eight 8-bit integers. Some of the built-in functions operate on
7754 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
7756 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7757 of two 32-bit floating point values.
7759 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7760 floating point values. Some instructions use a vector of four 32-bit
7761 integers, these use @code{V4SI}. Finally, some instructions operate on an
7762 entire vector register, interpreting it as a 128-bit integer, these use mode
7765 In 64-bit mode, the x86-64 family of processors uses additional built-in
7766 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7767 floating point and @code{TC} 128-bit complex floating point values.
7769 The following floating point built-in functions are available in 64-bit
7770 mode. All of them implement the function that is part of the name.
7773 __float128 __builtin_fabsq (__float128)
7774 __float128 __builtin_copysignq (__float128, __float128)
7777 The following floating point built-in functions are made available in the
7781 @item __float128 __builtin_infq (void)
7782 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7783 @findex __builtin_infq
7785 @item __float128 __builtin_huge_valq (void)
7786 Similar to @code{__builtin_huge_val}, except the return type is @code{__float128}.
7787 @findex __builtin_huge_valq
7790 The following built-in functions are made available by @option{-mmmx}.
7791 All of them generate the machine instruction that is part of the name.
7794 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7795 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7796 v2si __builtin_ia32_paddd (v2si, v2si)
7797 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7798 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7799 v2si __builtin_ia32_psubd (v2si, v2si)
7800 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7801 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7802 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7803 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7804 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7805 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7806 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7807 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7808 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7809 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7810 di __builtin_ia32_pand (di, di)
7811 di __builtin_ia32_pandn (di,di)
7812 di __builtin_ia32_por (di, di)
7813 di __builtin_ia32_pxor (di, di)
7814 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7815 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7816 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7817 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7818 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7819 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7820 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7821 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7822 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7823 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7824 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7825 v2si __builtin_ia32_punpckldq (v2si, v2si)
7826 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7827 v4hi __builtin_ia32_packssdw (v2si, v2si)
7828 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7830 v4hi __builtin_ia32_psllw (v4hi, v4hi)
7831 v2si __builtin_ia32_pslld (v2si, v2si)
7832 v1di __builtin_ia32_psllq (v1di, v1di)
7833 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
7834 v2si __builtin_ia32_psrld (v2si, v2si)
7835 v1di __builtin_ia32_psrlq (v1di, v1di)
7836 v4hi __builtin_ia32_psraw (v4hi, v4hi)
7837 v2si __builtin_ia32_psrad (v2si, v2si)
7838 v4hi __builtin_ia32_psllwi (v4hi, int)
7839 v2si __builtin_ia32_pslldi (v2si, int)
7840 v1di __builtin_ia32_psllqi (v1di, int)
7841 v4hi __builtin_ia32_psrlwi (v4hi, int)
7842 v2si __builtin_ia32_psrldi (v2si, int)
7843 v1di __builtin_ia32_psrlqi (v1di, int)
7844 v4hi __builtin_ia32_psrawi (v4hi, int)
7845 v2si __builtin_ia32_psradi (v2si, int)
7849 The following built-in functions are made available either with
7850 @option{-msse}, or with a combination of @option{-m3dnow} and
7851 @option{-march=athlon}. All of them generate the machine
7852 instruction that is part of the name.
7855 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7856 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7857 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7858 v1di __builtin_ia32_psadbw (v8qi, v8qi)
7859 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7860 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7861 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7862 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7863 int __builtin_ia32_pextrw (v4hi, int)
7864 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7865 int __builtin_ia32_pmovmskb (v8qi)
7866 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7867 void __builtin_ia32_movntq (di *, di)
7868 void __builtin_ia32_sfence (void)
7871 The following built-in functions are available when @option{-msse} is used.
7872 All of them generate the machine instruction that is part of the name.
7875 int __builtin_ia32_comieq (v4sf, v4sf)
7876 int __builtin_ia32_comineq (v4sf, v4sf)
7877 int __builtin_ia32_comilt (v4sf, v4sf)
7878 int __builtin_ia32_comile (v4sf, v4sf)
7879 int __builtin_ia32_comigt (v4sf, v4sf)
7880 int __builtin_ia32_comige (v4sf, v4sf)
7881 int __builtin_ia32_ucomieq (v4sf, v4sf)
7882 int __builtin_ia32_ucomineq (v4sf, v4sf)
7883 int __builtin_ia32_ucomilt (v4sf, v4sf)
7884 int __builtin_ia32_ucomile (v4sf, v4sf)
7885 int __builtin_ia32_ucomigt (v4sf, v4sf)
7886 int __builtin_ia32_ucomige (v4sf, v4sf)
7887 v4sf __builtin_ia32_addps (v4sf, v4sf)
7888 v4sf __builtin_ia32_subps (v4sf, v4sf)
7889 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7890 v4sf __builtin_ia32_divps (v4sf, v4sf)
7891 v4sf __builtin_ia32_addss (v4sf, v4sf)
7892 v4sf __builtin_ia32_subss (v4sf, v4sf)
7893 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7894 v4sf __builtin_ia32_divss (v4sf, v4sf)
7895 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7896 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7897 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7898 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7899 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7900 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7901 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7902 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7903 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7904 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7905 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7906 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7907 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7908 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7909 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7910 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7911 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7912 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7913 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7914 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7915 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7916 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7917 v4sf __builtin_ia32_minps (v4sf, v4sf)
7918 v4sf __builtin_ia32_minss (v4sf, v4sf)
7919 v4sf __builtin_ia32_andps (v4sf, v4sf)
7920 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7921 v4sf __builtin_ia32_orps (v4sf, v4sf)
7922 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7923 v4sf __builtin_ia32_movss (v4sf, v4sf)
7924 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7925 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7926 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7927 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7928 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7929 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7930 v2si __builtin_ia32_cvtps2pi (v4sf)
7931 int __builtin_ia32_cvtss2si (v4sf)
7932 v2si __builtin_ia32_cvttps2pi (v4sf)
7933 int __builtin_ia32_cvttss2si (v4sf)
7934 v4sf __builtin_ia32_rcpps (v4sf)
7935 v4sf __builtin_ia32_rsqrtps (v4sf)
7936 v4sf __builtin_ia32_sqrtps (v4sf)
7937 v4sf __builtin_ia32_rcpss (v4sf)
7938 v4sf __builtin_ia32_rsqrtss (v4sf)
7939 v4sf __builtin_ia32_sqrtss (v4sf)
7940 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7941 void __builtin_ia32_movntps (float *, v4sf)
7942 int __builtin_ia32_movmskps (v4sf)
7945 The following built-in functions are available when @option{-msse} is used.
7948 @item v4sf __builtin_ia32_loadaps (float *)
7949 Generates the @code{movaps} machine instruction as a load from memory.
7950 @item void __builtin_ia32_storeaps (float *, v4sf)
7951 Generates the @code{movaps} machine instruction as a store to memory.
7952 @item v4sf __builtin_ia32_loadups (float *)
7953 Generates the @code{movups} machine instruction as a load from memory.
7954 @item void __builtin_ia32_storeups (float *, v4sf)
7955 Generates the @code{movups} machine instruction as a store to memory.
7956 @item v4sf __builtin_ia32_loadsss (float *)
7957 Generates the @code{movss} machine instruction as a load from memory.
7958 @item void __builtin_ia32_storess (float *, v4sf)
7959 Generates the @code{movss} machine instruction as a store to memory.
7960 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
7961 Generates the @code{movhps} machine instruction as a load from memory.
7962 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
7963 Generates the @code{movlps} machine instruction as a load from memory
7964 @item void __builtin_ia32_storehps (v2sf *, v4sf)
7965 Generates the @code{movhps} machine instruction as a store to memory.
7966 @item void __builtin_ia32_storelps (v2sf *, v4sf)
7967 Generates the @code{movlps} machine instruction as a store to memory.
7970 The following built-in functions are available when @option{-msse2} is used.
7971 All of them generate the machine instruction that is part of the name.
7974 int __builtin_ia32_comisdeq (v2df, v2df)
7975 int __builtin_ia32_comisdlt (v2df, v2df)
7976 int __builtin_ia32_comisdle (v2df, v2df)
7977 int __builtin_ia32_comisdgt (v2df, v2df)
7978 int __builtin_ia32_comisdge (v2df, v2df)
7979 int __builtin_ia32_comisdneq (v2df, v2df)
7980 int __builtin_ia32_ucomisdeq (v2df, v2df)
7981 int __builtin_ia32_ucomisdlt (v2df, v2df)
7982 int __builtin_ia32_ucomisdle (v2df, v2df)
7983 int __builtin_ia32_ucomisdgt (v2df, v2df)
7984 int __builtin_ia32_ucomisdge (v2df, v2df)
7985 int __builtin_ia32_ucomisdneq (v2df, v2df)
7986 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7987 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7988 v2df __builtin_ia32_cmplepd (v2df, v2df)
7989 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7990 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7991 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7992 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7993 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7994 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7995 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7996 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7997 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7998 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7999 v2df __builtin_ia32_cmpltsd (v2df, v2df)
8000 v2df __builtin_ia32_cmplesd (v2df, v2df)
8001 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
8002 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
8003 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
8004 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
8005 v2df __builtin_ia32_cmpordsd (v2df, v2df)
8006 v2di __builtin_ia32_paddq (v2di, v2di)
8007 v2di __builtin_ia32_psubq (v2di, v2di)
8008 v2df __builtin_ia32_addpd (v2df, v2df)
8009 v2df __builtin_ia32_subpd (v2df, v2df)
8010 v2df __builtin_ia32_mulpd (v2df, v2df)
8011 v2df __builtin_ia32_divpd (v2df, v2df)
8012 v2df __builtin_ia32_addsd (v2df, v2df)
8013 v2df __builtin_ia32_subsd (v2df, v2df)
8014 v2df __builtin_ia32_mulsd (v2df, v2df)
8015 v2df __builtin_ia32_divsd (v2df, v2df)
8016 v2df __builtin_ia32_minpd (v2df, v2df)
8017 v2df __builtin_ia32_maxpd (v2df, v2df)
8018 v2df __builtin_ia32_minsd (v2df, v2df)
8019 v2df __builtin_ia32_maxsd (v2df, v2df)
8020 v2df __builtin_ia32_andpd (v2df, v2df)
8021 v2df __builtin_ia32_andnpd (v2df, v2df)
8022 v2df __builtin_ia32_orpd (v2df, v2df)
8023 v2df __builtin_ia32_xorpd (v2df, v2df)
8024 v2df __builtin_ia32_movsd (v2df, v2df)
8025 v2df __builtin_ia32_unpckhpd (v2df, v2df)
8026 v2df __builtin_ia32_unpcklpd (v2df, v2df)
8027 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
8028 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
8029 v4si __builtin_ia32_paddd128 (v4si, v4si)
8030 v2di __builtin_ia32_paddq128 (v2di, v2di)
8031 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
8032 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
8033 v4si __builtin_ia32_psubd128 (v4si, v4si)
8034 v2di __builtin_ia32_psubq128 (v2di, v2di)
8035 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
8036 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
8037 v2di __builtin_ia32_pand128 (v2di, v2di)
8038 v2di __builtin_ia32_pandn128 (v2di, v2di)
8039 v2di __builtin_ia32_por128 (v2di, v2di)
8040 v2di __builtin_ia32_pxor128 (v2di, v2di)
8041 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
8042 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
8043 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
8044 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
8045 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
8046 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
8047 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
8048 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
8049 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
8050 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
8051 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
8052 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8053 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8054 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8055 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8056 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8057 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8058 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8059 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8060 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8061 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8062 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8063 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8064 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8065 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8066 v2df __builtin_ia32_loadupd (double *)
8067 void __builtin_ia32_storeupd (double *, v2df)
8068 v2df __builtin_ia32_loadhpd (v2df, double const *)
8069 v2df __builtin_ia32_loadlpd (v2df, double const *)
8070 int __builtin_ia32_movmskpd (v2df)
8071 int __builtin_ia32_pmovmskb128 (v16qi)
8072 void __builtin_ia32_movnti (int *, int)
8073 void __builtin_ia32_movntpd (double *, v2df)
8074 void __builtin_ia32_movntdq (v2df *, v2df)
8075 v4si __builtin_ia32_pshufd (v4si, int)
8076 v8hi __builtin_ia32_pshuflw (v8hi, int)
8077 v8hi __builtin_ia32_pshufhw (v8hi, int)
8078 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
8079 v2df __builtin_ia32_sqrtpd (v2df)
8080 v2df __builtin_ia32_sqrtsd (v2df)
8081 v2df __builtin_ia32_shufpd (v2df, v2df, int)
8082 v2df __builtin_ia32_cvtdq2pd (v4si)
8083 v4sf __builtin_ia32_cvtdq2ps (v4si)
8084 v4si __builtin_ia32_cvtpd2dq (v2df)
8085 v2si __builtin_ia32_cvtpd2pi (v2df)
8086 v4sf __builtin_ia32_cvtpd2ps (v2df)
8087 v4si __builtin_ia32_cvttpd2dq (v2df)
8088 v2si __builtin_ia32_cvttpd2pi (v2df)
8089 v2df __builtin_ia32_cvtpi2pd (v2si)
8090 int __builtin_ia32_cvtsd2si (v2df)
8091 int __builtin_ia32_cvttsd2si (v2df)
8092 long long __builtin_ia32_cvtsd2si64 (v2df)
8093 long long __builtin_ia32_cvttsd2si64 (v2df)
8094 v4si __builtin_ia32_cvtps2dq (v4sf)
8095 v2df __builtin_ia32_cvtps2pd (v4sf)
8096 v4si __builtin_ia32_cvttps2dq (v4sf)
8097 v2df __builtin_ia32_cvtsi2sd (v2df, int)
8098 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
8099 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
8100 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
8101 void __builtin_ia32_clflush (const void *)
8102 void __builtin_ia32_lfence (void)
8103 void __builtin_ia32_mfence (void)
8104 v16qi __builtin_ia32_loaddqu (const char *)
8105 void __builtin_ia32_storedqu (char *, v16qi)
8106 v1di __builtin_ia32_pmuludq (v2si, v2si)
8107 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
8108 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
8109 v4si __builtin_ia32_pslld128 (v4si, v4si)
8110 v2di __builtin_ia32_psllq128 (v2di, v2di)
8111 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
8112 v4si __builtin_ia32_psrld128 (v4si, v4si)
8113 v2di __builtin_ia32_psrlq128 (v2di, v2di)
8114 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
8115 v4si __builtin_ia32_psrad128 (v4si, v4si)
8116 v2di __builtin_ia32_pslldqi128 (v2di, int)
8117 v8hi __builtin_ia32_psllwi128 (v8hi, int)
8118 v4si __builtin_ia32_pslldi128 (v4si, int)
8119 v2di __builtin_ia32_psllqi128 (v2di, int)
8120 v2di __builtin_ia32_psrldqi128 (v2di, int)
8121 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
8122 v4si __builtin_ia32_psrldi128 (v4si, int)
8123 v2di __builtin_ia32_psrlqi128 (v2di, int)
8124 v8hi __builtin_ia32_psrawi128 (v8hi, int)
8125 v4si __builtin_ia32_psradi128 (v4si, int)
8126 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
8127 v2di __builtin_ia32_movq128 (v2di)
8130 The following built-in functions are available when @option{-msse3} is used.
8131 All of them generate the machine instruction that is part of the name.
8134 v2df __builtin_ia32_addsubpd (v2df, v2df)
8135 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
8136 v2df __builtin_ia32_haddpd (v2df, v2df)
8137 v4sf __builtin_ia32_haddps (v4sf, v4sf)
8138 v2df __builtin_ia32_hsubpd (v2df, v2df)
8139 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
8140 v16qi __builtin_ia32_lddqu (char const *)
8141 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
8142 v2df __builtin_ia32_movddup (v2df)
8143 v4sf __builtin_ia32_movshdup (v4sf)
8144 v4sf __builtin_ia32_movsldup (v4sf)
8145 void __builtin_ia32_mwait (unsigned int, unsigned int)
8148 The following built-in functions are available when @option{-msse3} is used.
8151 @item v2df __builtin_ia32_loadddup (double const *)
8152 Generates the @code{movddup} machine instruction as a load from memory.
8155 The following built-in functions are available when @option{-mssse3} is used.
8156 All of them generate the machine instruction that is part of the name
8160 v2si __builtin_ia32_phaddd (v2si, v2si)
8161 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
8162 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
8163 v2si __builtin_ia32_phsubd (v2si, v2si)
8164 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
8165 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
8166 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
8167 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
8168 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
8169 v8qi __builtin_ia32_psignb (v8qi, v8qi)
8170 v2si __builtin_ia32_psignd (v2si, v2si)
8171 v4hi __builtin_ia32_psignw (v4hi, v4hi)
8172 v1di __builtin_ia32_palignr (v1di, v1di, int)
8173 v8qi __builtin_ia32_pabsb (v8qi)
8174 v2si __builtin_ia32_pabsd (v2si)
8175 v4hi __builtin_ia32_pabsw (v4hi)
8178 The following built-in functions are available when @option{-mssse3} is used.
8179 All of them generate the machine instruction that is part of the name
8183 v4si __builtin_ia32_phaddd128 (v4si, v4si)
8184 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
8185 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
8186 v4si __builtin_ia32_phsubd128 (v4si, v4si)
8187 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
8188 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
8189 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
8190 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
8191 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
8192 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
8193 v4si __builtin_ia32_psignd128 (v4si, v4si)
8194 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
8195 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
8196 v16qi __builtin_ia32_pabsb128 (v16qi)
8197 v4si __builtin_ia32_pabsd128 (v4si)
8198 v8hi __builtin_ia32_pabsw128 (v8hi)
8201 The following built-in functions are available when @option{-msse4.1} is
8202 used. All of them generate the machine instruction that is part of the
8206 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
8207 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
8208 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
8209 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
8210 v2df __builtin_ia32_dppd (v2df, v2df, const int)
8211 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
8212 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
8213 v2di __builtin_ia32_movntdqa (v2di *);
8214 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
8215 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
8216 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
8217 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
8218 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
8219 v8hi __builtin_ia32_phminposuw128 (v8hi)
8220 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
8221 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
8222 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
8223 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
8224 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
8225 v4si __builtin_ia32_pminsd128 (v4si, v4si)
8226 v4si __builtin_ia32_pminud128 (v4si, v4si)
8227 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
8228 v4si __builtin_ia32_pmovsxbd128 (v16qi)
8229 v2di __builtin_ia32_pmovsxbq128 (v16qi)
8230 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
8231 v2di __builtin_ia32_pmovsxdq128 (v4si)
8232 v4si __builtin_ia32_pmovsxwd128 (v8hi)
8233 v2di __builtin_ia32_pmovsxwq128 (v8hi)
8234 v4si __builtin_ia32_pmovzxbd128 (v16qi)
8235 v2di __builtin_ia32_pmovzxbq128 (v16qi)
8236 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
8237 v2di __builtin_ia32_pmovzxdq128 (v4si)
8238 v4si __builtin_ia32_pmovzxwd128 (v8hi)
8239 v2di __builtin_ia32_pmovzxwq128 (v8hi)
8240 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
8241 v4si __builtin_ia32_pmulld128 (v4si, v4si)
8242 int __builtin_ia32_ptestc128 (v2di, v2di)
8243 int __builtin_ia32_ptestnzc128 (v2di, v2di)
8244 int __builtin_ia32_ptestz128 (v2di, v2di)
8245 v2df __builtin_ia32_roundpd (v2df, const int)
8246 v4sf __builtin_ia32_roundps (v4sf, const int)
8247 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
8248 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
8251 The following built-in functions are available when @option{-msse4.1} is
8255 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
8256 Generates the @code{insertps} machine instruction.
8257 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
8258 Generates the @code{pextrb} machine instruction.
8259 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
8260 Generates the @code{pinsrb} machine instruction.
8261 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
8262 Generates the @code{pinsrd} machine instruction.
8263 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
8264 Generates the @code{pinsrq} machine instruction in 64bit mode.
8267 The following built-in functions are changed to generate new SSE4.1
8268 instructions when @option{-msse4.1} is used.
8271 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
8272 Generates the @code{extractps} machine instruction.
8273 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
8274 Generates the @code{pextrd} machine instruction.
8275 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
8276 Generates the @code{pextrq} machine instruction in 64bit mode.
8279 The following built-in functions are available when @option{-msse4.2} is
8280 used. All of them generate the machine instruction that is part of the
8284 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
8285 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
8286 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
8287 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
8288 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
8289 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
8290 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
8291 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
8292 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
8293 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
8294 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
8295 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
8296 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
8297 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
8298 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
8301 The following built-in functions are available when @option{-msse4.2} is
8305 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
8306 Generates the @code{crc32b} machine instruction.
8307 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
8308 Generates the @code{crc32w} machine instruction.
8309 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
8310 Generates the @code{crc32l} machine instruction.
8311 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
8314 The following built-in functions are changed to generate new SSE4.2
8315 instructions when @option{-msse4.2} is used.
8318 @item int __builtin_popcount (unsigned int)
8319 Generates the @code{popcntl} machine instruction.
8320 @item int __builtin_popcountl (unsigned long)
8321 Generates the @code{popcntl} or @code{popcntq} machine instruction,
8322 depending on the size of @code{unsigned long}.
8323 @item int __builtin_popcountll (unsigned long long)
8324 Generates the @code{popcntq} machine instruction.
8327 The following built-in functions are available when @option{-mavx} is
8328 used. All of them generate the machine instruction that is part of the
8332 v4df __builtin_ia32_addpd256 (v4df,v4df)
8333 v8sf __builtin_ia32_addps256 (v8sf,v8sf)
8334 v4df __builtin_ia32_addsubpd256 (v4df,v4df)
8335 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf)
8336 v4df __builtin_ia32_andnpd256 (v4df,v4df)
8337 v8sf __builtin_ia32_andnps256 (v8sf,v8sf)
8338 v4df __builtin_ia32_andpd256 (v4df,v4df)
8339 v8sf __builtin_ia32_andps256 (v8sf,v8sf)
8340 v4df __builtin_ia32_blendpd256 (v4df,v4df,int)
8341 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int)
8342 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df)
8343 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf)
8344 v2df __builtin_ia32_cmppd (v2df,v2df,int)
8345 v4df __builtin_ia32_cmppd256 (v4df,v4df,int)
8346 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int)
8347 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int)
8348 v2df __builtin_ia32_cmpsd (v2df,v2df,int)
8349 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int)
8350 v4df __builtin_ia32_cvtdq2pd256 (v4si)
8351 v8sf __builtin_ia32_cvtdq2ps256 (v8si)
8352 v4si __builtin_ia32_cvtpd2dq256 (v4df)
8353 v4sf __builtin_ia32_cvtpd2ps256 (v4df)
8354 v8si __builtin_ia32_cvtps2dq256 (v8sf)
8355 v4df __builtin_ia32_cvtps2pd256 (v4sf)
8356 v4si __builtin_ia32_cvttpd2dq256 (v4df)
8357 v8si __builtin_ia32_cvttps2dq256 (v8sf)
8358 v4df __builtin_ia32_divpd256 (v4df,v4df)
8359 v8sf __builtin_ia32_divps256 (v8sf,v8sf)
8360 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int)
8361 v4df __builtin_ia32_haddpd256 (v4df,v4df)
8362 v8sf __builtin_ia32_haddps256 (v8sf,v8sf)
8363 v4df __builtin_ia32_hsubpd256 (v4df,v4df)
8364 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf)
8365 v32qi __builtin_ia32_lddqu256 (pcchar)
8366 v32qi __builtin_ia32_loaddqu256 (pcchar)
8367 v4df __builtin_ia32_loadupd256 (pcdouble)
8368 v8sf __builtin_ia32_loadups256 (pcfloat)
8369 v2df __builtin_ia32_maskloadpd (pcv2df,v2df)
8370 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df)
8371 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf)
8372 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf)
8373 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df)
8374 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df)
8375 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf)
8376 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf)
8377 v4df __builtin_ia32_maxpd256 (v4df,v4df)
8378 v8sf __builtin_ia32_maxps256 (v8sf,v8sf)
8379 v4df __builtin_ia32_minpd256 (v4df,v4df)
8380 v8sf __builtin_ia32_minps256 (v8sf,v8sf)
8381 v4df __builtin_ia32_movddup256 (v4df)
8382 int __builtin_ia32_movmskpd256 (v4df)
8383 int __builtin_ia32_movmskps256 (v8sf)
8384 v8sf __builtin_ia32_movshdup256 (v8sf)
8385 v8sf __builtin_ia32_movsldup256 (v8sf)
8386 v4df __builtin_ia32_mulpd256 (v4df,v4df)
8387 v8sf __builtin_ia32_mulps256 (v8sf,v8sf)
8388 v4df __builtin_ia32_orpd256 (v4df,v4df)
8389 v8sf __builtin_ia32_orps256 (v8sf,v8sf)
8390 v2df __builtin_ia32_pd_pd256 (v4df)
8391 v4df __builtin_ia32_pd256_pd (v2df)
8392 v4sf __builtin_ia32_ps_ps256 (v8sf)
8393 v8sf __builtin_ia32_ps256_ps (v4sf)
8394 int __builtin_ia32_ptestc256 (v4di,v4di,ptest)
8395 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest)
8396 int __builtin_ia32_ptestz256 (v4di,v4di,ptest)
8397 v8sf __builtin_ia32_rcpps256 (v8sf)
8398 v4df __builtin_ia32_roundpd256 (v4df,int)
8399 v8sf __builtin_ia32_roundps256 (v8sf,int)
8400 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf)
8401 v8sf __builtin_ia32_rsqrtps256 (v8sf)
8402 v4df __builtin_ia32_shufpd256 (v4df,v4df,int)
8403 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int)
8404 v4si __builtin_ia32_si_si256 (v8si)
8405 v8si __builtin_ia32_si256_si (v4si)
8406 v4df __builtin_ia32_sqrtpd256 (v4df)
8407 v8sf __builtin_ia32_sqrtps_nr256 (v8sf)
8408 v8sf __builtin_ia32_sqrtps256 (v8sf)
8409 void __builtin_ia32_storedqu256 (pchar,v32qi)
8410 void __builtin_ia32_storeupd256 (pdouble,v4df)
8411 void __builtin_ia32_storeups256 (pfloat,v8sf)
8412 v4df __builtin_ia32_subpd256 (v4df,v4df)
8413 v8sf __builtin_ia32_subps256 (v8sf,v8sf)
8414 v4df __builtin_ia32_unpckhpd256 (v4df,v4df)
8415 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf)
8416 v4df __builtin_ia32_unpcklpd256 (v4df,v4df)
8417 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf)
8418 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df)
8419 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf)
8420 v4df __builtin_ia32_vbroadcastsd256 (pcdouble)
8421 v4sf __builtin_ia32_vbroadcastss (pcfloat)
8422 v8sf __builtin_ia32_vbroadcastss256 (pcfloat)
8423 v2df __builtin_ia32_vextractf128_pd256 (v4df,int)
8424 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int)
8425 v4si __builtin_ia32_vextractf128_si256 (v8si,int)
8426 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int)
8427 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int)
8428 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int)
8429 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int)
8430 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int)
8431 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int)
8432 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int)
8433 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int)
8434 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int)
8435 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int)
8436 v2df __builtin_ia32_vpermilpd (v2df,int)
8437 v4df __builtin_ia32_vpermilpd256 (v4df,int)
8438 v4sf __builtin_ia32_vpermilps (v4sf,int)
8439 v8sf __builtin_ia32_vpermilps256 (v8sf,int)
8440 v2df __builtin_ia32_vpermilvarpd (v2df,v2di)
8441 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di)
8442 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si)
8443 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si)
8444 int __builtin_ia32_vtestcpd (v2df,v2df,ptest)
8445 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest)
8446 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest)
8447 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest)
8448 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest)
8449 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest)
8450 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest)
8451 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest)
8452 int __builtin_ia32_vtestzpd (v2df,v2df,ptest)
8453 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest)
8454 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest)
8455 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest)
8456 void __builtin_ia32_vzeroall (void)
8457 void __builtin_ia32_vzeroupper (void)
8458 v4df __builtin_ia32_xorpd256 (v4df,v4df)
8459 v8sf __builtin_ia32_xorps256 (v8sf,v8sf)
8462 The following built-in functions are available when @option{-maes} is
8463 used. All of them generate the machine instruction that is part of the
8467 v2di __builtin_ia32_aesenc128 (v2di, v2di)
8468 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
8469 v2di __builtin_ia32_aesdec128 (v2di, v2di)
8470 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
8471 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
8472 v2di __builtin_ia32_aesimc128 (v2di)
8475 The following built-in function is available when @option{-mpclmul} is
8479 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
8480 Generates the @code{pclmulqdq} machine instruction.
8483 The following built-in functions are available when @option{-msse4a} is used.
8484 All of them generate the machine instruction that is part of the name.
8487 void __builtin_ia32_movntsd (double *, v2df)
8488 void __builtin_ia32_movntss (float *, v4sf)
8489 v2di __builtin_ia32_extrq (v2di, v16qi)
8490 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
8491 v2di __builtin_ia32_insertq (v2di, v2di)
8492 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
8495 The following built-in functions are available when @option{-msse5} is used.
8496 All of them generate the machine instruction that is part of the name
8500 v2df __builtin_ia32_comeqpd (v2df, v2df)
8501 v2df __builtin_ia32_comeqps (v2df, v2df)
8502 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
8503 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
8504 v2df __builtin_ia32_comfalsepd (v2df, v2df)
8505 v2df __builtin_ia32_comfalseps (v2df, v2df)
8506 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
8507 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
8508 v2df __builtin_ia32_comgepd (v2df, v2df)
8509 v2df __builtin_ia32_comgeps (v2df, v2df)
8510 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
8511 v4sf __builtin_ia32_comgess (v4sf, v4sf)
8512 v2df __builtin_ia32_comgtpd (v2df, v2df)
8513 v2df __builtin_ia32_comgtps (v2df, v2df)
8514 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
8515 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
8516 v2df __builtin_ia32_comlepd (v2df, v2df)
8517 v2df __builtin_ia32_comleps (v2df, v2df)
8518 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
8519 v4sf __builtin_ia32_comless (v4sf, v4sf)
8520 v2df __builtin_ia32_comltpd (v2df, v2df)
8521 v2df __builtin_ia32_comltps (v2df, v2df)
8522 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
8523 v4sf __builtin_ia32_comltss (v4sf, v4sf)
8524 v2df __builtin_ia32_comnepd (v2df, v2df)
8525 v2df __builtin_ia32_comneps (v2df, v2df)
8526 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
8527 v4sf __builtin_ia32_comness (v4sf, v4sf)
8528 v2df __builtin_ia32_comordpd (v2df, v2df)
8529 v2df __builtin_ia32_comordps (v2df, v2df)
8530 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
8531 v4sf __builtin_ia32_comordss (v4sf, v4sf)
8532 v2df __builtin_ia32_comtruepd (v2df, v2df)
8533 v2df __builtin_ia32_comtrueps (v2df, v2df)
8534 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
8535 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
8536 v2df __builtin_ia32_comueqpd (v2df, v2df)
8537 v2df __builtin_ia32_comueqps (v2df, v2df)
8538 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
8539 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
8540 v2df __builtin_ia32_comugepd (v2df, v2df)
8541 v2df __builtin_ia32_comugeps (v2df, v2df)
8542 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
8543 v4sf __builtin_ia32_comugess (v4sf, v4sf)
8544 v2df __builtin_ia32_comugtpd (v2df, v2df)
8545 v2df __builtin_ia32_comugtps (v2df, v2df)
8546 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
8547 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
8548 v2df __builtin_ia32_comulepd (v2df, v2df)
8549 v2df __builtin_ia32_comuleps (v2df, v2df)
8550 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
8551 v4sf __builtin_ia32_comuless (v4sf, v4sf)
8552 v2df __builtin_ia32_comultpd (v2df, v2df)
8553 v2df __builtin_ia32_comultps (v2df, v2df)
8554 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
8555 v4sf __builtin_ia32_comultss (v4sf, v4sf)
8556 v2df __builtin_ia32_comunepd (v2df, v2df)
8557 v2df __builtin_ia32_comuneps (v2df, v2df)
8558 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
8559 v4sf __builtin_ia32_comuness (v4sf, v4sf)
8560 v2df __builtin_ia32_comunordpd (v2df, v2df)
8561 v2df __builtin_ia32_comunordps (v2df, v2df)
8562 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
8563 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
8564 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
8565 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
8566 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
8567 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
8568 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
8569 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
8570 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
8571 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
8572 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
8573 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
8574 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
8575 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
8576 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
8577 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
8578 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
8579 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
8580 v2df __builtin_ia32_frczpd (v2df)
8581 v4sf __builtin_ia32_frczps (v4sf)
8582 v2df __builtin_ia32_frczsd (v2df, v2df)
8583 v4sf __builtin_ia32_frczss (v4sf, v4sf)
8584 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
8585 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
8586 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
8587 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
8588 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
8589 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
8590 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
8591 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
8592 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8593 v4si __builtin_ia32_pcomeqd (v4si, v4si)
8594 v2di __builtin_ia32_pcomeqq (v2di, v2di)
8595 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
8596 v4si __builtin_ia32_pcomequd (v4si, v4si)
8597 v2di __builtin_ia32_pcomequq (v2di, v2di)
8598 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
8599 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8600 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
8601 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
8602 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
8603 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
8604 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
8605 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
8606 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
8607 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
8608 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
8609 v4si __builtin_ia32_pcomged (v4si, v4si)
8610 v2di __builtin_ia32_pcomgeq (v2di, v2di)
8611 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
8612 v4si __builtin_ia32_pcomgeud (v4si, v4si)
8613 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
8614 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
8615 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
8616 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
8617 v4si __builtin_ia32_pcomgtd (v4si, v4si)
8618 v2di __builtin_ia32_pcomgtq (v2di, v2di)
8619 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
8620 v4si __builtin_ia32_pcomgtud (v4si, v4si)
8621 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
8622 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
8623 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
8624 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
8625 v4si __builtin_ia32_pcomled (v4si, v4si)
8626 v2di __builtin_ia32_pcomleq (v2di, v2di)
8627 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
8628 v4si __builtin_ia32_pcomleud (v4si, v4si)
8629 v2di __builtin_ia32_pcomleuq (v2di, v2di)
8630 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
8631 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
8632 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
8633 v4si __builtin_ia32_pcomltd (v4si, v4si)
8634 v2di __builtin_ia32_pcomltq (v2di, v2di)
8635 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
8636 v4si __builtin_ia32_pcomltud (v4si, v4si)
8637 v2di __builtin_ia32_pcomltuq (v2di, v2di)
8638 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
8639 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
8640 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
8641 v4si __builtin_ia32_pcomned (v4si, v4si)
8642 v2di __builtin_ia32_pcomneq (v2di, v2di)
8643 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
8644 v4si __builtin_ia32_pcomneud (v4si, v4si)
8645 v2di __builtin_ia32_pcomneuq (v2di, v2di)
8646 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
8647 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
8648 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
8649 v4si __builtin_ia32_pcomtrued (v4si, v4si)
8650 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
8651 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
8652 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
8653 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
8654 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
8655 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
8656 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
8657 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
8658 v4si __builtin_ia32_phaddbd (v16qi)
8659 v2di __builtin_ia32_phaddbq (v16qi)
8660 v8hi __builtin_ia32_phaddbw (v16qi)
8661 v2di __builtin_ia32_phadddq (v4si)
8662 v4si __builtin_ia32_phaddubd (v16qi)
8663 v2di __builtin_ia32_phaddubq (v16qi)
8664 v8hi __builtin_ia32_phaddubw (v16qi)
8665 v2di __builtin_ia32_phaddudq (v4si)
8666 v4si __builtin_ia32_phadduwd (v8hi)
8667 v2di __builtin_ia32_phadduwq (v8hi)
8668 v4si __builtin_ia32_phaddwd (v8hi)
8669 v2di __builtin_ia32_phaddwq (v8hi)
8670 v8hi __builtin_ia32_phsubbw (v16qi)
8671 v2di __builtin_ia32_phsubdq (v4si)
8672 v4si __builtin_ia32_phsubwd (v8hi)
8673 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
8674 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
8675 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
8676 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
8677 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
8678 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
8679 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
8680 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
8681 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
8682 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
8683 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
8684 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
8685 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
8686 v16qi __builtin_ia32_protb (v16qi, v16qi)
8687 v4si __builtin_ia32_protd (v4si, v4si)
8688 v2di __builtin_ia32_protq (v2di, v2di)
8689 v8hi __builtin_ia32_protw (v8hi, v8hi)
8690 v16qi __builtin_ia32_pshab (v16qi, v16qi)
8691 v4si __builtin_ia32_pshad (v4si, v4si)
8692 v2di __builtin_ia32_pshaq (v2di, v2di)
8693 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
8694 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
8695 v4si __builtin_ia32_pshld (v4si, v4si)
8696 v2di __builtin_ia32_pshlq (v2di, v2di)
8697 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
8700 The following builtin-in functions are available when @option{-msse5}
8701 is used. The second argument must be an integer constant and generate
8702 the machine instruction that is part of the name with the @samp{_imm}
8706 v16qi __builtin_ia32_protb_imm (v16qi, int)
8707 v4si __builtin_ia32_protd_imm (v4si, int)
8708 v2di __builtin_ia32_protq_imm (v2di, int)
8709 v8hi __builtin_ia32_protw_imm (v8hi, int)
8712 The following built-in functions are available when @option{-m3dnow} is used.
8713 All of them generate the machine instruction that is part of the name.
8716 void __builtin_ia32_femms (void)
8717 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
8718 v2si __builtin_ia32_pf2id (v2sf)
8719 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
8720 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
8721 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
8722 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
8723 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
8724 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
8725 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
8726 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
8727 v2sf __builtin_ia32_pfrcp (v2sf)
8728 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
8729 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
8730 v2sf __builtin_ia32_pfrsqrt (v2sf)
8731 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
8732 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
8733 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
8734 v2sf __builtin_ia32_pi2fd (v2si)
8735 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
8738 The following built-in functions are available when both @option{-m3dnow}
8739 and @option{-march=athlon} are used. All of them generate the machine
8740 instruction that is part of the name.
8743 v2si __builtin_ia32_pf2iw (v2sf)
8744 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
8745 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
8746 v2sf __builtin_ia32_pi2fw (v2si)
8747 v2sf __builtin_ia32_pswapdsf (v2sf)
8748 v2si __builtin_ia32_pswapdsi (v2si)
8751 @node MIPS DSP Built-in Functions
8752 @subsection MIPS DSP Built-in Functions
8754 The MIPS DSP Application-Specific Extension (ASE) includes new
8755 instructions that are designed to improve the performance of DSP and
8756 media applications. It provides instructions that operate on packed
8757 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
8759 GCC supports MIPS DSP operations using both the generic
8760 vector extensions (@pxref{Vector Extensions}) and a collection of
8761 MIPS-specific built-in functions. Both kinds of support are
8762 enabled by the @option{-mdsp} command-line option.
8764 Revision 2 of the ASE was introduced in the second half of 2006.
8765 This revision adds extra instructions to the original ASE, but is
8766 otherwise backwards-compatible with it. You can select revision 2
8767 using the command-line option @option{-mdspr2}; this option implies
8770 The SCOUNT and POS bits of the DSP control register are global. The
8771 WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and
8772 POS bits. During optimization, the compiler will not delete these
8773 instructions and it will not delete calls to functions containing
8776 At present, GCC only provides support for operations on 32-bit
8777 vectors. The vector type associated with 8-bit integer data is
8778 usually called @code{v4i8}, the vector type associated with Q7
8779 is usually called @code{v4q7}, the vector type associated with 16-bit
8780 integer data is usually called @code{v2i16}, and the vector type
8781 associated with Q15 is usually called @code{v2q15}. They can be
8782 defined in C as follows:
8785 typedef signed char v4i8 __attribute__ ((vector_size(4)));
8786 typedef signed char v4q7 __attribute__ ((vector_size(4)));
8787 typedef short v2i16 __attribute__ ((vector_size(4)));
8788 typedef short v2q15 __attribute__ ((vector_size(4)));
8791 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
8792 initialized in the same way as aggregates. For example:
8795 v4i8 a = @{1, 2, 3, 4@};
8797 b = (v4i8) @{5, 6, 7, 8@};
8799 v2q15 c = @{0x0fcb, 0x3a75@};
8801 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
8804 @emph{Note:} The CPU's endianness determines the order in which values
8805 are packed. On little-endian targets, the first value is the least
8806 significant and the last value is the most significant. The opposite
8807 order applies to big-endian targets. For example, the code above will
8808 set the lowest byte of @code{a} to @code{1} on little-endian targets
8809 and @code{4} on big-endian targets.
8811 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
8812 representation. As shown in this example, the integer representation
8813 of a Q7 value can be obtained by multiplying the fractional value by
8814 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
8815 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
8818 The table below lists the @code{v4i8} and @code{v2q15} operations for which
8819 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
8820 and @code{c} and @code{d} are @code{v2q15} values.
8822 @multitable @columnfractions .50 .50
8823 @item C code @tab MIPS instruction
8824 @item @code{a + b} @tab @code{addu.qb}
8825 @item @code{c + d} @tab @code{addq.ph}
8826 @item @code{a - b} @tab @code{subu.qb}
8827 @item @code{c - d} @tab @code{subq.ph}
8830 The table below lists the @code{v2i16} operation for which
8831 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
8832 @code{v2i16} values.
8834 @multitable @columnfractions .50 .50
8835 @item C code @tab MIPS instruction
8836 @item @code{e * f} @tab @code{mul.ph}
8839 It is easier to describe the DSP built-in functions if we first define
8840 the following types:
8845 typedef unsigned int ui32;
8846 typedef long long a64;
8849 @code{q31} and @code{i32} are actually the same as @code{int}, but we
8850 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
8851 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
8852 @code{long long}, but we use @code{a64} to indicate values that will
8853 be placed in one of the four DSP accumulators (@code{$ac0},
8854 @code{$ac1}, @code{$ac2} or @code{$ac3}).
8856 Also, some built-in functions prefer or require immediate numbers as
8857 parameters, because the corresponding DSP instructions accept both immediate
8858 numbers and register operands, or accept immediate numbers only. The
8859 immediate parameters are listed as follows.
8868 imm_n32_31: -32 to 31.
8869 imm_n512_511: -512 to 511.
8872 The following built-in functions map directly to a particular MIPS DSP
8873 instruction. Please refer to the architecture specification
8874 for details on what each instruction does.
8877 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
8878 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
8879 q31 __builtin_mips_addq_s_w (q31, q31)
8880 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
8881 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
8882 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
8883 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
8884 q31 __builtin_mips_subq_s_w (q31, q31)
8885 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
8886 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
8887 i32 __builtin_mips_addsc (i32, i32)
8888 i32 __builtin_mips_addwc (i32, i32)
8889 i32 __builtin_mips_modsub (i32, i32)
8890 i32 __builtin_mips_raddu_w_qb (v4i8)
8891 v2q15 __builtin_mips_absq_s_ph (v2q15)
8892 q31 __builtin_mips_absq_s_w (q31)
8893 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
8894 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
8895 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
8896 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
8897 q31 __builtin_mips_preceq_w_phl (v2q15)
8898 q31 __builtin_mips_preceq_w_phr (v2q15)
8899 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
8900 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
8901 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
8902 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
8903 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
8904 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
8905 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
8906 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
8907 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
8908 v4i8 __builtin_mips_shll_qb (v4i8, i32)
8909 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
8910 v2q15 __builtin_mips_shll_ph (v2q15, i32)
8911 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
8912 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
8913 q31 __builtin_mips_shll_s_w (q31, imm0_31)
8914 q31 __builtin_mips_shll_s_w (q31, i32)
8915 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
8916 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
8917 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
8918 v2q15 __builtin_mips_shra_ph (v2q15, i32)
8919 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
8920 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
8921 q31 __builtin_mips_shra_r_w (q31, imm0_31)
8922 q31 __builtin_mips_shra_r_w (q31, i32)
8923 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
8924 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
8925 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
8926 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
8927 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
8928 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
8929 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
8930 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
8931 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
8932 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
8933 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
8934 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
8935 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
8936 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
8937 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
8938 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
8939 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
8940 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
8941 i32 __builtin_mips_bitrev (i32)
8942 i32 __builtin_mips_insv (i32, i32)
8943 v4i8 __builtin_mips_repl_qb (imm0_255)
8944 v4i8 __builtin_mips_repl_qb (i32)
8945 v2q15 __builtin_mips_repl_ph (imm_n512_511)
8946 v2q15 __builtin_mips_repl_ph (i32)
8947 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
8948 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
8949 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
8950 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
8951 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
8952 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
8953 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
8954 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
8955 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
8956 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
8957 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
8958 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
8959 i32 __builtin_mips_extr_w (a64, imm0_31)
8960 i32 __builtin_mips_extr_w (a64, i32)
8961 i32 __builtin_mips_extr_r_w (a64, imm0_31)
8962 i32 __builtin_mips_extr_s_h (a64, i32)
8963 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
8964 i32 __builtin_mips_extr_rs_w (a64, i32)
8965 i32 __builtin_mips_extr_s_h (a64, imm0_31)
8966 i32 __builtin_mips_extr_r_w (a64, i32)
8967 i32 __builtin_mips_extp (a64, imm0_31)
8968 i32 __builtin_mips_extp (a64, i32)
8969 i32 __builtin_mips_extpdp (a64, imm0_31)
8970 i32 __builtin_mips_extpdp (a64, i32)
8971 a64 __builtin_mips_shilo (a64, imm_n32_31)
8972 a64 __builtin_mips_shilo (a64, i32)
8973 a64 __builtin_mips_mthlip (a64, i32)
8974 void __builtin_mips_wrdsp (i32, imm0_63)
8975 i32 __builtin_mips_rddsp (imm0_63)
8976 i32 __builtin_mips_lbux (void *, i32)
8977 i32 __builtin_mips_lhx (void *, i32)
8978 i32 __builtin_mips_lwx (void *, i32)
8979 i32 __builtin_mips_bposge32 (void)
8982 The following built-in functions map directly to a particular MIPS DSP REV 2
8983 instruction. Please refer to the architecture specification
8984 for details on what each instruction does.
8987 v4q7 __builtin_mips_absq_s_qb (v4q7);
8988 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
8989 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
8990 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
8991 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
8992 i32 __builtin_mips_append (i32, i32, imm0_31);
8993 i32 __builtin_mips_balign (i32, i32, imm0_3);
8994 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
8995 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
8996 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
8997 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
8998 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
8999 a64 __builtin_mips_madd (a64, i32, i32);
9000 a64 __builtin_mips_maddu (a64, ui32, ui32);
9001 a64 __builtin_mips_msub (a64, i32, i32);
9002 a64 __builtin_mips_msubu (a64, ui32, ui32);
9003 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
9004 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
9005 q31 __builtin_mips_mulq_rs_w (q31, q31);
9006 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
9007 q31 __builtin_mips_mulq_s_w (q31, q31);
9008 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
9009 a64 __builtin_mips_mult (i32, i32);
9010 a64 __builtin_mips_multu (ui32, ui32);
9011 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
9012 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
9013 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
9014 i32 __builtin_mips_prepend (i32, i32, imm0_31);
9015 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
9016 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
9017 v4i8 __builtin_mips_shra_qb (v4i8, i32);
9018 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
9019 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
9020 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
9021 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
9022 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
9023 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
9024 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
9025 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
9026 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
9027 q31 __builtin_mips_addqh_w (q31, q31);
9028 q31 __builtin_mips_addqh_r_w (q31, q31);
9029 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
9030 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
9031 q31 __builtin_mips_subqh_w (q31, q31);
9032 q31 __builtin_mips_subqh_r_w (q31, q31);
9033 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
9034 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
9035 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
9036 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
9037 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
9038 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
9042 @node MIPS Paired-Single Support
9043 @subsection MIPS Paired-Single Support
9045 The MIPS64 architecture includes a number of instructions that
9046 operate on pairs of single-precision floating-point values.
9047 Each pair is packed into a 64-bit floating-point register,
9048 with one element being designated the ``upper half'' and
9049 the other being designated the ``lower half''.
9051 GCC supports paired-single operations using both the generic
9052 vector extensions (@pxref{Vector Extensions}) and a collection of
9053 MIPS-specific built-in functions. Both kinds of support are
9054 enabled by the @option{-mpaired-single} command-line option.
9056 The vector type associated with paired-single values is usually
9057 called @code{v2sf}. It can be defined in C as follows:
9060 typedef float v2sf __attribute__ ((vector_size (8)));
9063 @code{v2sf} values are initialized in the same way as aggregates.
9067 v2sf a = @{1.5, 9.1@};
9070 b = (v2sf) @{e, f@};
9073 @emph{Note:} The CPU's endianness determines which value is stored in
9074 the upper half of a register and which value is stored in the lower half.
9075 On little-endian targets, the first value is the lower one and the second
9076 value is the upper one. The opposite order applies to big-endian targets.
9077 For example, the code above will set the lower half of @code{a} to
9078 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
9080 @node MIPS Loongson Built-in Functions
9081 @subsection MIPS Loongson Built-in Functions
9083 GCC provides intrinsics to access the SIMD instructions provided by the
9084 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
9085 available after inclusion of the @code{loongson.h} header file,
9086 operate on the following 64-bit vector types:
9089 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
9090 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
9091 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
9092 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
9093 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
9094 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
9097 The intrinsics provided are listed below; each is named after the
9098 machine instruction to which it corresponds, with suffixes added as
9099 appropriate to distinguish intrinsics that expand to the same machine
9100 instruction yet have different argument types. Refer to the architecture
9101 documentation for a description of the functionality of each
9105 int16x4_t packsswh (int32x2_t s, int32x2_t t);
9106 int8x8_t packsshb (int16x4_t s, int16x4_t t);
9107 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
9108 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
9109 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
9110 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
9111 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
9112 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
9113 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
9114 uint64_t paddd_u (uint64_t s, uint64_t t);
9115 int64_t paddd_s (int64_t s, int64_t t);
9116 int16x4_t paddsh (int16x4_t s, int16x4_t t);
9117 int8x8_t paddsb (int8x8_t s, int8x8_t t);
9118 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
9119 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
9120 uint64_t pandn_ud (uint64_t s, uint64_t t);
9121 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
9122 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
9123 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
9124 int64_t pandn_sd (int64_t s, int64_t t);
9125 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
9126 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
9127 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
9128 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
9129 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
9130 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
9131 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
9132 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
9133 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
9134 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
9135 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
9136 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
9137 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
9138 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
9139 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
9140 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
9141 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
9142 uint16x4_t pextrh_u (uint16x4_t s, int field);
9143 int16x4_t pextrh_s (int16x4_t s, int field);
9144 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
9145 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
9146 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
9147 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
9148 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
9149 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
9150 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
9151 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
9152 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
9153 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
9154 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
9155 int16x4_t pminsh (int16x4_t s, int16x4_t t);
9156 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
9157 uint8x8_t pmovmskb_u (uint8x8_t s);
9158 int8x8_t pmovmskb_s (int8x8_t s);
9159 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
9160 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
9161 int16x4_t pmullh (int16x4_t s, int16x4_t t);
9162 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
9163 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
9164 uint16x4_t biadd (uint8x8_t s);
9165 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
9166 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
9167 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
9168 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
9169 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
9170 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
9171 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
9172 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
9173 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
9174 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
9175 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
9176 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
9177 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
9178 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
9179 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
9180 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
9181 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
9182 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
9183 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
9184 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
9185 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
9186 uint64_t psubd_u (uint64_t s, uint64_t t);
9187 int64_t psubd_s (int64_t s, int64_t t);
9188 int16x4_t psubsh (int16x4_t s, int16x4_t t);
9189 int8x8_t psubsb (int8x8_t s, int8x8_t t);
9190 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
9191 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
9192 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
9193 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
9194 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
9195 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
9196 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
9197 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
9198 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
9199 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
9200 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
9201 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
9202 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
9203 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
9207 * Paired-Single Arithmetic::
9208 * Paired-Single Built-in Functions::
9209 * MIPS-3D Built-in Functions::
9212 @node Paired-Single Arithmetic
9213 @subsubsection Paired-Single Arithmetic
9215 The table below lists the @code{v2sf} operations for which hardware
9216 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
9217 values and @code{x} is an integral value.
9219 @multitable @columnfractions .50 .50
9220 @item C code @tab MIPS instruction
9221 @item @code{a + b} @tab @code{add.ps}
9222 @item @code{a - b} @tab @code{sub.ps}
9223 @item @code{-a} @tab @code{neg.ps}
9224 @item @code{a * b} @tab @code{mul.ps}
9225 @item @code{a * b + c} @tab @code{madd.ps}
9226 @item @code{a * b - c} @tab @code{msub.ps}
9227 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
9228 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
9229 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
9232 Note that the multiply-accumulate instructions can be disabled
9233 using the command-line option @code{-mno-fused-madd}.
9235 @node Paired-Single Built-in Functions
9236 @subsubsection Paired-Single Built-in Functions
9238 The following paired-single functions map directly to a particular
9239 MIPS instruction. Please refer to the architecture specification
9240 for details on what each instruction does.
9243 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
9244 Pair lower lower (@code{pll.ps}).
9246 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
9247 Pair upper lower (@code{pul.ps}).
9249 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
9250 Pair lower upper (@code{plu.ps}).
9252 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
9253 Pair upper upper (@code{puu.ps}).
9255 @item v2sf __builtin_mips_cvt_ps_s (float, float)
9256 Convert pair to paired single (@code{cvt.ps.s}).
9258 @item float __builtin_mips_cvt_s_pl (v2sf)
9259 Convert pair lower to single (@code{cvt.s.pl}).
9261 @item float __builtin_mips_cvt_s_pu (v2sf)
9262 Convert pair upper to single (@code{cvt.s.pu}).
9264 @item v2sf __builtin_mips_abs_ps (v2sf)
9265 Absolute value (@code{abs.ps}).
9267 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
9268 Align variable (@code{alnv.ps}).
9270 @emph{Note:} The value of the third parameter must be 0 or 4
9271 modulo 8, otherwise the result will be unpredictable. Please read the
9272 instruction description for details.
9275 The following multi-instruction functions are also available.
9276 In each case, @var{cond} can be any of the 16 floating-point conditions:
9277 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9278 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
9279 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9282 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9283 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9284 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
9285 @code{movt.ps}/@code{movf.ps}).
9287 The @code{movt} functions return the value @var{x} computed by:
9290 c.@var{cond}.ps @var{cc},@var{a},@var{b}
9291 mov.ps @var{x},@var{c}
9292 movt.ps @var{x},@var{d},@var{cc}
9295 The @code{movf} functions are similar but use @code{movf.ps} instead
9298 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9299 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9300 Comparison of two paired-single values (@code{c.@var{cond}.ps},
9301 @code{bc1t}/@code{bc1f}).
9303 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9304 and return either the upper or lower half of the result. For example:
9308 if (__builtin_mips_upper_c_eq_ps (a, b))
9309 upper_halves_are_equal ();
9311 upper_halves_are_unequal ();
9313 if (__builtin_mips_lower_c_eq_ps (a, b))
9314 lower_halves_are_equal ();
9316 lower_halves_are_unequal ();
9320 @node MIPS-3D Built-in Functions
9321 @subsubsection MIPS-3D Built-in Functions
9323 The MIPS-3D Application-Specific Extension (ASE) includes additional
9324 paired-single instructions that are designed to improve the performance
9325 of 3D graphics operations. Support for these instructions is controlled
9326 by the @option{-mips3d} command-line option.
9328 The functions listed below map directly to a particular MIPS-3D
9329 instruction. Please refer to the architecture specification for
9330 more details on what each instruction does.
9333 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
9334 Reduction add (@code{addr.ps}).
9336 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
9337 Reduction multiply (@code{mulr.ps}).
9339 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
9340 Convert paired single to paired word (@code{cvt.pw.ps}).
9342 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
9343 Convert paired word to paired single (@code{cvt.ps.pw}).
9345 @item float __builtin_mips_recip1_s (float)
9346 @itemx double __builtin_mips_recip1_d (double)
9347 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
9348 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
9350 @item float __builtin_mips_recip2_s (float, float)
9351 @itemx double __builtin_mips_recip2_d (double, double)
9352 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
9353 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
9355 @item float __builtin_mips_rsqrt1_s (float)
9356 @itemx double __builtin_mips_rsqrt1_d (double)
9357 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
9358 Reduced precision reciprocal square root (sequence step 1)
9359 (@code{rsqrt1.@var{fmt}}).
9361 @item float __builtin_mips_rsqrt2_s (float, float)
9362 @itemx double __builtin_mips_rsqrt2_d (double, double)
9363 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
9364 Reduced precision reciprocal square root (sequence step 2)
9365 (@code{rsqrt2.@var{fmt}}).
9368 The following multi-instruction functions are also available.
9369 In each case, @var{cond} can be any of the 16 floating-point conditions:
9370 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9371 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
9372 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9375 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
9376 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
9377 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
9378 @code{bc1t}/@code{bc1f}).
9380 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
9381 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
9386 if (__builtin_mips_cabs_eq_s (a, b))
9392 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9393 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9394 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
9395 @code{bc1t}/@code{bc1f}).
9397 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
9398 and return either the upper or lower half of the result. For example:
9402 if (__builtin_mips_upper_cabs_eq_ps (a, b))
9403 upper_halves_are_equal ();
9405 upper_halves_are_unequal ();
9407 if (__builtin_mips_lower_cabs_eq_ps (a, b))
9408 lower_halves_are_equal ();
9410 lower_halves_are_unequal ();
9413 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9414 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9415 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
9416 @code{movt.ps}/@code{movf.ps}).
9418 The @code{movt} functions return the value @var{x} computed by:
9421 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
9422 mov.ps @var{x},@var{c}
9423 movt.ps @var{x},@var{d},@var{cc}
9426 The @code{movf} functions are similar but use @code{movf.ps} instead
9429 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9430 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9431 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9432 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9433 Comparison of two paired-single values
9434 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9435 @code{bc1any2t}/@code{bc1any2f}).
9437 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9438 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
9439 result is true and the @code{all} forms return true if both results are true.
9444 if (__builtin_mips_any_c_eq_ps (a, b))
9449 if (__builtin_mips_all_c_eq_ps (a, b))
9455 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9456 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9457 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9458 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9459 Comparison of four paired-single values
9460 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9461 @code{bc1any4t}/@code{bc1any4f}).
9463 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
9464 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
9465 The @code{any} forms return true if any of the four results are true
9466 and the @code{all} forms return true if all four results are true.
9471 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
9476 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
9483 @node picoChip Built-in Functions
9484 @subsection picoChip Built-in Functions
9486 GCC provides an interface to selected machine instructions from the
9487 picoChip instruction set.
9490 @item int __builtin_sbc (int @var{value})
9491 Sign bit count. Return the number of consecutive bits in @var{value}
9492 which have the same value as the sign-bit. The result is the number of
9493 leading sign bits minus one, giving the number of redundant sign bits in
9496 @item int __builtin_byteswap (int @var{value})
9497 Byte swap. Return the result of swapping the upper and lower bytes of
9500 @item int __builtin_brev (int @var{value})
9501 Bit reversal. Return the result of reversing the bits in
9502 @var{value}. Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1,
9505 @item int __builtin_adds (int @var{x}, int @var{y})
9506 Saturating addition. Return the result of adding @var{x} and @var{y},
9507 storing the value 32767 if the result overflows.
9509 @item int __builtin_subs (int @var{x}, int @var{y})
9510 Saturating subtraction. Return the result of subtracting @var{y} from
9511 @var{x}, storing the value -32768 if the result overflows.
9513 @item void __builtin_halt (void)
9514 Halt. The processor will stop execution. This built-in is useful for
9515 implementing assertions.
9519 @node Other MIPS Built-in Functions
9520 @subsection Other MIPS Built-in Functions
9522 GCC provides other MIPS-specific built-in functions:
9525 @item void __builtin_mips_cache (int @var{op}, const volatile void *@var{addr})
9526 Insert a @samp{cache} instruction with operands @var{op} and @var{addr}.
9527 GCC defines the preprocessor macro @code{___GCC_HAVE_BUILTIN_MIPS_CACHE}
9528 when this function is available.
9531 @node PowerPC AltiVec Built-in Functions
9532 @subsection PowerPC AltiVec Built-in Functions
9534 GCC provides an interface for the PowerPC family of processors to access
9535 the AltiVec operations described in Motorola's AltiVec Programming
9536 Interface Manual. The interface is made available by including
9537 @code{<altivec.h>} and using @option{-maltivec} and
9538 @option{-mabi=altivec}. The interface supports the following vector
9542 vector unsigned char
9546 vector unsigned short
9557 GCC's implementation of the high-level language interface available from
9558 C and C++ code differs from Motorola's documentation in several ways.
9563 A vector constant is a list of constant expressions within curly braces.
9566 A vector initializer requires no cast if the vector constant is of the
9567 same type as the variable it is initializing.
9570 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9571 vector type is the default signedness of the base type. The default
9572 varies depending on the operating system, so a portable program should
9573 always specify the signedness.
9576 Compiling with @option{-maltivec} adds keywords @code{__vector},
9577 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
9578 @code{bool}. When compiling ISO C, the context-sensitive substitution
9579 of the keywords @code{vector}, @code{pixel} and @code{bool} is
9580 disabled. To use them, you must include @code{<altivec.h>} instead.
9583 GCC allows using a @code{typedef} name as the type specifier for a
9587 For C, overloaded functions are implemented with macros so the following
9591 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9594 Since @code{vec_add} is a macro, the vector constant in the example
9595 is treated as four separate arguments. Wrap the entire argument in
9596 parentheses for this to work.
9599 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
9600 Internally, GCC uses built-in functions to achieve the functionality in
9601 the aforementioned header file, but they are not supported and are
9602 subject to change without notice.
9604 The following interfaces are supported for the generic and specific
9605 AltiVec operations and the AltiVec predicates. In cases where there
9606 is a direct mapping between generic and specific operations, only the
9607 generic names are shown here, although the specific operations can also
9610 Arguments that are documented as @code{const int} require literal
9611 integral values within the range required for that operation.
9614 vector signed char vec_abs (vector signed char);
9615 vector signed short vec_abs (vector signed short);
9616 vector signed int vec_abs (vector signed int);
9617 vector float vec_abs (vector float);
9619 vector signed char vec_abss (vector signed char);
9620 vector signed short vec_abss (vector signed short);
9621 vector signed int vec_abss (vector signed int);
9623 vector signed char vec_add (vector bool char, vector signed char);
9624 vector signed char vec_add (vector signed char, vector bool char);
9625 vector signed char vec_add (vector signed char, vector signed char);
9626 vector unsigned char vec_add (vector bool char, vector unsigned char);
9627 vector unsigned char vec_add (vector unsigned char, vector bool char);
9628 vector unsigned char vec_add (vector unsigned char,
9629 vector unsigned char);
9630 vector signed short vec_add (vector bool short, vector signed short);
9631 vector signed short vec_add (vector signed short, vector bool short);
9632 vector signed short vec_add (vector signed short, vector signed short);
9633 vector unsigned short vec_add (vector bool short,
9634 vector unsigned short);
9635 vector unsigned short vec_add (vector unsigned short,
9637 vector unsigned short vec_add (vector unsigned short,
9638 vector unsigned short);
9639 vector signed int vec_add (vector bool int, vector signed int);
9640 vector signed int vec_add (vector signed int, vector bool int);
9641 vector signed int vec_add (vector signed int, vector signed int);
9642 vector unsigned int vec_add (vector bool int, vector unsigned int);
9643 vector unsigned int vec_add (vector unsigned int, vector bool int);
9644 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
9645 vector float vec_add (vector float, vector float);
9647 vector float vec_vaddfp (vector float, vector float);
9649 vector signed int vec_vadduwm (vector bool int, vector signed int);
9650 vector signed int vec_vadduwm (vector signed int, vector bool int);
9651 vector signed int vec_vadduwm (vector signed int, vector signed int);
9652 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
9653 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
9654 vector unsigned int vec_vadduwm (vector unsigned int,
9655 vector unsigned int);
9657 vector signed short vec_vadduhm (vector bool short,
9658 vector signed short);
9659 vector signed short vec_vadduhm (vector signed short,
9661 vector signed short vec_vadduhm (vector signed short,
9662 vector signed short);
9663 vector unsigned short vec_vadduhm (vector bool short,
9664 vector unsigned short);
9665 vector unsigned short vec_vadduhm (vector unsigned short,
9667 vector unsigned short vec_vadduhm (vector unsigned short,
9668 vector unsigned short);
9670 vector signed char vec_vaddubm (vector bool char, vector signed char);
9671 vector signed char vec_vaddubm (vector signed char, vector bool char);
9672 vector signed char vec_vaddubm (vector signed char, vector signed char);
9673 vector unsigned char vec_vaddubm (vector bool char,
9674 vector unsigned char);
9675 vector unsigned char vec_vaddubm (vector unsigned char,
9677 vector unsigned char vec_vaddubm (vector unsigned char,
9678 vector unsigned char);
9680 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
9682 vector unsigned char vec_adds (vector bool char, vector unsigned char);
9683 vector unsigned char vec_adds (vector unsigned char, vector bool char);
9684 vector unsigned char vec_adds (vector unsigned char,
9685 vector unsigned char);
9686 vector signed char vec_adds (vector bool char, vector signed char);
9687 vector signed char vec_adds (vector signed char, vector bool char);
9688 vector signed char vec_adds (vector signed char, vector signed char);
9689 vector unsigned short vec_adds (vector bool short,
9690 vector unsigned short);
9691 vector unsigned short vec_adds (vector unsigned short,
9693 vector unsigned short vec_adds (vector unsigned short,
9694 vector unsigned short);
9695 vector signed short vec_adds (vector bool short, vector signed short);
9696 vector signed short vec_adds (vector signed short, vector bool short);
9697 vector signed short vec_adds (vector signed short, vector signed short);
9698 vector unsigned int vec_adds (vector bool int, vector unsigned int);
9699 vector unsigned int vec_adds (vector unsigned int, vector bool int);
9700 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
9701 vector signed int vec_adds (vector bool int, vector signed int);
9702 vector signed int vec_adds (vector signed int, vector bool int);
9703 vector signed int vec_adds (vector signed int, vector signed int);
9705 vector signed int vec_vaddsws (vector bool int, vector signed int);
9706 vector signed int vec_vaddsws (vector signed int, vector bool int);
9707 vector signed int vec_vaddsws (vector signed int, vector signed int);
9709 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
9710 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
9711 vector unsigned int vec_vadduws (vector unsigned int,
9712 vector unsigned int);
9714 vector signed short vec_vaddshs (vector bool short,
9715 vector signed short);
9716 vector signed short vec_vaddshs (vector signed short,
9718 vector signed short vec_vaddshs (vector signed short,
9719 vector signed short);
9721 vector unsigned short vec_vadduhs (vector bool short,
9722 vector unsigned short);
9723 vector unsigned short vec_vadduhs (vector unsigned short,
9725 vector unsigned short vec_vadduhs (vector unsigned short,
9726 vector unsigned short);
9728 vector signed char vec_vaddsbs (vector bool char, vector signed char);
9729 vector signed char vec_vaddsbs (vector signed char, vector bool char);
9730 vector signed char vec_vaddsbs (vector signed char, vector signed char);
9732 vector unsigned char vec_vaddubs (vector bool char,
9733 vector unsigned char);
9734 vector unsigned char vec_vaddubs (vector unsigned char,
9736 vector unsigned char vec_vaddubs (vector unsigned char,
9737 vector unsigned char);
9739 vector float vec_and (vector float, vector float);
9740 vector float vec_and (vector float, vector bool int);
9741 vector float vec_and (vector bool int, vector float);
9742 vector bool int vec_and (vector bool int, vector bool int);
9743 vector signed int vec_and (vector bool int, vector signed int);
9744 vector signed int vec_and (vector signed int, vector bool int);
9745 vector signed int vec_and (vector signed int, vector signed int);
9746 vector unsigned int vec_and (vector bool int, vector unsigned int);
9747 vector unsigned int vec_and (vector unsigned int, vector bool int);
9748 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
9749 vector bool short vec_and (vector bool short, vector bool short);
9750 vector signed short vec_and (vector bool short, vector signed short);
9751 vector signed short vec_and (vector signed short, vector bool short);
9752 vector signed short vec_and (vector signed short, vector signed short);
9753 vector unsigned short vec_and (vector bool short,
9754 vector unsigned short);
9755 vector unsigned short vec_and (vector unsigned short,
9757 vector unsigned short vec_and (vector unsigned short,
9758 vector unsigned short);
9759 vector signed char vec_and (vector bool char, vector signed char);
9760 vector bool char vec_and (vector bool char, vector bool char);
9761 vector signed char vec_and (vector signed char, vector bool char);
9762 vector signed char vec_and (vector signed char, vector signed char);
9763 vector unsigned char vec_and (vector bool char, vector unsigned char);
9764 vector unsigned char vec_and (vector unsigned char, vector bool char);
9765 vector unsigned char vec_and (vector unsigned char,
9766 vector unsigned char);
9768 vector float vec_andc (vector float, vector float);
9769 vector float vec_andc (vector float, vector bool int);
9770 vector float vec_andc (vector bool int, vector float);
9771 vector bool int vec_andc (vector bool int, vector bool int);
9772 vector signed int vec_andc (vector bool int, vector signed int);
9773 vector signed int vec_andc (vector signed int, vector bool int);
9774 vector signed int vec_andc (vector signed int, vector signed int);
9775 vector unsigned int vec_andc (vector bool int, vector unsigned int);
9776 vector unsigned int vec_andc (vector unsigned int, vector bool int);
9777 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
9778 vector bool short vec_andc (vector bool short, vector bool short);
9779 vector signed short vec_andc (vector bool short, vector signed short);
9780 vector signed short vec_andc (vector signed short, vector bool short);
9781 vector signed short vec_andc (vector signed short, vector signed short);
9782 vector unsigned short vec_andc (vector bool short,
9783 vector unsigned short);
9784 vector unsigned short vec_andc (vector unsigned short,
9786 vector unsigned short vec_andc (vector unsigned short,
9787 vector unsigned short);
9788 vector signed char vec_andc (vector bool char, vector signed char);
9789 vector bool char vec_andc (vector bool char, vector bool char);
9790 vector signed char vec_andc (vector signed char, vector bool char);
9791 vector signed char vec_andc (vector signed char, vector signed char);
9792 vector unsigned char vec_andc (vector bool char, vector unsigned char);
9793 vector unsigned char vec_andc (vector unsigned char, vector bool char);
9794 vector unsigned char vec_andc (vector unsigned char,
9795 vector unsigned char);
9797 vector unsigned char vec_avg (vector unsigned char,
9798 vector unsigned char);
9799 vector signed char vec_avg (vector signed char, vector signed char);
9800 vector unsigned short vec_avg (vector unsigned short,
9801 vector unsigned short);
9802 vector signed short vec_avg (vector signed short, vector signed short);
9803 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
9804 vector signed int vec_avg (vector signed int, vector signed int);
9806 vector signed int vec_vavgsw (vector signed int, vector signed int);
9808 vector unsigned int vec_vavguw (vector unsigned int,
9809 vector unsigned int);
9811 vector signed short vec_vavgsh (vector signed short,
9812 vector signed short);
9814 vector unsigned short vec_vavguh (vector unsigned short,
9815 vector unsigned short);
9817 vector signed char vec_vavgsb (vector signed char, vector signed char);
9819 vector unsigned char vec_vavgub (vector unsigned char,
9820 vector unsigned char);
9822 vector float vec_ceil (vector float);
9824 vector signed int vec_cmpb (vector float, vector float);
9826 vector bool char vec_cmpeq (vector signed char, vector signed char);
9827 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
9828 vector bool short vec_cmpeq (vector signed short, vector signed short);
9829 vector bool short vec_cmpeq (vector unsigned short,
9830 vector unsigned short);
9831 vector bool int vec_cmpeq (vector signed int, vector signed int);
9832 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
9833 vector bool int vec_cmpeq (vector float, vector float);
9835 vector bool int vec_vcmpeqfp (vector float, vector float);
9837 vector bool int vec_vcmpequw (vector signed int, vector signed int);
9838 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
9840 vector bool short vec_vcmpequh (vector signed short,
9841 vector signed short);
9842 vector bool short vec_vcmpequh (vector unsigned short,
9843 vector unsigned short);
9845 vector bool char vec_vcmpequb (vector signed char, vector signed char);
9846 vector bool char vec_vcmpequb (vector unsigned char,
9847 vector unsigned char);
9849 vector bool int vec_cmpge (vector float, vector float);
9851 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
9852 vector bool char vec_cmpgt (vector signed char, vector signed char);
9853 vector bool short vec_cmpgt (vector unsigned short,
9854 vector unsigned short);
9855 vector bool short vec_cmpgt (vector signed short, vector signed short);
9856 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
9857 vector bool int vec_cmpgt (vector signed int, vector signed int);
9858 vector bool int vec_cmpgt (vector float, vector float);
9860 vector bool int vec_vcmpgtfp (vector float, vector float);
9862 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
9864 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
9866 vector bool short vec_vcmpgtsh (vector signed short,
9867 vector signed short);
9869 vector bool short vec_vcmpgtuh (vector unsigned short,
9870 vector unsigned short);
9872 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
9874 vector bool char vec_vcmpgtub (vector unsigned char,
9875 vector unsigned char);
9877 vector bool int vec_cmple (vector float, vector float);
9879 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
9880 vector bool char vec_cmplt (vector signed char, vector signed char);
9881 vector bool short vec_cmplt (vector unsigned short,
9882 vector unsigned short);
9883 vector bool short vec_cmplt (vector signed short, vector signed short);
9884 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
9885 vector bool int vec_cmplt (vector signed int, vector signed int);
9886 vector bool int vec_cmplt (vector float, vector float);
9888 vector float vec_ctf (vector unsigned int, const int);
9889 vector float vec_ctf (vector signed int, const int);
9891 vector float vec_vcfsx (vector signed int, const int);
9893 vector float vec_vcfux (vector unsigned int, const int);
9895 vector signed int vec_cts (vector float, const int);
9897 vector unsigned int vec_ctu (vector float, const int);
9899 void vec_dss (const int);
9901 void vec_dssall (void);
9903 void vec_dst (const vector unsigned char *, int, const int);
9904 void vec_dst (const vector signed char *, int, const int);
9905 void vec_dst (const vector bool char *, int, const int);
9906 void vec_dst (const vector unsigned short *, int, const int);
9907 void vec_dst (const vector signed short *, int, const int);
9908 void vec_dst (const vector bool short *, int, const int);
9909 void vec_dst (const vector pixel *, int, const int);
9910 void vec_dst (const vector unsigned int *, int, const int);
9911 void vec_dst (const vector signed int *, int, const int);
9912 void vec_dst (const vector bool int *, int, const int);
9913 void vec_dst (const vector float *, int, const int);
9914 void vec_dst (const unsigned char *, int, const int);
9915 void vec_dst (const signed char *, int, const int);
9916 void vec_dst (const unsigned short *, int, const int);
9917 void vec_dst (const short *, int, const int);
9918 void vec_dst (const unsigned int *, int, const int);
9919 void vec_dst (const int *, int, const int);
9920 void vec_dst (const unsigned long *, int, const int);
9921 void vec_dst (const long *, int, const int);
9922 void vec_dst (const float *, int, const int);
9924 void vec_dstst (const vector unsigned char *, int, const int);
9925 void vec_dstst (const vector signed char *, int, const int);
9926 void vec_dstst (const vector bool char *, int, const int);
9927 void vec_dstst (const vector unsigned short *, int, const int);
9928 void vec_dstst (const vector signed short *, int, const int);
9929 void vec_dstst (const vector bool short *, int, const int);
9930 void vec_dstst (const vector pixel *, int, const int);
9931 void vec_dstst (const vector unsigned int *, int, const int);
9932 void vec_dstst (const vector signed int *, int, const int);
9933 void vec_dstst (const vector bool int *, int, const int);
9934 void vec_dstst (const vector float *, int, const int);
9935 void vec_dstst (const unsigned char *, int, const int);
9936 void vec_dstst (const signed char *, int, const int);
9937 void vec_dstst (const unsigned short *, int, const int);
9938 void vec_dstst (const short *, int, const int);
9939 void vec_dstst (const unsigned int *, int, const int);
9940 void vec_dstst (const int *, int, const int);
9941 void vec_dstst (const unsigned long *, int, const int);
9942 void vec_dstst (const long *, int, const int);
9943 void vec_dstst (const float *, int, const int);
9945 void vec_dststt (const vector unsigned char *, int, const int);
9946 void vec_dststt (const vector signed char *, int, const int);
9947 void vec_dststt (const vector bool char *, int, const int);
9948 void vec_dststt (const vector unsigned short *, int, const int);
9949 void vec_dststt (const vector signed short *, int, const int);
9950 void vec_dststt (const vector bool short *, int, const int);
9951 void vec_dststt (const vector pixel *, int, const int);
9952 void vec_dststt (const vector unsigned int *, int, const int);
9953 void vec_dststt (const vector signed int *, int, const int);
9954 void vec_dststt (const vector bool int *, int, const int);
9955 void vec_dststt (const vector float *, int, const int);
9956 void vec_dststt (const unsigned char *, int, const int);
9957 void vec_dststt (const signed char *, int, const int);
9958 void vec_dststt (const unsigned short *, int, const int);
9959 void vec_dststt (const short *, int, const int);
9960 void vec_dststt (const unsigned int *, int, const int);
9961 void vec_dststt (const int *, int, const int);
9962 void vec_dststt (const unsigned long *, int, const int);
9963 void vec_dststt (const long *, int, const int);
9964 void vec_dststt (const float *, int, const int);
9966 void vec_dstt (const vector unsigned char *, int, const int);
9967 void vec_dstt (const vector signed char *, int, const int);
9968 void vec_dstt (const vector bool char *, int, const int);
9969 void vec_dstt (const vector unsigned short *, int, const int);
9970 void vec_dstt (const vector signed short *, int, const int);
9971 void vec_dstt (const vector bool short *, int, const int);
9972 void vec_dstt (const vector pixel *, int, const int);
9973 void vec_dstt (const vector unsigned int *, int, const int);
9974 void vec_dstt (const vector signed int *, int, const int);
9975 void vec_dstt (const vector bool int *, int, const int);
9976 void vec_dstt (const vector float *, int, const int);
9977 void vec_dstt (const unsigned char *, int, const int);
9978 void vec_dstt (const signed char *, int, const int);
9979 void vec_dstt (const unsigned short *, int, const int);
9980 void vec_dstt (const short *, int, const int);
9981 void vec_dstt (const unsigned int *, int, const int);
9982 void vec_dstt (const int *, int, const int);
9983 void vec_dstt (const unsigned long *, int, const int);
9984 void vec_dstt (const long *, int, const int);
9985 void vec_dstt (const float *, int, const int);
9987 vector float vec_expte (vector float);
9989 vector float vec_floor (vector float);
9991 vector float vec_ld (int, const vector float *);
9992 vector float vec_ld (int, const float *);
9993 vector bool int vec_ld (int, const vector bool int *);
9994 vector signed int vec_ld (int, const vector signed int *);
9995 vector signed int vec_ld (int, const int *);
9996 vector signed int vec_ld (int, const long *);
9997 vector unsigned int vec_ld (int, const vector unsigned int *);
9998 vector unsigned int vec_ld (int, const unsigned int *);
9999 vector unsigned int vec_ld (int, const unsigned long *);
10000 vector bool short vec_ld (int, const vector bool short *);
10001 vector pixel vec_ld (int, const vector pixel *);
10002 vector signed short vec_ld (int, const vector signed short *);
10003 vector signed short vec_ld (int, const short *);
10004 vector unsigned short vec_ld (int, const vector unsigned short *);
10005 vector unsigned short vec_ld (int, const unsigned short *);
10006 vector bool char vec_ld (int, const vector bool char *);
10007 vector signed char vec_ld (int, const vector signed char *);
10008 vector signed char vec_ld (int, const signed char *);
10009 vector unsigned char vec_ld (int, const vector unsigned char *);
10010 vector unsigned char vec_ld (int, const unsigned char *);
10012 vector signed char vec_lde (int, const signed char *);
10013 vector unsigned char vec_lde (int, const unsigned char *);
10014 vector signed short vec_lde (int, const short *);
10015 vector unsigned short vec_lde (int, const unsigned short *);
10016 vector float vec_lde (int, const float *);
10017 vector signed int vec_lde (int, const int *);
10018 vector unsigned int vec_lde (int, const unsigned int *);
10019 vector signed int vec_lde (int, const long *);
10020 vector unsigned int vec_lde (int, const unsigned long *);
10022 vector float vec_lvewx (int, float *);
10023 vector signed int vec_lvewx (int, int *);
10024 vector unsigned int vec_lvewx (int, unsigned int *);
10025 vector signed int vec_lvewx (int, long *);
10026 vector unsigned int vec_lvewx (int, unsigned long *);
10028 vector signed short vec_lvehx (int, short *);
10029 vector unsigned short vec_lvehx (int, unsigned short *);
10031 vector signed char vec_lvebx (int, char *);
10032 vector unsigned char vec_lvebx (int, unsigned char *);
10034 vector float vec_ldl (int, const vector float *);
10035 vector float vec_ldl (int, const float *);
10036 vector bool int vec_ldl (int, const vector bool int *);
10037 vector signed int vec_ldl (int, const vector signed int *);
10038 vector signed int vec_ldl (int, const int *);
10039 vector signed int vec_ldl (int, const long *);
10040 vector unsigned int vec_ldl (int, const vector unsigned int *);
10041 vector unsigned int vec_ldl (int, const unsigned int *);
10042 vector unsigned int vec_ldl (int, const unsigned long *);
10043 vector bool short vec_ldl (int, const vector bool short *);
10044 vector pixel vec_ldl (int, const vector pixel *);
10045 vector signed short vec_ldl (int, const vector signed short *);
10046 vector signed short vec_ldl (int, const short *);
10047 vector unsigned short vec_ldl (int, const vector unsigned short *);
10048 vector unsigned short vec_ldl (int, const unsigned short *);
10049 vector bool char vec_ldl (int, const vector bool char *);
10050 vector signed char vec_ldl (int, const vector signed char *);
10051 vector signed char vec_ldl (int, const signed char *);
10052 vector unsigned char vec_ldl (int, const vector unsigned char *);
10053 vector unsigned char vec_ldl (int, const unsigned char *);
10055 vector float vec_loge (vector float);
10057 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
10058 vector unsigned char vec_lvsl (int, const volatile signed char *);
10059 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
10060 vector unsigned char vec_lvsl (int, const volatile short *);
10061 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
10062 vector unsigned char vec_lvsl (int, const volatile int *);
10063 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
10064 vector unsigned char vec_lvsl (int, const volatile long *);
10065 vector unsigned char vec_lvsl (int, const volatile float *);
10067 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
10068 vector unsigned char vec_lvsr (int, const volatile signed char *);
10069 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
10070 vector unsigned char vec_lvsr (int, const volatile short *);
10071 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
10072 vector unsigned char vec_lvsr (int, const volatile int *);
10073 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
10074 vector unsigned char vec_lvsr (int, const volatile long *);
10075 vector unsigned char vec_lvsr (int, const volatile float *);
10077 vector float vec_madd (vector float, vector float, vector float);
10079 vector signed short vec_madds (vector signed short,
10080 vector signed short,
10081 vector signed short);
10083 vector unsigned char vec_max (vector bool char, vector unsigned char);
10084 vector unsigned char vec_max (vector unsigned char, vector bool char);
10085 vector unsigned char vec_max (vector unsigned char,
10086 vector unsigned char);
10087 vector signed char vec_max (vector bool char, vector signed char);
10088 vector signed char vec_max (vector signed char, vector bool char);
10089 vector signed char vec_max (vector signed char, vector signed char);
10090 vector unsigned short vec_max (vector bool short,
10091 vector unsigned short);
10092 vector unsigned short vec_max (vector unsigned short,
10093 vector bool short);
10094 vector unsigned short vec_max (vector unsigned short,
10095 vector unsigned short);
10096 vector signed short vec_max (vector bool short, vector signed short);
10097 vector signed short vec_max (vector signed short, vector bool short);
10098 vector signed short vec_max (vector signed short, vector signed short);
10099 vector unsigned int vec_max (vector bool int, vector unsigned int);
10100 vector unsigned int vec_max (vector unsigned int, vector bool int);
10101 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
10102 vector signed int vec_max (vector bool int, vector signed int);
10103 vector signed int vec_max (vector signed int, vector bool int);
10104 vector signed int vec_max (vector signed int, vector signed int);
10105 vector float vec_max (vector float, vector float);
10107 vector float vec_vmaxfp (vector float, vector float);
10109 vector signed int vec_vmaxsw (vector bool int, vector signed int);
10110 vector signed int vec_vmaxsw (vector signed int, vector bool int);
10111 vector signed int vec_vmaxsw (vector signed int, vector signed int);
10113 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
10114 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
10115 vector unsigned int vec_vmaxuw (vector unsigned int,
10116 vector unsigned int);
10118 vector signed short vec_vmaxsh (vector bool short, vector signed short);
10119 vector signed short vec_vmaxsh (vector signed short, vector bool short);
10120 vector signed short vec_vmaxsh (vector signed short,
10121 vector signed short);
10123 vector unsigned short vec_vmaxuh (vector bool short,
10124 vector unsigned short);
10125 vector unsigned short vec_vmaxuh (vector unsigned short,
10126 vector bool short);
10127 vector unsigned short vec_vmaxuh (vector unsigned short,
10128 vector unsigned short);
10130 vector signed char vec_vmaxsb (vector bool char, vector signed char);
10131 vector signed char vec_vmaxsb (vector signed char, vector bool char);
10132 vector signed char vec_vmaxsb (vector signed char, vector signed char);
10134 vector unsigned char vec_vmaxub (vector bool char,
10135 vector unsigned char);
10136 vector unsigned char vec_vmaxub (vector unsigned char,
10138 vector unsigned char vec_vmaxub (vector unsigned char,
10139 vector unsigned char);
10141 vector bool char vec_mergeh (vector bool char, vector bool char);
10142 vector signed char vec_mergeh (vector signed char, vector signed char);
10143 vector unsigned char vec_mergeh (vector unsigned char,
10144 vector unsigned char);
10145 vector bool short vec_mergeh (vector bool short, vector bool short);
10146 vector pixel vec_mergeh (vector pixel, vector pixel);
10147 vector signed short vec_mergeh (vector signed short,
10148 vector signed short);
10149 vector unsigned short vec_mergeh (vector unsigned short,
10150 vector unsigned short);
10151 vector float vec_mergeh (vector float, vector float);
10152 vector bool int vec_mergeh (vector bool int, vector bool int);
10153 vector signed int vec_mergeh (vector signed int, vector signed int);
10154 vector unsigned int vec_mergeh (vector unsigned int,
10155 vector unsigned int);
10157 vector float vec_vmrghw (vector float, vector float);
10158 vector bool int vec_vmrghw (vector bool int, vector bool int);
10159 vector signed int vec_vmrghw (vector signed int, vector signed int);
10160 vector unsigned int vec_vmrghw (vector unsigned int,
10161 vector unsigned int);
10163 vector bool short vec_vmrghh (vector bool short, vector bool short);
10164 vector signed short vec_vmrghh (vector signed short,
10165 vector signed short);
10166 vector unsigned short vec_vmrghh (vector unsigned short,
10167 vector unsigned short);
10168 vector pixel vec_vmrghh (vector pixel, vector pixel);
10170 vector bool char vec_vmrghb (vector bool char, vector bool char);
10171 vector signed char vec_vmrghb (vector signed char, vector signed char);
10172 vector unsigned char vec_vmrghb (vector unsigned char,
10173 vector unsigned char);
10175 vector bool char vec_mergel (vector bool char, vector bool char);
10176 vector signed char vec_mergel (vector signed char, vector signed char);
10177 vector unsigned char vec_mergel (vector unsigned char,
10178 vector unsigned char);
10179 vector bool short vec_mergel (vector bool short, vector bool short);
10180 vector pixel vec_mergel (vector pixel, vector pixel);
10181 vector signed short vec_mergel (vector signed short,
10182 vector signed short);
10183 vector unsigned short vec_mergel (vector unsigned short,
10184 vector unsigned short);
10185 vector float vec_mergel (vector float, vector float);
10186 vector bool int vec_mergel (vector bool int, vector bool int);
10187 vector signed int vec_mergel (vector signed int, vector signed int);
10188 vector unsigned int vec_mergel (vector unsigned int,
10189 vector unsigned int);
10191 vector float vec_vmrglw (vector float, vector float);
10192 vector signed int vec_vmrglw (vector signed int, vector signed int);
10193 vector unsigned int vec_vmrglw (vector unsigned int,
10194 vector unsigned int);
10195 vector bool int vec_vmrglw (vector bool int, vector bool int);
10197 vector bool short vec_vmrglh (vector bool short, vector bool short);
10198 vector signed short vec_vmrglh (vector signed short,
10199 vector signed short);
10200 vector unsigned short vec_vmrglh (vector unsigned short,
10201 vector unsigned short);
10202 vector pixel vec_vmrglh (vector pixel, vector pixel);
10204 vector bool char vec_vmrglb (vector bool char, vector bool char);
10205 vector signed char vec_vmrglb (vector signed char, vector signed char);
10206 vector unsigned char vec_vmrglb (vector unsigned char,
10207 vector unsigned char);
10209 vector unsigned short vec_mfvscr (void);
10211 vector unsigned char vec_min (vector bool char, vector unsigned char);
10212 vector unsigned char vec_min (vector unsigned char, vector bool char);
10213 vector unsigned char vec_min (vector unsigned char,
10214 vector unsigned char);
10215 vector signed char vec_min (vector bool char, vector signed char);
10216 vector signed char vec_min (vector signed char, vector bool char);
10217 vector signed char vec_min (vector signed char, vector signed char);
10218 vector unsigned short vec_min (vector bool short,
10219 vector unsigned short);
10220 vector unsigned short vec_min (vector unsigned short,
10221 vector bool short);
10222 vector unsigned short vec_min (vector unsigned short,
10223 vector unsigned short);
10224 vector signed short vec_min (vector bool short, vector signed short);
10225 vector signed short vec_min (vector signed short, vector bool short);
10226 vector signed short vec_min (vector signed short, vector signed short);
10227 vector unsigned int vec_min (vector bool int, vector unsigned int);
10228 vector unsigned int vec_min (vector unsigned int, vector bool int);
10229 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
10230 vector signed int vec_min (vector bool int, vector signed int);
10231 vector signed int vec_min (vector signed int, vector bool int);
10232 vector signed int vec_min (vector signed int, vector signed int);
10233 vector float vec_min (vector float, vector float);
10235 vector float vec_vminfp (vector float, vector float);
10237 vector signed int vec_vminsw (vector bool int, vector signed int);
10238 vector signed int vec_vminsw (vector signed int, vector bool int);
10239 vector signed int vec_vminsw (vector signed int, vector signed int);
10241 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
10242 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
10243 vector unsigned int vec_vminuw (vector unsigned int,
10244 vector unsigned int);
10246 vector signed short vec_vminsh (vector bool short, vector signed short);
10247 vector signed short vec_vminsh (vector signed short, vector bool short);
10248 vector signed short vec_vminsh (vector signed short,
10249 vector signed short);
10251 vector unsigned short vec_vminuh (vector bool short,
10252 vector unsigned short);
10253 vector unsigned short vec_vminuh (vector unsigned short,
10254 vector bool short);
10255 vector unsigned short vec_vminuh (vector unsigned short,
10256 vector unsigned short);
10258 vector signed char vec_vminsb (vector bool char, vector signed char);
10259 vector signed char vec_vminsb (vector signed char, vector bool char);
10260 vector signed char vec_vminsb (vector signed char, vector signed char);
10262 vector unsigned char vec_vminub (vector bool char,
10263 vector unsigned char);
10264 vector unsigned char vec_vminub (vector unsigned char,
10266 vector unsigned char vec_vminub (vector unsigned char,
10267 vector unsigned char);
10269 vector signed short vec_mladd (vector signed short,
10270 vector signed short,
10271 vector signed short);
10272 vector signed short vec_mladd (vector signed short,
10273 vector unsigned short,
10274 vector unsigned short);
10275 vector signed short vec_mladd (vector unsigned short,
10276 vector signed short,
10277 vector signed short);
10278 vector unsigned short vec_mladd (vector unsigned short,
10279 vector unsigned short,
10280 vector unsigned short);
10282 vector signed short vec_mradds (vector signed short,
10283 vector signed short,
10284 vector signed short);
10286 vector unsigned int vec_msum (vector unsigned char,
10287 vector unsigned char,
10288 vector unsigned int);
10289 vector signed int vec_msum (vector signed char,
10290 vector unsigned char,
10291 vector signed int);
10292 vector unsigned int vec_msum (vector unsigned short,
10293 vector unsigned short,
10294 vector unsigned int);
10295 vector signed int vec_msum (vector signed short,
10296 vector signed short,
10297 vector signed int);
10299 vector signed int vec_vmsumshm (vector signed short,
10300 vector signed short,
10301 vector signed int);
10303 vector unsigned int vec_vmsumuhm (vector unsigned short,
10304 vector unsigned short,
10305 vector unsigned int);
10307 vector signed int vec_vmsummbm (vector signed char,
10308 vector unsigned char,
10309 vector signed int);
10311 vector unsigned int vec_vmsumubm (vector unsigned char,
10312 vector unsigned char,
10313 vector unsigned int);
10315 vector unsigned int vec_msums (vector unsigned short,
10316 vector unsigned short,
10317 vector unsigned int);
10318 vector signed int vec_msums (vector signed short,
10319 vector signed short,
10320 vector signed int);
10322 vector signed int vec_vmsumshs (vector signed short,
10323 vector signed short,
10324 vector signed int);
10326 vector unsigned int vec_vmsumuhs (vector unsigned short,
10327 vector unsigned short,
10328 vector unsigned int);
10330 void vec_mtvscr (vector signed int);
10331 void vec_mtvscr (vector unsigned int);
10332 void vec_mtvscr (vector bool int);
10333 void vec_mtvscr (vector signed short);
10334 void vec_mtvscr (vector unsigned short);
10335 void vec_mtvscr (vector bool short);
10336 void vec_mtvscr (vector pixel);
10337 void vec_mtvscr (vector signed char);
10338 void vec_mtvscr (vector unsigned char);
10339 void vec_mtvscr (vector bool char);
10341 vector unsigned short vec_mule (vector unsigned char,
10342 vector unsigned char);
10343 vector signed short vec_mule (vector signed char,
10344 vector signed char);
10345 vector unsigned int vec_mule (vector unsigned short,
10346 vector unsigned short);
10347 vector signed int vec_mule (vector signed short, vector signed short);
10349 vector signed int vec_vmulesh (vector signed short,
10350 vector signed short);
10352 vector unsigned int vec_vmuleuh (vector unsigned short,
10353 vector unsigned short);
10355 vector signed short vec_vmulesb (vector signed char,
10356 vector signed char);
10358 vector unsigned short vec_vmuleub (vector unsigned char,
10359 vector unsigned char);
10361 vector unsigned short vec_mulo (vector unsigned char,
10362 vector unsigned char);
10363 vector signed short vec_mulo (vector signed char, vector signed char);
10364 vector unsigned int vec_mulo (vector unsigned short,
10365 vector unsigned short);
10366 vector signed int vec_mulo (vector signed short, vector signed short);
10368 vector signed int vec_vmulosh (vector signed short,
10369 vector signed short);
10371 vector unsigned int vec_vmulouh (vector unsigned short,
10372 vector unsigned short);
10374 vector signed short vec_vmulosb (vector signed char,
10375 vector signed char);
10377 vector unsigned short vec_vmuloub (vector unsigned char,
10378 vector unsigned char);
10380 vector float vec_nmsub (vector float, vector float, vector float);
10382 vector float vec_nor (vector float, vector float);
10383 vector signed int vec_nor (vector signed int, vector signed int);
10384 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
10385 vector bool int vec_nor (vector bool int, vector bool int);
10386 vector signed short vec_nor (vector signed short, vector signed short);
10387 vector unsigned short vec_nor (vector unsigned short,
10388 vector unsigned short);
10389 vector bool short vec_nor (vector bool short, vector bool short);
10390 vector signed char vec_nor (vector signed char, vector signed char);
10391 vector unsigned char vec_nor (vector unsigned char,
10392 vector unsigned char);
10393 vector bool char vec_nor (vector bool char, vector bool char);
10395 vector float vec_or (vector float, vector float);
10396 vector float vec_or (vector float, vector bool int);
10397 vector float vec_or (vector bool int, vector float);
10398 vector bool int vec_or (vector bool int, vector bool int);
10399 vector signed int vec_or (vector bool int, vector signed int);
10400 vector signed int vec_or (vector signed int, vector bool int);
10401 vector signed int vec_or (vector signed int, vector signed int);
10402 vector unsigned int vec_or (vector bool int, vector unsigned int);
10403 vector unsigned int vec_or (vector unsigned int, vector bool int);
10404 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
10405 vector bool short vec_or (vector bool short, vector bool short);
10406 vector signed short vec_or (vector bool short, vector signed short);
10407 vector signed short vec_or (vector signed short, vector bool short);
10408 vector signed short vec_or (vector signed short, vector signed short);
10409 vector unsigned short vec_or (vector bool short, vector unsigned short);
10410 vector unsigned short vec_or (vector unsigned short, vector bool short);
10411 vector unsigned short vec_or (vector unsigned short,
10412 vector unsigned short);
10413 vector signed char vec_or (vector bool char, vector signed char);
10414 vector bool char vec_or (vector bool char, vector bool char);
10415 vector signed char vec_or (vector signed char, vector bool char);
10416 vector signed char vec_or (vector signed char, vector signed char);
10417 vector unsigned char vec_or (vector bool char, vector unsigned char);
10418 vector unsigned char vec_or (vector unsigned char, vector bool char);
10419 vector unsigned char vec_or (vector unsigned char,
10420 vector unsigned char);
10422 vector signed char vec_pack (vector signed short, vector signed short);
10423 vector unsigned char vec_pack (vector unsigned short,
10424 vector unsigned short);
10425 vector bool char vec_pack (vector bool short, vector bool short);
10426 vector signed short vec_pack (vector signed int, vector signed int);
10427 vector unsigned short vec_pack (vector unsigned int,
10428 vector unsigned int);
10429 vector bool short vec_pack (vector bool int, vector bool int);
10431 vector bool short vec_vpkuwum (vector bool int, vector bool int);
10432 vector signed short vec_vpkuwum (vector signed int, vector signed int);
10433 vector unsigned short vec_vpkuwum (vector unsigned int,
10434 vector unsigned int);
10436 vector bool char vec_vpkuhum (vector bool short, vector bool short);
10437 vector signed char vec_vpkuhum (vector signed short,
10438 vector signed short);
10439 vector unsigned char vec_vpkuhum (vector unsigned short,
10440 vector unsigned short);
10442 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
10444 vector unsigned char vec_packs (vector unsigned short,
10445 vector unsigned short);
10446 vector signed char vec_packs (vector signed short, vector signed short);
10447 vector unsigned short vec_packs (vector unsigned int,
10448 vector unsigned int);
10449 vector signed short vec_packs (vector signed int, vector signed int);
10451 vector signed short vec_vpkswss (vector signed int, vector signed int);
10453 vector unsigned short vec_vpkuwus (vector unsigned int,
10454 vector unsigned int);
10456 vector signed char vec_vpkshss (vector signed short,
10457 vector signed short);
10459 vector unsigned char vec_vpkuhus (vector unsigned short,
10460 vector unsigned short);
10462 vector unsigned char vec_packsu (vector unsigned short,
10463 vector unsigned short);
10464 vector unsigned char vec_packsu (vector signed short,
10465 vector signed short);
10466 vector unsigned short vec_packsu (vector unsigned int,
10467 vector unsigned int);
10468 vector unsigned short vec_packsu (vector signed int, vector signed int);
10470 vector unsigned short vec_vpkswus (vector signed int,
10471 vector signed int);
10473 vector unsigned char vec_vpkshus (vector signed short,
10474 vector signed short);
10476 vector float vec_perm (vector float,
10478 vector unsigned char);
10479 vector signed int vec_perm (vector signed int,
10481 vector unsigned char);
10482 vector unsigned int vec_perm (vector unsigned int,
10483 vector unsigned int,
10484 vector unsigned char);
10485 vector bool int vec_perm (vector bool int,
10487 vector unsigned char);
10488 vector signed short vec_perm (vector signed short,
10489 vector signed short,
10490 vector unsigned char);
10491 vector unsigned short vec_perm (vector unsigned short,
10492 vector unsigned short,
10493 vector unsigned char);
10494 vector bool short vec_perm (vector bool short,
10496 vector unsigned char);
10497 vector pixel vec_perm (vector pixel,
10499 vector unsigned char);
10500 vector signed char vec_perm (vector signed char,
10501 vector signed char,
10502 vector unsigned char);
10503 vector unsigned char vec_perm (vector unsigned char,
10504 vector unsigned char,
10505 vector unsigned char);
10506 vector bool char vec_perm (vector bool char,
10508 vector unsigned char);
10510 vector float vec_re (vector float);
10512 vector signed char vec_rl (vector signed char,
10513 vector unsigned char);
10514 vector unsigned char vec_rl (vector unsigned char,
10515 vector unsigned char);
10516 vector signed short vec_rl (vector signed short, vector unsigned short);
10517 vector unsigned short vec_rl (vector unsigned short,
10518 vector unsigned short);
10519 vector signed int vec_rl (vector signed int, vector unsigned int);
10520 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
10522 vector signed int vec_vrlw (vector signed int, vector unsigned int);
10523 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
10525 vector signed short vec_vrlh (vector signed short,
10526 vector unsigned short);
10527 vector unsigned short vec_vrlh (vector unsigned short,
10528 vector unsigned short);
10530 vector signed char vec_vrlb (vector signed char, vector unsigned char);
10531 vector unsigned char vec_vrlb (vector unsigned char,
10532 vector unsigned char);
10534 vector float vec_round (vector float);
10536 vector float vec_rsqrte (vector float);
10538 vector float vec_sel (vector float, vector float, vector bool int);
10539 vector float vec_sel (vector float, vector float, vector unsigned int);
10540 vector signed int vec_sel (vector signed int,
10543 vector signed int vec_sel (vector signed int,
10545 vector unsigned int);
10546 vector unsigned int vec_sel (vector unsigned int,
10547 vector unsigned int,
10549 vector unsigned int vec_sel (vector unsigned int,
10550 vector unsigned int,
10551 vector unsigned int);
10552 vector bool int vec_sel (vector bool int,
10555 vector bool int vec_sel (vector bool int,
10557 vector unsigned int);
10558 vector signed short vec_sel (vector signed short,
10559 vector signed short,
10560 vector bool short);
10561 vector signed short vec_sel (vector signed short,
10562 vector signed short,
10563 vector unsigned short);
10564 vector unsigned short vec_sel (vector unsigned short,
10565 vector unsigned short,
10566 vector bool short);
10567 vector unsigned short vec_sel (vector unsigned short,
10568 vector unsigned short,
10569 vector unsigned short);
10570 vector bool short vec_sel (vector bool short,
10572 vector bool short);
10573 vector bool short vec_sel (vector bool short,
10575 vector unsigned short);
10576 vector signed char vec_sel (vector signed char,
10577 vector signed char,
10579 vector signed char vec_sel (vector signed char,
10580 vector signed char,
10581 vector unsigned char);
10582 vector unsigned char vec_sel (vector unsigned char,
10583 vector unsigned char,
10585 vector unsigned char vec_sel (vector unsigned char,
10586 vector unsigned char,
10587 vector unsigned char);
10588 vector bool char vec_sel (vector bool char,
10591 vector bool char vec_sel (vector bool char,
10593 vector unsigned char);
10595 vector signed char vec_sl (vector signed char,
10596 vector unsigned char);
10597 vector unsigned char vec_sl (vector unsigned char,
10598 vector unsigned char);
10599 vector signed short vec_sl (vector signed short, vector unsigned short);
10600 vector unsigned short vec_sl (vector unsigned short,
10601 vector unsigned short);
10602 vector signed int vec_sl (vector signed int, vector unsigned int);
10603 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
10605 vector signed int vec_vslw (vector signed int, vector unsigned int);
10606 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
10608 vector signed short vec_vslh (vector signed short,
10609 vector unsigned short);
10610 vector unsigned short vec_vslh (vector unsigned short,
10611 vector unsigned short);
10613 vector signed char vec_vslb (vector signed char, vector unsigned char);
10614 vector unsigned char vec_vslb (vector unsigned char,
10615 vector unsigned char);
10617 vector float vec_sld (vector float, vector float, const int);
10618 vector signed int vec_sld (vector signed int,
10621 vector unsigned int vec_sld (vector unsigned int,
10622 vector unsigned int,
10624 vector bool int vec_sld (vector bool int,
10627 vector signed short vec_sld (vector signed short,
10628 vector signed short,
10630 vector unsigned short vec_sld (vector unsigned short,
10631 vector unsigned short,
10633 vector bool short vec_sld (vector bool short,
10636 vector pixel vec_sld (vector pixel,
10639 vector signed char vec_sld (vector signed char,
10640 vector signed char,
10642 vector unsigned char vec_sld (vector unsigned char,
10643 vector unsigned char,
10645 vector bool char vec_sld (vector bool char,
10649 vector signed int vec_sll (vector signed int,
10650 vector unsigned int);
10651 vector signed int vec_sll (vector signed int,
10652 vector unsigned short);
10653 vector signed int vec_sll (vector signed int,
10654 vector unsigned char);
10655 vector unsigned int vec_sll (vector unsigned int,
10656 vector unsigned int);
10657 vector unsigned int vec_sll (vector unsigned int,
10658 vector unsigned short);
10659 vector unsigned int vec_sll (vector unsigned int,
10660 vector unsigned char);
10661 vector bool int vec_sll (vector bool int,
10662 vector unsigned int);
10663 vector bool int vec_sll (vector bool int,
10664 vector unsigned short);
10665 vector bool int vec_sll (vector bool int,
10666 vector unsigned char);
10667 vector signed short vec_sll (vector signed short,
10668 vector unsigned int);
10669 vector signed short vec_sll (vector signed short,
10670 vector unsigned short);
10671 vector signed short vec_sll (vector signed short,
10672 vector unsigned char);
10673 vector unsigned short vec_sll (vector unsigned short,
10674 vector unsigned int);
10675 vector unsigned short vec_sll (vector unsigned short,
10676 vector unsigned short);
10677 vector unsigned short vec_sll (vector unsigned short,
10678 vector unsigned char);
10679 vector bool short vec_sll (vector bool short, vector unsigned int);
10680 vector bool short vec_sll (vector bool short, vector unsigned short);
10681 vector bool short vec_sll (vector bool short, vector unsigned char);
10682 vector pixel vec_sll (vector pixel, vector unsigned int);
10683 vector pixel vec_sll (vector pixel, vector unsigned short);
10684 vector pixel vec_sll (vector pixel, vector unsigned char);
10685 vector signed char vec_sll (vector signed char, vector unsigned int);
10686 vector signed char vec_sll (vector signed char, vector unsigned short);
10687 vector signed char vec_sll (vector signed char, vector unsigned char);
10688 vector unsigned char vec_sll (vector unsigned char,
10689 vector unsigned int);
10690 vector unsigned char vec_sll (vector unsigned char,
10691 vector unsigned short);
10692 vector unsigned char vec_sll (vector unsigned char,
10693 vector unsigned char);
10694 vector bool char vec_sll (vector bool char, vector unsigned int);
10695 vector bool char vec_sll (vector bool char, vector unsigned short);
10696 vector bool char vec_sll (vector bool char, vector unsigned char);
10698 vector float vec_slo (vector float, vector signed char);
10699 vector float vec_slo (vector float, vector unsigned char);
10700 vector signed int vec_slo (vector signed int, vector signed char);
10701 vector signed int vec_slo (vector signed int, vector unsigned char);
10702 vector unsigned int vec_slo (vector unsigned int, vector signed char);
10703 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
10704 vector signed short vec_slo (vector signed short, vector signed char);
10705 vector signed short vec_slo (vector signed short, vector unsigned char);
10706 vector unsigned short vec_slo (vector unsigned short,
10707 vector signed char);
10708 vector unsigned short vec_slo (vector unsigned short,
10709 vector unsigned char);
10710 vector pixel vec_slo (vector pixel, vector signed char);
10711 vector pixel vec_slo (vector pixel, vector unsigned char);
10712 vector signed char vec_slo (vector signed char, vector signed char);
10713 vector signed char vec_slo (vector signed char, vector unsigned char);
10714 vector unsigned char vec_slo (vector unsigned char, vector signed char);
10715 vector unsigned char vec_slo (vector unsigned char,
10716 vector unsigned char);
10718 vector signed char vec_splat (vector signed char, const int);
10719 vector unsigned char vec_splat (vector unsigned char, const int);
10720 vector bool char vec_splat (vector bool char, const int);
10721 vector signed short vec_splat (vector signed short, const int);
10722 vector unsigned short vec_splat (vector unsigned short, const int);
10723 vector bool short vec_splat (vector bool short, const int);
10724 vector pixel vec_splat (vector pixel, const int);
10725 vector float vec_splat (vector float, const int);
10726 vector signed int vec_splat (vector signed int, const int);
10727 vector unsigned int vec_splat (vector unsigned int, const int);
10728 vector bool int vec_splat (vector bool int, const int);
10730 vector float vec_vspltw (vector float, const int);
10731 vector signed int vec_vspltw (vector signed int, const int);
10732 vector unsigned int vec_vspltw (vector unsigned int, const int);
10733 vector bool int vec_vspltw (vector bool int, const int);
10735 vector bool short vec_vsplth (vector bool short, const int);
10736 vector signed short vec_vsplth (vector signed short, const int);
10737 vector unsigned short vec_vsplth (vector unsigned short, const int);
10738 vector pixel vec_vsplth (vector pixel, const int);
10740 vector signed char vec_vspltb (vector signed char, const int);
10741 vector unsigned char vec_vspltb (vector unsigned char, const int);
10742 vector bool char vec_vspltb (vector bool char, const int);
10744 vector signed char vec_splat_s8 (const int);
10746 vector signed short vec_splat_s16 (const int);
10748 vector signed int vec_splat_s32 (const int);
10750 vector unsigned char vec_splat_u8 (const int);
10752 vector unsigned short vec_splat_u16 (const int);
10754 vector unsigned int vec_splat_u32 (const int);
10756 vector signed char vec_sr (vector signed char, vector unsigned char);
10757 vector unsigned char vec_sr (vector unsigned char,
10758 vector unsigned char);
10759 vector signed short vec_sr (vector signed short,
10760 vector unsigned short);
10761 vector unsigned short vec_sr (vector unsigned short,
10762 vector unsigned short);
10763 vector signed int vec_sr (vector signed int, vector unsigned int);
10764 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
10766 vector signed int vec_vsrw (vector signed int, vector unsigned int);
10767 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
10769 vector signed short vec_vsrh (vector signed short,
10770 vector unsigned short);
10771 vector unsigned short vec_vsrh (vector unsigned short,
10772 vector unsigned short);
10774 vector signed char vec_vsrb (vector signed char, vector unsigned char);
10775 vector unsigned char vec_vsrb (vector unsigned char,
10776 vector unsigned char);
10778 vector signed char vec_sra (vector signed char, vector unsigned char);
10779 vector unsigned char vec_sra (vector unsigned char,
10780 vector unsigned char);
10781 vector signed short vec_sra (vector signed short,
10782 vector unsigned short);
10783 vector unsigned short vec_sra (vector unsigned short,
10784 vector unsigned short);
10785 vector signed int vec_sra (vector signed int, vector unsigned int);
10786 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
10788 vector signed int vec_vsraw (vector signed int, vector unsigned int);
10789 vector unsigned int vec_vsraw (vector unsigned int,
10790 vector unsigned int);
10792 vector signed short vec_vsrah (vector signed short,
10793 vector unsigned short);
10794 vector unsigned short vec_vsrah (vector unsigned short,
10795 vector unsigned short);
10797 vector signed char vec_vsrab (vector signed char, vector unsigned char);
10798 vector unsigned char vec_vsrab (vector unsigned char,
10799 vector unsigned char);
10801 vector signed int vec_srl (vector signed int, vector unsigned int);
10802 vector signed int vec_srl (vector signed int, vector unsigned short);
10803 vector signed int vec_srl (vector signed int, vector unsigned char);
10804 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
10805 vector unsigned int vec_srl (vector unsigned int,
10806 vector unsigned short);
10807 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
10808 vector bool int vec_srl (vector bool int, vector unsigned int);
10809 vector bool int vec_srl (vector bool int, vector unsigned short);
10810 vector bool int vec_srl (vector bool int, vector unsigned char);
10811 vector signed short vec_srl (vector signed short, vector unsigned int);
10812 vector signed short vec_srl (vector signed short,
10813 vector unsigned short);
10814 vector signed short vec_srl (vector signed short, vector unsigned char);
10815 vector unsigned short vec_srl (vector unsigned short,
10816 vector unsigned int);
10817 vector unsigned short vec_srl (vector unsigned short,
10818 vector unsigned short);
10819 vector unsigned short vec_srl (vector unsigned short,
10820 vector unsigned char);
10821 vector bool short vec_srl (vector bool short, vector unsigned int);
10822 vector bool short vec_srl (vector bool short, vector unsigned short);
10823 vector bool short vec_srl (vector bool short, vector unsigned char);
10824 vector pixel vec_srl (vector pixel, vector unsigned int);
10825 vector pixel vec_srl (vector pixel, vector unsigned short);
10826 vector pixel vec_srl (vector pixel, vector unsigned char);
10827 vector signed char vec_srl (vector signed char, vector unsigned int);
10828 vector signed char vec_srl (vector signed char, vector unsigned short);
10829 vector signed char vec_srl (vector signed char, vector unsigned char);
10830 vector unsigned char vec_srl (vector unsigned char,
10831 vector unsigned int);
10832 vector unsigned char vec_srl (vector unsigned char,
10833 vector unsigned short);
10834 vector unsigned char vec_srl (vector unsigned char,
10835 vector unsigned char);
10836 vector bool char vec_srl (vector bool char, vector unsigned int);
10837 vector bool char vec_srl (vector bool char, vector unsigned short);
10838 vector bool char vec_srl (vector bool char, vector unsigned char);
10840 vector float vec_sro (vector float, vector signed char);
10841 vector float vec_sro (vector float, vector unsigned char);
10842 vector signed int vec_sro (vector signed int, vector signed char);
10843 vector signed int vec_sro (vector signed int, vector unsigned char);
10844 vector unsigned int vec_sro (vector unsigned int, vector signed char);
10845 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
10846 vector signed short vec_sro (vector signed short, vector signed char);
10847 vector signed short vec_sro (vector signed short, vector unsigned char);
10848 vector unsigned short vec_sro (vector unsigned short,
10849 vector signed char);
10850 vector unsigned short vec_sro (vector unsigned short,
10851 vector unsigned char);
10852 vector pixel vec_sro (vector pixel, vector signed char);
10853 vector pixel vec_sro (vector pixel, vector unsigned char);
10854 vector signed char vec_sro (vector signed char, vector signed char);
10855 vector signed char vec_sro (vector signed char, vector unsigned char);
10856 vector unsigned char vec_sro (vector unsigned char, vector signed char);
10857 vector unsigned char vec_sro (vector unsigned char,
10858 vector unsigned char);
10860 void vec_st (vector float, int, vector float *);
10861 void vec_st (vector float, int, float *);
10862 void vec_st (vector signed int, int, vector signed int *);
10863 void vec_st (vector signed int, int, int *);
10864 void vec_st (vector unsigned int, int, vector unsigned int *);
10865 void vec_st (vector unsigned int, int, unsigned int *);
10866 void vec_st (vector bool int, int, vector bool int *);
10867 void vec_st (vector bool int, int, unsigned int *);
10868 void vec_st (vector bool int, int, int *);
10869 void vec_st (vector signed short, int, vector signed short *);
10870 void vec_st (vector signed short, int, short *);
10871 void vec_st (vector unsigned short, int, vector unsigned short *);
10872 void vec_st (vector unsigned short, int, unsigned short *);
10873 void vec_st (vector bool short, int, vector bool short *);
10874 void vec_st (vector bool short, int, unsigned short *);
10875 void vec_st (vector pixel, int, vector pixel *);
10876 void vec_st (vector pixel, int, unsigned short *);
10877 void vec_st (vector pixel, int, short *);
10878 void vec_st (vector bool short, int, short *);
10879 void vec_st (vector signed char, int, vector signed char *);
10880 void vec_st (vector signed char, int, signed char *);
10881 void vec_st (vector unsigned char, int, vector unsigned char *);
10882 void vec_st (vector unsigned char, int, unsigned char *);
10883 void vec_st (vector bool char, int, vector bool char *);
10884 void vec_st (vector bool char, int, unsigned char *);
10885 void vec_st (vector bool char, int, signed char *);
10887 void vec_ste (vector signed char, int, signed char *);
10888 void vec_ste (vector unsigned char, int, unsigned char *);
10889 void vec_ste (vector bool char, int, signed char *);
10890 void vec_ste (vector bool char, int, unsigned char *);
10891 void vec_ste (vector signed short, int, short *);
10892 void vec_ste (vector unsigned short, int, unsigned short *);
10893 void vec_ste (vector bool short, int, short *);
10894 void vec_ste (vector bool short, int, unsigned short *);
10895 void vec_ste (vector pixel, int, short *);
10896 void vec_ste (vector pixel, int, unsigned short *);
10897 void vec_ste (vector float, int, float *);
10898 void vec_ste (vector signed int, int, int *);
10899 void vec_ste (vector unsigned int, int, unsigned int *);
10900 void vec_ste (vector bool int, int, int *);
10901 void vec_ste (vector bool int, int, unsigned int *);
10903 void vec_stvewx (vector float, int, float *);
10904 void vec_stvewx (vector signed int, int, int *);
10905 void vec_stvewx (vector unsigned int, int, unsigned int *);
10906 void vec_stvewx (vector bool int, int, int *);
10907 void vec_stvewx (vector bool int, int, unsigned int *);
10909 void vec_stvehx (vector signed short, int, short *);
10910 void vec_stvehx (vector unsigned short, int, unsigned short *);
10911 void vec_stvehx (vector bool short, int, short *);
10912 void vec_stvehx (vector bool short, int, unsigned short *);
10913 void vec_stvehx (vector pixel, int, short *);
10914 void vec_stvehx (vector pixel, int, unsigned short *);
10916 void vec_stvebx (vector signed char, int, signed char *);
10917 void vec_stvebx (vector unsigned char, int, unsigned char *);
10918 void vec_stvebx (vector bool char, int, signed char *);
10919 void vec_stvebx (vector bool char, int, unsigned char *);
10921 void vec_stl (vector float, int, vector float *);
10922 void vec_stl (vector float, int, float *);
10923 void vec_stl (vector signed int, int, vector signed int *);
10924 void vec_stl (vector signed int, int, int *);
10925 void vec_stl (vector unsigned int, int, vector unsigned int *);
10926 void vec_stl (vector unsigned int, int, unsigned int *);
10927 void vec_stl (vector bool int, int, vector bool int *);
10928 void vec_stl (vector bool int, int, unsigned int *);
10929 void vec_stl (vector bool int, int, int *);
10930 void vec_stl (vector signed short, int, vector signed short *);
10931 void vec_stl (vector signed short, int, short *);
10932 void vec_stl (vector unsigned short, int, vector unsigned short *);
10933 void vec_stl (vector unsigned short, int, unsigned short *);
10934 void vec_stl (vector bool short, int, vector bool short *);
10935 void vec_stl (vector bool short, int, unsigned short *);
10936 void vec_stl (vector bool short, int, short *);
10937 void vec_stl (vector pixel, int, vector pixel *);
10938 void vec_stl (vector pixel, int, unsigned short *);
10939 void vec_stl (vector pixel, int, short *);
10940 void vec_stl (vector signed char, int, vector signed char *);
10941 void vec_stl (vector signed char, int, signed char *);
10942 void vec_stl (vector unsigned char, int, vector unsigned char *);
10943 void vec_stl (vector unsigned char, int, unsigned char *);
10944 void vec_stl (vector bool char, int, vector bool char *);
10945 void vec_stl (vector bool char, int, unsigned char *);
10946 void vec_stl (vector bool char, int, signed char *);
10948 vector signed char vec_sub (vector bool char, vector signed char);
10949 vector signed char vec_sub (vector signed char, vector bool char);
10950 vector signed char vec_sub (vector signed char, vector signed char);
10951 vector unsigned char vec_sub (vector bool char, vector unsigned char);
10952 vector unsigned char vec_sub (vector unsigned char, vector bool char);
10953 vector unsigned char vec_sub (vector unsigned char,
10954 vector unsigned char);
10955 vector signed short vec_sub (vector bool short, vector signed short);
10956 vector signed short vec_sub (vector signed short, vector bool short);
10957 vector signed short vec_sub (vector signed short, vector signed short);
10958 vector unsigned short vec_sub (vector bool short,
10959 vector unsigned short);
10960 vector unsigned short vec_sub (vector unsigned short,
10961 vector bool short);
10962 vector unsigned short vec_sub (vector unsigned short,
10963 vector unsigned short);
10964 vector signed int vec_sub (vector bool int, vector signed int);
10965 vector signed int vec_sub (vector signed int, vector bool int);
10966 vector signed int vec_sub (vector signed int, vector signed int);
10967 vector unsigned int vec_sub (vector bool int, vector unsigned int);
10968 vector unsigned int vec_sub (vector unsigned int, vector bool int);
10969 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
10970 vector float vec_sub (vector float, vector float);
10972 vector float vec_vsubfp (vector float, vector float);
10974 vector signed int vec_vsubuwm (vector bool int, vector signed int);
10975 vector signed int vec_vsubuwm (vector signed int, vector bool int);
10976 vector signed int vec_vsubuwm (vector signed int, vector signed int);
10977 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
10978 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
10979 vector unsigned int vec_vsubuwm (vector unsigned int,
10980 vector unsigned int);
10982 vector signed short vec_vsubuhm (vector bool short,
10983 vector signed short);
10984 vector signed short vec_vsubuhm (vector signed short,
10985 vector bool short);
10986 vector signed short vec_vsubuhm (vector signed short,
10987 vector signed short);
10988 vector unsigned short vec_vsubuhm (vector bool short,
10989 vector unsigned short);
10990 vector unsigned short vec_vsubuhm (vector unsigned short,
10991 vector bool short);
10992 vector unsigned short vec_vsubuhm (vector unsigned short,
10993 vector unsigned short);
10995 vector signed char vec_vsububm (vector bool char, vector signed char);
10996 vector signed char vec_vsububm (vector signed char, vector bool char);
10997 vector signed char vec_vsububm (vector signed char, vector signed char);
10998 vector unsigned char vec_vsububm (vector bool char,
10999 vector unsigned char);
11000 vector unsigned char vec_vsububm (vector unsigned char,
11002 vector unsigned char vec_vsububm (vector unsigned char,
11003 vector unsigned char);
11005 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
11007 vector unsigned char vec_subs (vector bool char, vector unsigned char);
11008 vector unsigned char vec_subs (vector unsigned char, vector bool char);
11009 vector unsigned char vec_subs (vector unsigned char,
11010 vector unsigned char);
11011 vector signed char vec_subs (vector bool char, vector signed char);
11012 vector signed char vec_subs (vector signed char, vector bool char);
11013 vector signed char vec_subs (vector signed char, vector signed char);
11014 vector unsigned short vec_subs (vector bool short,
11015 vector unsigned short);
11016 vector unsigned short vec_subs (vector unsigned short,
11017 vector bool short);
11018 vector unsigned short vec_subs (vector unsigned short,
11019 vector unsigned short);
11020 vector signed short vec_subs (vector bool short, vector signed short);
11021 vector signed short vec_subs (vector signed short, vector bool short);
11022 vector signed short vec_subs (vector signed short, vector signed short);
11023 vector unsigned int vec_subs (vector bool int, vector unsigned int);
11024 vector unsigned int vec_subs (vector unsigned int, vector bool int);
11025 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
11026 vector signed int vec_subs (vector bool int, vector signed int);
11027 vector signed int vec_subs (vector signed int, vector bool int);
11028 vector signed int vec_subs (vector signed int, vector signed int);
11030 vector signed int vec_vsubsws (vector bool int, vector signed int);
11031 vector signed int vec_vsubsws (vector signed int, vector bool int);
11032 vector signed int vec_vsubsws (vector signed int, vector signed int);
11034 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
11035 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
11036 vector unsigned int vec_vsubuws (vector unsigned int,
11037 vector unsigned int);
11039 vector signed short vec_vsubshs (vector bool short,
11040 vector signed short);
11041 vector signed short vec_vsubshs (vector signed short,
11042 vector bool short);
11043 vector signed short vec_vsubshs (vector signed short,
11044 vector signed short);
11046 vector unsigned short vec_vsubuhs (vector bool short,
11047 vector unsigned short);
11048 vector unsigned short vec_vsubuhs (vector unsigned short,
11049 vector bool short);
11050 vector unsigned short vec_vsubuhs (vector unsigned short,
11051 vector unsigned short);
11053 vector signed char vec_vsubsbs (vector bool char, vector signed char);
11054 vector signed char vec_vsubsbs (vector signed char, vector bool char);
11055 vector signed char vec_vsubsbs (vector signed char, vector signed char);
11057 vector unsigned char vec_vsububs (vector bool char,
11058 vector unsigned char);
11059 vector unsigned char vec_vsububs (vector unsigned char,
11061 vector unsigned char vec_vsububs (vector unsigned char,
11062 vector unsigned char);
11064 vector unsigned int vec_sum4s (vector unsigned char,
11065 vector unsigned int);
11066 vector signed int vec_sum4s (vector signed char, vector signed int);
11067 vector signed int vec_sum4s (vector signed short, vector signed int);
11069 vector signed int vec_vsum4shs (vector signed short, vector signed int);
11071 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
11073 vector unsigned int vec_vsum4ubs (vector unsigned char,
11074 vector unsigned int);
11076 vector signed int vec_sum2s (vector signed int, vector signed int);
11078 vector signed int vec_sums (vector signed int, vector signed int);
11080 vector float vec_trunc (vector float);
11082 vector signed short vec_unpackh (vector signed char);
11083 vector bool short vec_unpackh (vector bool char);
11084 vector signed int vec_unpackh (vector signed short);
11085 vector bool int vec_unpackh (vector bool short);
11086 vector unsigned int vec_unpackh (vector pixel);
11088 vector bool int vec_vupkhsh (vector bool short);
11089 vector signed int vec_vupkhsh (vector signed short);
11091 vector unsigned int vec_vupkhpx (vector pixel);
11093 vector bool short vec_vupkhsb (vector bool char);
11094 vector signed short vec_vupkhsb (vector signed char);
11096 vector signed short vec_unpackl (vector signed char);
11097 vector bool short vec_unpackl (vector bool char);
11098 vector unsigned int vec_unpackl (vector pixel);
11099 vector signed int vec_unpackl (vector signed short);
11100 vector bool int vec_unpackl (vector bool short);
11102 vector unsigned int vec_vupklpx (vector pixel);
11104 vector bool int vec_vupklsh (vector bool short);
11105 vector signed int vec_vupklsh (vector signed short);
11107 vector bool short vec_vupklsb (vector bool char);
11108 vector signed short vec_vupklsb (vector signed char);
11110 vector float vec_xor (vector float, vector float);
11111 vector float vec_xor (vector float, vector bool int);
11112 vector float vec_xor (vector bool int, vector float);
11113 vector bool int vec_xor (vector bool int, vector bool int);
11114 vector signed int vec_xor (vector bool int, vector signed int);
11115 vector signed int vec_xor (vector signed int, vector bool int);
11116 vector signed int vec_xor (vector signed int, vector signed int);
11117 vector unsigned int vec_xor (vector bool int, vector unsigned int);
11118 vector unsigned int vec_xor (vector unsigned int, vector bool int);
11119 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
11120 vector bool short vec_xor (vector bool short, vector bool short);
11121 vector signed short vec_xor (vector bool short, vector signed short);
11122 vector signed short vec_xor (vector signed short, vector bool short);
11123 vector signed short vec_xor (vector signed short, vector signed short);
11124 vector unsigned short vec_xor (vector bool short,
11125 vector unsigned short);
11126 vector unsigned short vec_xor (vector unsigned short,
11127 vector bool short);
11128 vector unsigned short vec_xor (vector unsigned short,
11129 vector unsigned short);
11130 vector signed char vec_xor (vector bool char, vector signed char);
11131 vector bool char vec_xor (vector bool char, vector bool char);
11132 vector signed char vec_xor (vector signed char, vector bool char);
11133 vector signed char vec_xor (vector signed char, vector signed char);
11134 vector unsigned char vec_xor (vector bool char, vector unsigned char);
11135 vector unsigned char vec_xor (vector unsigned char, vector bool char);
11136 vector unsigned char vec_xor (vector unsigned char,
11137 vector unsigned char);
11139 int vec_all_eq (vector signed char, vector bool char);
11140 int vec_all_eq (vector signed char, vector signed char);
11141 int vec_all_eq (vector unsigned char, vector bool char);
11142 int vec_all_eq (vector unsigned char, vector unsigned char);
11143 int vec_all_eq (vector bool char, vector bool char);
11144 int vec_all_eq (vector bool char, vector unsigned char);
11145 int vec_all_eq (vector bool char, vector signed char);
11146 int vec_all_eq (vector signed short, vector bool short);
11147 int vec_all_eq (vector signed short, vector signed short);
11148 int vec_all_eq (vector unsigned short, vector bool short);
11149 int vec_all_eq (vector unsigned short, vector unsigned short);
11150 int vec_all_eq (vector bool short, vector bool short);
11151 int vec_all_eq (vector bool short, vector unsigned short);
11152 int vec_all_eq (vector bool short, vector signed short);
11153 int vec_all_eq (vector pixel, vector pixel);
11154 int vec_all_eq (vector signed int, vector bool int);
11155 int vec_all_eq (vector signed int, vector signed int);
11156 int vec_all_eq (vector unsigned int, vector bool int);
11157 int vec_all_eq (vector unsigned int, vector unsigned int);
11158 int vec_all_eq (vector bool int, vector bool int);
11159 int vec_all_eq (vector bool int, vector unsigned int);
11160 int vec_all_eq (vector bool int, vector signed int);
11161 int vec_all_eq (vector float, vector float);
11163 int vec_all_ge (vector bool char, vector unsigned char);
11164 int vec_all_ge (vector unsigned char, vector bool char);
11165 int vec_all_ge (vector unsigned char, vector unsigned char);
11166 int vec_all_ge (vector bool char, vector signed char);
11167 int vec_all_ge (vector signed char, vector bool char);
11168 int vec_all_ge (vector signed char, vector signed char);
11169 int vec_all_ge (vector bool short, vector unsigned short);
11170 int vec_all_ge (vector unsigned short, vector bool short);
11171 int vec_all_ge (vector unsigned short, vector unsigned short);
11172 int vec_all_ge (vector signed short, vector signed short);
11173 int vec_all_ge (vector bool short, vector signed short);
11174 int vec_all_ge (vector signed short, vector bool short);
11175 int vec_all_ge (vector bool int, vector unsigned int);
11176 int vec_all_ge (vector unsigned int, vector bool int);
11177 int vec_all_ge (vector unsigned int, vector unsigned int);
11178 int vec_all_ge (vector bool int, vector signed int);
11179 int vec_all_ge (vector signed int, vector bool int);
11180 int vec_all_ge (vector signed int, vector signed int);
11181 int vec_all_ge (vector float, vector float);
11183 int vec_all_gt (vector bool char, vector unsigned char);
11184 int vec_all_gt (vector unsigned char, vector bool char);
11185 int vec_all_gt (vector unsigned char, vector unsigned char);
11186 int vec_all_gt (vector bool char, vector signed char);
11187 int vec_all_gt (vector signed char, vector bool char);
11188 int vec_all_gt (vector signed char, vector signed char);
11189 int vec_all_gt (vector bool short, vector unsigned short);
11190 int vec_all_gt (vector unsigned short, vector bool short);
11191 int vec_all_gt (vector unsigned short, vector unsigned short);
11192 int vec_all_gt (vector bool short, vector signed short);
11193 int vec_all_gt (vector signed short, vector bool short);
11194 int vec_all_gt (vector signed short, vector signed short);
11195 int vec_all_gt (vector bool int, vector unsigned int);
11196 int vec_all_gt (vector unsigned int, vector bool int);
11197 int vec_all_gt (vector unsigned int, vector unsigned int);
11198 int vec_all_gt (vector bool int, vector signed int);
11199 int vec_all_gt (vector signed int, vector bool int);
11200 int vec_all_gt (vector signed int, vector signed int);
11201 int vec_all_gt (vector float, vector float);
11203 int vec_all_in (vector float, vector float);
11205 int vec_all_le (vector bool char, vector unsigned char);
11206 int vec_all_le (vector unsigned char, vector bool char);
11207 int vec_all_le (vector unsigned char, vector unsigned char);
11208 int vec_all_le (vector bool char, vector signed char);
11209 int vec_all_le (vector signed char, vector bool char);
11210 int vec_all_le (vector signed char, vector signed char);
11211 int vec_all_le (vector bool short, vector unsigned short);
11212 int vec_all_le (vector unsigned short, vector bool short);
11213 int vec_all_le (vector unsigned short, vector unsigned short);
11214 int vec_all_le (vector bool short, vector signed short);
11215 int vec_all_le (vector signed short, vector bool short);
11216 int vec_all_le (vector signed short, vector signed short);
11217 int vec_all_le (vector bool int, vector unsigned int);
11218 int vec_all_le (vector unsigned int, vector bool int);
11219 int vec_all_le (vector unsigned int, vector unsigned int);
11220 int vec_all_le (vector bool int, vector signed int);
11221 int vec_all_le (vector signed int, vector bool int);
11222 int vec_all_le (vector signed int, vector signed int);
11223 int vec_all_le (vector float, vector float);
11225 int vec_all_lt (vector bool char, vector unsigned char);
11226 int vec_all_lt (vector unsigned char, vector bool char);
11227 int vec_all_lt (vector unsigned char, vector unsigned char);
11228 int vec_all_lt (vector bool char, vector signed char);
11229 int vec_all_lt (vector signed char, vector bool char);
11230 int vec_all_lt (vector signed char, vector signed char);
11231 int vec_all_lt (vector bool short, vector unsigned short);
11232 int vec_all_lt (vector unsigned short, vector bool short);
11233 int vec_all_lt (vector unsigned short, vector unsigned short);
11234 int vec_all_lt (vector bool short, vector signed short);
11235 int vec_all_lt (vector signed short, vector bool short);
11236 int vec_all_lt (vector signed short, vector signed short);
11237 int vec_all_lt (vector bool int, vector unsigned int);
11238 int vec_all_lt (vector unsigned int, vector bool int);
11239 int vec_all_lt (vector unsigned int, vector unsigned int);
11240 int vec_all_lt (vector bool int, vector signed int);
11241 int vec_all_lt (vector signed int, vector bool int);
11242 int vec_all_lt (vector signed int, vector signed int);
11243 int vec_all_lt (vector float, vector float);
11245 int vec_all_nan (vector float);
11247 int vec_all_ne (vector signed char, vector bool char);
11248 int vec_all_ne (vector signed char, vector signed char);
11249 int vec_all_ne (vector unsigned char, vector bool char);
11250 int vec_all_ne (vector unsigned char, vector unsigned char);
11251 int vec_all_ne (vector bool char, vector bool char);
11252 int vec_all_ne (vector bool char, vector unsigned char);
11253 int vec_all_ne (vector bool char, vector signed char);
11254 int vec_all_ne (vector signed short, vector bool short);
11255 int vec_all_ne (vector signed short, vector signed short);
11256 int vec_all_ne (vector unsigned short, vector bool short);
11257 int vec_all_ne (vector unsigned short, vector unsigned short);
11258 int vec_all_ne (vector bool short, vector bool short);
11259 int vec_all_ne (vector bool short, vector unsigned short);
11260 int vec_all_ne (vector bool short, vector signed short);
11261 int vec_all_ne (vector pixel, vector pixel);
11262 int vec_all_ne (vector signed int, vector bool int);
11263 int vec_all_ne (vector signed int, vector signed int);
11264 int vec_all_ne (vector unsigned int, vector bool int);
11265 int vec_all_ne (vector unsigned int, vector unsigned int);
11266 int vec_all_ne (vector bool int, vector bool int);
11267 int vec_all_ne (vector bool int, vector unsigned int);
11268 int vec_all_ne (vector bool int, vector signed int);
11269 int vec_all_ne (vector float, vector float);
11271 int vec_all_nge (vector float, vector float);
11273 int vec_all_ngt (vector float, vector float);
11275 int vec_all_nle (vector float, vector float);
11277 int vec_all_nlt (vector float, vector float);
11279 int vec_all_numeric (vector float);
11281 int vec_any_eq (vector signed char, vector bool char);
11282 int vec_any_eq (vector signed char, vector signed char);
11283 int vec_any_eq (vector unsigned char, vector bool char);
11284 int vec_any_eq (vector unsigned char, vector unsigned char);
11285 int vec_any_eq (vector bool char, vector bool char);
11286 int vec_any_eq (vector bool char, vector unsigned char);
11287 int vec_any_eq (vector bool char, vector signed char);
11288 int vec_any_eq (vector signed short, vector bool short);
11289 int vec_any_eq (vector signed short, vector signed short);
11290 int vec_any_eq (vector unsigned short, vector bool short);
11291 int vec_any_eq (vector unsigned short, vector unsigned short);
11292 int vec_any_eq (vector bool short, vector bool short);
11293 int vec_any_eq (vector bool short, vector unsigned short);
11294 int vec_any_eq (vector bool short, vector signed short);
11295 int vec_any_eq (vector pixel, vector pixel);
11296 int vec_any_eq (vector signed int, vector bool int);
11297 int vec_any_eq (vector signed int, vector signed int);
11298 int vec_any_eq (vector unsigned int, vector bool int);
11299 int vec_any_eq (vector unsigned int, vector unsigned int);
11300 int vec_any_eq (vector bool int, vector bool int);
11301 int vec_any_eq (vector bool int, vector unsigned int);
11302 int vec_any_eq (vector bool int, vector signed int);
11303 int vec_any_eq (vector float, vector float);
11305 int vec_any_ge (vector signed char, vector bool char);
11306 int vec_any_ge (vector unsigned char, vector bool char);
11307 int vec_any_ge (vector unsigned char, vector unsigned char);
11308 int vec_any_ge (vector signed char, vector signed char);
11309 int vec_any_ge (vector bool char, vector unsigned char);
11310 int vec_any_ge (vector bool char, vector signed char);
11311 int vec_any_ge (vector unsigned short, vector bool short);
11312 int vec_any_ge (vector unsigned short, vector unsigned short);
11313 int vec_any_ge (vector signed short, vector signed short);
11314 int vec_any_ge (vector signed short, vector bool short);
11315 int vec_any_ge (vector bool short, vector unsigned short);
11316 int vec_any_ge (vector bool short, vector signed short);
11317 int vec_any_ge (vector signed int, vector bool int);
11318 int vec_any_ge (vector unsigned int, vector bool int);
11319 int vec_any_ge (vector unsigned int, vector unsigned int);
11320 int vec_any_ge (vector signed int, vector signed int);
11321 int vec_any_ge (vector bool int, vector unsigned int);
11322 int vec_any_ge (vector bool int, vector signed int);
11323 int vec_any_ge (vector float, vector float);
11325 int vec_any_gt (vector bool char, vector unsigned char);
11326 int vec_any_gt (vector unsigned char, vector bool char);
11327 int vec_any_gt (vector unsigned char, vector unsigned char);
11328 int vec_any_gt (vector bool char, vector signed char);
11329 int vec_any_gt (vector signed char, vector bool char);
11330 int vec_any_gt (vector signed char, vector signed char);
11331 int vec_any_gt (vector bool short, vector unsigned short);
11332 int vec_any_gt (vector unsigned short, vector bool short);
11333 int vec_any_gt (vector unsigned short, vector unsigned short);
11334 int vec_any_gt (vector bool short, vector signed short);
11335 int vec_any_gt (vector signed short, vector bool short);
11336 int vec_any_gt (vector signed short, vector signed short);
11337 int vec_any_gt (vector bool int, vector unsigned int);
11338 int vec_any_gt (vector unsigned int, vector bool int);
11339 int vec_any_gt (vector unsigned int, vector unsigned int);
11340 int vec_any_gt (vector bool int, vector signed int);
11341 int vec_any_gt (vector signed int, vector bool int);
11342 int vec_any_gt (vector signed int, vector signed int);
11343 int vec_any_gt (vector float, vector float);
11345 int vec_any_le (vector bool char, vector unsigned char);
11346 int vec_any_le (vector unsigned char, vector bool char);
11347 int vec_any_le (vector unsigned char, vector unsigned char);
11348 int vec_any_le (vector bool char, vector signed char);
11349 int vec_any_le (vector signed char, vector bool char);
11350 int vec_any_le (vector signed char, vector signed char);
11351 int vec_any_le (vector bool short, vector unsigned short);
11352 int vec_any_le (vector unsigned short, vector bool short);
11353 int vec_any_le (vector unsigned short, vector unsigned short);
11354 int vec_any_le (vector bool short, vector signed short);
11355 int vec_any_le (vector signed short, vector bool short);
11356 int vec_any_le (vector signed short, vector signed short);
11357 int vec_any_le (vector bool int, vector unsigned int);
11358 int vec_any_le (vector unsigned int, vector bool int);
11359 int vec_any_le (vector unsigned int, vector unsigned int);
11360 int vec_any_le (vector bool int, vector signed int);
11361 int vec_any_le (vector signed int, vector bool int);
11362 int vec_any_le (vector signed int, vector signed int);
11363 int vec_any_le (vector float, vector float);
11365 int vec_any_lt (vector bool char, vector unsigned char);
11366 int vec_any_lt (vector unsigned char, vector bool char);
11367 int vec_any_lt (vector unsigned char, vector unsigned char);
11368 int vec_any_lt (vector bool char, vector signed char);
11369 int vec_any_lt (vector signed char, vector bool char);
11370 int vec_any_lt (vector signed char, vector signed char);
11371 int vec_any_lt (vector bool short, vector unsigned short);
11372 int vec_any_lt (vector unsigned short, vector bool short);
11373 int vec_any_lt (vector unsigned short, vector unsigned short);
11374 int vec_any_lt (vector bool short, vector signed short);
11375 int vec_any_lt (vector signed short, vector bool short);
11376 int vec_any_lt (vector signed short, vector signed short);
11377 int vec_any_lt (vector bool int, vector unsigned int);
11378 int vec_any_lt (vector unsigned int, vector bool int);
11379 int vec_any_lt (vector unsigned int, vector unsigned int);
11380 int vec_any_lt (vector bool int, vector signed int);
11381 int vec_any_lt (vector signed int, vector bool int);
11382 int vec_any_lt (vector signed int, vector signed int);
11383 int vec_any_lt (vector float, vector float);
11385 int vec_any_nan (vector float);
11387 int vec_any_ne (vector signed char, vector bool char);
11388 int vec_any_ne (vector signed char, vector signed char);
11389 int vec_any_ne (vector unsigned char, vector bool char);
11390 int vec_any_ne (vector unsigned char, vector unsigned char);
11391 int vec_any_ne (vector bool char, vector bool char);
11392 int vec_any_ne (vector bool char, vector unsigned char);
11393 int vec_any_ne (vector bool char, vector signed char);
11394 int vec_any_ne (vector signed short, vector bool short);
11395 int vec_any_ne (vector signed short, vector signed short);
11396 int vec_any_ne (vector unsigned short, vector bool short);
11397 int vec_any_ne (vector unsigned short, vector unsigned short);
11398 int vec_any_ne (vector bool short, vector bool short);
11399 int vec_any_ne (vector bool short, vector unsigned short);
11400 int vec_any_ne (vector bool short, vector signed short);
11401 int vec_any_ne (vector pixel, vector pixel);
11402 int vec_any_ne (vector signed int, vector bool int);
11403 int vec_any_ne (vector signed int, vector signed int);
11404 int vec_any_ne (vector unsigned int, vector bool int);
11405 int vec_any_ne (vector unsigned int, vector unsigned int);
11406 int vec_any_ne (vector bool int, vector bool int);
11407 int vec_any_ne (vector bool int, vector unsigned int);
11408 int vec_any_ne (vector bool int, vector signed int);
11409 int vec_any_ne (vector float, vector float);
11411 int vec_any_nge (vector float, vector float);
11413 int vec_any_ngt (vector float, vector float);
11415 int vec_any_nle (vector float, vector float);
11417 int vec_any_nlt (vector float, vector float);
11419 int vec_any_numeric (vector float);
11421 int vec_any_out (vector float, vector float);
11424 @node SPARC VIS Built-in Functions
11425 @subsection SPARC VIS Built-in Functions
11427 GCC supports SIMD operations on the SPARC using both the generic vector
11428 extensions (@pxref{Vector Extensions}) as well as built-in functions for
11429 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
11430 switch, the VIS extension is exposed as the following built-in functions:
11433 typedef int v2si __attribute__ ((vector_size (8)));
11434 typedef short v4hi __attribute__ ((vector_size (8)));
11435 typedef short v2hi __attribute__ ((vector_size (4)));
11436 typedef char v8qi __attribute__ ((vector_size (8)));
11437 typedef char v4qi __attribute__ ((vector_size (4)));
11439 void * __builtin_vis_alignaddr (void *, long);
11440 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
11441 v2si __builtin_vis_faligndatav2si (v2si, v2si);
11442 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
11443 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
11445 v4hi __builtin_vis_fexpand (v4qi);
11447 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
11448 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
11449 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
11450 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
11451 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
11452 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
11453 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
11455 v4qi __builtin_vis_fpack16 (v4hi);
11456 v8qi __builtin_vis_fpack32 (v2si, v2si);
11457 v2hi __builtin_vis_fpackfix (v2si);
11458 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
11460 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
11463 @node SPU Built-in Functions
11464 @subsection SPU Built-in Functions
11466 GCC provides extensions for the SPU processor as described in the
11467 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
11468 found at @uref{http://cell.scei.co.jp/} or
11469 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
11470 implementation differs in several ways.
11475 The optional extension of specifying vector constants in parentheses is
11479 A vector initializer requires no cast if the vector constant is of the
11480 same type as the variable it is initializing.
11483 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11484 vector type is the default signedness of the base type. The default
11485 varies depending on the operating system, so a portable program should
11486 always specify the signedness.
11489 By default, the keyword @code{__vector} is added. The macro
11490 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
11494 GCC allows using a @code{typedef} name as the type specifier for a
11498 For C, overloaded functions are implemented with macros so the following
11502 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11505 Since @code{spu_add} is a macro, the vector constant in the example
11506 is treated as four separate arguments. Wrap the entire argument in
11507 parentheses for this to work.
11510 The extended version of @code{__builtin_expect} is not supported.
11514 @emph{Note:} Only the interface described in the aforementioned
11515 specification is supported. Internally, GCC uses built-in functions to
11516 implement the required functionality, but these are not supported and
11517 are subject to change without notice.
11519 @node Target Format Checks
11520 @section Format Checks Specific to Particular Target Machines
11522 For some target machines, GCC supports additional options to the
11524 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
11527 * Solaris Format Checks::
11530 @node Solaris Format Checks
11531 @subsection Solaris Format Checks
11533 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
11534 check. @code{cmn_err} accepts a subset of the standard @code{printf}
11535 conversions, and the two-argument @code{%b} conversion for displaying
11536 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
11539 @section Pragmas Accepted by GCC
11543 GCC supports several types of pragmas, primarily in order to compile
11544 code originally written for other compilers. Note that in general
11545 we do not recommend the use of pragmas; @xref{Function Attributes},
11546 for further explanation.
11551 * RS/6000 and PowerPC Pragmas::
11553 * Solaris Pragmas::
11554 * Symbol-Renaming Pragmas::
11555 * Structure-Packing Pragmas::
11557 * Diagnostic Pragmas::
11558 * Visibility Pragmas::
11559 * Push/Pop Macro Pragmas::
11560 * Function Specific Option Pragmas::
11564 @subsection ARM Pragmas
11566 The ARM target defines pragmas for controlling the default addition of
11567 @code{long_call} and @code{short_call} attributes to functions.
11568 @xref{Function Attributes}, for information about the effects of these
11573 @cindex pragma, long_calls
11574 Set all subsequent functions to have the @code{long_call} attribute.
11576 @item no_long_calls
11577 @cindex pragma, no_long_calls
11578 Set all subsequent functions to have the @code{short_call} attribute.
11580 @item long_calls_off
11581 @cindex pragma, long_calls_off
11582 Do not affect the @code{long_call} or @code{short_call} attributes of
11583 subsequent functions.
11587 @subsection M32C Pragmas
11590 @item memregs @var{number}
11591 @cindex pragma, memregs
11592 Overrides the command line option @code{-memregs=} for the current
11593 file. Use with care! This pragma must be before any function in the
11594 file, and mixing different memregs values in different objects may
11595 make them incompatible. This pragma is useful when a
11596 performance-critical function uses a memreg for temporary values,
11597 as it may allow you to reduce the number of memregs used.
11601 @node RS/6000 and PowerPC Pragmas
11602 @subsection RS/6000 and PowerPC Pragmas
11604 The RS/6000 and PowerPC targets define one pragma for controlling
11605 whether or not the @code{longcall} attribute is added to function
11606 declarations by default. This pragma overrides the @option{-mlongcall}
11607 option, but not the @code{longcall} and @code{shortcall} attributes.
11608 @xref{RS/6000 and PowerPC Options}, for more information about when long
11609 calls are and are not necessary.
11613 @cindex pragma, longcall
11614 Apply the @code{longcall} attribute to all subsequent function
11618 Do not apply the @code{longcall} attribute to subsequent function
11622 @c Describe h8300 pragmas here.
11623 @c Describe sh pragmas here.
11624 @c Describe v850 pragmas here.
11626 @node Darwin Pragmas
11627 @subsection Darwin Pragmas
11629 The following pragmas are available for all architectures running the
11630 Darwin operating system. These are useful for compatibility with other
11634 @item mark @var{tokens}@dots{}
11635 @cindex pragma, mark
11636 This pragma is accepted, but has no effect.
11638 @item options align=@var{alignment}
11639 @cindex pragma, options align
11640 This pragma sets the alignment of fields in structures. The values of
11641 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
11642 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
11643 properly; to restore the previous setting, use @code{reset} for the
11646 @item segment @var{tokens}@dots{}
11647 @cindex pragma, segment
11648 This pragma is accepted, but has no effect.
11650 @item unused (@var{var} [, @var{var}]@dots{})
11651 @cindex pragma, unused
11652 This pragma declares variables to be possibly unused. GCC will not
11653 produce warnings for the listed variables. The effect is similar to
11654 that of the @code{unused} attribute, except that this pragma may appear
11655 anywhere within the variables' scopes.
11658 @node Solaris Pragmas
11659 @subsection Solaris Pragmas
11661 The Solaris target supports @code{#pragma redefine_extname}
11662 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
11663 @code{#pragma} directives for compatibility with the system compiler.
11666 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
11667 @cindex pragma, align
11669 Increase the minimum alignment of each @var{variable} to @var{alignment}.
11670 This is the same as GCC's @code{aligned} attribute @pxref{Variable
11671 Attributes}). Macro expansion occurs on the arguments to this pragma
11672 when compiling C and Objective-C@. It does not currently occur when
11673 compiling C++, but this is a bug which may be fixed in a future
11676 @item fini (@var{function} [, @var{function}]...)
11677 @cindex pragma, fini
11679 This pragma causes each listed @var{function} to be called after
11680 main, or during shared module unloading, by adding a call to the
11681 @code{.fini} section.
11683 @item init (@var{function} [, @var{function}]...)
11684 @cindex pragma, init
11686 This pragma causes each listed @var{function} to be called during
11687 initialization (before @code{main}) or during shared module loading, by
11688 adding a call to the @code{.init} section.
11692 @node Symbol-Renaming Pragmas
11693 @subsection Symbol-Renaming Pragmas
11695 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
11696 supports two @code{#pragma} directives which change the name used in
11697 assembly for a given declaration. These pragmas are only available on
11698 platforms whose system headers need them. To get this effect on all
11699 platforms supported by GCC, use the asm labels extension (@pxref{Asm
11703 @item redefine_extname @var{oldname} @var{newname}
11704 @cindex pragma, redefine_extname
11706 This pragma gives the C function @var{oldname} the assembly symbol
11707 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
11708 will be defined if this pragma is available (currently only on
11711 @item extern_prefix @var{string}
11712 @cindex pragma, extern_prefix
11714 This pragma causes all subsequent external function and variable
11715 declarations to have @var{string} prepended to their assembly symbols.
11716 This effect may be terminated with another @code{extern_prefix} pragma
11717 whose argument is an empty string. The preprocessor macro
11718 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
11719 available (currently only on Tru64 UNIX)@.
11722 These pragmas and the asm labels extension interact in a complicated
11723 manner. Here are some corner cases you may want to be aware of.
11726 @item Both pragmas silently apply only to declarations with external
11727 linkage. Asm labels do not have this restriction.
11729 @item In C++, both pragmas silently apply only to declarations with
11730 ``C'' linkage. Again, asm labels do not have this restriction.
11732 @item If any of the three ways of changing the assembly name of a
11733 declaration is applied to a declaration whose assembly name has
11734 already been determined (either by a previous use of one of these
11735 features, or because the compiler needed the assembly name in order to
11736 generate code), and the new name is different, a warning issues and
11737 the name does not change.
11739 @item The @var{oldname} used by @code{#pragma redefine_extname} is
11740 always the C-language name.
11742 @item If @code{#pragma extern_prefix} is in effect, and a declaration
11743 occurs with an asm label attached, the prefix is silently ignored for
11746 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
11747 apply to the same declaration, whichever triggered first wins, and a
11748 warning issues if they contradict each other. (We would like to have
11749 @code{#pragma redefine_extname} always win, for consistency with asm
11750 labels, but if @code{#pragma extern_prefix} triggers first we have no
11751 way of knowing that that happened.)
11754 @node Structure-Packing Pragmas
11755 @subsection Structure-Packing Pragmas
11757 For compatibility with Microsoft Windows compilers, GCC supports a
11758 set of @code{#pragma} directives which change the maximum alignment of
11759 members of structures (other than zero-width bitfields), unions, and
11760 classes subsequently defined. The @var{n} value below always is required
11761 to be a small power of two and specifies the new alignment in bytes.
11764 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
11765 @item @code{#pragma pack()} sets the alignment to the one that was in
11766 effect when compilation started (see also command line option
11767 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
11768 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
11769 setting on an internal stack and then optionally sets the new alignment.
11770 @item @code{#pragma pack(pop)} restores the alignment setting to the one
11771 saved at the top of the internal stack (and removes that stack entry).
11772 Note that @code{#pragma pack([@var{n}])} does not influence this internal
11773 stack; thus it is possible to have @code{#pragma pack(push)} followed by
11774 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
11775 @code{#pragma pack(pop)}.
11778 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
11779 @code{#pragma} which lays out a structure as the documented
11780 @code{__attribute__ ((ms_struct))}.
11782 @item @code{#pragma ms_struct on} turns on the layout for structures
11784 @item @code{#pragma ms_struct off} turns off the layout for structures
11786 @item @code{#pragma ms_struct reset} goes back to the default layout.
11790 @subsection Weak Pragmas
11792 For compatibility with SVR4, GCC supports a set of @code{#pragma}
11793 directives for declaring symbols to be weak, and defining weak
11797 @item #pragma weak @var{symbol}
11798 @cindex pragma, weak
11799 This pragma declares @var{symbol} to be weak, as if the declaration
11800 had the attribute of the same name. The pragma may appear before
11801 or after the declaration of @var{symbol}, but must appear before
11802 either its first use or its definition. It is not an error for
11803 @var{symbol} to never be defined at all.
11805 @item #pragma weak @var{symbol1} = @var{symbol2}
11806 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
11807 It is an error if @var{symbol2} is not defined in the current
11811 @node Diagnostic Pragmas
11812 @subsection Diagnostic Pragmas
11814 GCC allows the user to selectively enable or disable certain types of
11815 diagnostics, and change the kind of the diagnostic. For example, a
11816 project's policy might require that all sources compile with
11817 @option{-Werror} but certain files might have exceptions allowing
11818 specific types of warnings. Or, a project might selectively enable
11819 diagnostics and treat them as errors depending on which preprocessor
11820 macros are defined.
11823 @item #pragma GCC diagnostic @var{kind} @var{option}
11824 @cindex pragma, diagnostic
11826 Modifies the disposition of a diagnostic. Note that not all
11827 diagnostics are modifiable; at the moment only warnings (normally
11828 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
11829 Use @option{-fdiagnostics-show-option} to determine which diagnostics
11830 are controllable and which option controls them.
11832 @var{kind} is @samp{error} to treat this diagnostic as an error,
11833 @samp{warning} to treat it like a warning (even if @option{-Werror} is
11834 in effect), or @samp{ignored} if the diagnostic is to be ignored.
11835 @var{option} is a double quoted string which matches the command line
11839 #pragma GCC diagnostic warning "-Wformat"
11840 #pragma GCC diagnostic error "-Wformat"
11841 #pragma GCC diagnostic ignored "-Wformat"
11844 Note that these pragmas override any command line options. Also,
11845 while it is syntactically valid to put these pragmas anywhere in your
11846 sources, the only supported location for them is before any data or
11847 functions are defined. Doing otherwise may result in unpredictable
11848 results depending on how the optimizer manages your sources. If the
11849 same option is listed multiple times, the last one specified is the
11850 one that is in effect. This pragma is not intended to be a general
11851 purpose replacement for command line options, but for implementing
11852 strict control over project policies.
11856 GCC also offers a simple mechanism for printing messages during
11860 @item #pragma message @var{string}
11861 @cindex pragma, diagnostic
11863 Prints @var{string} as a compiler message on compilation. The message
11864 is informational only, and is neither a compilation warning nor an error.
11867 #pragma message "Compiling " __FILE__ "..."
11870 @var{string} may be parenthesized, and is printed with location
11871 information. For example,
11874 #define DO_PRAGMA(x) _Pragma (#x)
11875 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
11877 TODO(Remember to fix this)
11880 prints @samp{/tmp/file.c:4: note: #pragma message:
11881 TODO - Remember to fix this}.
11885 @node Visibility Pragmas
11886 @subsection Visibility Pragmas
11889 @item #pragma GCC visibility push(@var{visibility})
11890 @itemx #pragma GCC visibility pop
11891 @cindex pragma, visibility
11893 This pragma allows the user to set the visibility for multiple
11894 declarations without having to give each a visibility attribute
11895 @xref{Function Attributes}, for more information about visibility and
11896 the attribute syntax.
11898 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
11899 declarations. Class members and template specializations are not
11900 affected; if you want to override the visibility for a particular
11901 member or instantiation, you must use an attribute.
11906 @node Push/Pop Macro Pragmas
11907 @subsection Push/Pop Macro Pragmas
11909 For compatibility with Microsoft Windows compilers, GCC supports
11910 @samp{#pragma push_macro(@var{"macro_name"})}
11911 and @samp{#pragma pop_macro(@var{"macro_name"})}.
11914 @item #pragma push_macro(@var{"macro_name"})
11915 @cindex pragma, push_macro
11916 This pragma saves the value of the macro named as @var{macro_name} to
11917 the top of the stack for this macro.
11919 @item #pragma pop_macro(@var{"macro_name"})
11920 @cindex pragma, pop_macro
11921 This pragma sets the value of the macro named as @var{macro_name} to
11922 the value on top of the stack for this macro. If the stack for
11923 @var{macro_name} is empty, the value of the macro remains unchanged.
11930 #pragma push_macro("X")
11933 #pragma pop_macro("X")
11937 In this example, the definition of X as 1 is saved by @code{#pragma
11938 push_macro} and restored by @code{#pragma pop_macro}.
11940 @node Function Specific Option Pragmas
11941 @subsection Function Specific Option Pragmas
11944 @item #pragma GCC target (@var{"string"}...)
11945 @cindex pragma GCC target
11947 This pragma allows you to set target specific options for functions
11948 defined later in the source file. One or more strings can be
11949 specified. Each function that is defined after this point will be as
11950 if @code{attribute((target("STRING")))} was specified for that
11951 function. The parenthesis around the options is optional.
11952 @xref{Function Attributes}, for more information about the
11953 @code{target} attribute and the attribute syntax.
11955 The @samp{#pragma GCC target} pragma is not implemented in GCC
11956 versions earlier than 4.4, and is currently only implemented for the
11957 386 and x86_64 backends.
11961 @item #pragma GCC optimize (@var{"string"}...)
11962 @cindex pragma GCC optimize
11964 This pragma allows you to set global optimization options for functions
11965 defined later in the source file. One or more strings can be
11966 specified. Each function that is defined after this point will be as
11967 if @code{attribute((optimize("STRING")))} was specified for that
11968 function. The parenthesis around the options is optional.
11969 @xref{Function Attributes}, for more information about the
11970 @code{optimize} attribute and the attribute syntax.
11972 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
11973 versions earlier than 4.4.
11977 @item #pragma GCC push_options
11978 @itemx #pragma GCC pop_options
11979 @cindex pragma GCC push_options
11980 @cindex pragma GCC pop_options
11982 These pragmas maintain a stack of the current target and optimization
11983 options. It is intended for include files where you temporarily want
11984 to switch to using a different @samp{#pragma GCC target} or
11985 @samp{#pragma GCC optimize} and then to pop back to the previous
11988 The @samp{#pragma GCC push_options} and @samp{#pragma GCC pop_options}
11989 pragmas are not implemented in GCC versions earlier than 4.4.
11993 @item #pragma GCC reset_options
11994 @cindex pragma GCC reset_options
11996 This pragma clears the current @code{#pragma GCC target} and
11997 @code{#pragma GCC optimize} to use the default switches as specified
11998 on the command line.
12000 The @samp{#pragma GCC reset_options} pragma is not implemented in GCC
12001 versions earlier than 4.4.
12004 @node Unnamed Fields
12005 @section Unnamed struct/union fields within structs/unions
12009 For compatibility with other compilers, GCC allows you to define
12010 a structure or union that contains, as fields, structures and unions
12011 without names. For example:
12024 In this example, the user would be able to access members of the unnamed
12025 union with code like @samp{foo.b}. Note that only unnamed structs and
12026 unions are allowed, you may not have, for example, an unnamed
12029 You must never create such structures that cause ambiguous field definitions.
12030 For example, this structure:
12041 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
12042 Such constructs are not supported and must be avoided. In the future,
12043 such constructs may be detected and treated as compilation errors.
12045 @opindex fms-extensions
12046 Unless @option{-fms-extensions} is used, the unnamed field must be a
12047 structure or union definition without a tag (for example, @samp{struct
12048 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
12049 also be a definition with a tag such as @samp{struct foo @{ int a;
12050 @};}, a reference to a previously defined structure or union such as
12051 @samp{struct foo;}, or a reference to a @code{typedef} name for a
12052 previously defined structure or union type.
12055 @section Thread-Local Storage
12056 @cindex Thread-Local Storage
12057 @cindex @acronym{TLS}
12060 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
12061 are allocated such that there is one instance of the variable per extant
12062 thread. The run-time model GCC uses to implement this originates
12063 in the IA-64 processor-specific ABI, but has since been migrated
12064 to other processors as well. It requires significant support from
12065 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
12066 system libraries (@file{libc.so} and @file{libpthread.so}), so it
12067 is not available everywhere.
12069 At the user level, the extension is visible with a new storage
12070 class keyword: @code{__thread}. For example:
12074 extern __thread struct state s;
12075 static __thread char *p;
12078 The @code{__thread} specifier may be used alone, with the @code{extern}
12079 or @code{static} specifiers, but with no other storage class specifier.
12080 When used with @code{extern} or @code{static}, @code{__thread} must appear
12081 immediately after the other storage class specifier.
12083 The @code{__thread} specifier may be applied to any global, file-scoped
12084 static, function-scoped static, or static data member of a class. It may
12085 not be applied to block-scoped automatic or non-static data member.
12087 When the address-of operator is applied to a thread-local variable, it is
12088 evaluated at run-time and returns the address of the current thread's
12089 instance of that variable. An address so obtained may be used by any
12090 thread. When a thread terminates, any pointers to thread-local variables
12091 in that thread become invalid.
12093 No static initialization may refer to the address of a thread-local variable.
12095 In C++, if an initializer is present for a thread-local variable, it must
12096 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
12099 See @uref{http://people.redhat.com/drepper/tls.pdf,
12100 ELF Handling For Thread-Local Storage} for a detailed explanation of
12101 the four thread-local storage addressing models, and how the run-time
12102 is expected to function.
12105 * C99 Thread-Local Edits::
12106 * C++98 Thread-Local Edits::
12109 @node C99 Thread-Local Edits
12110 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
12112 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
12113 that document the exact semantics of the language extension.
12117 @cite{5.1.2 Execution environments}
12119 Add new text after paragraph 1
12122 Within either execution environment, a @dfn{thread} is a flow of
12123 control within a program. It is implementation defined whether
12124 or not there may be more than one thread associated with a program.
12125 It is implementation defined how threads beyond the first are
12126 created, the name and type of the function called at thread
12127 startup, and how threads may be terminated. However, objects
12128 with thread storage duration shall be initialized before thread
12133 @cite{6.2.4 Storage durations of objects}
12135 Add new text before paragraph 3
12138 An object whose identifier is declared with the storage-class
12139 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
12140 Its lifetime is the entire execution of the thread, and its
12141 stored value is initialized only once, prior to thread startup.
12145 @cite{6.4.1 Keywords}
12147 Add @code{__thread}.
12150 @cite{6.7.1 Storage-class specifiers}
12152 Add @code{__thread} to the list of storage class specifiers in
12155 Change paragraph 2 to
12158 With the exception of @code{__thread}, at most one storage-class
12159 specifier may be given [@dots{}]. The @code{__thread} specifier may
12160 be used alone, or immediately following @code{extern} or
12164 Add new text after paragraph 6
12167 The declaration of an identifier for a variable that has
12168 block scope that specifies @code{__thread} shall also
12169 specify either @code{extern} or @code{static}.
12171 The @code{__thread} specifier shall be used only with
12176 @node C++98 Thread-Local Edits
12177 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
12179 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
12180 that document the exact semantics of the language extension.
12184 @b{[intro.execution]}
12186 New text after paragraph 4
12189 A @dfn{thread} is a flow of control within the abstract machine.
12190 It is implementation defined whether or not there may be more than
12194 New text after paragraph 7
12197 It is unspecified whether additional action must be taken to
12198 ensure when and whether side effects are visible to other threads.
12204 Add @code{__thread}.
12207 @b{[basic.start.main]}
12209 Add after paragraph 5
12212 The thread that begins execution at the @code{main} function is called
12213 the @dfn{main thread}. It is implementation defined how functions
12214 beginning threads other than the main thread are designated or typed.
12215 A function so designated, as well as the @code{main} function, is called
12216 a @dfn{thread startup function}. It is implementation defined what
12217 happens if a thread startup function returns. It is implementation
12218 defined what happens to other threads when any thread calls @code{exit}.
12222 @b{[basic.start.init]}
12224 Add after paragraph 4
12227 The storage for an object of thread storage duration shall be
12228 statically initialized before the first statement of the thread startup
12229 function. An object of thread storage duration shall not require
12230 dynamic initialization.
12234 @b{[basic.start.term]}
12236 Add after paragraph 3
12239 The type of an object with thread storage duration shall not have a
12240 non-trivial destructor, nor shall it be an array type whose elements
12241 (directly or indirectly) have non-trivial destructors.
12247 Add ``thread storage duration'' to the list in paragraph 1.
12252 Thread, static, and automatic storage durations are associated with
12253 objects introduced by declarations [@dots{}].
12256 Add @code{__thread} to the list of specifiers in paragraph 3.
12259 @b{[basic.stc.thread]}
12261 New section before @b{[basic.stc.static]}
12264 The keyword @code{__thread} applied to a non-local object gives the
12265 object thread storage duration.
12267 A local variable or class data member declared both @code{static}
12268 and @code{__thread} gives the variable or member thread storage
12273 @b{[basic.stc.static]}
12278 All objects which have neither thread storage duration, dynamic
12279 storage duration nor are local [@dots{}].
12285 Add @code{__thread} to the list in paragraph 1.
12290 With the exception of @code{__thread}, at most one
12291 @var{storage-class-specifier} shall appear in a given
12292 @var{decl-specifier-seq}. The @code{__thread} specifier may
12293 be used alone, or immediately following the @code{extern} or
12294 @code{static} specifiers. [@dots{}]
12297 Add after paragraph 5
12300 The @code{__thread} specifier can be applied only to the names of objects
12301 and to anonymous unions.
12307 Add after paragraph 6
12310 Non-@code{static} members shall not be @code{__thread}.
12314 @node Binary constants
12315 @section Binary constants using the @samp{0b} prefix
12316 @cindex Binary constants using the @samp{0b} prefix
12318 Integer constants can be written as binary constants, consisting of a
12319 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
12320 @samp{0B}. This is particularly useful in environments that operate a
12321 lot on the bit-level (like microcontrollers).
12323 The following statements are identical:
12332 The type of these constants follows the same rules as for octal or
12333 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
12336 @node C++ Extensions
12337 @chapter Extensions to the C++ Language
12338 @cindex extensions, C++ language
12339 @cindex C++ language extensions
12341 The GNU compiler provides these extensions to the C++ language (and you
12342 can also use most of the C language extensions in your C++ programs). If you
12343 want to write code that checks whether these features are available, you can
12344 test for the GNU compiler the same way as for C programs: check for a
12345 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
12346 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
12347 Predefined Macros,cpp,The GNU C Preprocessor}).
12350 * Volatiles:: What constitutes an access to a volatile object.
12351 * Restricted Pointers:: C99 restricted pointers and references.
12352 * Vague Linkage:: Where G++ puts inlines, vtables and such.
12353 * C++ Interface:: You can use a single C++ header file for both
12354 declarations and definitions.
12355 * Template Instantiation:: Methods for ensuring that exactly one copy of
12356 each needed template instantiation is emitted.
12357 * Bound member functions:: You can extract a function pointer to the
12358 method denoted by a @samp{->*} or @samp{.*} expression.
12359 * C++ Attributes:: Variable, function, and type attributes for C++ only.
12360 * Namespace Association:: Strong using-directives for namespace association.
12361 * Type Traits:: Compiler support for type traits
12362 * Java Exceptions:: Tweaking exception handling to work with Java.
12363 * Deprecated Features:: Things will disappear from g++.
12364 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
12368 @section When is a Volatile Object Accessed?
12369 @cindex accessing volatiles
12370 @cindex volatile read
12371 @cindex volatile write
12372 @cindex volatile access
12374 Both the C and C++ standard have the concept of volatile objects. These
12375 are normally accessed by pointers and used for accessing hardware. The
12376 standards encourage compilers to refrain from optimizations concerning
12377 accesses to volatile objects. The C standard leaves it implementation
12378 defined as to what constitutes a volatile access. The C++ standard omits
12379 to specify this, except to say that C++ should behave in a similar manner
12380 to C with respect to volatiles, where possible. The minimum either
12381 standard specifies is that at a sequence point all previous accesses to
12382 volatile objects have stabilized and no subsequent accesses have
12383 occurred. Thus an implementation is free to reorder and combine
12384 volatile accesses which occur between sequence points, but cannot do so
12385 for accesses across a sequence point. The use of volatiles does not
12386 allow you to violate the restriction on updating objects multiple times
12387 within a sequence point.
12389 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
12391 The behavior differs slightly between C and C++ in the non-obvious cases:
12394 volatile int *src = @var{somevalue};
12398 With C, such expressions are rvalues, and GCC interprets this either as a
12399 read of the volatile object being pointed to or only as request to evaluate
12400 the side-effects. The C++ standard specifies that such expressions do not
12401 undergo lvalue to rvalue conversion, and that the type of the dereferenced
12402 object may be incomplete. The C++ standard does not specify explicitly
12403 that it is this lvalue to rvalue conversion which may be responsible for
12404 causing an access. However, there is reason to believe that it is,
12405 because otherwise certain simple expressions become undefined. However,
12406 because it would surprise most programmers, G++ treats dereferencing a
12407 pointer to volatile object of complete type when the value is unused as
12408 GCC would do for an equivalent type in C@. When the object has incomplete
12409 type, G++ issues a warning; if you wish to force an error, you must
12410 force a conversion to rvalue with, for instance, a static cast.
12412 When using a reference to volatile, G++ does not treat equivalent
12413 expressions as accesses to volatiles, but instead issues a warning that
12414 no volatile is accessed. The rationale for this is that otherwise it
12415 becomes difficult to determine where volatile access occur, and not
12416 possible to ignore the return value from functions returning volatile
12417 references. Again, if you wish to force a read, cast the reference to
12420 @node Restricted Pointers
12421 @section Restricting Pointer Aliasing
12422 @cindex restricted pointers
12423 @cindex restricted references
12424 @cindex restricted this pointer
12426 As with the C front end, G++ understands the C99 feature of restricted pointers,
12427 specified with the @code{__restrict__}, or @code{__restrict} type
12428 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
12429 language flag, @code{restrict} is not a keyword in C++.
12431 In addition to allowing restricted pointers, you can specify restricted
12432 references, which indicate that the reference is not aliased in the local
12436 void fn (int *__restrict__ rptr, int &__restrict__ rref)
12443 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
12444 @var{rref} refers to a (different) unaliased integer.
12446 You may also specify whether a member function's @var{this} pointer is
12447 unaliased by using @code{__restrict__} as a member function qualifier.
12450 void T::fn () __restrict__
12457 Within the body of @code{T::fn}, @var{this} will have the effective
12458 definition @code{T *__restrict__ const this}. Notice that the
12459 interpretation of a @code{__restrict__} member function qualifier is
12460 different to that of @code{const} or @code{volatile} qualifier, in that it
12461 is applied to the pointer rather than the object. This is consistent with
12462 other compilers which implement restricted pointers.
12464 As with all outermost parameter qualifiers, @code{__restrict__} is
12465 ignored in function definition matching. This means you only need to
12466 specify @code{__restrict__} in a function definition, rather than
12467 in a function prototype as well.
12469 @node Vague Linkage
12470 @section Vague Linkage
12471 @cindex vague linkage
12473 There are several constructs in C++ which require space in the object
12474 file but are not clearly tied to a single translation unit. We say that
12475 these constructs have ``vague linkage''. Typically such constructs are
12476 emitted wherever they are needed, though sometimes we can be more
12480 @item Inline Functions
12481 Inline functions are typically defined in a header file which can be
12482 included in many different compilations. Hopefully they can usually be
12483 inlined, but sometimes an out-of-line copy is necessary, if the address
12484 of the function is taken or if inlining fails. In general, we emit an
12485 out-of-line copy in all translation units where one is needed. As an
12486 exception, we only emit inline virtual functions with the vtable, since
12487 it will always require a copy.
12489 Local static variables and string constants used in an inline function
12490 are also considered to have vague linkage, since they must be shared
12491 between all inlined and out-of-line instances of the function.
12495 C++ virtual functions are implemented in most compilers using a lookup
12496 table, known as a vtable. The vtable contains pointers to the virtual
12497 functions provided by a class, and each object of the class contains a
12498 pointer to its vtable (or vtables, in some multiple-inheritance
12499 situations). If the class declares any non-inline, non-pure virtual
12500 functions, the first one is chosen as the ``key method'' for the class,
12501 and the vtable is only emitted in the translation unit where the key
12504 @emph{Note:} If the chosen key method is later defined as inline, the
12505 vtable will still be emitted in every translation unit which defines it.
12506 Make sure that any inline virtuals are declared inline in the class
12507 body, even if they are not defined there.
12509 @item type_info objects
12512 C++ requires information about types to be written out in order to
12513 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
12514 For polymorphic classes (classes with virtual functions), the type_info
12515 object is written out along with the vtable so that @samp{dynamic_cast}
12516 can determine the dynamic type of a class object at runtime. For all
12517 other types, we write out the type_info object when it is used: when
12518 applying @samp{typeid} to an expression, throwing an object, or
12519 referring to a type in a catch clause or exception specification.
12521 @item Template Instantiations
12522 Most everything in this section also applies to template instantiations,
12523 but there are other options as well.
12524 @xref{Template Instantiation,,Where's the Template?}.
12528 When used with GNU ld version 2.8 or later on an ELF system such as
12529 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
12530 these constructs will be discarded at link time. This is known as
12533 On targets that don't support COMDAT, but do support weak symbols, GCC
12534 will use them. This way one copy will override all the others, but
12535 the unused copies will still take up space in the executable.
12537 For targets which do not support either COMDAT or weak symbols,
12538 most entities with vague linkage will be emitted as local symbols to
12539 avoid duplicate definition errors from the linker. This will not happen
12540 for local statics in inlines, however, as having multiple copies will
12541 almost certainly break things.
12543 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
12544 another way to control placement of these constructs.
12546 @node C++ Interface
12547 @section #pragma interface and implementation
12549 @cindex interface and implementation headers, C++
12550 @cindex C++ interface and implementation headers
12551 @cindex pragmas, interface and implementation
12553 @code{#pragma interface} and @code{#pragma implementation} provide the
12554 user with a way of explicitly directing the compiler to emit entities
12555 with vague linkage (and debugging information) in a particular
12558 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
12559 most cases, because of COMDAT support and the ``key method'' heuristic
12560 mentioned in @ref{Vague Linkage}. Using them can actually cause your
12561 program to grow due to unnecessary out-of-line copies of inline
12562 functions. Currently (3.4) the only benefit of these
12563 @code{#pragma}s is reduced duplication of debugging information, and
12564 that should be addressed soon on DWARF 2 targets with the use of
12568 @item #pragma interface
12569 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
12570 @kindex #pragma interface
12571 Use this directive in @emph{header files} that define object classes, to save
12572 space in most of the object files that use those classes. Normally,
12573 local copies of certain information (backup copies of inline member
12574 functions, debugging information, and the internal tables that implement
12575 virtual functions) must be kept in each object file that includes class
12576 definitions. You can use this pragma to avoid such duplication. When a
12577 header file containing @samp{#pragma interface} is included in a
12578 compilation, this auxiliary information will not be generated (unless
12579 the main input source file itself uses @samp{#pragma implementation}).
12580 Instead, the object files will contain references to be resolved at link
12583 The second form of this directive is useful for the case where you have
12584 multiple headers with the same name in different directories. If you
12585 use this form, you must specify the same string to @samp{#pragma
12588 @item #pragma implementation
12589 @itemx #pragma implementation "@var{objects}.h"
12590 @kindex #pragma implementation
12591 Use this pragma in a @emph{main input file}, when you want full output from
12592 included header files to be generated (and made globally visible). The
12593 included header file, in turn, should use @samp{#pragma interface}.
12594 Backup copies of inline member functions, debugging information, and the
12595 internal tables used to implement virtual functions are all generated in
12596 implementation files.
12598 @cindex implied @code{#pragma implementation}
12599 @cindex @code{#pragma implementation}, implied
12600 @cindex naming convention, implementation headers
12601 If you use @samp{#pragma implementation} with no argument, it applies to
12602 an include file with the same basename@footnote{A file's @dfn{basename}
12603 was the name stripped of all leading path information and of trailing
12604 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
12605 file. For example, in @file{allclass.cc}, giving just
12606 @samp{#pragma implementation}
12607 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
12609 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
12610 an implementation file whenever you would include it from
12611 @file{allclass.cc} even if you never specified @samp{#pragma
12612 implementation}. This was deemed to be more trouble than it was worth,
12613 however, and disabled.
12615 Use the string argument if you want a single implementation file to
12616 include code from multiple header files. (You must also use
12617 @samp{#include} to include the header file; @samp{#pragma
12618 implementation} only specifies how to use the file---it doesn't actually
12621 There is no way to split up the contents of a single header file into
12622 multiple implementation files.
12625 @cindex inlining and C++ pragmas
12626 @cindex C++ pragmas, effect on inlining
12627 @cindex pragmas in C++, effect on inlining
12628 @samp{#pragma implementation} and @samp{#pragma interface} also have an
12629 effect on function inlining.
12631 If you define a class in a header file marked with @samp{#pragma
12632 interface}, the effect on an inline function defined in that class is
12633 similar to an explicit @code{extern} declaration---the compiler emits
12634 no code at all to define an independent version of the function. Its
12635 definition is used only for inlining with its callers.
12637 @opindex fno-implement-inlines
12638 Conversely, when you include the same header file in a main source file
12639 that declares it as @samp{#pragma implementation}, the compiler emits
12640 code for the function itself; this defines a version of the function
12641 that can be found via pointers (or by callers compiled without
12642 inlining). If all calls to the function can be inlined, you can avoid
12643 emitting the function by compiling with @option{-fno-implement-inlines}.
12644 If any calls were not inlined, you will get linker errors.
12646 @node Template Instantiation
12647 @section Where's the Template?
12648 @cindex template instantiation
12650 C++ templates are the first language feature to require more
12651 intelligence from the environment than one usually finds on a UNIX
12652 system. Somehow the compiler and linker have to make sure that each
12653 template instance occurs exactly once in the executable if it is needed,
12654 and not at all otherwise. There are two basic approaches to this
12655 problem, which are referred to as the Borland model and the Cfront model.
12658 @item Borland model
12659 Borland C++ solved the template instantiation problem by adding the code
12660 equivalent of common blocks to their linker; the compiler emits template
12661 instances in each translation unit that uses them, and the linker
12662 collapses them together. The advantage of this model is that the linker
12663 only has to consider the object files themselves; there is no external
12664 complexity to worry about. This disadvantage is that compilation time
12665 is increased because the template code is being compiled repeatedly.
12666 Code written for this model tends to include definitions of all
12667 templates in the header file, since they must be seen to be
12671 The AT&T C++ translator, Cfront, solved the template instantiation
12672 problem by creating the notion of a template repository, an
12673 automatically maintained place where template instances are stored. A
12674 more modern version of the repository works as follows: As individual
12675 object files are built, the compiler places any template definitions and
12676 instantiations encountered in the repository. At link time, the link
12677 wrapper adds in the objects in the repository and compiles any needed
12678 instances that were not previously emitted. The advantages of this
12679 model are more optimal compilation speed and the ability to use the
12680 system linker; to implement the Borland model a compiler vendor also
12681 needs to replace the linker. The disadvantages are vastly increased
12682 complexity, and thus potential for error; for some code this can be
12683 just as transparent, but in practice it can been very difficult to build
12684 multiple programs in one directory and one program in multiple
12685 directories. Code written for this model tends to separate definitions
12686 of non-inline member templates into a separate file, which should be
12687 compiled separately.
12690 When used with GNU ld version 2.8 or later on an ELF system such as
12691 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
12692 Borland model. On other systems, G++ implements neither automatic
12695 A future version of G++ will support a hybrid model whereby the compiler
12696 will emit any instantiations for which the template definition is
12697 included in the compile, and store template definitions and
12698 instantiation context information into the object file for the rest.
12699 The link wrapper will extract that information as necessary and invoke
12700 the compiler to produce the remaining instantiations. The linker will
12701 then combine duplicate instantiations.
12703 In the mean time, you have the following options for dealing with
12704 template instantiations:
12709 Compile your template-using code with @option{-frepo}. The compiler will
12710 generate files with the extension @samp{.rpo} listing all of the
12711 template instantiations used in the corresponding object files which
12712 could be instantiated there; the link wrapper, @samp{collect2}, will
12713 then update the @samp{.rpo} files to tell the compiler where to place
12714 those instantiations and rebuild any affected object files. The
12715 link-time overhead is negligible after the first pass, as the compiler
12716 will continue to place the instantiations in the same files.
12718 This is your best option for application code written for the Borland
12719 model, as it will just work. Code written for the Cfront model will
12720 need to be modified so that the template definitions are available at
12721 one or more points of instantiation; usually this is as simple as adding
12722 @code{#include <tmethods.cc>} to the end of each template header.
12724 For library code, if you want the library to provide all of the template
12725 instantiations it needs, just try to link all of its object files
12726 together; the link will fail, but cause the instantiations to be
12727 generated as a side effect. Be warned, however, that this may cause
12728 conflicts if multiple libraries try to provide the same instantiations.
12729 For greater control, use explicit instantiation as described in the next
12733 @opindex fno-implicit-templates
12734 Compile your code with @option{-fno-implicit-templates} to disable the
12735 implicit generation of template instances, and explicitly instantiate
12736 all the ones you use. This approach requires more knowledge of exactly
12737 which instances you need than do the others, but it's less
12738 mysterious and allows greater control. You can scatter the explicit
12739 instantiations throughout your program, perhaps putting them in the
12740 translation units where the instances are used or the translation units
12741 that define the templates themselves; you can put all of the explicit
12742 instantiations you need into one big file; or you can create small files
12749 template class Foo<int>;
12750 template ostream& operator <<
12751 (ostream&, const Foo<int>&);
12754 for each of the instances you need, and create a template instantiation
12755 library from those.
12757 If you are using Cfront-model code, you can probably get away with not
12758 using @option{-fno-implicit-templates} when compiling files that don't
12759 @samp{#include} the member template definitions.
12761 If you use one big file to do the instantiations, you may want to
12762 compile it without @option{-fno-implicit-templates} so you get all of the
12763 instances required by your explicit instantiations (but not by any
12764 other files) without having to specify them as well.
12766 G++ has extended the template instantiation syntax given in the ISO
12767 standard to allow forward declaration of explicit instantiations
12768 (with @code{extern}), instantiation of the compiler support data for a
12769 template class (i.e.@: the vtable) without instantiating any of its
12770 members (with @code{inline}), and instantiation of only the static data
12771 members of a template class, without the support data or member
12772 functions (with (@code{static}):
12775 extern template int max (int, int);
12776 inline template class Foo<int>;
12777 static template class Foo<int>;
12781 Do nothing. Pretend G++ does implement automatic instantiation
12782 management. Code written for the Borland model will work fine, but
12783 each translation unit will contain instances of each of the templates it
12784 uses. In a large program, this can lead to an unacceptable amount of code
12788 @node Bound member functions
12789 @section Extracting the function pointer from a bound pointer to member function
12791 @cindex pointer to member function
12792 @cindex bound pointer to member function
12794 In C++, pointer to member functions (PMFs) are implemented using a wide
12795 pointer of sorts to handle all the possible call mechanisms; the PMF
12796 needs to store information about how to adjust the @samp{this} pointer,
12797 and if the function pointed to is virtual, where to find the vtable, and
12798 where in the vtable to look for the member function. If you are using
12799 PMFs in an inner loop, you should really reconsider that decision. If
12800 that is not an option, you can extract the pointer to the function that
12801 would be called for a given object/PMF pair and call it directly inside
12802 the inner loop, to save a bit of time.
12804 Note that you will still be paying the penalty for the call through a
12805 function pointer; on most modern architectures, such a call defeats the
12806 branch prediction features of the CPU@. This is also true of normal
12807 virtual function calls.
12809 The syntax for this extension is
12813 extern int (A::*fp)();
12814 typedef int (*fptr)(A *);
12816 fptr p = (fptr)(a.*fp);
12819 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
12820 no object is needed to obtain the address of the function. They can be
12821 converted to function pointers directly:
12824 fptr p1 = (fptr)(&A::foo);
12827 @opindex Wno-pmf-conversions
12828 You must specify @option{-Wno-pmf-conversions} to use this extension.
12830 @node C++ Attributes
12831 @section C++-Specific Variable, Function, and Type Attributes
12833 Some attributes only make sense for C++ programs.
12836 @item init_priority (@var{priority})
12837 @cindex init_priority attribute
12840 In Standard C++, objects defined at namespace scope are guaranteed to be
12841 initialized in an order in strict accordance with that of their definitions
12842 @emph{in a given translation unit}. No guarantee is made for initializations
12843 across translation units. However, GNU C++ allows users to control the
12844 order of initialization of objects defined at namespace scope with the
12845 @code{init_priority} attribute by specifying a relative @var{priority},
12846 a constant integral expression currently bounded between 101 and 65535
12847 inclusive. Lower numbers indicate a higher priority.
12849 In the following example, @code{A} would normally be created before
12850 @code{B}, but the @code{init_priority} attribute has reversed that order:
12853 Some_Class A __attribute__ ((init_priority (2000)));
12854 Some_Class B __attribute__ ((init_priority (543)));
12858 Note that the particular values of @var{priority} do not matter; only their
12861 @item java_interface
12862 @cindex java_interface attribute
12864 This type attribute informs C++ that the class is a Java interface. It may
12865 only be applied to classes declared within an @code{extern "Java"} block.
12866 Calls to methods declared in this interface will be dispatched using GCJ's
12867 interface table mechanism, instead of regular virtual table dispatch.
12871 See also @ref{Namespace Association}.
12873 @node Namespace Association
12874 @section Namespace Association
12876 @strong{Caution:} The semantics of this extension are not fully
12877 defined. Users should refrain from using this extension as its
12878 semantics may change subtly over time. It is possible that this
12879 extension will be removed in future versions of G++.
12881 A using-directive with @code{__attribute ((strong))} is stronger
12882 than a normal using-directive in two ways:
12886 Templates from the used namespace can be specialized and explicitly
12887 instantiated as though they were members of the using namespace.
12890 The using namespace is considered an associated namespace of all
12891 templates in the used namespace for purposes of argument-dependent
12895 The used namespace must be nested within the using namespace so that
12896 normal unqualified lookup works properly.
12898 This is useful for composing a namespace transparently from
12899 implementation namespaces. For example:
12904 template <class T> struct A @{ @};
12906 using namespace debug __attribute ((__strong__));
12907 template <> struct A<int> @{ @}; // @r{ok to specialize}
12909 template <class T> void f (A<T>);
12914 f (std::A<float>()); // @r{lookup finds} std::f
12920 @section Type Traits
12922 The C++ front-end implements syntactic extensions that allow to
12923 determine at compile time various characteristics of a type (or of a
12927 @item __has_nothrow_assign (type)
12928 If @code{type} is const qualified or is a reference type then the trait is
12929 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
12930 is true, else if @code{type} is a cv class or union type with copy assignment
12931 operators that are known not to throw an exception then the trait is true,
12932 else it is false. Requires: @code{type} shall be a complete type, an array
12933 type of unknown bound, or is a @code{void} type.
12935 @item __has_nothrow_copy (type)
12936 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
12937 @code{type} is a cv class or union type with copy constructors that
12938 are known not to throw an exception then the trait is true, else it is false.
12939 Requires: @code{type} shall be a complete type, an array type of
12940 unknown bound, or is a @code{void} type.
12942 @item __has_nothrow_constructor (type)
12943 If @code{__has_trivial_constructor (type)} is true then the trait is
12944 true, else if @code{type} is a cv class or union type (or array
12945 thereof) with a default constructor that is known not to throw an
12946 exception then the trait is true, else it is false. Requires:
12947 @code{type} shall be a complete type, an array type of unknown bound,
12948 or is a @code{void} type.
12950 @item __has_trivial_assign (type)
12951 If @code{type} is const qualified or is a reference type then the trait is
12952 false. Otherwise if @code{__is_pod (type)} is true then the trait is
12953 true, else if @code{type} is a cv class or union type with a trivial
12954 copy assignment ([class.copy]) then the trait is true, else it is
12955 false. Requires: @code{type} shall be a complete type, an array type
12956 of unknown bound, or is a @code{void} type.
12958 @item __has_trivial_copy (type)
12959 If @code{__is_pod (type)} is true or @code{type} is a reference type
12960 then the trait is true, else if @code{type} is a cv class or union type
12961 with a trivial copy constructor ([class.copy]) then the trait
12962 is true, else it is false. Requires: @code{type} shall be a complete
12963 type, an array type of unknown bound, or is a @code{void} type.
12965 @item __has_trivial_constructor (type)
12966 If @code{__is_pod (type)} is true then the trait is true, else if
12967 @code{type} is a cv class or union type (or array thereof) with a
12968 trivial default constructor ([class.ctor]) then the trait is true,
12969 else it is false. Requires: @code{type} shall be a complete type, an
12970 array type of unknown bound, or is a @code{void} type.
12972 @item __has_trivial_destructor (type)
12973 If @code{__is_pod (type)} is true or @code{type} is a reference type then
12974 the trait is true, else if @code{type} is a cv class or union type (or
12975 array thereof) with a trivial destructor ([class.dtor]) then the trait
12976 is true, else it is false. Requires: @code{type} shall be a complete
12977 type, an array type of unknown bound, or is a @code{void} type.
12979 @item __has_virtual_destructor (type)
12980 If @code{type} is a class type with a virtual destructor
12981 ([class.dtor]) then the trait is true, else it is false. Requires:
12982 @code{type} shall be a complete type, an array type of unknown bound,
12983 or is a @code{void} type.
12985 @item __is_abstract (type)
12986 If @code{type} is an abstract class ([class.abstract]) then the trait
12987 is true, else it is false. Requires: @code{type} shall be a complete
12988 type, an array type of unknown bound, or is a @code{void} type.
12990 @item __is_base_of (base_type, derived_type)
12991 If @code{base_type} is a base class of @code{derived_type}
12992 ([class.derived]) then the trait is true, otherwise it is false.
12993 Top-level cv qualifications of @code{base_type} and
12994 @code{derived_type} are ignored. For the purposes of this trait, a
12995 class type is considered is own base. Requires: if @code{__is_class
12996 (base_type)} and @code{__is_class (derived_type)} are true and
12997 @code{base_type} and @code{derived_type} are not the same type
12998 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
12999 type. Diagnostic is produced if this requirement is not met.
13001 @item __is_class (type)
13002 If @code{type} is a cv class type, and not a union type
13003 ([basic.compound]) the trait is true, else it is false.
13005 @item __is_empty (type)
13006 If @code{__is_class (type)} is false then the trait is false.
13007 Otherwise @code{type} is considered empty if and only if: @code{type}
13008 has no non-static data members, or all non-static data members, if
13009 any, are bit-fields of length 0, and @code{type} has no virtual
13010 members, and @code{type} has no virtual base classes, and @code{type}
13011 has no base classes @code{base_type} for which
13012 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
13013 be a complete type, an array type of unknown bound, or is a
13016 @item __is_enum (type)
13017 If @code{type} is a cv enumeration type ([basic.compound]) the trait is
13018 true, else it is false.
13020 @item __is_pod (type)
13021 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
13022 else it is false. Requires: @code{type} shall be a complete type,
13023 an array type of unknown bound, or is a @code{void} type.
13025 @item __is_polymorphic (type)
13026 If @code{type} is a polymorphic class ([class.virtual]) then the trait
13027 is true, else it is false. Requires: @code{type} shall be a complete
13028 type, an array type of unknown bound, or is a @code{void} type.
13030 @item __is_union (type)
13031 If @code{type} is a cv union type ([basic.compound]) the trait is
13032 true, else it is false.
13036 @node Java Exceptions
13037 @section Java Exceptions
13039 The Java language uses a slightly different exception handling model
13040 from C++. Normally, GNU C++ will automatically detect when you are
13041 writing C++ code that uses Java exceptions, and handle them
13042 appropriately. However, if C++ code only needs to execute destructors
13043 when Java exceptions are thrown through it, GCC will guess incorrectly.
13044 Sample problematic code is:
13047 struct S @{ ~S(); @};
13048 extern void bar(); // @r{is written in Java, and may throw exceptions}
13057 The usual effect of an incorrect guess is a link failure, complaining of
13058 a missing routine called @samp{__gxx_personality_v0}.
13060 You can inform the compiler that Java exceptions are to be used in a
13061 translation unit, irrespective of what it might think, by writing
13062 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
13063 @samp{#pragma} must appear before any functions that throw or catch
13064 exceptions, or run destructors when exceptions are thrown through them.
13066 You cannot mix Java and C++ exceptions in the same translation unit. It
13067 is believed to be safe to throw a C++ exception from one file through
13068 another file compiled for the Java exception model, or vice versa, but
13069 there may be bugs in this area.
13071 @node Deprecated Features
13072 @section Deprecated Features
13074 In the past, the GNU C++ compiler was extended to experiment with new
13075 features, at a time when the C++ language was still evolving. Now that
13076 the C++ standard is complete, some of those features are superseded by
13077 superior alternatives. Using the old features might cause a warning in
13078 some cases that the feature will be dropped in the future. In other
13079 cases, the feature might be gone already.
13081 While the list below is not exhaustive, it documents some of the options
13082 that are now deprecated:
13085 @item -fexternal-templates
13086 @itemx -falt-external-templates
13087 These are two of the many ways for G++ to implement template
13088 instantiation. @xref{Template Instantiation}. The C++ standard clearly
13089 defines how template definitions have to be organized across
13090 implementation units. G++ has an implicit instantiation mechanism that
13091 should work just fine for standard-conforming code.
13093 @item -fstrict-prototype
13094 @itemx -fno-strict-prototype
13095 Previously it was possible to use an empty prototype parameter list to
13096 indicate an unspecified number of parameters (like C), rather than no
13097 parameters, as C++ demands. This feature has been removed, except where
13098 it is required for backwards compatibility. @xref{Backwards Compatibility}.
13101 G++ allows a virtual function returning @samp{void *} to be overridden
13102 by one returning a different pointer type. This extension to the
13103 covariant return type rules is now deprecated and will be removed from a
13106 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
13107 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
13108 and are now removed from G++. Code using these operators should be
13109 modified to use @code{std::min} and @code{std::max} instead.
13111 The named return value extension has been deprecated, and is now
13114 The use of initializer lists with new expressions has been deprecated,
13115 and is now removed from G++.
13117 Floating and complex non-type template parameters have been deprecated,
13118 and are now removed from G++.
13120 The implicit typename extension has been deprecated and is now
13123 The use of default arguments in function pointers, function typedefs
13124 and other places where they are not permitted by the standard is
13125 deprecated and will be removed from a future version of G++.
13127 G++ allows floating-point literals to appear in integral constant expressions,
13128 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
13129 This extension is deprecated and will be removed from a future version.
13131 G++ allows static data members of const floating-point type to be declared
13132 with an initializer in a class definition. The standard only allows
13133 initializers for static members of const integral types and const
13134 enumeration types so this extension has been deprecated and will be removed
13135 from a future version.
13137 @node Backwards Compatibility
13138 @section Backwards Compatibility
13139 @cindex Backwards Compatibility
13140 @cindex ARM [Annotated C++ Reference Manual]
13142 Now that there is a definitive ISO standard C++, G++ has a specification
13143 to adhere to. The C++ language evolved over time, and features that
13144 used to be acceptable in previous drafts of the standard, such as the ARM
13145 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
13146 compilation of C++ written to such drafts, G++ contains some backwards
13147 compatibilities. @emph{All such backwards compatibility features are
13148 liable to disappear in future versions of G++.} They should be considered
13149 deprecated. @xref{Deprecated Features}.
13153 If a variable is declared at for scope, it used to remain in scope until
13154 the end of the scope which contained the for statement (rather than just
13155 within the for scope). G++ retains this, but issues a warning, if such a
13156 variable is accessed outside the for scope.
13158 @item Implicit C language
13159 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
13160 scope to set the language. On such systems, all header files are
13161 implicitly scoped inside a C language scope. Also, an empty prototype
13162 @code{()} will be treated as an unspecified number of arguments, rather
13163 than no arguments, as C++ demands.