1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Floating Types:: Additional Floating Types.
37 * Decimal Float:: Decimal Floating Types.
38 * Hex Floats:: Hexadecimal floating-point constants.
39 * Fixed-Point:: Fixed-Point Types.
40 * Zero Length:: Zero-length arrays.
41 * Variable Length:: Arrays whose length is computed at run time.
42 * Empty Structures:: Structures with no members.
43 * Variadic Macros:: Macros with a variable number of arguments.
44 * Escaped Newlines:: Slightly looser rules for escaped newlines.
45 * Subscripting:: Any array can be subscripted, even if not an lvalue.
46 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
47 * Initializers:: Non-constant initializers.
48 * Compound Literals:: Compound literals give structures, unions
50 * Designated Inits:: Labeling elements of initializers.
51 * Cast to Union:: Casting to union type from any member of the union.
52 * Case Ranges:: `case 1 ... 9' and such.
53 * Mixed Declarations:: Mixing declarations and code.
54 * Function Attributes:: Declaring that functions have no side effects,
55 or that they can never return.
56 * Attribute Syntax:: Formal syntax for attributes.
57 * Function Prototypes:: Prototype declarations and old-style definitions.
58 * C++ Comments:: C++ comments are recognized.
59 * Dollar Signs:: Dollar sign is allowed in identifiers.
60 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
61 * Variable Attributes:: Specifying attributes of variables.
62 * Type Attributes:: Specifying attributes of types.
63 * Alignment:: Inquiring about the alignment of a type or variable.
64 * Inline:: Defining inline functions (as fast as macros).
65 * Extended Asm:: Assembler instructions with C expressions as operands.
66 (With them you can define ``built-in'' functions.)
67 * Constraints:: Constraints for asm operands
68 * Asm Labels:: Specifying the assembler name to use for a C symbol.
69 * Explicit Reg Vars:: Defining variables residing in specified registers.
70 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
71 * Incomplete Enums:: @code{enum foo;}, with details to follow.
72 * Function Names:: Printable strings which are the name of the current
74 * Return Address:: Getting the return or frame address of a function.
75 * Vector Extensions:: Using vector instructions through built-in functions.
76 * Offsetof:: Special syntax for implementing @code{offsetof}.
77 * Atomic Builtins:: Built-in functions for atomic memory access.
78 * Object Size Checking:: Built-in functions for limited buffer overflow
80 * Other Builtins:: Other built-in functions.
81 * Target Builtins:: Built-in functions specific to particular targets.
82 * Target Format Checks:: Format checks specific to particular targets.
83 * Pragmas:: Pragmas accepted by GCC.
84 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
85 * Thread-Local:: Per-thread variables.
86 * Binary constants:: Binary constants using the @samp{0b} prefix.
90 @section Statements and Declarations in Expressions
91 @cindex statements inside expressions
92 @cindex declarations inside expressions
93 @cindex expressions containing statements
94 @cindex macros, statements in expressions
96 @c the above section title wrapped and causes an underfull hbox.. i
97 @c changed it from "within" to "in". --mew 4feb93
98 A compound statement enclosed in parentheses may appear as an expression
99 in GNU C@. This allows you to use loops, switches, and local variables
100 within an expression.
102 Recall that a compound statement is a sequence of statements surrounded
103 by braces; in this construct, parentheses go around the braces. For
107 (@{ int y = foo (); int z;
114 is a valid (though slightly more complex than necessary) expression
115 for the absolute value of @code{foo ()}.
117 The last thing in the compound statement should be an expression
118 followed by a semicolon; the value of this subexpression serves as the
119 value of the entire construct. (If you use some other kind of statement
120 last within the braces, the construct has type @code{void}, and thus
121 effectively no value.)
123 This feature is especially useful in making macro definitions ``safe'' (so
124 that they evaluate each operand exactly once). For example, the
125 ``maximum'' function is commonly defined as a macro in standard C as
129 #define max(a,b) ((a) > (b) ? (a) : (b))
133 @cindex side effects, macro argument
134 But this definition computes either @var{a} or @var{b} twice, with bad
135 results if the operand has side effects. In GNU C, if you know the
136 type of the operands (here taken as @code{int}), you can define
137 the macro safely as follows:
140 #define maxint(a,b) \
141 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
144 Embedded statements are not allowed in constant expressions, such as
145 the value of an enumeration constant, the width of a bit-field, or
146 the initial value of a static variable.
148 If you don't know the type of the operand, you can still do this, but you
149 must use @code{typeof} (@pxref{Typeof}).
151 In G++, the result value of a statement expression undergoes array and
152 function pointer decay, and is returned by value to the enclosing
153 expression. For instance, if @code{A} is a class, then
162 will construct a temporary @code{A} object to hold the result of the
163 statement expression, and that will be used to invoke @code{Foo}.
164 Therefore the @code{this} pointer observed by @code{Foo} will not be the
167 Any temporaries created within a statement within a statement expression
168 will be destroyed at the statement's end. This makes statement
169 expressions inside macros slightly different from function calls. In
170 the latter case temporaries introduced during argument evaluation will
171 be destroyed at the end of the statement that includes the function
172 call. In the statement expression case they will be destroyed during
173 the statement expression. For instance,
176 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
177 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
187 will have different places where temporaries are destroyed. For the
188 @code{macro} case, the temporary @code{X} will be destroyed just after
189 the initialization of @code{b}. In the @code{function} case that
190 temporary will be destroyed when the function returns.
192 These considerations mean that it is probably a bad idea to use
193 statement-expressions of this form in header files that are designed to
194 work with C++. (Note that some versions of the GNU C Library contained
195 header files using statement-expression that lead to precisely this
198 Jumping into a statement expression with @code{goto} or using a
199 @code{switch} statement outside the statement expression with a
200 @code{case} or @code{default} label inside the statement expression is
201 not permitted. Jumping into a statement expression with a computed
202 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
203 Jumping out of a statement expression is permitted, but if the
204 statement expression is part of a larger expression then it is
205 unspecified which other subexpressions of that expression have been
206 evaluated except where the language definition requires certain
207 subexpressions to be evaluated before or after the statement
208 expression. In any case, as with a function call the evaluation of a
209 statement expression is not interleaved with the evaluation of other
210 parts of the containing expression. For example,
213 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
217 will call @code{foo} and @code{bar1} and will not call @code{baz} but
218 may or may not call @code{bar2}. If @code{bar2} is called, it will be
219 called after @code{foo} and before @code{bar1}
222 @section Locally Declared Labels
224 @cindex macros, local labels
226 GCC allows you to declare @dfn{local labels} in any nested block
227 scope. A local label is just like an ordinary label, but you can
228 only reference it (with a @code{goto} statement, or by taking its
229 address) within the block in which it was declared.
231 A local label declaration looks like this:
234 __label__ @var{label};
241 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
244 Local label declarations must come at the beginning of the block,
245 before any ordinary declarations or statements.
247 The label declaration defines the label @emph{name}, but does not define
248 the label itself. You must do this in the usual way, with
249 @code{@var{label}:}, within the statements of the statement expression.
251 The local label feature is useful for complex macros. If a macro
252 contains nested loops, a @code{goto} can be useful for breaking out of
253 them. However, an ordinary label whose scope is the whole function
254 cannot be used: if the macro can be expanded several times in one
255 function, the label will be multiply defined in that function. A
256 local label avoids this problem. For example:
259 #define SEARCH(value, array, target) \
262 typeof (target) _SEARCH_target = (target); \
263 typeof (*(array)) *_SEARCH_array = (array); \
266 for (i = 0; i < max; i++) \
267 for (j = 0; j < max; j++) \
268 if (_SEARCH_array[i][j] == _SEARCH_target) \
269 @{ (value) = i; goto found; @} \
275 This could also be written using a statement-expression:
278 #define SEARCH(array, target) \
281 typeof (target) _SEARCH_target = (target); \
282 typeof (*(array)) *_SEARCH_array = (array); \
285 for (i = 0; i < max; i++) \
286 for (j = 0; j < max; j++) \
287 if (_SEARCH_array[i][j] == _SEARCH_target) \
288 @{ value = i; goto found; @} \
295 Local label declarations also make the labels they declare visible to
296 nested functions, if there are any. @xref{Nested Functions}, for details.
298 @node Labels as Values
299 @section Labels as Values
300 @cindex labels as values
301 @cindex computed gotos
302 @cindex goto with computed label
303 @cindex address of a label
305 You can get the address of a label defined in the current function
306 (or a containing function) with the unary operator @samp{&&}. The
307 value has type @code{void *}. This value is a constant and can be used
308 wherever a constant of that type is valid. For example:
316 To use these values, you need to be able to jump to one. This is done
317 with the computed goto statement@footnote{The analogous feature in
318 Fortran is called an assigned goto, but that name seems inappropriate in
319 C, where one can do more than simply store label addresses in label
320 variables.}, @code{goto *@var{exp};}. For example,
327 Any expression of type @code{void *} is allowed.
329 One way of using these constants is in initializing a static array that
330 will serve as a jump table:
333 static void *array[] = @{ &&foo, &&bar, &&hack @};
336 Then you can select a label with indexing, like this:
343 Note that this does not check whether the subscript is in bounds---array
344 indexing in C never does that.
346 Such an array of label values serves a purpose much like that of the
347 @code{switch} statement. The @code{switch} statement is cleaner, so
348 use that rather than an array unless the problem does not fit a
349 @code{switch} statement very well.
351 Another use of label values is in an interpreter for threaded code.
352 The labels within the interpreter function can be stored in the
353 threaded code for super-fast dispatching.
355 You may not use this mechanism to jump to code in a different function.
356 If you do that, totally unpredictable things will happen. The best way to
357 avoid this is to store the label address only in automatic variables and
358 never pass it as an argument.
360 An alternate way to write the above example is
363 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
365 goto *(&&foo + array[i]);
369 This is more friendly to code living in shared libraries, as it reduces
370 the number of dynamic relocations that are needed, and by consequence,
371 allows the data to be read-only.
373 The @code{&&foo} expressions for the same label might have different values
374 if the containing function is inlined or cloned. If a program relies on
375 them being always the same, @code{__attribute__((__noinline__))} should
376 be used to prevent inlining. If @code{&&foo} is used
377 in a static variable initializer, inlining is forbidden.
379 @node Nested Functions
380 @section Nested Functions
381 @cindex nested functions
382 @cindex downward funargs
385 A @dfn{nested function} is a function defined inside another function.
386 (Nested functions are not supported for GNU C++.) The nested function's
387 name is local to the block where it is defined. For example, here we
388 define a nested function named @code{square}, and call it twice:
392 foo (double a, double b)
394 double square (double z) @{ return z * z; @}
396 return square (a) + square (b);
401 The nested function can access all the variables of the containing
402 function that are visible at the point of its definition. This is
403 called @dfn{lexical scoping}. For example, here we show a nested
404 function which uses an inherited variable named @code{offset}:
408 bar (int *array, int offset, int size)
410 int access (int *array, int index)
411 @{ return array[index + offset]; @}
414 for (i = 0; i < size; i++)
415 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
420 Nested function definitions are permitted within functions in the places
421 where variable definitions are allowed; that is, in any block, mixed
422 with the other declarations and statements in the block.
424 It is possible to call the nested function from outside the scope of its
425 name by storing its address or passing the address to another function:
428 hack (int *array, int size)
430 void store (int index, int value)
431 @{ array[index] = value; @}
433 intermediate (store, size);
437 Here, the function @code{intermediate} receives the address of
438 @code{store} as an argument. If @code{intermediate} calls @code{store},
439 the arguments given to @code{store} are used to store into @code{array}.
440 But this technique works only so long as the containing function
441 (@code{hack}, in this example) does not exit.
443 If you try to call the nested function through its address after the
444 containing function has exited, all hell will break loose. If you try
445 to call it after a containing scope level has exited, and if it refers
446 to some of the variables that are no longer in scope, you may be lucky,
447 but it's not wise to take the risk. If, however, the nested function
448 does not refer to anything that has gone out of scope, you should be
451 GCC implements taking the address of a nested function using a technique
452 called @dfn{trampolines}. A paper describing them is available as
455 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
457 A nested function can jump to a label inherited from a containing
458 function, provided the label was explicitly declared in the containing
459 function (@pxref{Local Labels}). Such a jump returns instantly to the
460 containing function, exiting the nested function which did the
461 @code{goto} and any intermediate functions as well. Here is an example:
465 bar (int *array, int offset, int size)
468 int access (int *array, int index)
472 return array[index + offset];
476 for (i = 0; i < size; i++)
477 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
481 /* @r{Control comes here from @code{access}
482 if it detects an error.} */
489 A nested function always has no linkage. Declaring one with
490 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
491 before its definition, use @code{auto} (which is otherwise meaningless
492 for function declarations).
495 bar (int *array, int offset, int size)
498 auto int access (int *, int);
500 int access (int *array, int index)
504 return array[index + offset];
510 @node Constructing Calls
511 @section Constructing Function Calls
512 @cindex constructing calls
513 @cindex forwarding calls
515 Using the built-in functions described below, you can record
516 the arguments a function received, and call another function
517 with the same arguments, without knowing the number or types
520 You can also record the return value of that function call,
521 and later return that value, without knowing what data type
522 the function tried to return (as long as your caller expects
525 However, these built-in functions may interact badly with some
526 sophisticated features or other extensions of the language. It
527 is, therefore, not recommended to use them outside very simple
528 functions acting as mere forwarders for their arguments.
530 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
531 This built-in function returns a pointer to data
532 describing how to perform a call with the same arguments as were passed
533 to the current function.
535 The function saves the arg pointer register, structure value address,
536 and all registers that might be used to pass arguments to a function
537 into a block of memory allocated on the stack. Then it returns the
538 address of that block.
541 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
542 This built-in function invokes @var{function}
543 with a copy of the parameters described by @var{arguments}
546 The value of @var{arguments} should be the value returned by
547 @code{__builtin_apply_args}. The argument @var{size} specifies the size
548 of the stack argument data, in bytes.
550 This function returns a pointer to data describing
551 how to return whatever value was returned by @var{function}. The data
552 is saved in a block of memory allocated on the stack.
554 It is not always simple to compute the proper value for @var{size}. The
555 value is used by @code{__builtin_apply} to compute the amount of data
556 that should be pushed on the stack and copied from the incoming argument
560 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
561 This built-in function returns the value described by @var{result} from
562 the containing function. You should specify, for @var{result}, a value
563 returned by @code{__builtin_apply}.
566 @deftypefn {Built-in Function} __builtin_va_arg_pack ()
567 This built-in function represents all anonymous arguments of an inline
568 function. It can be used only in inline functions which will be always
569 inlined, never compiled as a separate function, such as those using
570 @code{__attribute__ ((__always_inline__))} or
571 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
572 It must be only passed as last argument to some other function
573 with variable arguments. This is useful for writing small wrapper
574 inlines for variable argument functions, when using preprocessor
575 macros is undesirable. For example:
577 extern int myprintf (FILE *f, const char *format, ...);
578 extern inline __attribute__ ((__gnu_inline__)) int
579 myprintf (FILE *f, const char *format, ...)
581 int r = fprintf (f, "myprintf: ");
584 int s = fprintf (f, format, __builtin_va_arg_pack ());
592 @deftypefn {Built-in Function} __builtin_va_arg_pack_len ()
593 This built-in function returns the number of anonymous arguments of
594 an inline function. It can be used only in inline functions which
595 will be always inlined, never compiled as a separate function, such
596 as those using @code{__attribute__ ((__always_inline__))} or
597 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
598 For example following will do link or runtime checking of open
599 arguments for optimized code:
602 extern inline __attribute__((__gnu_inline__)) int
603 myopen (const char *path, int oflag, ...)
605 if (__builtin_va_arg_pack_len () > 1)
606 warn_open_too_many_arguments ();
608 if (__builtin_constant_p (oflag))
610 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
612 warn_open_missing_mode ();
613 return __open_2 (path, oflag);
615 return open (path, oflag, __builtin_va_arg_pack ());
618 if (__builtin_va_arg_pack_len () < 1)
619 return __open_2 (path, oflag);
621 return open (path, oflag, __builtin_va_arg_pack ());
628 @section Referring to a Type with @code{typeof}
631 @cindex macros, types of arguments
633 Another way to refer to the type of an expression is with @code{typeof}.
634 The syntax of using of this keyword looks like @code{sizeof}, but the
635 construct acts semantically like a type name defined with @code{typedef}.
637 There are two ways of writing the argument to @code{typeof}: with an
638 expression or with a type. Here is an example with an expression:
645 This assumes that @code{x} is an array of pointers to functions;
646 the type described is that of the values of the functions.
648 Here is an example with a typename as the argument:
655 Here the type described is that of pointers to @code{int}.
657 If you are writing a header file that must work when included in ISO C
658 programs, write @code{__typeof__} instead of @code{typeof}.
659 @xref{Alternate Keywords}.
661 A @code{typeof}-construct can be used anywhere a typedef name could be
662 used. For example, you can use it in a declaration, in a cast, or inside
663 of @code{sizeof} or @code{typeof}.
665 @code{typeof} is often useful in conjunction with the
666 statements-within-expressions feature. Here is how the two together can
667 be used to define a safe ``maximum'' macro that operates on any
668 arithmetic type and evaluates each of its arguments exactly once:
672 (@{ typeof (a) _a = (a); \
673 typeof (b) _b = (b); \
674 _a > _b ? _a : _b; @})
677 @cindex underscores in variables in macros
678 @cindex @samp{_} in variables in macros
679 @cindex local variables in macros
680 @cindex variables, local, in macros
681 @cindex macros, local variables in
683 The reason for using names that start with underscores for the local
684 variables is to avoid conflicts with variable names that occur within the
685 expressions that are substituted for @code{a} and @code{b}. Eventually we
686 hope to design a new form of declaration syntax that allows you to declare
687 variables whose scopes start only after their initializers; this will be a
688 more reliable way to prevent such conflicts.
691 Some more examples of the use of @code{typeof}:
695 This declares @code{y} with the type of what @code{x} points to.
702 This declares @code{y} as an array of such values.
709 This declares @code{y} as an array of pointers to characters:
712 typeof (typeof (char *)[4]) y;
716 It is equivalent to the following traditional C declaration:
722 To see the meaning of the declaration using @code{typeof}, and why it
723 might be a useful way to write, rewrite it with these macros:
726 #define pointer(T) typeof(T *)
727 #define array(T, N) typeof(T [N])
731 Now the declaration can be rewritten this way:
734 array (pointer (char), 4) y;
738 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
739 pointers to @code{char}.
742 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
743 a more limited extension which permitted one to write
746 typedef @var{T} = @var{expr};
750 with the effect of declaring @var{T} to have the type of the expression
751 @var{expr}. This extension does not work with GCC 3 (versions between
752 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
753 relies on it should be rewritten to use @code{typeof}:
756 typedef typeof(@var{expr}) @var{T};
760 This will work with all versions of GCC@.
763 @section Conditionals with Omitted Operands
764 @cindex conditional expressions, extensions
765 @cindex omitted middle-operands
766 @cindex middle-operands, omitted
767 @cindex extensions, @code{?:}
768 @cindex @code{?:} extensions
770 The middle operand in a conditional expression may be omitted. Then
771 if the first operand is nonzero, its value is the value of the conditional
774 Therefore, the expression
781 has the value of @code{x} if that is nonzero; otherwise, the value of
784 This example is perfectly equivalent to
790 @cindex side effect in ?:
791 @cindex ?: side effect
793 In this simple case, the ability to omit the middle operand is not
794 especially useful. When it becomes useful is when the first operand does,
795 or may (if it is a macro argument), contain a side effect. Then repeating
796 the operand in the middle would perform the side effect twice. Omitting
797 the middle operand uses the value already computed without the undesirable
798 effects of recomputing it.
801 @section Double-Word Integers
802 @cindex @code{long long} data types
803 @cindex double-word arithmetic
804 @cindex multiprecision arithmetic
805 @cindex @code{LL} integer suffix
806 @cindex @code{ULL} integer suffix
808 ISO C99 supports data types for integers that are at least 64 bits wide,
809 and as an extension GCC supports them in C89 mode and in C++.
810 Simply write @code{long long int} for a signed integer, or
811 @code{unsigned long long int} for an unsigned integer. To make an
812 integer constant of type @code{long long int}, add the suffix @samp{LL}
813 to the integer. To make an integer constant of type @code{unsigned long
814 long int}, add the suffix @samp{ULL} to the integer.
816 You can use these types in arithmetic like any other integer types.
817 Addition, subtraction, and bitwise boolean operations on these types
818 are open-coded on all types of machines. Multiplication is open-coded
819 if the machine supports fullword-to-doubleword a widening multiply
820 instruction. Division and shifts are open-coded only on machines that
821 provide special support. The operations that are not open-coded use
822 special library routines that come with GCC@.
824 There may be pitfalls when you use @code{long long} types for function
825 arguments, unless you declare function prototypes. If a function
826 expects type @code{int} for its argument, and you pass a value of type
827 @code{long long int}, confusion will result because the caller and the
828 subroutine will disagree about the number of bytes for the argument.
829 Likewise, if the function expects @code{long long int} and you pass
830 @code{int}. The best way to avoid such problems is to use prototypes.
833 @section Complex Numbers
834 @cindex complex numbers
835 @cindex @code{_Complex} keyword
836 @cindex @code{__complex__} keyword
838 ISO C99 supports complex floating data types, and as an extension GCC
839 supports them in C89 mode and in C++, and supports complex integer data
840 types which are not part of ISO C99. You can declare complex types
841 using the keyword @code{_Complex}. As an extension, the older GNU
842 keyword @code{__complex__} is also supported.
844 For example, @samp{_Complex double x;} declares @code{x} as a
845 variable whose real part and imaginary part are both of type
846 @code{double}. @samp{_Complex short int y;} declares @code{y} to
847 have real and imaginary parts of type @code{short int}; this is not
848 likely to be useful, but it shows that the set of complex types is
851 To write a constant with a complex data type, use the suffix @samp{i} or
852 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
853 has type @code{_Complex float} and @code{3i} has type
854 @code{_Complex int}. Such a constant always has a pure imaginary
855 value, but you can form any complex value you like by adding one to a
856 real constant. This is a GNU extension; if you have an ISO C99
857 conforming C library (such as GNU libc), and want to construct complex
858 constants of floating type, you should include @code{<complex.h>} and
859 use the macros @code{I} or @code{_Complex_I} instead.
861 @cindex @code{__real__} keyword
862 @cindex @code{__imag__} keyword
863 To extract the real part of a complex-valued expression @var{exp}, write
864 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
865 extract the imaginary part. This is a GNU extension; for values of
866 floating type, you should use the ISO C99 functions @code{crealf},
867 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
868 @code{cimagl}, declared in @code{<complex.h>} and also provided as
869 built-in functions by GCC@.
871 @cindex complex conjugation
872 The operator @samp{~} performs complex conjugation when used on a value
873 with a complex type. This is a GNU extension; for values of
874 floating type, you should use the ISO C99 functions @code{conjf},
875 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
876 provided as built-in functions by GCC@.
878 GCC can allocate complex automatic variables in a noncontiguous
879 fashion; it's even possible for the real part to be in a register while
880 the imaginary part is on the stack (or vice-versa). Only the DWARF2
881 debug info format can represent this, so use of DWARF2 is recommended.
882 If you are using the stabs debug info format, GCC describes a noncontiguous
883 complex variable as if it were two separate variables of noncomplex type.
884 If the variable's actual name is @code{foo}, the two fictitious
885 variables are named @code{foo$real} and @code{foo$imag}. You can
886 examine and set these two fictitious variables with your debugger.
889 @section Additional Floating Types
890 @cindex additional floating types
891 @cindex @code{__float80} data type
892 @cindex @code{__float128} data type
893 @cindex @code{w} floating point suffix
894 @cindex @code{q} floating point suffix
895 @cindex @code{W} floating point suffix
896 @cindex @code{Q} floating point suffix
898 As an extension, the GNU C compiler supports additional floating
899 types, @code{__float80} and @code{__float128} to support 80bit
900 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
901 Support for additional types includes the arithmetic operators:
902 add, subtract, multiply, divide; unary arithmetic operators;
903 relational operators; equality operators; and conversions to and from
904 integer and other floating types. Use a suffix @samp{w} or @samp{W}
905 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
906 for @code{_float128}. You can declare complex types using the
907 corresponding internal complex type, @code{XCmode} for @code{__float80}
908 type and @code{TCmode} for @code{__float128} type:
911 typedef _Complex float __attribute__((mode(TC))) _Complex128;
912 typedef _Complex float __attribute__((mode(XC))) _Complex80;
915 Not all targets support additional floating point types. @code{__float80}
916 is supported on i386, x86_64 and ia64 targets and target @code{__float128}
917 is supported on x86_64 and ia64 targets.
920 @section Decimal Floating Types
921 @cindex decimal floating types
922 @cindex @code{_Decimal32} data type
923 @cindex @code{_Decimal64} data type
924 @cindex @code{_Decimal128} data type
925 @cindex @code{df} integer suffix
926 @cindex @code{dd} integer suffix
927 @cindex @code{dl} integer suffix
928 @cindex @code{DF} integer suffix
929 @cindex @code{DD} integer suffix
930 @cindex @code{DL} integer suffix
932 As an extension, the GNU C compiler supports decimal floating types as
933 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
934 floating types in GCC will evolve as the draft technical report changes.
935 Calling conventions for any target might also change. Not all targets
936 support decimal floating types.
938 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
939 @code{_Decimal128}. They use a radix of ten, unlike the floating types
940 @code{float}, @code{double}, and @code{long double} whose radix is not
941 specified by the C standard but is usually two.
943 Support for decimal floating types includes the arithmetic operators
944 add, subtract, multiply, divide; unary arithmetic operators;
945 relational operators; equality operators; and conversions to and from
946 integer and other floating types. Use a suffix @samp{df} or
947 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
948 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
951 GCC support of decimal float as specified by the draft technical report
956 Translation time data type (TTDT) is not supported.
959 When the value of a decimal floating type cannot be represented in the
960 integer type to which it is being converted, the result is undefined
961 rather than the result value specified by the draft technical report.
964 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
965 are supported by the DWARF2 debug information format.
971 ISO C99 supports floating-point numbers written not only in the usual
972 decimal notation, such as @code{1.55e1}, but also numbers such as
973 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
974 supports this in C89 mode (except in some cases when strictly
975 conforming) and in C++. In that format the
976 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
977 mandatory. The exponent is a decimal number that indicates the power of
978 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
985 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
986 is the same as @code{1.55e1}.
988 Unlike for floating-point numbers in the decimal notation the exponent
989 is always required in the hexadecimal notation. Otherwise the compiler
990 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
991 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
992 extension for floating-point constants of type @code{float}.
995 @section Fixed-Point Types
996 @cindex fixed-point types
997 @cindex @code{_Fract} data type
998 @cindex @code{_Accum} data type
999 @cindex @code{_Sat} data type
1000 @cindex @code{hr} fixed-suffix
1001 @cindex @code{r} fixed-suffix
1002 @cindex @code{lr} fixed-suffix
1003 @cindex @code{llr} fixed-suffix
1004 @cindex @code{uhr} fixed-suffix
1005 @cindex @code{ur} fixed-suffix
1006 @cindex @code{ulr} fixed-suffix
1007 @cindex @code{ullr} fixed-suffix
1008 @cindex @code{hk} fixed-suffix
1009 @cindex @code{k} fixed-suffix
1010 @cindex @code{lk} fixed-suffix
1011 @cindex @code{llk} fixed-suffix
1012 @cindex @code{uhk} fixed-suffix
1013 @cindex @code{uk} fixed-suffix
1014 @cindex @code{ulk} fixed-suffix
1015 @cindex @code{ullk} fixed-suffix
1016 @cindex @code{HR} fixed-suffix
1017 @cindex @code{R} fixed-suffix
1018 @cindex @code{LR} fixed-suffix
1019 @cindex @code{LLR} fixed-suffix
1020 @cindex @code{UHR} fixed-suffix
1021 @cindex @code{UR} fixed-suffix
1022 @cindex @code{ULR} fixed-suffix
1023 @cindex @code{ULLR} fixed-suffix
1024 @cindex @code{HK} fixed-suffix
1025 @cindex @code{K} fixed-suffix
1026 @cindex @code{LK} fixed-suffix
1027 @cindex @code{LLK} fixed-suffix
1028 @cindex @code{UHK} fixed-suffix
1029 @cindex @code{UK} fixed-suffix
1030 @cindex @code{ULK} fixed-suffix
1031 @cindex @code{ULLK} fixed-suffix
1033 As an extension, the GNU C compiler supports fixed-point types as
1034 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
1035 types in GCC will evolve as the draft technical report changes.
1036 Calling conventions for any target might also change. Not all targets
1037 support fixed-point types.
1039 The fixed-point types are
1040 @code{short _Fract},
1043 @code{long long _Fract},
1044 @code{unsigned short _Fract},
1045 @code{unsigned _Fract},
1046 @code{unsigned long _Fract},
1047 @code{unsigned long long _Fract},
1048 @code{_Sat short _Fract},
1050 @code{_Sat long _Fract},
1051 @code{_Sat long long _Fract},
1052 @code{_Sat unsigned short _Fract},
1053 @code{_Sat unsigned _Fract},
1054 @code{_Sat unsigned long _Fract},
1055 @code{_Sat unsigned long long _Fract},
1056 @code{short _Accum},
1059 @code{long long _Accum},
1060 @code{unsigned short _Accum},
1061 @code{unsigned _Accum},
1062 @code{unsigned long _Accum},
1063 @code{unsigned long long _Accum},
1064 @code{_Sat short _Accum},
1066 @code{_Sat long _Accum},
1067 @code{_Sat long long _Accum},
1068 @code{_Sat unsigned short _Accum},
1069 @code{_Sat unsigned _Accum},
1070 @code{_Sat unsigned long _Accum},
1071 @code{_Sat unsigned long long _Accum}.
1072 Fixed-point data values contain fractional and optional integral parts.
1073 The format of fixed-point data varies and depends on the target machine.
1075 Support for fixed-point types includes prefix and postfix increment
1076 and decrement operators (@code{++}, @code{--}); unary arithmetic operators
1077 (@code{+}, @code{-}, @code{!}); binary arithmetic operators (@code{+},
1078 @code{-}, @code{*}, @code{/}); binary shift operators (@code{<<}, @code{>>});
1079 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>});
1080 equality operators (@code{==}, @code{!=}); assignment operators
1081 (@code{+=}, @code{-=}, @code{*=}, @code{/=}, @code{<<=}, @code{>>=});
1082 and conversions to and from integer, floating-point, or fixed-point types.
1084 Use a suffix @samp{hr} or @samp{HR} in a literal constant of type
1085 @code{short _Fract} and @code{_Sat short _Fract},
1086 @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract},
1087 @samp{lr} or @samp{LR} for @code{long _Fract} and @code{_Sat long _Fract},
1088 @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1089 @code{_Sat long long _Fract},
1090 @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1091 @code{_Sat unsigned short _Fract},
1092 @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1093 @code{_Sat unsigned _Fract},
1094 @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1095 @code{_Sat unsigned long _Fract},
1096 @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1097 and @code{_Sat unsigned long long _Fract},
1098 @samp{hk} or @samp{HK} for @code{short _Accum} and @code{_Sat short _Accum},
1099 @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum},
1100 @samp{lk} or @samp{LK} for @code{long _Accum} and @code{_Sat long _Accum},
1101 @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1102 @code{_Sat long long _Accum},
1103 @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1104 @code{_Sat unsigned short _Accum},
1105 @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1106 @code{_Sat unsigned _Accum},
1107 @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1108 @code{_Sat unsigned long _Accum},
1109 and @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1110 and @code{_Sat unsigned long long _Accum}.
1112 GCC support of fixed-point types as specified by the draft technical report
1117 Pragmas to control overflow and rounding behaviors are not implemented.
1120 Fixed-point types are supported by the DWARF2 debug information format.
1123 @section Arrays of Length Zero
1124 @cindex arrays of length zero
1125 @cindex zero-length arrays
1126 @cindex length-zero arrays
1127 @cindex flexible array members
1129 Zero-length arrays are allowed in GNU C@. They are very useful as the
1130 last element of a structure which is really a header for a variable-length
1139 struct line *thisline = (struct line *)
1140 malloc (sizeof (struct line) + this_length);
1141 thisline->length = this_length;
1144 In ISO C90, you would have to give @code{contents} a length of 1, which
1145 means either you waste space or complicate the argument to @code{malloc}.
1147 In ISO C99, you would use a @dfn{flexible array member}, which is
1148 slightly different in syntax and semantics:
1152 Flexible array members are written as @code{contents[]} without
1156 Flexible array members have incomplete type, and so the @code{sizeof}
1157 operator may not be applied. As a quirk of the original implementation
1158 of zero-length arrays, @code{sizeof} evaluates to zero.
1161 Flexible array members may only appear as the last member of a
1162 @code{struct} that is otherwise non-empty.
1165 A structure containing a flexible array member, or a union containing
1166 such a structure (possibly recursively), may not be a member of a
1167 structure or an element of an array. (However, these uses are
1168 permitted by GCC as extensions.)
1171 GCC versions before 3.0 allowed zero-length arrays to be statically
1172 initialized, as if they were flexible arrays. In addition to those
1173 cases that were useful, it also allowed initializations in situations
1174 that would corrupt later data. Non-empty initialization of zero-length
1175 arrays is now treated like any case where there are more initializer
1176 elements than the array holds, in that a suitable warning about "excess
1177 elements in array" is given, and the excess elements (all of them, in
1178 this case) are ignored.
1180 Instead GCC allows static initialization of flexible array members.
1181 This is equivalent to defining a new structure containing the original
1182 structure followed by an array of sufficient size to contain the data.
1183 I.e.@: in the following, @code{f1} is constructed as if it were declared
1189 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1192 struct f1 f1; int data[3];
1193 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1197 The convenience of this extension is that @code{f1} has the desired
1198 type, eliminating the need to consistently refer to @code{f2.f1}.
1200 This has symmetry with normal static arrays, in that an array of
1201 unknown size is also written with @code{[]}.
1203 Of course, this extension only makes sense if the extra data comes at
1204 the end of a top-level object, as otherwise we would be overwriting
1205 data at subsequent offsets. To avoid undue complication and confusion
1206 with initialization of deeply nested arrays, we simply disallow any
1207 non-empty initialization except when the structure is the top-level
1208 object. For example:
1211 struct foo @{ int x; int y[]; @};
1212 struct bar @{ struct foo z; @};
1214 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1215 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1216 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1217 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1220 @node Empty Structures
1221 @section Structures With No Members
1222 @cindex empty structures
1223 @cindex zero-size structures
1225 GCC permits a C structure to have no members:
1232 The structure will have size zero. In C++, empty structures are part
1233 of the language. G++ treats empty structures as if they had a single
1234 member of type @code{char}.
1236 @node Variable Length
1237 @section Arrays of Variable Length
1238 @cindex variable-length arrays
1239 @cindex arrays of variable length
1242 Variable-length automatic arrays are allowed in ISO C99, and as an
1243 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1244 implementation of variable-length arrays does not yet conform in detail
1245 to the ISO C99 standard.) These arrays are
1246 declared like any other automatic arrays, but with a length that is not
1247 a constant expression. The storage is allocated at the point of
1248 declaration and deallocated when the brace-level is exited. For
1253 concat_fopen (char *s1, char *s2, char *mode)
1255 char str[strlen (s1) + strlen (s2) + 1];
1258 return fopen (str, mode);
1262 @cindex scope of a variable length array
1263 @cindex variable-length array scope
1264 @cindex deallocating variable length arrays
1265 Jumping or breaking out of the scope of the array name deallocates the
1266 storage. Jumping into the scope is not allowed; you get an error
1269 @cindex @code{alloca} vs variable-length arrays
1270 You can use the function @code{alloca} to get an effect much like
1271 variable-length arrays. The function @code{alloca} is available in
1272 many other C implementations (but not in all). On the other hand,
1273 variable-length arrays are more elegant.
1275 There are other differences between these two methods. Space allocated
1276 with @code{alloca} exists until the containing @emph{function} returns.
1277 The space for a variable-length array is deallocated as soon as the array
1278 name's scope ends. (If you use both variable-length arrays and
1279 @code{alloca} in the same function, deallocation of a variable-length array
1280 will also deallocate anything more recently allocated with @code{alloca}.)
1282 You can also use variable-length arrays as arguments to functions:
1286 tester (int len, char data[len][len])
1292 The length of an array is computed once when the storage is allocated
1293 and is remembered for the scope of the array in case you access it with
1296 If you want to pass the array first and the length afterward, you can
1297 use a forward declaration in the parameter list---another GNU extension.
1301 tester (int len; char data[len][len], int len)
1307 @cindex parameter forward declaration
1308 The @samp{int len} before the semicolon is a @dfn{parameter forward
1309 declaration}, and it serves the purpose of making the name @code{len}
1310 known when the declaration of @code{data} is parsed.
1312 You can write any number of such parameter forward declarations in the
1313 parameter list. They can be separated by commas or semicolons, but the
1314 last one must end with a semicolon, which is followed by the ``real''
1315 parameter declarations. Each forward declaration must match a ``real''
1316 declaration in parameter name and data type. ISO C99 does not support
1317 parameter forward declarations.
1319 @node Variadic Macros
1320 @section Macros with a Variable Number of Arguments.
1321 @cindex variable number of arguments
1322 @cindex macro with variable arguments
1323 @cindex rest argument (in macro)
1324 @cindex variadic macros
1326 In the ISO C standard of 1999, a macro can be declared to accept a
1327 variable number of arguments much as a function can. The syntax for
1328 defining the macro is similar to that of a function. Here is an
1332 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1335 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1336 such a macro, it represents the zero or more tokens until the closing
1337 parenthesis that ends the invocation, including any commas. This set of
1338 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1339 wherever it appears. See the CPP manual for more information.
1341 GCC has long supported variadic macros, and used a different syntax that
1342 allowed you to give a name to the variable arguments just like any other
1343 argument. Here is an example:
1346 #define debug(format, args...) fprintf (stderr, format, args)
1349 This is in all ways equivalent to the ISO C example above, but arguably
1350 more readable and descriptive.
1352 GNU CPP has two further variadic macro extensions, and permits them to
1353 be used with either of the above forms of macro definition.
1355 In standard C, you are not allowed to leave the variable argument out
1356 entirely; but you are allowed to pass an empty argument. For example,
1357 this invocation is invalid in ISO C, because there is no comma after
1364 GNU CPP permits you to completely omit the variable arguments in this
1365 way. In the above examples, the compiler would complain, though since
1366 the expansion of the macro still has the extra comma after the format
1369 To help solve this problem, CPP behaves specially for variable arguments
1370 used with the token paste operator, @samp{##}. If instead you write
1373 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1376 and if the variable arguments are omitted or empty, the @samp{##}
1377 operator causes the preprocessor to remove the comma before it. If you
1378 do provide some variable arguments in your macro invocation, GNU CPP
1379 does not complain about the paste operation and instead places the
1380 variable arguments after the comma. Just like any other pasted macro
1381 argument, these arguments are not macro expanded.
1383 @node Escaped Newlines
1384 @section Slightly Looser Rules for Escaped Newlines
1385 @cindex escaped newlines
1386 @cindex newlines (escaped)
1388 Recently, the preprocessor has relaxed its treatment of escaped
1389 newlines. Previously, the newline had to immediately follow a
1390 backslash. The current implementation allows whitespace in the form
1391 of spaces, horizontal and vertical tabs, and form feeds between the
1392 backslash and the subsequent newline. The preprocessor issues a
1393 warning, but treats it as a valid escaped newline and combines the two
1394 lines to form a single logical line. This works within comments and
1395 tokens, as well as between tokens. Comments are @emph{not} treated as
1396 whitespace for the purposes of this relaxation, since they have not
1397 yet been replaced with spaces.
1400 @section Non-Lvalue Arrays May Have Subscripts
1401 @cindex subscripting
1402 @cindex arrays, non-lvalue
1404 @cindex subscripting and function values
1405 In ISO C99, arrays that are not lvalues still decay to pointers, and
1406 may be subscripted, although they may not be modified or used after
1407 the next sequence point and the unary @samp{&} operator may not be
1408 applied to them. As an extension, GCC allows such arrays to be
1409 subscripted in C89 mode, though otherwise they do not decay to
1410 pointers outside C99 mode. For example,
1411 this is valid in GNU C though not valid in C89:
1415 struct foo @{int a[4];@};
1421 return f().a[index];
1427 @section Arithmetic on @code{void}- and Function-Pointers
1428 @cindex void pointers, arithmetic
1429 @cindex void, size of pointer to
1430 @cindex function pointers, arithmetic
1431 @cindex function, size of pointer to
1433 In GNU C, addition and subtraction operations are supported on pointers to
1434 @code{void} and on pointers to functions. This is done by treating the
1435 size of a @code{void} or of a function as 1.
1437 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1438 and on function types, and returns 1.
1440 @opindex Wpointer-arith
1441 The option @option{-Wpointer-arith} requests a warning if these extensions
1445 @section Non-Constant Initializers
1446 @cindex initializers, non-constant
1447 @cindex non-constant initializers
1449 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1450 automatic variable are not required to be constant expressions in GNU C@.
1451 Here is an example of an initializer with run-time varying elements:
1454 foo (float f, float g)
1456 float beat_freqs[2] = @{ f-g, f+g @};
1461 @node Compound Literals
1462 @section Compound Literals
1463 @cindex constructor expressions
1464 @cindex initializations in expressions
1465 @cindex structures, constructor expression
1466 @cindex expressions, constructor
1467 @cindex compound literals
1468 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1470 ISO C99 supports compound literals. A compound literal looks like
1471 a cast containing an initializer. Its value is an object of the
1472 type specified in the cast, containing the elements specified in
1473 the initializer; it is an lvalue. As an extension, GCC supports
1474 compound literals in C89 mode and in C++.
1476 Usually, the specified type is a structure. Assume that
1477 @code{struct foo} and @code{structure} are declared as shown:
1480 struct foo @{int a; char b[2];@} structure;
1484 Here is an example of constructing a @code{struct foo} with a compound literal:
1487 structure = ((struct foo) @{x + y, 'a', 0@});
1491 This is equivalent to writing the following:
1495 struct foo temp = @{x + y, 'a', 0@};
1500 You can also construct an array. If all the elements of the compound literal
1501 are (made up of) simple constant expressions, suitable for use in
1502 initializers of objects of static storage duration, then the compound
1503 literal can be coerced to a pointer to its first element and used in
1504 such an initializer, as shown here:
1507 char **foo = (char *[]) @{ "x", "y", "z" @};
1510 Compound literals for scalar types and union types are is
1511 also allowed, but then the compound literal is equivalent
1514 As a GNU extension, GCC allows initialization of objects with static storage
1515 duration by compound literals (which is not possible in ISO C99, because
1516 the initializer is not a constant).
1517 It is handled as if the object was initialized only with the bracket
1518 enclosed list if the types of the compound literal and the object match.
1519 The initializer list of the compound literal must be constant.
1520 If the object being initialized has array type of unknown size, the size is
1521 determined by compound literal size.
1524 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1525 static int y[] = (int []) @{1, 2, 3@};
1526 static int z[] = (int [3]) @{1@};
1530 The above lines are equivalent to the following:
1532 static struct foo x = @{1, 'a', 'b'@};
1533 static int y[] = @{1, 2, 3@};
1534 static int z[] = @{1, 0, 0@};
1537 @node Designated Inits
1538 @section Designated Initializers
1539 @cindex initializers with labeled elements
1540 @cindex labeled elements in initializers
1541 @cindex case labels in initializers
1542 @cindex designated initializers
1544 Standard C89 requires the elements of an initializer to appear in a fixed
1545 order, the same as the order of the elements in the array or structure
1548 In ISO C99 you can give the elements in any order, specifying the array
1549 indices or structure field names they apply to, and GNU C allows this as
1550 an extension in C89 mode as well. This extension is not
1551 implemented in GNU C++.
1553 To specify an array index, write
1554 @samp{[@var{index}] =} before the element value. For example,
1557 int a[6] = @{ [4] = 29, [2] = 15 @};
1564 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1568 The index values must be constant expressions, even if the array being
1569 initialized is automatic.
1571 An alternative syntax for this which has been obsolete since GCC 2.5 but
1572 GCC still accepts is to write @samp{[@var{index}]} before the element
1573 value, with no @samp{=}.
1575 To initialize a range of elements to the same value, write
1576 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1577 extension. For example,
1580 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1584 If the value in it has side-effects, the side-effects will happen only once,
1585 not for each initialized field by the range initializer.
1588 Note that the length of the array is the highest value specified
1591 In a structure initializer, specify the name of a field to initialize
1592 with @samp{.@var{fieldname} =} before the element value. For example,
1593 given the following structure,
1596 struct point @{ int x, y; @};
1600 the following initialization
1603 struct point p = @{ .y = yvalue, .x = xvalue @};
1610 struct point p = @{ xvalue, yvalue @};
1613 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1614 @samp{@var{fieldname}:}, as shown here:
1617 struct point p = @{ y: yvalue, x: xvalue @};
1621 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1622 @dfn{designator}. You can also use a designator (or the obsolete colon
1623 syntax) when initializing a union, to specify which element of the union
1624 should be used. For example,
1627 union foo @{ int i; double d; @};
1629 union foo f = @{ .d = 4 @};
1633 will convert 4 to a @code{double} to store it in the union using
1634 the second element. By contrast, casting 4 to type @code{union foo}
1635 would store it into the union as the integer @code{i}, since it is
1636 an integer. (@xref{Cast to Union}.)
1638 You can combine this technique of naming elements with ordinary C
1639 initialization of successive elements. Each initializer element that
1640 does not have a designator applies to the next consecutive element of the
1641 array or structure. For example,
1644 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1651 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1654 Labeling the elements of an array initializer is especially useful
1655 when the indices are characters or belong to an @code{enum} type.
1660 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1661 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1664 @cindex designator lists
1665 You can also write a series of @samp{.@var{fieldname}} and
1666 @samp{[@var{index}]} designators before an @samp{=} to specify a
1667 nested subobject to initialize; the list is taken relative to the
1668 subobject corresponding to the closest surrounding brace pair. For
1669 example, with the @samp{struct point} declaration above:
1672 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1676 If the same field is initialized multiple times, it will have value from
1677 the last initialization. If any such overridden initialization has
1678 side-effect, it is unspecified whether the side-effect happens or not.
1679 Currently, GCC will discard them and issue a warning.
1682 @section Case Ranges
1684 @cindex ranges in case statements
1686 You can specify a range of consecutive values in a single @code{case} label,
1690 case @var{low} ... @var{high}:
1694 This has the same effect as the proper number of individual @code{case}
1695 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1697 This feature is especially useful for ranges of ASCII character codes:
1703 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1704 it may be parsed wrong when you use it with integer values. For example,
1719 @section Cast to a Union Type
1720 @cindex cast to a union
1721 @cindex union, casting to a
1723 A cast to union type is similar to other casts, except that the type
1724 specified is a union type. You can specify the type either with
1725 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1726 a constructor though, not a cast, and hence does not yield an lvalue like
1727 normal casts. (@xref{Compound Literals}.)
1729 The types that may be cast to the union type are those of the members
1730 of the union. Thus, given the following union and variables:
1733 union foo @{ int i; double d; @};
1739 both @code{x} and @code{y} can be cast to type @code{union foo}.
1741 Using the cast as the right-hand side of an assignment to a variable of
1742 union type is equivalent to storing in a member of the union:
1747 u = (union foo) x @equiv{} u.i = x
1748 u = (union foo) y @equiv{} u.d = y
1751 You can also use the union cast as a function argument:
1754 void hack (union foo);
1756 hack ((union foo) x);
1759 @node Mixed Declarations
1760 @section Mixed Declarations and Code
1761 @cindex mixed declarations and code
1762 @cindex declarations, mixed with code
1763 @cindex code, mixed with declarations
1765 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1766 within compound statements. As an extension, GCC also allows this in
1767 C89 mode. For example, you could do:
1776 Each identifier is visible from where it is declared until the end of
1777 the enclosing block.
1779 @node Function Attributes
1780 @section Declaring Attributes of Functions
1781 @cindex function attributes
1782 @cindex declaring attributes of functions
1783 @cindex functions that never return
1784 @cindex functions that return more than once
1785 @cindex functions that have no side effects
1786 @cindex functions in arbitrary sections
1787 @cindex functions that behave like malloc
1788 @cindex @code{volatile} applied to function
1789 @cindex @code{const} applied to function
1790 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1791 @cindex functions with non-null pointer arguments
1792 @cindex functions that are passed arguments in registers on the 386
1793 @cindex functions that pop the argument stack on the 386
1794 @cindex functions that do not pop the argument stack on the 386
1795 @cindex functions that have different compilation options on the 386
1796 @cindex functions that have different optimization options
1798 In GNU C, you declare certain things about functions called in your program
1799 which help the compiler optimize function calls and check your code more
1802 The keyword @code{__attribute__} allows you to specify special
1803 attributes when making a declaration. This keyword is followed by an
1804 attribute specification inside double parentheses. The following
1805 attributes are currently defined for functions on all targets:
1806 @code{aligned}, @code{alloc_size}, @code{noreturn},
1807 @code{returns_twice}, @code{noinline}, @code{always_inline},
1808 @code{flatten}, @code{pure}, @code{const}, @code{nothrow},
1809 @code{sentinel}, @code{format}, @code{format_arg},
1810 @code{no_instrument_function}, @code{section}, @code{constructor},
1811 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1812 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1813 @code{nonnull}, @code{gnu_inline}, @code{externally_visible},
1814 @code{hot}, @code{cold}, @code{artificial}, @code{error}
1816 Several other attributes are defined for functions on particular
1817 target systems. Other attributes, including @code{section} are
1818 supported for variables declarations (@pxref{Variable Attributes}) and
1819 for types (@pxref{Type Attributes}).
1821 You may also specify attributes with @samp{__} preceding and following
1822 each keyword. This allows you to use them in header files without
1823 being concerned about a possible macro of the same name. For example,
1824 you may use @code{__noreturn__} instead of @code{noreturn}.
1826 @xref{Attribute Syntax}, for details of the exact syntax for using
1830 @c Keep this table alphabetized by attribute name. Treat _ as space.
1832 @item alias ("@var{target}")
1833 @cindex @code{alias} attribute
1834 The @code{alias} attribute causes the declaration to be emitted as an
1835 alias for another symbol, which must be specified. For instance,
1838 void __f () @{ /* @r{Do something.} */; @}
1839 void f () __attribute__ ((weak, alias ("__f")));
1842 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1843 mangled name for the target must be used. It is an error if @samp{__f}
1844 is not defined in the same translation unit.
1846 Not all target machines support this attribute.
1848 @item aligned (@var{alignment})
1849 @cindex @code{aligned} attribute
1850 This attribute specifies a minimum alignment for the function,
1853 You cannot use this attribute to decrease the alignment of a function,
1854 only to increase it. However, when you explicitly specify a function
1855 alignment this will override the effect of the
1856 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1859 Note that the effectiveness of @code{aligned} attributes may be
1860 limited by inherent limitations in your linker. On many systems, the
1861 linker is only able to arrange for functions to be aligned up to a
1862 certain maximum alignment. (For some linkers, the maximum supported
1863 alignment may be very very small.) See your linker documentation for
1864 further information.
1866 The @code{aligned} attribute can also be used for variables and fields
1867 (@pxref{Variable Attributes}.)
1870 @cindex @code{alloc_size} attribute
1871 The @code{alloc_size} attribute is used to tell the compiler that the
1872 function return value points to memory, where the size is given by
1873 one or two of the functions parameters. GCC uses this
1874 information to improve the correctness of @code{__builtin_object_size}.
1876 The function parameter(s) denoting the allocated size are specified by
1877 one or two integer arguments supplied to the attribute. The allocated size
1878 is either the value of the single function argument specified or the product
1879 of the two function arguments specified. Argument numbering starts at
1885 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1886 void my_realloc(void*, size_t) __attribute__((alloc_size(2)))
1889 declares that my_calloc will return memory of the size given by
1890 the product of parameter 1 and 2 and that my_realloc will return memory
1891 of the size given by parameter 2.
1894 @cindex @code{always_inline} function attribute
1895 Generally, functions are not inlined unless optimization is specified.
1896 For functions declared inline, this attribute inlines the function even
1897 if no optimization level was specified.
1900 @cindex @code{gnu_inline} function attribute
1901 This attribute should be used with a function which is also declared
1902 with the @code{inline} keyword. It directs GCC to treat the function
1903 as if it were defined in gnu89 mode even when compiling in C99 or
1906 If the function is declared @code{extern}, then this definition of the
1907 function is used only for inlining. In no case is the function
1908 compiled as a standalone function, not even if you take its address
1909 explicitly. Such an address becomes an external reference, as if you
1910 had only declared the function, and had not defined it. This has
1911 almost the effect of a macro. The way to use this is to put a
1912 function definition in a header file with this attribute, and put
1913 another copy of the function, without @code{extern}, in a library
1914 file. The definition in the header file will cause most calls to the
1915 function to be inlined. If any uses of the function remain, they will
1916 refer to the single copy in the library. Note that the two
1917 definitions of the functions need not be precisely the same, although
1918 if they do not have the same effect your program may behave oddly.
1920 In C, if the function is neither @code{extern} nor @code{static}, then
1921 the function is compiled as a standalone function, as well as being
1922 inlined where possible.
1924 This is how GCC traditionally handled functions declared
1925 @code{inline}. Since ISO C99 specifies a different semantics for
1926 @code{inline}, this function attribute is provided as a transition
1927 measure and as a useful feature in its own right. This attribute is
1928 available in GCC 4.1.3 and later. It is available if either of the
1929 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1930 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1931 Function is As Fast As a Macro}.
1933 In C++, this attribute does not depend on @code{extern} in any way,
1934 but it still requires the @code{inline} keyword to enable its special
1937 @cindex @code{artificial} function attribute
1939 This attribute is useful for small inline wrappers which if possible
1940 should appear during debugging as a unit, depending on the debug
1941 info format it will either mean marking the function as artificial
1942 or using the caller location for all instructions within the inlined
1945 @cindex @code{flatten} function attribute
1947 Generally, inlining into a function is limited. For a function marked with
1948 this attribute, every call inside this function will be inlined, if possible.
1949 Whether the function itself is considered for inlining depends on its size and
1950 the current inlining parameters.
1952 @item error ("@var{message}")
1953 @cindex @code{error} function attribute
1954 If this attribute is used on a function declaration and a call to such a function
1955 is not eliminated through dead code elimination or other optimizations, an error
1956 which will include @var{message} will be diagnosed. This is useful
1957 for compile time checking, especially together with @code{__builtin_constant_p}
1958 and inline functions where checking the inline function arguments is not
1959 possible through @code{extern char [(condition) ? 1 : -1];} tricks.
1960 While it is possible to leave the function undefined and thus invoke
1961 a link failure, when using this attribute the problem will be diagnosed
1962 earlier and with exact location of the call even in presence of inline
1963 functions or when not emitting debugging information.
1965 @item warning ("@var{message}")
1966 @cindex @code{warning} function attribute
1967 If this attribute is used on a function declaration and a call to such a function
1968 is not eliminated through dead code elimination or other optimizations, a warning
1969 which will include @var{message} will be diagnosed. This is useful
1970 for compile time checking, especially together with @code{__builtin_constant_p}
1971 and inline functions. While it is possible to define the function with
1972 a message in @code{.gnu.warning*} section, when using this attribute the problem
1973 will be diagnosed earlier and with exact location of the call even in presence
1974 of inline functions or when not emitting debugging information.
1977 @cindex functions that do pop the argument stack on the 386
1979 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1980 assume that the calling function will pop off the stack space used to
1981 pass arguments. This is
1982 useful to override the effects of the @option{-mrtd} switch.
1985 @cindex @code{const} function attribute
1986 Many functions do not examine any values except their arguments, and
1987 have no effects except the return value. Basically this is just slightly
1988 more strict class than the @code{pure} attribute below, since function is not
1989 allowed to read global memory.
1991 @cindex pointer arguments
1992 Note that a function that has pointer arguments and examines the data
1993 pointed to must @emph{not} be declared @code{const}. Likewise, a
1994 function that calls a non-@code{const} function usually must not be
1995 @code{const}. It does not make sense for a @code{const} function to
1998 The attribute @code{const} is not implemented in GCC versions earlier
1999 than 2.5. An alternative way to declare that a function has no side
2000 effects, which works in the current version and in some older versions,
2004 typedef int intfn ();
2006 extern const intfn square;
2009 This approach does not work in GNU C++ from 2.6.0 on, since the language
2010 specifies that the @samp{const} must be attached to the return value.
2014 @itemx constructor (@var{priority})
2015 @itemx destructor (@var{priority})
2016 @cindex @code{constructor} function attribute
2017 @cindex @code{destructor} function attribute
2018 The @code{constructor} attribute causes the function to be called
2019 automatically before execution enters @code{main ()}. Similarly, the
2020 @code{destructor} attribute causes the function to be called
2021 automatically after @code{main ()} has completed or @code{exit ()} has
2022 been called. Functions with these attributes are useful for
2023 initializing data that will be used implicitly during the execution of
2026 You may provide an optional integer priority to control the order in
2027 which constructor and destructor functions are run. A constructor
2028 with a smaller priority number runs before a constructor with a larger
2029 priority number; the opposite relationship holds for destructors. So,
2030 if you have a constructor that allocates a resource and a destructor
2031 that deallocates the same resource, both functions typically have the
2032 same priority. The priorities for constructor and destructor
2033 functions are the same as those specified for namespace-scope C++
2034 objects (@pxref{C++ Attributes}).
2036 These attributes are not currently implemented for Objective-C@.
2039 @cindex @code{deprecated} attribute.
2040 The @code{deprecated} attribute results in a warning if the function
2041 is used anywhere in the source file. This is useful when identifying
2042 functions that are expected to be removed in a future version of a
2043 program. The warning also includes the location of the declaration
2044 of the deprecated function, to enable users to easily find further
2045 information about why the function is deprecated, or what they should
2046 do instead. Note that the warnings only occurs for uses:
2049 int old_fn () __attribute__ ((deprecated));
2051 int (*fn_ptr)() = old_fn;
2054 results in a warning on line 3 but not line 2.
2056 The @code{deprecated} attribute can also be used for variables and
2057 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2060 @cindex @code{__declspec(dllexport)}
2061 On Microsoft Windows targets and Symbian OS targets the
2062 @code{dllexport} attribute causes the compiler to provide a global
2063 pointer to a pointer in a DLL, so that it can be referenced with the
2064 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
2065 name is formed by combining @code{_imp__} and the function or variable
2068 You can use @code{__declspec(dllexport)} as a synonym for
2069 @code{__attribute__ ((dllexport))} for compatibility with other
2072 On systems that support the @code{visibility} attribute, this
2073 attribute also implies ``default'' visibility. It is an error to
2074 explicitly specify any other visibility.
2076 Currently, the @code{dllexport} attribute is ignored for inlined
2077 functions, unless the @option{-fkeep-inline-functions} flag has been
2078 used. The attribute is also ignored for undefined symbols.
2080 When applied to C++ classes, the attribute marks defined non-inlined
2081 member functions and static data members as exports. Static consts
2082 initialized in-class are not marked unless they are also defined
2085 For Microsoft Windows targets there are alternative methods for
2086 including the symbol in the DLL's export table such as using a
2087 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2088 the @option{--export-all} linker flag.
2091 @cindex @code{__declspec(dllimport)}
2092 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2093 attribute causes the compiler to reference a function or variable via
2094 a global pointer to a pointer that is set up by the DLL exporting the
2095 symbol. The attribute implies @code{extern}. On Microsoft Windows
2096 targets, the pointer name is formed by combining @code{_imp__} and the
2097 function or variable name.
2099 You can use @code{__declspec(dllimport)} as a synonym for
2100 @code{__attribute__ ((dllimport))} for compatibility with other
2103 On systems that support the @code{visibility} attribute, this
2104 attribute also implies ``default'' visibility. It is an error to
2105 explicitly specify any other visibility.
2107 Currently, the attribute is ignored for inlined functions. If the
2108 attribute is applied to a symbol @emph{definition}, an error is reported.
2109 If a symbol previously declared @code{dllimport} is later defined, the
2110 attribute is ignored in subsequent references, and a warning is emitted.
2111 The attribute is also overridden by a subsequent declaration as
2114 When applied to C++ classes, the attribute marks non-inlined
2115 member functions and static data members as imports. However, the
2116 attribute is ignored for virtual methods to allow creation of vtables
2119 On the SH Symbian OS target the @code{dllimport} attribute also has
2120 another affect---it can cause the vtable and run-time type information
2121 for a class to be exported. This happens when the class has a
2122 dllimport'ed constructor or a non-inline, non-pure virtual function
2123 and, for either of those two conditions, the class also has a inline
2124 constructor or destructor and has a key function that is defined in
2125 the current translation unit.
2127 For Microsoft Windows based targets the use of the @code{dllimport}
2128 attribute on functions is not necessary, but provides a small
2129 performance benefit by eliminating a thunk in the DLL@. The use of the
2130 @code{dllimport} attribute on imported variables was required on older
2131 versions of the GNU linker, but can now be avoided by passing the
2132 @option{--enable-auto-import} switch to the GNU linker. As with
2133 functions, using the attribute for a variable eliminates a thunk in
2136 One drawback to using this attribute is that a pointer to a
2137 @emph{variable} marked as @code{dllimport} cannot be used as a constant
2138 address. However, a pointer to a @emph{function} with the
2139 @code{dllimport} attribute can be used as a constant initializer; in
2140 this case, the address of a stub function in the import lib is
2141 referenced. On Microsoft Windows targets, the attribute can be disabled
2142 for functions by setting the @option{-mnop-fun-dllimport} flag.
2145 @cindex eight bit data on the H8/300, H8/300H, and H8S
2146 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2147 variable should be placed into the eight bit data section.
2148 The compiler will generate more efficient code for certain operations
2149 on data in the eight bit data area. Note the eight bit data area is limited to
2152 You must use GAS and GLD from GNU binutils version 2.7 or later for
2153 this attribute to work correctly.
2155 @item exception_handler
2156 @cindex exception handler functions on the Blackfin processor
2157 Use this attribute on the Blackfin to indicate that the specified function
2158 is an exception handler. The compiler will generate function entry and
2159 exit sequences suitable for use in an exception handler when this
2160 attribute is present.
2163 @cindex functions which handle memory bank switching
2164 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2165 use a calling convention that takes care of switching memory banks when
2166 entering and leaving a function. This calling convention is also the
2167 default when using the @option{-mlong-calls} option.
2169 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2170 to call and return from a function.
2172 On 68HC11 the compiler will generate a sequence of instructions
2173 to invoke a board-specific routine to switch the memory bank and call the
2174 real function. The board-specific routine simulates a @code{call}.
2175 At the end of a function, it will jump to a board-specific routine
2176 instead of using @code{rts}. The board-specific return routine simulates
2180 @cindex functions that pop the argument stack on the 386
2181 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2182 pass the first argument (if of integral type) in the register ECX and
2183 the second argument (if of integral type) in the register EDX@. Subsequent
2184 and other typed arguments are passed on the stack. The called function will
2185 pop the arguments off the stack. If the number of arguments is variable all
2186 arguments are pushed on the stack.
2188 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2189 @cindex @code{format} function attribute
2191 The @code{format} attribute specifies that a function takes @code{printf},
2192 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2193 should be type-checked against a format string. For example, the
2198 my_printf (void *my_object, const char *my_format, ...)
2199 __attribute__ ((format (printf, 2, 3)));
2203 causes the compiler to check the arguments in calls to @code{my_printf}
2204 for consistency with the @code{printf} style format string argument
2207 The parameter @var{archetype} determines how the format string is
2208 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
2209 @code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
2210 @code{strfmon}. (You can also use @code{__printf__},
2211 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) On
2212 MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
2213 @code{ms_strftime} are also present.
2214 @var{archtype} values such as @code{printf} refer to the formats accepted
2215 by the system's C run-time library, while @code{gnu_} values always refer
2216 to the formats accepted by the GNU C Library. On Microsoft Windows
2217 targets, @code{ms_} values refer to the formats accepted by the
2218 @file{msvcrt.dll} library.
2219 The parameter @var{string-index}
2220 specifies which argument is the format string argument (starting
2221 from 1), while @var{first-to-check} is the number of the first
2222 argument to check against the format string. For functions
2223 where the arguments are not available to be checked (such as
2224 @code{vprintf}), specify the third parameter as zero. In this case the
2225 compiler only checks the format string for consistency. For
2226 @code{strftime} formats, the third parameter is required to be zero.
2227 Since non-static C++ methods have an implicit @code{this} argument, the
2228 arguments of such methods should be counted from two, not one, when
2229 giving values for @var{string-index} and @var{first-to-check}.
2231 In the example above, the format string (@code{my_format}) is the second
2232 argument of the function @code{my_print}, and the arguments to check
2233 start with the third argument, so the correct parameters for the format
2234 attribute are 2 and 3.
2236 @opindex ffreestanding
2237 @opindex fno-builtin
2238 The @code{format} attribute allows you to identify your own functions
2239 which take format strings as arguments, so that GCC can check the
2240 calls to these functions for errors. The compiler always (unless
2241 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2242 for the standard library functions @code{printf}, @code{fprintf},
2243 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2244 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2245 warnings are requested (using @option{-Wformat}), so there is no need to
2246 modify the header file @file{stdio.h}. In C99 mode, the functions
2247 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2248 @code{vsscanf} are also checked. Except in strictly conforming C
2249 standard modes, the X/Open function @code{strfmon} is also checked as
2250 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2251 @xref{C Dialect Options,,Options Controlling C Dialect}.
2253 The target may provide additional types of format checks.
2254 @xref{Target Format Checks,,Format Checks Specific to Particular
2257 @item format_arg (@var{string-index})
2258 @cindex @code{format_arg} function attribute
2259 @opindex Wformat-nonliteral
2260 The @code{format_arg} attribute specifies that a function takes a format
2261 string for a @code{printf}, @code{scanf}, @code{strftime} or
2262 @code{strfmon} style function and modifies it (for example, to translate
2263 it into another language), so the result can be passed to a
2264 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2265 function (with the remaining arguments to the format function the same
2266 as they would have been for the unmodified string). For example, the
2271 my_dgettext (char *my_domain, const char *my_format)
2272 __attribute__ ((format_arg (2)));
2276 causes the compiler to check the arguments in calls to a @code{printf},
2277 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2278 format string argument is a call to the @code{my_dgettext} function, for
2279 consistency with the format string argument @code{my_format}. If the
2280 @code{format_arg} attribute had not been specified, all the compiler
2281 could tell in such calls to format functions would be that the format
2282 string argument is not constant; this would generate a warning when
2283 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2284 without the attribute.
2286 The parameter @var{string-index} specifies which argument is the format
2287 string argument (starting from one). Since non-static C++ methods have
2288 an implicit @code{this} argument, the arguments of such methods should
2289 be counted from two.
2291 The @code{format-arg} attribute allows you to identify your own
2292 functions which modify format strings, so that GCC can check the
2293 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2294 type function whose operands are a call to one of your own function.
2295 The compiler always treats @code{gettext}, @code{dgettext}, and
2296 @code{dcgettext} in this manner except when strict ISO C support is
2297 requested by @option{-ansi} or an appropriate @option{-std} option, or
2298 @option{-ffreestanding} or @option{-fno-builtin}
2299 is used. @xref{C Dialect Options,,Options
2300 Controlling C Dialect}.
2302 @item function_vector
2303 @cindex calling functions through the function vector on H8/300, M16C, M32C and SH2A processors
2304 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2305 function should be called through the function vector. Calling a
2306 function through the function vector will reduce code size, however;
2307 the function vector has a limited size (maximum 128 entries on the H8/300
2308 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2310 In SH2A target, this attribute declares a function to be called using the
2311 TBR relative addressing mode. The argument to this attribute is the entry
2312 number of the same function in a vector table containing all the TBR
2313 relative addressable functions. For the successful jump, register TBR
2314 should contain the start address of this TBR relative vector table.
2315 In the startup routine of the user application, user needs to care of this
2316 TBR register initialization. The TBR relative vector table can have at
2317 max 256 function entries. The jumps to these functions will be generated
2318 using a SH2A specific, non delayed branch instruction JSR/N @@(disp8,TBR).
2319 You must use GAS and GLD from GNU binutils version 2.7 or later for
2320 this attribute to work correctly.
2322 Please refer the example of M16C target, to see the use of this
2323 attribute while declaring a function,
2325 In an application, for a function being called once, this attribute will
2326 save at least 8 bytes of code; and if other successive calls are being
2327 made to the same function, it will save 2 bytes of code per each of these
2330 On M16C/M32C targets, the @code{function_vector} attribute declares a
2331 special page subroutine call function. Use of this attribute reduces
2332 the code size by 2 bytes for each call generated to the
2333 subroutine. The argument to the attribute is the vector number entry
2334 from the special page vector table which contains the 16 low-order
2335 bits of the subroutine's entry address. Each vector table has special
2336 page number (18 to 255) which are used in @code{jsrs} instruction.
2337 Jump addresses of the routines are generated by adding 0x0F0000 (in
2338 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2339 byte addresses set in the vector table. Therefore you need to ensure
2340 that all the special page vector routines should get mapped within the
2341 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2344 In the following example 2 bytes will be saved for each call to
2345 function @code{foo}.
2348 void foo (void) __attribute__((function_vector(0x18)));
2359 If functions are defined in one file and are called in another file,
2360 then be sure to write this declaration in both files.
2362 This attribute is ignored for R8C target.
2365 @cindex interrupt handler functions
2366 Use this attribute on the ARM, AVR, CRX, M32C, M32R/D, m68k,
2367 and Xstormy16 ports to indicate that the specified function is an
2368 interrupt handler. The compiler will generate function entry and exit
2369 sequences suitable for use in an interrupt handler when this attribute
2372 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2373 SH processors can be specified via the @code{interrupt_handler} attribute.
2375 Note, on the AVR, interrupts will be enabled inside the function.
2377 Note, for the ARM, you can specify the kind of interrupt to be handled by
2378 adding an optional parameter to the interrupt attribute like this:
2381 void f () __attribute__ ((interrupt ("IRQ")));
2384 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2386 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2387 may be called with a word aligned stack pointer.
2389 @item interrupt_handler
2390 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2391 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2392 indicate that the specified function is an interrupt handler. The compiler
2393 will generate function entry and exit sequences suitable for use in an
2394 interrupt handler when this attribute is present.
2396 @item interrupt_thread
2397 @cindex interrupt thread functions on fido
2398 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2399 that the specified function is an interrupt handler that is designed
2400 to run as a thread. The compiler omits generate prologue/epilogue
2401 sequences and replaces the return instruction with a @code{sleep}
2402 instruction. This attribute is available only on fido.
2405 @cindex User stack pointer in interrupts on the Blackfin
2406 When used together with @code{interrupt_handler}, @code{exception_handler}
2407 or @code{nmi_handler}, code will be generated to load the stack pointer
2408 from the USP register in the function prologue.
2411 @cindex @code{l1_text} function attribute
2412 This attribute specifies a function to be placed into L1 Instruction
2413 SRAM@. The function will be put into a specific section named @code{.l1.text}.
2414 With @option{-mfdpic}, function calls with a such function as the callee
2415 or caller will use inlined PLT.
2417 @item long_call/short_call
2418 @cindex indirect calls on ARM
2419 This attribute specifies how a particular function is called on
2420 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2421 command line switch and @code{#pragma long_calls} settings. The
2422 @code{long_call} attribute indicates that the function might be far
2423 away from the call site and require a different (more expensive)
2424 calling sequence. The @code{short_call} attribute always places
2425 the offset to the function from the call site into the @samp{BL}
2426 instruction directly.
2428 @item longcall/shortcall
2429 @cindex functions called via pointer on the RS/6000 and PowerPC
2430 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2431 indicates that the function might be far away from the call site and
2432 require a different (more expensive) calling sequence. The
2433 @code{shortcall} attribute indicates that the function is always close
2434 enough for the shorter calling sequence to be used. These attributes
2435 override both the @option{-mlongcall} switch and, on the RS/6000 and
2436 PowerPC, the @code{#pragma longcall} setting.
2438 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2439 calls are necessary.
2441 @item long_call/near/far
2442 @cindex indirect calls on MIPS
2443 These attributes specify how a particular function is called on MIPS@.
2444 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2445 command-line switch. The @code{long_call} and @code{far} attributes are
2446 synonyms, and cause the compiler to always call
2447 the function by first loading its address into a register, and then using
2448 the contents of that register. The @code{near} attribute has the opposite
2449 effect; it specifies that non-PIC calls should be made using the more
2450 efficient @code{jal} instruction.
2453 @cindex @code{malloc} attribute
2454 The @code{malloc} attribute is used to tell the compiler that a function
2455 may be treated as if any non-@code{NULL} pointer it returns cannot
2456 alias any other pointer valid when the function returns.
2457 This will often improve optimization.
2458 Standard functions with this property include @code{malloc} and
2459 @code{calloc}. @code{realloc}-like functions have this property as
2460 long as the old pointer is never referred to (including comparing it
2461 to the new pointer) after the function returns a non-@code{NULL}
2464 @item mips16/nomips16
2465 @cindex @code{mips16} attribute
2466 @cindex @code{nomips16} attribute
2468 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2469 function attributes to locally select or turn off MIPS16 code generation.
2470 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2471 while MIPS16 code generation is disabled for functions with the
2472 @code{nomips16} attribute. These attributes override the
2473 @option{-mips16} and @option{-mno-mips16} options on the command line
2474 (@pxref{MIPS Options}).
2476 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2477 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2478 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2479 may interact badly with some GCC extensions such as @code{__builtin_apply}
2480 (@pxref{Constructing Calls}).
2482 @item model (@var{model-name})
2483 @cindex function addressability on the M32R/D
2484 @cindex variable addressability on the IA-64
2486 On the M32R/D, use this attribute to set the addressability of an
2487 object, and of the code generated for a function. The identifier
2488 @var{model-name} is one of @code{small}, @code{medium}, or
2489 @code{large}, representing each of the code models.
2491 Small model objects live in the lower 16MB of memory (so that their
2492 addresses can be loaded with the @code{ld24} instruction), and are
2493 callable with the @code{bl} instruction.
2495 Medium model objects may live anywhere in the 32-bit address space (the
2496 compiler will generate @code{seth/add3} instructions to load their addresses),
2497 and are callable with the @code{bl} instruction.
2499 Large model objects may live anywhere in the 32-bit address space (the
2500 compiler will generate @code{seth/add3} instructions to load their addresses),
2501 and may not be reachable with the @code{bl} instruction (the compiler will
2502 generate the much slower @code{seth/add3/jl} instruction sequence).
2504 On IA-64, use this attribute to set the addressability of an object.
2505 At present, the only supported identifier for @var{model-name} is
2506 @code{small}, indicating addressability via ``small'' (22-bit)
2507 addresses (so that their addresses can be loaded with the @code{addl}
2508 instruction). Caveat: such addressing is by definition not position
2509 independent and hence this attribute must not be used for objects
2510 defined by shared libraries.
2512 @item ms_abi/sysv_abi
2513 @cindex @code{ms_abi} attribute
2514 @cindex @code{sysv_abi} attribute
2516 On 64-bit x86_65-*-* targets, you can use an ABI attribute to indicate
2517 which calling convention should be used for a function. The @code{ms_abi}
2518 attribute tells the compiler to use the Microsoft ABI, while the
2519 @code{sysv_abi} attribute tells the compiler to use the ABI used on
2520 GNU/Linux and other systems. The default is to use the Microsoft ABI
2521 when targeting Windows. On all other systems, the default is the AMD ABI.
2523 Note, This feature is currently sorried out for Windows targets trying to
2526 @cindex function without a prologue/epilogue code
2527 Use this attribute on the ARM, AVR, IP2K and SPU ports to indicate that
2528 the specified function does not need prologue/epilogue sequences generated by
2529 the compiler. It is up to the programmer to provide these sequences. The
2530 only statements that can be safely included in naked functions are
2531 @code{asm} statements that do not have operands. All other statements,
2532 including declarations of local variables, @code{if} statements, and so
2533 forth, should be avoided. Naked functions should be used to implement the
2534 body of an assembly function, while allowing the compiler to construct
2535 the requisite function declaration for the assembler.
2538 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2539 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2540 use the normal calling convention based on @code{jsr} and @code{rts}.
2541 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2545 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2546 Use this attribute together with @code{interrupt_handler},
2547 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2548 entry code should enable nested interrupts or exceptions.
2551 @cindex NMI handler functions on the Blackfin processor
2552 Use this attribute on the Blackfin to indicate that the specified function
2553 is an NMI handler. The compiler will generate function entry and
2554 exit sequences suitable for use in an NMI handler when this
2555 attribute is present.
2557 @item no_instrument_function
2558 @cindex @code{no_instrument_function} function attribute
2559 @opindex finstrument-functions
2560 If @option{-finstrument-functions} is given, profiling function calls will
2561 be generated at entry and exit of most user-compiled functions.
2562 Functions with this attribute will not be so instrumented.
2565 @cindex @code{noinline} function attribute
2566 This function attribute prevents a function from being considered for
2568 @c Don't enumerate the optimizations by name here; we try to be
2569 @c future-compatible with this mechanism.
2570 If the function does not have side-effects, there are optimizations
2571 other than inlining that causes function calls to be optimized away,
2572 although the function call is live. To keep such calls from being
2577 (@pxref{Extended Asm}) in the called function, to serve as a special
2580 @item nonnull (@var{arg-index}, @dots{})
2581 @cindex @code{nonnull} function attribute
2582 The @code{nonnull} attribute specifies that some function parameters should
2583 be non-null pointers. For instance, the declaration:
2587 my_memcpy (void *dest, const void *src, size_t len)
2588 __attribute__((nonnull (1, 2)));
2592 causes the compiler to check that, in calls to @code{my_memcpy},
2593 arguments @var{dest} and @var{src} are non-null. If the compiler
2594 determines that a null pointer is passed in an argument slot marked
2595 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2596 is issued. The compiler may also choose to make optimizations based
2597 on the knowledge that certain function arguments will not be null.
2599 If no argument index list is given to the @code{nonnull} attribute,
2600 all pointer arguments are marked as non-null. To illustrate, the
2601 following declaration is equivalent to the previous example:
2605 my_memcpy (void *dest, const void *src, size_t len)
2606 __attribute__((nonnull));
2610 @cindex @code{noreturn} function attribute
2611 A few standard library functions, such as @code{abort} and @code{exit},
2612 cannot return. GCC knows this automatically. Some programs define
2613 their own functions that never return. You can declare them
2614 @code{noreturn} to tell the compiler this fact. For example,
2618 void fatal () __attribute__ ((noreturn));
2621 fatal (/* @r{@dots{}} */)
2623 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2629 The @code{noreturn} keyword tells the compiler to assume that
2630 @code{fatal} cannot return. It can then optimize without regard to what
2631 would happen if @code{fatal} ever did return. This makes slightly
2632 better code. More importantly, it helps avoid spurious warnings of
2633 uninitialized variables.
2635 The @code{noreturn} keyword does not affect the exceptional path when that
2636 applies: a @code{noreturn}-marked function may still return to the caller
2637 by throwing an exception or calling @code{longjmp}.
2639 Do not assume that registers saved by the calling function are
2640 restored before calling the @code{noreturn} function.
2642 It does not make sense for a @code{noreturn} function to have a return
2643 type other than @code{void}.
2645 The attribute @code{noreturn} is not implemented in GCC versions
2646 earlier than 2.5. An alternative way to declare that a function does
2647 not return, which works in the current version and in some older
2648 versions, is as follows:
2651 typedef void voidfn ();
2653 volatile voidfn fatal;
2656 This approach does not work in GNU C++.
2659 @cindex @code{nothrow} function attribute
2660 The @code{nothrow} attribute is used to inform the compiler that a
2661 function cannot throw an exception. For example, most functions in
2662 the standard C library can be guaranteed not to throw an exception
2663 with the notable exceptions of @code{qsort} and @code{bsearch} that
2664 take function pointer arguments. The @code{nothrow} attribute is not
2665 implemented in GCC versions earlier than 3.3.
2668 @cindex @code{option} function attribute
2669 The @code{option} attribute is used to specify that a function is to
2670 be compiled with different target options than specified on the
2671 command line. This can be used for instance to have functions
2672 compiled with a different ISA (instruction set architecture) than the
2673 default. You can also use the @samp{#pragma GCC option} pragma to set
2674 more than one function to be compiled with specific target options.
2675 @xref{Function Specific Option Pragmas}, for details about the
2676 @samp{#pragma GCC option} pragma.
2678 For instance on a 386, you could compile one function with
2679 @code{option("sse4.1,arch=core2")} and another with
2680 @code{option("sse4a,arch=amdfam10")} that would be equivalent to
2681 compiling the first function with @option{-msse4.1} and
2682 @option{-march=core2} options, and the second function with
2683 @option{-msse4a} and @option{-march=amdfam10} options. It is up to the
2684 user to make sure that a function is only invoked on a machine that
2685 supports the particular ISA it was compiled for (for example by using
2686 @code{cpuid} on 386 to determine what feature bits and architecture
2690 int core2_func (void) __attribute__ ((__option__ ("arch=core2")));
2691 int sse3_func (void) __attribute__ ((__option__ ("sse3")));
2694 On the 386, the following options are allowed:
2699 @cindex option("abm")
2700 Enable/disable the generation of the advanced bit instructions.
2704 @cindex @code{option("aes")} attribute
2705 Enable/disable the generation of the AES instructions.
2709 @cindex @code{option("mmx")} attribute
2710 Enable/disable the generation of the MMX instructions.
2714 @cindex @code{option("pclmul")} attribute
2715 Enable/disable the generation of the PCLMUL instructions.
2719 @cindex @code{option("popcnt")} attribute
2720 Enable/disable the generation of the POPCNT instruction.
2724 @cindex @code{option("sse")} attribute
2725 Enable/disable the generation of the SSE instructions.
2729 @cindex @code{option("sse2")} attribute
2730 Enable/disable the generation of the SSE2 instructions.
2734 @cindex @code{option("sse3")} attribute
2735 Enable/disable the generation of the SSE3 instructions.
2739 @cindex @code{option("sse4")} attribute
2740 Enable/disable the generation of the SSE4 instructions (both SSE4.1
2745 @cindex @code{option("sse4.1")} attribute
2746 Enable/disable the generation of the sse4.1 instructions.
2750 @cindex @code{option("sse4.2")} attribute
2751 Enable/disable the generation of the sse4.2 instructions.
2755 @cindex @code{option("sse4a")} attribute
2756 Enable/disable the generation of the SSE4A instructions.
2760 @cindex @code{option("sse5")} attribute
2761 Enable/disable the generation of the SSE5 instructions.
2765 @cindex @code{option("ssse3")} attribute
2766 Enable/disable the generation of the SSSE3 instructions.
2770 @cindex @code{option("cld")} attribute
2771 Enable/disable the generation of the CLD before string moves.
2773 @item fancy-math-387
2774 @itemx no-fancy-math-387
2775 @cindex @code{option("fancy-math-387")} attribute
2776 Enable/disable the generation of the @code{sin}, @code{cos}, and
2777 @code{sqrt} instructions on the 387 floating point unit.
2780 @itemx no-fused-madd
2781 @cindex @code{option("fused-madd")} attribute
2782 Enable/disable the generation of the fused multiply/add instructions.
2786 @cindex @code{option("ieee-fp")} attribute
2787 Enable/disable the generation of floating point that depends on IEEE arithmetic.
2789 @item inline-all-stringops
2790 @itemx no-inline-all-stringops
2791 @cindex @code{option("inline-all-stringops")} attribute
2792 Enable/disable inlining of string operations.
2794 @item inline-stringops-dynamically
2795 @itemx no-inline-stringops-dynamically
2796 @cindex @code{option("inline-stringops-dynamically")} attribute
2797 Enable/disable the generation of the inline code to do small string
2798 operations and calling the library routines for large operations.
2800 @item align-stringops
2801 @itemx no-align-stringops
2802 @cindex @code{option("align-stringops")} attribute
2803 Do/do not align destination of inlined string operations.
2807 @cindex @code{option("recip")} attribute
2808 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
2809 instructions followed an additional Newton-Rhapson step instead of
2810 doing a floating point division.
2812 @item arch=@var{ARCH}
2813 @cindex @code{option("arch=@var{ARCH}")} attribute
2814 Specify the architecture to generate code for in compiling the function.
2816 @item tune=@var{TUNE}
2817 @cindex @code{option("tune=@var{TUNE}")} attribute
2818 Specify the architecture to tune for in compiling the function.
2820 @item fpmath=@var{FPMATH}
2821 @cindex @code{option("fpmath=@var{FPMATH}")} attribute
2822 Specify which floating point unit to use. The
2823 @code{option("fpmath=sse,387")} option must be specified as
2824 @code{option("fpmath=sse+387")} because the comma would separate
2828 On the 386, you can use either multiple strings to specify multiple
2829 options, or you can separate the option with a comma (@code{,}).
2831 On the 386, the inliner will not inline a function that has different
2832 target options than the caller, unless the callee has a subset of the
2833 target options of the caller. For example a function declared with
2834 @code{option("sse5")} can inline a function with
2835 @code{option("sse2")}, since @code{-msse5} implies @code{-msse2}.
2837 The @code{option} attribute is not implemented in GCC versions earlier
2838 than 4.4, and at present only the 386 uses it.
2841 @cindex @code{optimize} function attribute
2842 The @code{optimize} attribute is used to specify that a function is to
2843 be compiled with different optimization options than specified on the
2844 command line. Arguments can either be numbers or strings. Numbers
2845 are assumed to be an optimization level. Strings that begin with
2846 @code{O} are assumed to be an optimization option, while other options
2847 are assumed to be used with a @code{-f} prefix. You can also use the
2848 @samp{#pragma GCC optimize} pragma to set the optimization options
2849 that affect more than one function.
2850 @xref{Function Specific Option Pragmas}, for details about the
2851 @samp{#pragma GCC option} pragma.
2853 This can be used for instance to have frequently executed functions
2854 compiled with more aggressive optimization options that produce faster
2855 and larger code, while other functions can be called with less
2856 aggressive options. On some targets, the @code{hot} attribute implies
2857 @code{optimize("O3")}, and @code{cold} attribute implies
2858 @code{optimize("Os")}.
2861 int fast_func (void) __attribute__ ((__optimize__ ("O3,unroll-loops")));
2862 int slow_func (void) __attribute__ ((__optimize__ ("Os")));
2865 The inliner will not inline functions with a higher optimization level
2866 than the caller or different space/time trade offs.
2869 @cindex @code{pure} function attribute
2870 Many functions have no effects except the return value and their
2871 return value depends only on the parameters and/or global variables.
2872 Such a function can be subject
2873 to common subexpression elimination and loop optimization just as an
2874 arithmetic operator would be. These functions should be declared
2875 with the attribute @code{pure}. For example,
2878 int square (int) __attribute__ ((pure));
2882 says that the hypothetical function @code{square} is safe to call
2883 fewer times than the program says.
2885 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2886 Interesting non-pure functions are functions with infinite loops or those
2887 depending on volatile memory or other system resource, that may change between
2888 two consecutive calls (such as @code{feof} in a multithreading environment).
2890 The attribute @code{pure} is not implemented in GCC versions earlier
2894 @cindex @code{hot} function attribute
2895 The @code{hot} attribute is used to inform the compiler that a function is a
2896 hot spot of the compiled program. The function is optimized more aggressively
2897 and on many target it is placed into special subsection of the text section so
2898 all hot functions appears close together improving locality.
2900 When profile feedback is available, via @option{-fprofile-use}, hot functions
2901 are automatically detected and this attribute is ignored.
2903 The @code{hot} attribute is not implemented in GCC versions earlier
2906 Starting with GCC 4.4, the @code{cold} attribute sets
2907 @code{optimize("O3")} to turn on more aggressive optimization on the
2908 the i386, x86_64, and IA-64 targets.
2911 @cindex @code{cold} function attribute
2912 The @code{cold} attribute is used to inform the compiler that a function is
2913 unlikely executed. The function is optimized for size rather than speed and on
2914 many targets it is placed into special subsection of the text section so all
2915 cold functions appears close together improving code locality of non-cold parts
2916 of program. The paths leading to call of cold functions within code are marked
2917 as unlikely by the branch prediction mechanism. It is thus useful to mark
2918 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2919 improve optimization of hot functions that do call marked functions in rare
2922 When profile feedback is available, via @option{-fprofile-use}, hot functions
2923 are automatically detected and this attribute is ignored.
2925 The @code{cold} attribute is not implemented in GCC versions earlier than 4.3.
2927 Starting with GCC 4.4, the @code{cold} attribute sets
2928 @code{optimize("Os")} to save space on the the i386, x86_64, and IA-64
2931 @item regparm (@var{number})
2932 @cindex @code{regparm} attribute
2933 @cindex functions that are passed arguments in registers on the 386
2934 On the Intel 386, the @code{regparm} attribute causes the compiler to
2935 pass arguments number one to @var{number} if they are of integral type
2936 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2937 take a variable number of arguments will continue to be passed all of their
2938 arguments on the stack.
2940 Beware that on some ELF systems this attribute is unsuitable for
2941 global functions in shared libraries with lazy binding (which is the
2942 default). Lazy binding will send the first call via resolving code in
2943 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2944 per the standard calling conventions. Solaris 8 is affected by this.
2945 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2946 safe since the loaders there save all registers. (Lazy binding can be
2947 disabled with the linker or the loader if desired, to avoid the
2951 @cindex @code{sseregparm} attribute
2952 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2953 causes the compiler to pass up to 3 floating point arguments in
2954 SSE registers instead of on the stack. Functions that take a
2955 variable number of arguments will continue to pass all of their
2956 floating point arguments on the stack.
2958 @item force_align_arg_pointer
2959 @cindex @code{force_align_arg_pointer} attribute
2960 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2961 applied to individual function definitions, generating an alternate
2962 prologue and epilogue that realigns the runtime stack if necessary.
2963 This supports mixing legacy codes that run with a 4-byte aligned stack
2964 with modern codes that keep a 16-byte stack for SSE compatibility.
2967 @cindex @code{resbank} attribute
2968 On the SH2A target, this attribute enables the high-speed register
2969 saving and restoration using a register bank for @code{interrupt_handler}
2970 routines. Saving to the bank is performed automatcially after the CPU
2971 accepts an interrupt that uses a register bank.
2973 The nineteen 32-bit registers comprising general register R0 to R14,
2974 control register GBR, and system registers MACH, MACL, and PR and the
2975 vector table address offset are saved into a register bank. Register
2976 banks are stacked in first-in last-out (FILO) sequence. Restoration
2977 from the bank is executed by issuing a RESBANK instruction.
2980 @cindex @code{returns_twice} attribute
2981 The @code{returns_twice} attribute tells the compiler that a function may
2982 return more than one time. The compiler will ensure that all registers
2983 are dead before calling such a function and will emit a warning about
2984 the variables that may be clobbered after the second return from the
2985 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2986 The @code{longjmp}-like counterpart of such function, if any, might need
2987 to be marked with the @code{noreturn} attribute.
2990 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2991 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2992 all registers except the stack pointer should be saved in the prologue
2993 regardless of whether they are used or not.
2995 @item section ("@var{section-name}")
2996 @cindex @code{section} function attribute
2997 Normally, the compiler places the code it generates in the @code{text} section.
2998 Sometimes, however, you need additional sections, or you need certain
2999 particular functions to appear in special sections. The @code{section}
3000 attribute specifies that a function lives in a particular section.
3001 For example, the declaration:
3004 extern void foobar (void) __attribute__ ((section ("bar")));
3008 puts the function @code{foobar} in the @code{bar} section.
3010 Some file formats do not support arbitrary sections so the @code{section}
3011 attribute is not available on all platforms.
3012 If you need to map the entire contents of a module to a particular
3013 section, consider using the facilities of the linker instead.
3016 @cindex @code{sentinel} function attribute
3017 This function attribute ensures that a parameter in a function call is
3018 an explicit @code{NULL}. The attribute is only valid on variadic
3019 functions. By default, the sentinel is located at position zero, the
3020 last parameter of the function call. If an optional integer position
3021 argument P is supplied to the attribute, the sentinel must be located at
3022 position P counting backwards from the end of the argument list.
3025 __attribute__ ((sentinel))
3027 __attribute__ ((sentinel(0)))
3030 The attribute is automatically set with a position of 0 for the built-in
3031 functions @code{execl} and @code{execlp}. The built-in function
3032 @code{execle} has the attribute set with a position of 1.
3034 A valid @code{NULL} in this context is defined as zero with any pointer
3035 type. If your system defines the @code{NULL} macro with an integer type
3036 then you need to add an explicit cast. GCC replaces @code{stddef.h}
3037 with a copy that redefines NULL appropriately.
3039 The warnings for missing or incorrect sentinels are enabled with
3043 See long_call/short_call.
3046 See longcall/shortcall.
3049 @cindex signal handler functions on the AVR processors
3050 Use this attribute on the AVR to indicate that the specified
3051 function is a signal handler. The compiler will generate function
3052 entry and exit sequences suitable for use in a signal handler when this
3053 attribute is present. Interrupts will be disabled inside the function.
3056 Use this attribute on the SH to indicate an @code{interrupt_handler}
3057 function should switch to an alternate stack. It expects a string
3058 argument that names a global variable holding the address of the
3063 void f () __attribute__ ((interrupt_handler,
3064 sp_switch ("alt_stack")));
3068 @cindex functions that pop the argument stack on the 386
3069 On the Intel 386, the @code{stdcall} attribute causes the compiler to
3070 assume that the called function will pop off the stack space used to
3071 pass arguments, unless it takes a variable number of arguments.
3074 @cindex tiny data section on the H8/300H and H8S
3075 Use this attribute on the H8/300H and H8S to indicate that the specified
3076 variable should be placed into the tiny data section.
3077 The compiler will generate more efficient code for loads and stores
3078 on data in the tiny data section. Note the tiny data area is limited to
3079 slightly under 32kbytes of data.
3082 Use this attribute on the SH for an @code{interrupt_handler} to return using
3083 @code{trapa} instead of @code{rte}. This attribute expects an integer
3084 argument specifying the trap number to be used.
3087 @cindex @code{unused} attribute.
3088 This attribute, attached to a function, means that the function is meant
3089 to be possibly unused. GCC will not produce a warning for this
3093 @cindex @code{used} attribute.
3094 This attribute, attached to a function, means that code must be emitted
3095 for the function even if it appears that the function is not referenced.
3096 This is useful, for example, when the function is referenced only in
3100 @cindex @code{version_id} attribute on IA64 HP-UX
3101 This attribute, attached to a global variable or function, renames a
3102 symbol to contain a version string, thus allowing for function level
3103 versioning. HP-UX system header files may use version level functioning
3104 for some system calls.
3107 extern int foo () __attribute__((version_id ("20040821")));
3110 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
3112 @item visibility ("@var{visibility_type}")
3113 @cindex @code{visibility} attribute
3114 This attribute affects the linkage of the declaration to which it is attached.
3115 There are four supported @var{visibility_type} values: default,
3116 hidden, protected or internal visibility.
3119 void __attribute__ ((visibility ("protected")))
3120 f () @{ /* @r{Do something.} */; @}
3121 int i __attribute__ ((visibility ("hidden")));
3124 The possible values of @var{visibility_type} correspond to the
3125 visibility settings in the ELF gABI.
3128 @c keep this list of visibilities in alphabetical order.
3131 Default visibility is the normal case for the object file format.
3132 This value is available for the visibility attribute to override other
3133 options that may change the assumed visibility of entities.
3135 On ELF, default visibility means that the declaration is visible to other
3136 modules and, in shared libraries, means that the declared entity may be
3139 On Darwin, default visibility means that the declaration is visible to
3142 Default visibility corresponds to ``external linkage'' in the language.
3145 Hidden visibility indicates that the entity declared will have a new
3146 form of linkage, which we'll call ``hidden linkage''. Two
3147 declarations of an object with hidden linkage refer to the same object
3148 if they are in the same shared object.
3151 Internal visibility is like hidden visibility, but with additional
3152 processor specific semantics. Unless otherwise specified by the
3153 psABI, GCC defines internal visibility to mean that a function is
3154 @emph{never} called from another module. Compare this with hidden
3155 functions which, while they cannot be referenced directly by other
3156 modules, can be referenced indirectly via function pointers. By
3157 indicating that a function cannot be called from outside the module,
3158 GCC may for instance omit the load of a PIC register since it is known
3159 that the calling function loaded the correct value.
3162 Protected visibility is like default visibility except that it
3163 indicates that references within the defining module will bind to the
3164 definition in that module. That is, the declared entity cannot be
3165 overridden by another module.
3169 All visibilities are supported on many, but not all, ELF targets
3170 (supported when the assembler supports the @samp{.visibility}
3171 pseudo-op). Default visibility is supported everywhere. Hidden
3172 visibility is supported on Darwin targets.
3174 The visibility attribute should be applied only to declarations which
3175 would otherwise have external linkage. The attribute should be applied
3176 consistently, so that the same entity should not be declared with
3177 different settings of the attribute.
3179 In C++, the visibility attribute applies to types as well as functions
3180 and objects, because in C++ types have linkage. A class must not have
3181 greater visibility than its non-static data member types and bases,
3182 and class members default to the visibility of their class. Also, a
3183 declaration without explicit visibility is limited to the visibility
3186 In C++, you can mark member functions and static member variables of a
3187 class with the visibility attribute. This is useful if if you know a
3188 particular method or static member variable should only be used from
3189 one shared object; then you can mark it hidden while the rest of the
3190 class has default visibility. Care must be taken to avoid breaking
3191 the One Definition Rule; for example, it is usually not useful to mark
3192 an inline method as hidden without marking the whole class as hidden.
3194 A C++ namespace declaration can also have the visibility attribute.
3195 This attribute applies only to the particular namespace body, not to
3196 other definitions of the same namespace; it is equivalent to using
3197 @samp{#pragma GCC visibility} before and after the namespace
3198 definition (@pxref{Visibility Pragmas}).
3200 In C++, if a template argument has limited visibility, this
3201 restriction is implicitly propagated to the template instantiation.
3202 Otherwise, template instantiations and specializations default to the
3203 visibility of their template.
3205 If both the template and enclosing class have explicit visibility, the
3206 visibility from the template is used.
3208 @item warn_unused_result
3209 @cindex @code{warn_unused_result} attribute
3210 The @code{warn_unused_result} attribute causes a warning to be emitted
3211 if a caller of the function with this attribute does not use its
3212 return value. This is useful for functions where not checking
3213 the result is either a security problem or always a bug, such as
3217 int fn () __attribute__ ((warn_unused_result));
3220 if (fn () < 0) return -1;
3226 results in warning on line 5.
3229 @cindex @code{weak} attribute
3230 The @code{weak} attribute causes the declaration to be emitted as a weak
3231 symbol rather than a global. This is primarily useful in defining
3232 library functions which can be overridden in user code, though it can
3233 also be used with non-function declarations. Weak symbols are supported
3234 for ELF targets, and also for a.out targets when using the GNU assembler
3238 @itemx weakref ("@var{target}")
3239 @cindex @code{weakref} attribute
3240 The @code{weakref} attribute marks a declaration as a weak reference.
3241 Without arguments, it should be accompanied by an @code{alias} attribute
3242 naming the target symbol. Optionally, the @var{target} may be given as
3243 an argument to @code{weakref} itself. In either case, @code{weakref}
3244 implicitly marks the declaration as @code{weak}. Without a
3245 @var{target}, given as an argument to @code{weakref} or to @code{alias},
3246 @code{weakref} is equivalent to @code{weak}.
3249 static int x() __attribute__ ((weakref ("y")));
3250 /* is equivalent to... */
3251 static int x() __attribute__ ((weak, weakref, alias ("y")));
3253 static int x() __attribute__ ((weakref));
3254 static int x() __attribute__ ((alias ("y")));
3257 A weak reference is an alias that does not by itself require a
3258 definition to be given for the target symbol. If the target symbol is
3259 only referenced through weak references, then the becomes a @code{weak}
3260 undefined symbol. If it is directly referenced, however, then such
3261 strong references prevail, and a definition will be required for the
3262 symbol, not necessarily in the same translation unit.
3264 The effect is equivalent to moving all references to the alias to a
3265 separate translation unit, renaming the alias to the aliased symbol,
3266 declaring it as weak, compiling the two separate translation units and
3267 performing a reloadable link on them.
3269 At present, a declaration to which @code{weakref} is attached can
3270 only be @code{static}.
3272 @item externally_visible
3273 @cindex @code{externally_visible} attribute.
3274 This attribute, attached to a global variable or function nullify
3275 effect of @option{-fwhole-program} command line option, so the object
3276 remain visible outside the current compilation unit
3280 You can specify multiple attributes in a declaration by separating them
3281 by commas within the double parentheses or by immediately following an
3282 attribute declaration with another attribute declaration.
3284 @cindex @code{#pragma}, reason for not using
3285 @cindex pragma, reason for not using
3286 Some people object to the @code{__attribute__} feature, suggesting that
3287 ISO C's @code{#pragma} should be used instead. At the time
3288 @code{__attribute__} was designed, there were two reasons for not doing
3293 It is impossible to generate @code{#pragma} commands from a macro.
3296 There is no telling what the same @code{#pragma} might mean in another
3300 These two reasons applied to almost any application that might have been
3301 proposed for @code{#pragma}. It was basically a mistake to use
3302 @code{#pragma} for @emph{anything}.
3304 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
3305 to be generated from macros. In addition, a @code{#pragma GCC}
3306 namespace is now in use for GCC-specific pragmas. However, it has been
3307 found convenient to use @code{__attribute__} to achieve a natural
3308 attachment of attributes to their corresponding declarations, whereas
3309 @code{#pragma GCC} is of use for constructs that do not naturally form
3310 part of the grammar. @xref{Other Directives,,Miscellaneous
3311 Preprocessing Directives, cpp, The GNU C Preprocessor}.
3313 @node Attribute Syntax
3314 @section Attribute Syntax
3315 @cindex attribute syntax
3317 This section describes the syntax with which @code{__attribute__} may be
3318 used, and the constructs to which attribute specifiers bind, for the C
3319 language. Some details may vary for C++ and Objective-C@. Because of
3320 infelicities in the grammar for attributes, some forms described here
3321 may not be successfully parsed in all cases.
3323 There are some problems with the semantics of attributes in C++. For
3324 example, there are no manglings for attributes, although they may affect
3325 code generation, so problems may arise when attributed types are used in
3326 conjunction with templates or overloading. Similarly, @code{typeid}
3327 does not distinguish between types with different attributes. Support
3328 for attributes in C++ may be restricted in future to attributes on
3329 declarations only, but not on nested declarators.
3331 @xref{Function Attributes}, for details of the semantics of attributes
3332 applying to functions. @xref{Variable Attributes}, for details of the
3333 semantics of attributes applying to variables. @xref{Type Attributes},
3334 for details of the semantics of attributes applying to structure, union
3335 and enumerated types.
3337 An @dfn{attribute specifier} is of the form
3338 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
3339 is a possibly empty comma-separated sequence of @dfn{attributes}, where
3340 each attribute is one of the following:
3344 Empty. Empty attributes are ignored.
3347 A word (which may be an identifier such as @code{unused}, or a reserved
3348 word such as @code{const}).
3351 A word, followed by, in parentheses, parameters for the attribute.
3352 These parameters take one of the following forms:
3356 An identifier. For example, @code{mode} attributes use this form.
3359 An identifier followed by a comma and a non-empty comma-separated list
3360 of expressions. For example, @code{format} attributes use this form.
3363 A possibly empty comma-separated list of expressions. For example,
3364 @code{format_arg} attributes use this form with the list being a single
3365 integer constant expression, and @code{alias} attributes use this form
3366 with the list being a single string constant.
3370 An @dfn{attribute specifier list} is a sequence of one or more attribute
3371 specifiers, not separated by any other tokens.
3373 In GNU C, an attribute specifier list may appear after the colon following a
3374 label, other than a @code{case} or @code{default} label. The only
3375 attribute it makes sense to use after a label is @code{unused}. This
3376 feature is intended for code generated by programs which contains labels
3377 that may be unused but which is compiled with @option{-Wall}. It would
3378 not normally be appropriate to use in it human-written code, though it
3379 could be useful in cases where the code that jumps to the label is
3380 contained within an @code{#ifdef} conditional. GNU C++ does not permit
3381 such placement of attribute lists, as it is permissible for a
3382 declaration, which could begin with an attribute list, to be labelled in
3383 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
3384 does not arise there.
3386 An attribute specifier list may appear as part of a @code{struct},
3387 @code{union} or @code{enum} specifier. It may go either immediately
3388 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3389 the closing brace. The former syntax is preferred.
3390 Where attribute specifiers follow the closing brace, they are considered
3391 to relate to the structure, union or enumerated type defined, not to any
3392 enclosing declaration the type specifier appears in, and the type
3393 defined is not complete until after the attribute specifiers.
3394 @c Otherwise, there would be the following problems: a shift/reduce
3395 @c conflict between attributes binding the struct/union/enum and
3396 @c binding to the list of specifiers/qualifiers; and "aligned"
3397 @c attributes could use sizeof for the structure, but the size could be
3398 @c changed later by "packed" attributes.
3400 Otherwise, an attribute specifier appears as part of a declaration,
3401 counting declarations of unnamed parameters and type names, and relates
3402 to that declaration (which may be nested in another declaration, for
3403 example in the case of a parameter declaration), or to a particular declarator
3404 within a declaration. Where an
3405 attribute specifier is applied to a parameter declared as a function or
3406 an array, it should apply to the function or array rather than the
3407 pointer to which the parameter is implicitly converted, but this is not
3408 yet correctly implemented.
3410 Any list of specifiers and qualifiers at the start of a declaration may
3411 contain attribute specifiers, whether or not such a list may in that
3412 context contain storage class specifiers. (Some attributes, however,
3413 are essentially in the nature of storage class specifiers, and only make
3414 sense where storage class specifiers may be used; for example,
3415 @code{section}.) There is one necessary limitation to this syntax: the
3416 first old-style parameter declaration in a function definition cannot
3417 begin with an attribute specifier, because such an attribute applies to
3418 the function instead by syntax described below (which, however, is not
3419 yet implemented in this case). In some other cases, attribute
3420 specifiers are permitted by this grammar but not yet supported by the
3421 compiler. All attribute specifiers in this place relate to the
3422 declaration as a whole. In the obsolescent usage where a type of
3423 @code{int} is implied by the absence of type specifiers, such a list of
3424 specifiers and qualifiers may be an attribute specifier list with no
3425 other specifiers or qualifiers.
3427 At present, the first parameter in a function prototype must have some
3428 type specifier which is not an attribute specifier; this resolves an
3429 ambiguity in the interpretation of @code{void f(int
3430 (__attribute__((foo)) x))}, but is subject to change. At present, if
3431 the parentheses of a function declarator contain only attributes then
3432 those attributes are ignored, rather than yielding an error or warning
3433 or implying a single parameter of type int, but this is subject to
3436 An attribute specifier list may appear immediately before a declarator
3437 (other than the first) in a comma-separated list of declarators in a
3438 declaration of more than one identifier using a single list of
3439 specifiers and qualifiers. Such attribute specifiers apply
3440 only to the identifier before whose declarator they appear. For
3444 __attribute__((noreturn)) void d0 (void),
3445 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3450 the @code{noreturn} attribute applies to all the functions
3451 declared; the @code{format} attribute only applies to @code{d1}.
3453 An attribute specifier list may appear immediately before the comma,
3454 @code{=} or semicolon terminating the declaration of an identifier other
3455 than a function definition. Such attribute specifiers apply
3456 to the declared object or function. Where an
3457 assembler name for an object or function is specified (@pxref{Asm
3458 Labels}), the attribute must follow the @code{asm}
3461 An attribute specifier list may, in future, be permitted to appear after
3462 the declarator in a function definition (before any old-style parameter
3463 declarations or the function body).
3465 Attribute specifiers may be mixed with type qualifiers appearing inside
3466 the @code{[]} of a parameter array declarator, in the C99 construct by
3467 which such qualifiers are applied to the pointer to which the array is
3468 implicitly converted. Such attribute specifiers apply to the pointer,
3469 not to the array, but at present this is not implemented and they are
3472 An attribute specifier list may appear at the start of a nested
3473 declarator. At present, there are some limitations in this usage: the
3474 attributes correctly apply to the declarator, but for most individual
3475 attributes the semantics this implies are not implemented.
3476 When attribute specifiers follow the @code{*} of a pointer
3477 declarator, they may be mixed with any type qualifiers present.
3478 The following describes the formal semantics of this syntax. It will make the
3479 most sense if you are familiar with the formal specification of
3480 declarators in the ISO C standard.
3482 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3483 D1}, where @code{T} contains declaration specifiers that specify a type
3484 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3485 contains an identifier @var{ident}. The type specified for @var{ident}
3486 for derived declarators whose type does not include an attribute
3487 specifier is as in the ISO C standard.
3489 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3490 and the declaration @code{T D} specifies the type
3491 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3492 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3493 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3495 If @code{D1} has the form @code{*
3496 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3497 declaration @code{T D} specifies the type
3498 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3499 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3500 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3506 void (__attribute__((noreturn)) ****f) (void);
3510 specifies the type ``pointer to pointer to pointer to pointer to
3511 non-returning function returning @code{void}''. As another example,
3514 char *__attribute__((aligned(8))) *f;
3518 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3519 Note again that this does not work with most attributes; for example,
3520 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3521 is not yet supported.
3523 For compatibility with existing code written for compiler versions that
3524 did not implement attributes on nested declarators, some laxity is
3525 allowed in the placing of attributes. If an attribute that only applies
3526 to types is applied to a declaration, it will be treated as applying to
3527 the type of that declaration. If an attribute that only applies to
3528 declarations is applied to the type of a declaration, it will be treated
3529 as applying to that declaration; and, for compatibility with code
3530 placing the attributes immediately before the identifier declared, such
3531 an attribute applied to a function return type will be treated as
3532 applying to the function type, and such an attribute applied to an array
3533 element type will be treated as applying to the array type. If an
3534 attribute that only applies to function types is applied to a
3535 pointer-to-function type, it will be treated as applying to the pointer
3536 target type; if such an attribute is applied to a function return type
3537 that is not a pointer-to-function type, it will be treated as applying
3538 to the function type.
3540 @node Function Prototypes
3541 @section Prototypes and Old-Style Function Definitions
3542 @cindex function prototype declarations
3543 @cindex old-style function definitions
3544 @cindex promotion of formal parameters
3546 GNU C extends ISO C to allow a function prototype to override a later
3547 old-style non-prototype definition. Consider the following example:
3550 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3557 /* @r{Prototype function declaration.} */
3558 int isroot P((uid_t));
3560 /* @r{Old-style function definition.} */
3562 isroot (x) /* @r{??? lossage here ???} */
3569 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3570 not allow this example, because subword arguments in old-style
3571 non-prototype definitions are promoted. Therefore in this example the
3572 function definition's argument is really an @code{int}, which does not
3573 match the prototype argument type of @code{short}.
3575 This restriction of ISO C makes it hard to write code that is portable
3576 to traditional C compilers, because the programmer does not know
3577 whether the @code{uid_t} type is @code{short}, @code{int}, or
3578 @code{long}. Therefore, in cases like these GNU C allows a prototype
3579 to override a later old-style definition. More precisely, in GNU C, a
3580 function prototype argument type overrides the argument type specified
3581 by a later old-style definition if the former type is the same as the
3582 latter type before promotion. Thus in GNU C the above example is
3583 equivalent to the following:
3596 GNU C++ does not support old-style function definitions, so this
3597 extension is irrelevant.
3600 @section C++ Style Comments
3602 @cindex C++ comments
3603 @cindex comments, C++ style
3605 In GNU C, you may use C++ style comments, which start with @samp{//} and
3606 continue until the end of the line. Many other C implementations allow
3607 such comments, and they are included in the 1999 C standard. However,
3608 C++ style comments are not recognized if you specify an @option{-std}
3609 option specifying a version of ISO C before C99, or @option{-ansi}
3610 (equivalent to @option{-std=c89}).
3613 @section Dollar Signs in Identifier Names
3615 @cindex dollar signs in identifier names
3616 @cindex identifier names, dollar signs in
3618 In GNU C, you may normally use dollar signs in identifier names.
3619 This is because many traditional C implementations allow such identifiers.
3620 However, dollar signs in identifiers are not supported on a few target
3621 machines, typically because the target assembler does not allow them.
3623 @node Character Escapes
3624 @section The Character @key{ESC} in Constants
3626 You can use the sequence @samp{\e} in a string or character constant to
3627 stand for the ASCII character @key{ESC}.
3630 @section Inquiring on Alignment of Types or Variables
3632 @cindex type alignment
3633 @cindex variable alignment
3635 The keyword @code{__alignof__} allows you to inquire about how an object
3636 is aligned, or the minimum alignment usually required by a type. Its
3637 syntax is just like @code{sizeof}.
3639 For example, if the target machine requires a @code{double} value to be
3640 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3641 This is true on many RISC machines. On more traditional machine
3642 designs, @code{__alignof__ (double)} is 4 or even 2.
3644 Some machines never actually require alignment; they allow reference to any
3645 data type even at an odd address. For these machines, @code{__alignof__}
3646 reports the smallest alignment that GCC will give the data type, usually as
3647 mandated by the target ABI.
3649 If the operand of @code{__alignof__} is an lvalue rather than a type,
3650 its value is the required alignment for its type, taking into account
3651 any minimum alignment specified with GCC's @code{__attribute__}
3652 extension (@pxref{Variable Attributes}). For example, after this
3656 struct foo @{ int x; char y; @} foo1;
3660 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3661 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3663 It is an error to ask for the alignment of an incomplete type.
3665 @node Variable Attributes
3666 @section Specifying Attributes of Variables
3667 @cindex attribute of variables
3668 @cindex variable attributes
3670 The keyword @code{__attribute__} allows you to specify special
3671 attributes of variables or structure fields. This keyword is followed
3672 by an attribute specification inside double parentheses. Some
3673 attributes are currently defined generically for variables.
3674 Other attributes are defined for variables on particular target
3675 systems. Other attributes are available for functions
3676 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3677 Other front ends might define more attributes
3678 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3680 You may also specify attributes with @samp{__} preceding and following
3681 each keyword. This allows you to use them in header files without
3682 being concerned about a possible macro of the same name. For example,
3683 you may use @code{__aligned__} instead of @code{aligned}.
3685 @xref{Attribute Syntax}, for details of the exact syntax for using
3689 @cindex @code{aligned} attribute
3690 @item aligned (@var{alignment})
3691 This attribute specifies a minimum alignment for the variable or
3692 structure field, measured in bytes. For example, the declaration:
3695 int x __attribute__ ((aligned (16))) = 0;
3699 causes the compiler to allocate the global variable @code{x} on a
3700 16-byte boundary. On a 68040, this could be used in conjunction with
3701 an @code{asm} expression to access the @code{move16} instruction which
3702 requires 16-byte aligned operands.
3704 You can also specify the alignment of structure fields. For example, to
3705 create a double-word aligned @code{int} pair, you could write:
3708 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3712 This is an alternative to creating a union with a @code{double} member
3713 that forces the union to be double-word aligned.
3715 As in the preceding examples, you can explicitly specify the alignment
3716 (in bytes) that you wish the compiler to use for a given variable or
3717 structure field. Alternatively, you can leave out the alignment factor
3718 and just ask the compiler to align a variable or field to the maximum
3719 useful alignment for the target machine you are compiling for. For
3720 example, you could write:
3723 short array[3] __attribute__ ((aligned));
3726 Whenever you leave out the alignment factor in an @code{aligned} attribute
3727 specification, the compiler automatically sets the alignment for the declared
3728 variable or field to the largest alignment which is ever used for any data
3729 type on the target machine you are compiling for. Doing this can often make
3730 copy operations more efficient, because the compiler can use whatever
3731 instructions copy the biggest chunks of memory when performing copies to
3732 or from the variables or fields that you have aligned this way.
3734 When used on a struct, or struct member, the @code{aligned} attribute can
3735 only increase the alignment; in order to decrease it, the @code{packed}
3736 attribute must be specified as well. When used as part of a typedef, the
3737 @code{aligned} attribute can both increase and decrease alignment, and
3738 specifying the @code{packed} attribute will generate a warning.
3740 Note that the effectiveness of @code{aligned} attributes may be limited
3741 by inherent limitations in your linker. On many systems, the linker is
3742 only able to arrange for variables to be aligned up to a certain maximum
3743 alignment. (For some linkers, the maximum supported alignment may
3744 be very very small.) If your linker is only able to align variables
3745 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3746 in an @code{__attribute__} will still only provide you with 8 byte
3747 alignment. See your linker documentation for further information.
3749 The @code{aligned} attribute can also be used for functions
3750 (@pxref{Function Attributes}.)
3752 @item cleanup (@var{cleanup_function})
3753 @cindex @code{cleanup} attribute
3754 The @code{cleanup} attribute runs a function when the variable goes
3755 out of scope. This attribute can only be applied to auto function
3756 scope variables; it may not be applied to parameters or variables
3757 with static storage duration. The function must take one parameter,
3758 a pointer to a type compatible with the variable. The return value
3759 of the function (if any) is ignored.
3761 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3762 will be run during the stack unwinding that happens during the
3763 processing of the exception. Note that the @code{cleanup} attribute
3764 does not allow the exception to be caught, only to perform an action.
3765 It is undefined what happens if @var{cleanup_function} does not
3770 @cindex @code{common} attribute
3771 @cindex @code{nocommon} attribute
3774 The @code{common} attribute requests GCC to place a variable in
3775 ``common'' storage. The @code{nocommon} attribute requests the
3776 opposite---to allocate space for it directly.
3778 These attributes override the default chosen by the
3779 @option{-fno-common} and @option{-fcommon} flags respectively.
3782 @cindex @code{deprecated} attribute
3783 The @code{deprecated} attribute results in a warning if the variable
3784 is used anywhere in the source file. This is useful when identifying
3785 variables that are expected to be removed in a future version of a
3786 program. The warning also includes the location of the declaration
3787 of the deprecated variable, to enable users to easily find further
3788 information about why the variable is deprecated, or what they should
3789 do instead. Note that the warning only occurs for uses:
3792 extern int old_var __attribute__ ((deprecated));
3794 int new_fn () @{ return old_var; @}
3797 results in a warning on line 3 but not line 2.
3799 The @code{deprecated} attribute can also be used for functions and
3800 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3802 @item mode (@var{mode})
3803 @cindex @code{mode} attribute
3804 This attribute specifies the data type for the declaration---whichever
3805 type corresponds to the mode @var{mode}. This in effect lets you
3806 request an integer or floating point type according to its width.
3808 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3809 indicate the mode corresponding to a one-byte integer, @samp{word} or
3810 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3811 or @samp{__pointer__} for the mode used to represent pointers.
3814 @cindex @code{packed} attribute
3815 The @code{packed} attribute specifies that a variable or structure field
3816 should have the smallest possible alignment---one byte for a variable,
3817 and one bit for a field, unless you specify a larger value with the
3818 @code{aligned} attribute.
3820 Here is a structure in which the field @code{x} is packed, so that it
3821 immediately follows @code{a}:
3827 int x[2] __attribute__ ((packed));
3831 @item section ("@var{section-name}")
3832 @cindex @code{section} variable attribute
3833 Normally, the compiler places the objects it generates in sections like
3834 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3835 or you need certain particular variables to appear in special sections,
3836 for example to map to special hardware. The @code{section}
3837 attribute specifies that a variable (or function) lives in a particular
3838 section. For example, this small program uses several specific section names:
3841 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3842 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3843 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3844 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3848 /* @r{Initialize stack pointer} */
3849 init_sp (stack + sizeof (stack));
3851 /* @r{Initialize initialized data} */
3852 memcpy (&init_data, &data, &edata - &data);
3854 /* @r{Turn on the serial ports} */
3861 Use the @code{section} attribute with an @emph{initialized} definition
3862 of a @emph{global} variable, as shown in the example. GCC issues
3863 a warning and otherwise ignores the @code{section} attribute in
3864 uninitialized variable declarations.
3866 You may only use the @code{section} attribute with a fully initialized
3867 global definition because of the way linkers work. The linker requires
3868 each object be defined once, with the exception that uninitialized
3869 variables tentatively go in the @code{common} (or @code{bss}) section
3870 and can be multiply ``defined''. You can force a variable to be
3871 initialized with the @option{-fno-common} flag or the @code{nocommon}
3874 Some file formats do not support arbitrary sections so the @code{section}
3875 attribute is not available on all platforms.
3876 If you need to map the entire contents of a module to a particular
3877 section, consider using the facilities of the linker instead.
3880 @cindex @code{shared} variable attribute
3881 On Microsoft Windows, in addition to putting variable definitions in a named
3882 section, the section can also be shared among all running copies of an
3883 executable or DLL@. For example, this small program defines shared data
3884 by putting it in a named section @code{shared} and marking the section
3888 int foo __attribute__((section ("shared"), shared)) = 0;
3893 /* @r{Read and write foo. All running
3894 copies see the same value.} */
3900 You may only use the @code{shared} attribute along with @code{section}
3901 attribute with a fully initialized global definition because of the way
3902 linkers work. See @code{section} attribute for more information.
3904 The @code{shared} attribute is only available on Microsoft Windows@.
3906 @item tls_model ("@var{tls_model}")
3907 @cindex @code{tls_model} attribute
3908 The @code{tls_model} attribute sets thread-local storage model
3909 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3910 overriding @option{-ftls-model=} command line switch on a per-variable
3912 The @var{tls_model} argument should be one of @code{global-dynamic},
3913 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3915 Not all targets support this attribute.
3918 This attribute, attached to a variable, means that the variable is meant
3919 to be possibly unused. GCC will not produce a warning for this
3923 This attribute, attached to a variable, means that the variable must be
3924 emitted even if it appears that the variable is not referenced.
3926 @item vector_size (@var{bytes})
3927 This attribute specifies the vector size for the variable, measured in
3928 bytes. For example, the declaration:
3931 int foo __attribute__ ((vector_size (16)));
3935 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3936 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3937 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3939 This attribute is only applicable to integral and float scalars,
3940 although arrays, pointers, and function return values are allowed in
3941 conjunction with this construct.
3943 Aggregates with this attribute are invalid, even if they are of the same
3944 size as a corresponding scalar. For example, the declaration:
3947 struct S @{ int a; @};
3948 struct S __attribute__ ((vector_size (16))) foo;
3952 is invalid even if the size of the structure is the same as the size of
3956 The @code{selectany} attribute causes an initialized global variable to
3957 have link-once semantics. When multiple definitions of the variable are
3958 encountered by the linker, the first is selected and the remainder are
3959 discarded. Following usage by the Microsoft compiler, the linker is told
3960 @emph{not} to warn about size or content differences of the multiple
3963 Although the primary usage of this attribute is for POD types, the
3964 attribute can also be applied to global C++ objects that are initialized
3965 by a constructor. In this case, the static initialization and destruction
3966 code for the object is emitted in each translation defining the object,
3967 but the calls to the constructor and destructor are protected by a
3968 link-once guard variable.
3970 The @code{selectany} attribute is only available on Microsoft Windows
3971 targets. You can use @code{__declspec (selectany)} as a synonym for
3972 @code{__attribute__ ((selectany))} for compatibility with other
3976 The @code{weak} attribute is described in @ref{Function Attributes}.
3979 The @code{dllimport} attribute is described in @ref{Function Attributes}.
3982 The @code{dllexport} attribute is described in @ref{Function Attributes}.
3986 @subsection Blackfin Variable Attributes
3988 Three attributes are currently defined for the Blackfin.
3994 @cindex @code{l1_data} variable attribute
3995 @cindex @code{l1_data_A} variable attribute
3996 @cindex @code{l1_data_B} variable attribute
3997 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
3998 Variables with @code{l1_data} attribute will be put into the specific section
3999 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
4000 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
4001 attribute will be put into the specific section named @code{.l1.data.B}.
4004 @subsection M32R/D Variable Attributes
4006 One attribute is currently defined for the M32R/D@.
4009 @item model (@var{model-name})
4010 @cindex variable addressability on the M32R/D
4011 Use this attribute on the M32R/D to set the addressability of an object.
4012 The identifier @var{model-name} is one of @code{small}, @code{medium},
4013 or @code{large}, representing each of the code models.
4015 Small model objects live in the lower 16MB of memory (so that their
4016 addresses can be loaded with the @code{ld24} instruction).
4018 Medium and large model objects may live anywhere in the 32-bit address space
4019 (the compiler will generate @code{seth/add3} instructions to load their
4023 @anchor{i386 Variable Attributes}
4024 @subsection i386 Variable Attributes
4026 Two attributes are currently defined for i386 configurations:
4027 @code{ms_struct} and @code{gcc_struct}
4032 @cindex @code{ms_struct} attribute
4033 @cindex @code{gcc_struct} attribute
4035 If @code{packed} is used on a structure, or if bit-fields are used
4036 it may be that the Microsoft ABI packs them differently
4037 than GCC would normally pack them. Particularly when moving packed
4038 data between functions compiled with GCC and the native Microsoft compiler
4039 (either via function call or as data in a file), it may be necessary to access
4042 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4043 compilers to match the native Microsoft compiler.
4045 The Microsoft structure layout algorithm is fairly simple with the exception
4046 of the bitfield packing:
4048 The padding and alignment of members of structures and whether a bit field
4049 can straddle a storage-unit boundary
4052 @item Structure members are stored sequentially in the order in which they are
4053 declared: the first member has the lowest memory address and the last member
4056 @item Every data object has an alignment-requirement. The alignment-requirement
4057 for all data except structures, unions, and arrays is either the size of the
4058 object or the current packing size (specified with either the aligned attribute
4059 or the pack pragma), whichever is less. For structures, unions, and arrays,
4060 the alignment-requirement is the largest alignment-requirement of its members.
4061 Every object is allocated an offset so that:
4063 offset % alignment-requirement == 0
4065 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
4066 unit if the integral types are the same size and if the next bit field fits
4067 into the current allocation unit without crossing the boundary imposed by the
4068 common alignment requirements of the bit fields.
4071 Handling of zero-length bitfields:
4073 MSVC interprets zero-length bitfields in the following ways:
4076 @item If a zero-length bitfield is inserted between two bitfields that would
4077 normally be coalesced, the bitfields will not be coalesced.
4084 unsigned long bf_1 : 12;
4086 unsigned long bf_2 : 12;
4090 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
4091 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
4093 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
4094 alignment of the zero-length bitfield is greater than the member that follows it,
4095 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
4115 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
4116 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
4117 bitfield will not affect the alignment of @code{bar} or, as a result, the size
4120 Taking this into account, it is important to note the following:
4123 @item If a zero-length bitfield follows a normal bitfield, the type of the
4124 zero-length bitfield may affect the alignment of the structure as whole. For
4125 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
4126 normal bitfield, and is of type short.
4128 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
4129 still affect the alignment of the structure:
4139 Here, @code{t4} will take up 4 bytes.
4142 @item Zero-length bitfields following non-bitfield members are ignored:
4153 Here, @code{t5} will take up 2 bytes.
4157 @subsection PowerPC Variable Attributes
4159 Three attributes currently are defined for PowerPC configurations:
4160 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4162 For full documentation of the struct attributes please see the
4163 documentation in @ref{i386 Variable Attributes}.
4165 For documentation of @code{altivec} attribute please see the
4166 documentation in @ref{PowerPC Type Attributes}.
4168 @subsection SPU Variable Attributes
4170 The SPU supports the @code{spu_vector} attribute for variables. For
4171 documentation of this attribute please see the documentation in
4172 @ref{SPU Type Attributes}.
4174 @subsection Xstormy16 Variable Attributes
4176 One attribute is currently defined for xstormy16 configurations:
4181 @cindex @code{below100} attribute
4183 If a variable has the @code{below100} attribute (@code{BELOW100} is
4184 allowed also), GCC will place the variable in the first 0x100 bytes of
4185 memory and use special opcodes to access it. Such variables will be
4186 placed in either the @code{.bss_below100} section or the
4187 @code{.data_below100} section.
4191 @subsection AVR Variable Attributes
4195 @cindex @code{progmem} variable attribute
4196 The @code{progmem} attribute is used on the AVR to place data in the Program
4197 Memory address space. The AVR is a Harvard Architecture processor and data
4198 normally resides in the Data Memory address space.
4201 @node Type Attributes
4202 @section Specifying Attributes of Types
4203 @cindex attribute of types
4204 @cindex type attributes
4206 The keyword @code{__attribute__} allows you to specify special
4207 attributes of @code{struct} and @code{union} types when you define
4208 such types. This keyword is followed by an attribute specification
4209 inside double parentheses. Seven attributes are currently defined for
4210 types: @code{aligned}, @code{packed}, @code{transparent_union},
4211 @code{unused}, @code{deprecated}, @code{visibility}, and
4212 @code{may_alias}. Other attributes are defined for functions
4213 (@pxref{Function Attributes}) and for variables (@pxref{Variable
4216 You may also specify any one of these attributes with @samp{__}
4217 preceding and following its keyword. This allows you to use these
4218 attributes in header files without being concerned about a possible
4219 macro of the same name. For example, you may use @code{__aligned__}
4220 instead of @code{aligned}.
4222 You may specify type attributes in an enum, struct or union type
4223 declaration or definition, or for other types in a @code{typedef}
4226 For an enum, struct or union type, you may specify attributes either
4227 between the enum, struct or union tag and the name of the type, or
4228 just past the closing curly brace of the @emph{definition}. The
4229 former syntax is preferred.
4231 @xref{Attribute Syntax}, for details of the exact syntax for using
4235 @cindex @code{aligned} attribute
4236 @item aligned (@var{alignment})
4237 This attribute specifies a minimum alignment (in bytes) for variables
4238 of the specified type. For example, the declarations:
4241 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
4242 typedef int more_aligned_int __attribute__ ((aligned (8)));
4246 force the compiler to insure (as far as it can) that each variable whose
4247 type is @code{struct S} or @code{more_aligned_int} will be allocated and
4248 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
4249 variables of type @code{struct S} aligned to 8-byte boundaries allows
4250 the compiler to use the @code{ldd} and @code{std} (doubleword load and
4251 store) instructions when copying one variable of type @code{struct S} to
4252 another, thus improving run-time efficiency.
4254 Note that the alignment of any given @code{struct} or @code{union} type
4255 is required by the ISO C standard to be at least a perfect multiple of
4256 the lowest common multiple of the alignments of all of the members of
4257 the @code{struct} or @code{union} in question. This means that you @emph{can}
4258 effectively adjust the alignment of a @code{struct} or @code{union}
4259 type by attaching an @code{aligned} attribute to any one of the members
4260 of such a type, but the notation illustrated in the example above is a
4261 more obvious, intuitive, and readable way to request the compiler to
4262 adjust the alignment of an entire @code{struct} or @code{union} type.
4264 As in the preceding example, you can explicitly specify the alignment
4265 (in bytes) that you wish the compiler to use for a given @code{struct}
4266 or @code{union} type. Alternatively, you can leave out the alignment factor
4267 and just ask the compiler to align a type to the maximum
4268 useful alignment for the target machine you are compiling for. For
4269 example, you could write:
4272 struct S @{ short f[3]; @} __attribute__ ((aligned));
4275 Whenever you leave out the alignment factor in an @code{aligned}
4276 attribute specification, the compiler automatically sets the alignment
4277 for the type to the largest alignment which is ever used for any data
4278 type on the target machine you are compiling for. Doing this can often
4279 make copy operations more efficient, because the compiler can use
4280 whatever instructions copy the biggest chunks of memory when performing
4281 copies to or from the variables which have types that you have aligned
4284 In the example above, if the size of each @code{short} is 2 bytes, then
4285 the size of the entire @code{struct S} type is 6 bytes. The smallest
4286 power of two which is greater than or equal to that is 8, so the
4287 compiler sets the alignment for the entire @code{struct S} type to 8
4290 Note that although you can ask the compiler to select a time-efficient
4291 alignment for a given type and then declare only individual stand-alone
4292 objects of that type, the compiler's ability to select a time-efficient
4293 alignment is primarily useful only when you plan to create arrays of
4294 variables having the relevant (efficiently aligned) type. If you
4295 declare or use arrays of variables of an efficiently-aligned type, then
4296 it is likely that your program will also be doing pointer arithmetic (or
4297 subscripting, which amounts to the same thing) on pointers to the
4298 relevant type, and the code that the compiler generates for these
4299 pointer arithmetic operations will often be more efficient for
4300 efficiently-aligned types than for other types.
4302 The @code{aligned} attribute can only increase the alignment; but you
4303 can decrease it by specifying @code{packed} as well. See below.
4305 Note that the effectiveness of @code{aligned} attributes may be limited
4306 by inherent limitations in your linker. On many systems, the linker is
4307 only able to arrange for variables to be aligned up to a certain maximum
4308 alignment. (For some linkers, the maximum supported alignment may
4309 be very very small.) If your linker is only able to align variables
4310 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
4311 in an @code{__attribute__} will still only provide you with 8 byte
4312 alignment. See your linker documentation for further information.
4315 This attribute, attached to @code{struct} or @code{union} type
4316 definition, specifies that each member (other than zero-width bitfields)
4317 of the structure or union is placed to minimize the memory required. When
4318 attached to an @code{enum} definition, it indicates that the smallest
4319 integral type should be used.
4321 @opindex fshort-enums
4322 Specifying this attribute for @code{struct} and @code{union} types is
4323 equivalent to specifying the @code{packed} attribute on each of the
4324 structure or union members. Specifying the @option{-fshort-enums}
4325 flag on the line is equivalent to specifying the @code{packed}
4326 attribute on all @code{enum} definitions.
4328 In the following example @code{struct my_packed_struct}'s members are
4329 packed closely together, but the internal layout of its @code{s} member
4330 is not packed---to do that, @code{struct my_unpacked_struct} would need to
4334 struct my_unpacked_struct
4340 struct __attribute__ ((__packed__)) my_packed_struct
4344 struct my_unpacked_struct s;
4348 You may only specify this attribute on the definition of a @code{enum},
4349 @code{struct} or @code{union}, not on a @code{typedef} which does not
4350 also define the enumerated type, structure or union.
4352 @item transparent_union
4353 This attribute, attached to a @code{union} type definition, indicates
4354 that any function parameter having that union type causes calls to that
4355 function to be treated in a special way.
4357 First, the argument corresponding to a transparent union type can be of
4358 any type in the union; no cast is required. Also, if the union contains
4359 a pointer type, the corresponding argument can be a null pointer
4360 constant or a void pointer expression; and if the union contains a void
4361 pointer type, the corresponding argument can be any pointer expression.
4362 If the union member type is a pointer, qualifiers like @code{const} on
4363 the referenced type must be respected, just as with normal pointer
4366 Second, the argument is passed to the function using the calling
4367 conventions of the first member of the transparent union, not the calling
4368 conventions of the union itself. All members of the union must have the
4369 same machine representation; this is necessary for this argument passing
4372 Transparent unions are designed for library functions that have multiple
4373 interfaces for compatibility reasons. For example, suppose the
4374 @code{wait} function must accept either a value of type @code{int *} to
4375 comply with Posix, or a value of type @code{union wait *} to comply with
4376 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4377 @code{wait} would accept both kinds of arguments, but it would also
4378 accept any other pointer type and this would make argument type checking
4379 less useful. Instead, @code{<sys/wait.h>} might define the interface
4383 typedef union __attribute__ ((__transparent_union__))
4387 @} wait_status_ptr_t;
4389 pid_t wait (wait_status_ptr_t);
4392 This interface allows either @code{int *} or @code{union wait *}
4393 arguments to be passed, using the @code{int *} calling convention.
4394 The program can call @code{wait} with arguments of either type:
4397 int w1 () @{ int w; return wait (&w); @}
4398 int w2 () @{ union wait w; return wait (&w); @}
4401 With this interface, @code{wait}'s implementation might look like this:
4404 pid_t wait (wait_status_ptr_t p)
4406 return waitpid (-1, p.__ip, 0);
4411 When attached to a type (including a @code{union} or a @code{struct}),
4412 this attribute means that variables of that type are meant to appear
4413 possibly unused. GCC will not produce a warning for any variables of
4414 that type, even if the variable appears to do nothing. This is often
4415 the case with lock or thread classes, which are usually defined and then
4416 not referenced, but contain constructors and destructors that have
4417 nontrivial bookkeeping functions.
4420 The @code{deprecated} attribute results in a warning if the type
4421 is used anywhere in the source file. This is useful when identifying
4422 types that are expected to be removed in a future version of a program.
4423 If possible, the warning also includes the location of the declaration
4424 of the deprecated type, to enable users to easily find further
4425 information about why the type is deprecated, or what they should do
4426 instead. Note that the warnings only occur for uses and then only
4427 if the type is being applied to an identifier that itself is not being
4428 declared as deprecated.
4431 typedef int T1 __attribute__ ((deprecated));
4435 typedef T1 T3 __attribute__ ((deprecated));
4436 T3 z __attribute__ ((deprecated));
4439 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4440 warning is issued for line 4 because T2 is not explicitly
4441 deprecated. Line 5 has no warning because T3 is explicitly
4442 deprecated. Similarly for line 6.
4444 The @code{deprecated} attribute can also be used for functions and
4445 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4448 Accesses through pointers to types with this attribute are not subject
4449 to type-based alias analysis, but are instead assumed to be able to alias
4450 any other type of objects. In the context of 6.5/7 an lvalue expression
4451 dereferencing such a pointer is treated like having a character type.
4452 See @option{-fstrict-aliasing} for more information on aliasing issues.
4453 This extension exists to support some vector APIs, in which pointers to
4454 one vector type are permitted to alias pointers to a different vector type.
4456 Note that an object of a type with this attribute does not have any
4462 typedef short __attribute__((__may_alias__)) short_a;
4468 short_a *b = (short_a *) &a;
4472 if (a == 0x12345678)
4479 If you replaced @code{short_a} with @code{short} in the variable
4480 declaration, the above program would abort when compiled with
4481 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4482 above in recent GCC versions.
4485 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4486 applied to class, struct, union and enum types. Unlike other type
4487 attributes, the attribute must appear between the initial keyword and
4488 the name of the type; it cannot appear after the body of the type.
4490 Note that the type visibility is applied to vague linkage entities
4491 associated with the class (vtable, typeinfo node, etc.). In
4492 particular, if a class is thrown as an exception in one shared object
4493 and caught in another, the class must have default visibility.
4494 Otherwise the two shared objects will be unable to use the same
4495 typeinfo node and exception handling will break.
4499 @subsection ARM Type Attributes
4501 On those ARM targets that support @code{dllimport} (such as Symbian
4502 OS), you can use the @code{notshared} attribute to indicate that the
4503 virtual table and other similar data for a class should not be
4504 exported from a DLL@. For example:
4507 class __declspec(notshared) C @{
4509 __declspec(dllimport) C();
4513 __declspec(dllexport)
4517 In this code, @code{C::C} is exported from the current DLL, but the
4518 virtual table for @code{C} is not exported. (You can use
4519 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4520 most Symbian OS code uses @code{__declspec}.)
4522 @anchor{i386 Type Attributes}
4523 @subsection i386 Type Attributes
4525 Two attributes are currently defined for i386 configurations:
4526 @code{ms_struct} and @code{gcc_struct}.
4532 @cindex @code{ms_struct}
4533 @cindex @code{gcc_struct}
4535 If @code{packed} is used on a structure, or if bit-fields are used
4536 it may be that the Microsoft ABI packs them differently
4537 than GCC would normally pack them. Particularly when moving packed
4538 data between functions compiled with GCC and the native Microsoft compiler
4539 (either via function call or as data in a file), it may be necessary to access
4542 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4543 compilers to match the native Microsoft compiler.
4546 To specify multiple attributes, separate them by commas within the
4547 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4550 @anchor{PowerPC Type Attributes}
4551 @subsection PowerPC Type Attributes
4553 Three attributes currently are defined for PowerPC configurations:
4554 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4556 For full documentation of the @code{ms_struct} and @code{gcc_struct}
4557 attributes please see the documentation in @ref{i386 Type Attributes}.
4559 The @code{altivec} attribute allows one to declare AltiVec vector data
4560 types supported by the AltiVec Programming Interface Manual. The
4561 attribute requires an argument to specify one of three vector types:
4562 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4563 and @code{bool__} (always followed by unsigned).
4566 __attribute__((altivec(vector__)))
4567 __attribute__((altivec(pixel__))) unsigned short
4568 __attribute__((altivec(bool__))) unsigned
4571 These attributes mainly are intended to support the @code{__vector},
4572 @code{__pixel}, and @code{__bool} AltiVec keywords.
4574 @anchor{SPU Type Attributes}
4575 @subsection SPU Type Attributes
4577 The SPU supports the @code{spu_vector} attribute for types. This attribute
4578 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
4579 Language Extensions Specification. It is intended to support the
4580 @code{__vector} keyword.
4584 @section An Inline Function is As Fast As a Macro
4585 @cindex inline functions
4586 @cindex integrating function code
4588 @cindex macros, inline alternative
4590 By declaring a function inline, you can direct GCC to make
4591 calls to that function faster. One way GCC can achieve this is to
4592 integrate that function's code into the code for its callers. This
4593 makes execution faster by eliminating the function-call overhead; in
4594 addition, if any of the actual argument values are constant, their
4595 known values may permit simplifications at compile time so that not
4596 all of the inline function's code needs to be included. The effect on
4597 code size is less predictable; object code may be larger or smaller
4598 with function inlining, depending on the particular case. You can
4599 also direct GCC to try to integrate all ``simple enough'' functions
4600 into their callers with the option @option{-finline-functions}.
4602 GCC implements three different semantics of declaring a function
4603 inline. One is available with @option{-std=gnu89} or
4604 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
4605 on all inline declarations, another when @option{-std=c99} or
4606 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
4607 is used when compiling C++.
4609 To declare a function inline, use the @code{inline} keyword in its
4610 declaration, like this:
4620 If you are writing a header file to be included in ISO C89 programs, write
4621 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4623 The three types of inlining behave similarly in two important cases:
4624 when the @code{inline} keyword is used on a @code{static} function,
4625 like the example above, and when a function is first declared without
4626 using the @code{inline} keyword and then is defined with
4627 @code{inline}, like this:
4630 extern int inc (int *a);
4638 In both of these common cases, the program behaves the same as if you
4639 had not used the @code{inline} keyword, except for its speed.
4641 @cindex inline functions, omission of
4642 @opindex fkeep-inline-functions
4643 When a function is both inline and @code{static}, if all calls to the
4644 function are integrated into the caller, and the function's address is
4645 never used, then the function's own assembler code is never referenced.
4646 In this case, GCC does not actually output assembler code for the
4647 function, unless you specify the option @option{-fkeep-inline-functions}.
4648 Some calls cannot be integrated for various reasons (in particular,
4649 calls that precede the function's definition cannot be integrated, and
4650 neither can recursive calls within the definition). If there is a
4651 nonintegrated call, then the function is compiled to assembler code as
4652 usual. The function must also be compiled as usual if the program
4653 refers to its address, because that can't be inlined.
4656 Note that certain usages in a function definition can make it unsuitable
4657 for inline substitution. Among these usages are: use of varargs, use of
4658 alloca, use of variable sized data types (@pxref{Variable Length}),
4659 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4660 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4661 will warn when a function marked @code{inline} could not be substituted,
4662 and will give the reason for the failure.
4664 @cindex automatic @code{inline} for C++ member fns
4665 @cindex @code{inline} automatic for C++ member fns
4666 @cindex member fns, automatically @code{inline}
4667 @cindex C++ member fns, automatically @code{inline}
4668 @opindex fno-default-inline
4669 As required by ISO C++, GCC considers member functions defined within
4670 the body of a class to be marked inline even if they are
4671 not explicitly declared with the @code{inline} keyword. You can
4672 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4673 Options,,Options Controlling C++ Dialect}.
4675 GCC does not inline any functions when not optimizing unless you specify
4676 the @samp{always_inline} attribute for the function, like this:
4679 /* @r{Prototype.} */
4680 inline void foo (const char) __attribute__((always_inline));
4683 The remainder of this section is specific to GNU C89 inlining.
4685 @cindex non-static inline function
4686 When an inline function is not @code{static}, then the compiler must assume
4687 that there may be calls from other source files; since a global symbol can
4688 be defined only once in any program, the function must not be defined in
4689 the other source files, so the calls therein cannot be integrated.
4690 Therefore, a non-@code{static} inline function is always compiled on its
4691 own in the usual fashion.
4693 If you specify both @code{inline} and @code{extern} in the function
4694 definition, then the definition is used only for inlining. In no case
4695 is the function compiled on its own, not even if you refer to its
4696 address explicitly. Such an address becomes an external reference, as
4697 if you had only declared the function, and had not defined it.
4699 This combination of @code{inline} and @code{extern} has almost the
4700 effect of a macro. The way to use it is to put a function definition in
4701 a header file with these keywords, and put another copy of the
4702 definition (lacking @code{inline} and @code{extern}) in a library file.
4703 The definition in the header file will cause most calls to the function
4704 to be inlined. If any uses of the function remain, they will refer to
4705 the single copy in the library.
4708 @section Assembler Instructions with C Expression Operands
4709 @cindex extended @code{asm}
4710 @cindex @code{asm} expressions
4711 @cindex assembler instructions
4714 In an assembler instruction using @code{asm}, you can specify the
4715 operands of the instruction using C expressions. This means you need not
4716 guess which registers or memory locations will contain the data you want
4719 You must specify an assembler instruction template much like what
4720 appears in a machine description, plus an operand constraint string for
4723 For example, here is how to use the 68881's @code{fsinx} instruction:
4726 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4730 Here @code{angle} is the C expression for the input operand while
4731 @code{result} is that of the output operand. Each has @samp{"f"} as its
4732 operand constraint, saying that a floating point register is required.
4733 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4734 output operands' constraints must use @samp{=}. The constraints use the
4735 same language used in the machine description (@pxref{Constraints}).
4737 Each operand is described by an operand-constraint string followed by
4738 the C expression in parentheses. A colon separates the assembler
4739 template from the first output operand and another separates the last
4740 output operand from the first input, if any. Commas separate the
4741 operands within each group. The total number of operands is currently
4742 limited to 30; this limitation may be lifted in some future version of
4745 If there are no output operands but there are input operands, you must
4746 place two consecutive colons surrounding the place where the output
4749 As of GCC version 3.1, it is also possible to specify input and output
4750 operands using symbolic names which can be referenced within the
4751 assembler code. These names are specified inside square brackets
4752 preceding the constraint string, and can be referenced inside the
4753 assembler code using @code{%[@var{name}]} instead of a percentage sign
4754 followed by the operand number. Using named operands the above example
4758 asm ("fsinx %[angle],%[output]"
4759 : [output] "=f" (result)
4760 : [angle] "f" (angle));
4764 Note that the symbolic operand names have no relation whatsoever to
4765 other C identifiers. You may use any name you like, even those of
4766 existing C symbols, but you must ensure that no two operands within the same
4767 assembler construct use the same symbolic name.
4769 Output operand expressions must be lvalues; the compiler can check this.
4770 The input operands need not be lvalues. The compiler cannot check
4771 whether the operands have data types that are reasonable for the
4772 instruction being executed. It does not parse the assembler instruction
4773 template and does not know what it means or even whether it is valid
4774 assembler input. The extended @code{asm} feature is most often used for
4775 machine instructions the compiler itself does not know exist. If
4776 the output expression cannot be directly addressed (for example, it is a
4777 bit-field), your constraint must allow a register. In that case, GCC
4778 will use the register as the output of the @code{asm}, and then store
4779 that register into the output.
4781 The ordinary output operands must be write-only; GCC will assume that
4782 the values in these operands before the instruction are dead and need
4783 not be generated. Extended asm supports input-output or read-write
4784 operands. Use the constraint character @samp{+} to indicate such an
4785 operand and list it with the output operands. You should only use
4786 read-write operands when the constraints for the operand (or the
4787 operand in which only some of the bits are to be changed) allow a
4790 You may, as an alternative, logically split its function into two
4791 separate operands, one input operand and one write-only output
4792 operand. The connection between them is expressed by constraints
4793 which say they need to be in the same location when the instruction
4794 executes. You can use the same C expression for both operands, or
4795 different expressions. For example, here we write the (fictitious)
4796 @samp{combine} instruction with @code{bar} as its read-only source
4797 operand and @code{foo} as its read-write destination:
4800 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4804 The constraint @samp{"0"} for operand 1 says that it must occupy the
4805 same location as operand 0. A number in constraint is allowed only in
4806 an input operand and it must refer to an output operand.
4808 Only a number in the constraint can guarantee that one operand will be in
4809 the same place as another. The mere fact that @code{foo} is the value
4810 of both operands is not enough to guarantee that they will be in the
4811 same place in the generated assembler code. The following would not
4815 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4818 Various optimizations or reloading could cause operands 0 and 1 to be in
4819 different registers; GCC knows no reason not to do so. For example, the
4820 compiler might find a copy of the value of @code{foo} in one register and
4821 use it for operand 1, but generate the output operand 0 in a different
4822 register (copying it afterward to @code{foo}'s own address). Of course,
4823 since the register for operand 1 is not even mentioned in the assembler
4824 code, the result will not work, but GCC can't tell that.
4826 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4827 the operand number for a matching constraint. For example:
4830 asm ("cmoveq %1,%2,%[result]"
4831 : [result] "=r"(result)
4832 : "r" (test), "r"(new), "[result]"(old));
4835 Sometimes you need to make an @code{asm} operand be a specific register,
4836 but there's no matching constraint letter for that register @emph{by
4837 itself}. To force the operand into that register, use a local variable
4838 for the operand and specify the register in the variable declaration.
4839 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4840 register constraint letter that matches the register:
4843 register int *p1 asm ("r0") = @dots{};
4844 register int *p2 asm ("r1") = @dots{};
4845 register int *result asm ("r0");
4846 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4849 @anchor{Example of asm with clobbered asm reg}
4850 In the above example, beware that a register that is call-clobbered by
4851 the target ABI will be overwritten by any function call in the
4852 assignment, including library calls for arithmetic operators.
4853 Assuming it is a call-clobbered register, this may happen to @code{r0}
4854 above by the assignment to @code{p2}. If you have to use such a
4855 register, use temporary variables for expressions between the register
4860 register int *p1 asm ("r0") = @dots{};
4861 register int *p2 asm ("r1") = t1;
4862 register int *result asm ("r0");
4863 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4866 Some instructions clobber specific hard registers. To describe this,
4867 write a third colon after the input operands, followed by the names of
4868 the clobbered hard registers (given as strings). Here is a realistic
4869 example for the VAX:
4872 asm volatile ("movc3 %0,%1,%2"
4873 : /* @r{no outputs} */
4874 : "g" (from), "g" (to), "g" (count)
4875 : "r0", "r1", "r2", "r3", "r4", "r5");
4878 You may not write a clobber description in a way that overlaps with an
4879 input or output operand. For example, you may not have an operand
4880 describing a register class with one member if you mention that register
4881 in the clobber list. Variables declared to live in specific registers
4882 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4883 have no part mentioned in the clobber description.
4884 There is no way for you to specify that an input
4885 operand is modified without also specifying it as an output
4886 operand. Note that if all the output operands you specify are for this
4887 purpose (and hence unused), you will then also need to specify
4888 @code{volatile} for the @code{asm} construct, as described below, to
4889 prevent GCC from deleting the @code{asm} statement as unused.
4891 If you refer to a particular hardware register from the assembler code,
4892 you will probably have to list the register after the third colon to
4893 tell the compiler the register's value is modified. In some assemblers,
4894 the register names begin with @samp{%}; to produce one @samp{%} in the
4895 assembler code, you must write @samp{%%} in the input.
4897 If your assembler instruction can alter the condition code register, add
4898 @samp{cc} to the list of clobbered registers. GCC on some machines
4899 represents the condition codes as a specific hardware register;
4900 @samp{cc} serves to name this register. On other machines, the
4901 condition code is handled differently, and specifying @samp{cc} has no
4902 effect. But it is valid no matter what the machine.
4904 If your assembler instructions access memory in an unpredictable
4905 fashion, add @samp{memory} to the list of clobbered registers. This
4906 will cause GCC to not keep memory values cached in registers across the
4907 assembler instruction and not optimize stores or loads to that memory.
4908 You will also want to add the @code{volatile} keyword if the memory
4909 affected is not listed in the inputs or outputs of the @code{asm}, as
4910 the @samp{memory} clobber does not count as a side-effect of the
4911 @code{asm}. If you know how large the accessed memory is, you can add
4912 it as input or output but if this is not known, you should add
4913 @samp{memory}. As an example, if you access ten bytes of a string, you
4914 can use a memory input like:
4917 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4920 Note that in the following example the memory input is necessary,
4921 otherwise GCC might optimize the store to @code{x} away:
4928 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4929 "=&d" (r) : "a" (y), "m" (*y));
4934 You can put multiple assembler instructions together in a single
4935 @code{asm} template, separated by the characters normally used in assembly
4936 code for the system. A combination that works in most places is a newline
4937 to break the line, plus a tab character to move to the instruction field
4938 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4939 assembler allows semicolons as a line-breaking character. Note that some
4940 assembler dialects use semicolons to start a comment.
4941 The input operands are guaranteed not to use any of the clobbered
4942 registers, and neither will the output operands' addresses, so you can
4943 read and write the clobbered registers as many times as you like. Here
4944 is an example of multiple instructions in a template; it assumes the
4945 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4948 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4950 : "g" (from), "g" (to)
4954 Unless an output operand has the @samp{&} constraint modifier, GCC
4955 may allocate it in the same register as an unrelated input operand, on
4956 the assumption the inputs are consumed before the outputs are produced.
4957 This assumption may be false if the assembler code actually consists of
4958 more than one instruction. In such a case, use @samp{&} for each output
4959 operand that may not overlap an input. @xref{Modifiers}.
4961 If you want to test the condition code produced by an assembler
4962 instruction, you must include a branch and a label in the @code{asm}
4963 construct, as follows:
4966 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4972 This assumes your assembler supports local labels, as the GNU assembler
4973 and most Unix assemblers do.
4975 Speaking of labels, jumps from one @code{asm} to another are not
4976 supported. The compiler's optimizers do not know about these jumps, and
4977 therefore they cannot take account of them when deciding how to
4980 @cindex macros containing @code{asm}
4981 Usually the most convenient way to use these @code{asm} instructions is to
4982 encapsulate them in macros that look like functions. For example,
4986 (@{ double __value, __arg = (x); \
4987 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4992 Here the variable @code{__arg} is used to make sure that the instruction
4993 operates on a proper @code{double} value, and to accept only those
4994 arguments @code{x} which can convert automatically to a @code{double}.
4996 Another way to make sure the instruction operates on the correct data
4997 type is to use a cast in the @code{asm}. This is different from using a
4998 variable @code{__arg} in that it converts more different types. For
4999 example, if the desired type were @code{int}, casting the argument to
5000 @code{int} would accept a pointer with no complaint, while assigning the
5001 argument to an @code{int} variable named @code{__arg} would warn about
5002 using a pointer unless the caller explicitly casts it.
5004 If an @code{asm} has output operands, GCC assumes for optimization
5005 purposes the instruction has no side effects except to change the output
5006 operands. This does not mean instructions with a side effect cannot be
5007 used, but you must be careful, because the compiler may eliminate them
5008 if the output operands aren't used, or move them out of loops, or
5009 replace two with one if they constitute a common subexpression. Also,
5010 if your instruction does have a side effect on a variable that otherwise
5011 appears not to change, the old value of the variable may be reused later
5012 if it happens to be found in a register.
5014 You can prevent an @code{asm} instruction from being deleted
5015 by writing the keyword @code{volatile} after
5016 the @code{asm}. For example:
5019 #define get_and_set_priority(new) \
5021 asm volatile ("get_and_set_priority %0, %1" \
5022 : "=g" (__old) : "g" (new)); \
5027 The @code{volatile} keyword indicates that the instruction has
5028 important side-effects. GCC will not delete a volatile @code{asm} if
5029 it is reachable. (The instruction can still be deleted if GCC can
5030 prove that control-flow will never reach the location of the
5031 instruction.) Note that even a volatile @code{asm} instruction
5032 can be moved relative to other code, including across jump
5033 instructions. For example, on many targets there is a system
5034 register which can be set to control the rounding mode of
5035 floating point operations. You might try
5036 setting it with a volatile @code{asm}, like this PowerPC example:
5039 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
5044 This will not work reliably, as the compiler may move the addition back
5045 before the volatile @code{asm}. To make it work you need to add an
5046 artificial dependency to the @code{asm} referencing a variable in the code
5047 you don't want moved, for example:
5050 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
5054 Similarly, you can't expect a
5055 sequence of volatile @code{asm} instructions to remain perfectly
5056 consecutive. If you want consecutive output, use a single @code{asm}.
5057 Also, GCC will perform some optimizations across a volatile @code{asm}
5058 instruction; GCC does not ``forget everything'' when it encounters
5059 a volatile @code{asm} instruction the way some other compilers do.
5061 An @code{asm} instruction without any output operands will be treated
5062 identically to a volatile @code{asm} instruction.
5064 It is a natural idea to look for a way to give access to the condition
5065 code left by the assembler instruction. However, when we attempted to
5066 implement this, we found no way to make it work reliably. The problem
5067 is that output operands might need reloading, which would result in
5068 additional following ``store'' instructions. On most machines, these
5069 instructions would alter the condition code before there was time to
5070 test it. This problem doesn't arise for ordinary ``test'' and
5071 ``compare'' instructions because they don't have any output operands.
5073 For reasons similar to those described above, it is not possible to give
5074 an assembler instruction access to the condition code left by previous
5077 If you are writing a header file that should be includable in ISO C
5078 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
5081 @subsection Size of an @code{asm}
5083 Some targets require that GCC track the size of each instruction used in
5084 order to generate correct code. Because the final length of an
5085 @code{asm} is only known by the assembler, GCC must make an estimate as
5086 to how big it will be. The estimate is formed by counting the number of
5087 statements in the pattern of the @code{asm} and multiplying that by the
5088 length of the longest instruction on that processor. Statements in the
5089 @code{asm} are identified by newline characters and whatever statement
5090 separator characters are supported by the assembler; on most processors
5091 this is the `@code{;}' character.
5093 Normally, GCC's estimate is perfectly adequate to ensure that correct
5094 code is generated, but it is possible to confuse the compiler if you use
5095 pseudo instructions or assembler macros that expand into multiple real
5096 instructions or if you use assembler directives that expand to more
5097 space in the object file than would be needed for a single instruction.
5098 If this happens then the assembler will produce a diagnostic saying that
5099 a label is unreachable.
5101 @subsection i386 floating point asm operands
5103 There are several rules on the usage of stack-like regs in
5104 asm_operands insns. These rules apply only to the operands that are
5109 Given a set of input regs that die in an asm_operands, it is
5110 necessary to know which are implicitly popped by the asm, and
5111 which must be explicitly popped by gcc.
5113 An input reg that is implicitly popped by the asm must be
5114 explicitly clobbered, unless it is constrained to match an
5118 For any input reg that is implicitly popped by an asm, it is
5119 necessary to know how to adjust the stack to compensate for the pop.
5120 If any non-popped input is closer to the top of the reg-stack than
5121 the implicitly popped reg, it would not be possible to know what the
5122 stack looked like---it's not clear how the rest of the stack ``slides
5125 All implicitly popped input regs must be closer to the top of
5126 the reg-stack than any input that is not implicitly popped.
5128 It is possible that if an input dies in an insn, reload might
5129 use the input reg for an output reload. Consider this example:
5132 asm ("foo" : "=t" (a) : "f" (b));
5135 This asm says that input B is not popped by the asm, and that
5136 the asm pushes a result onto the reg-stack, i.e., the stack is one
5137 deeper after the asm than it was before. But, it is possible that
5138 reload will think that it can use the same reg for both the input and
5139 the output, if input B dies in this insn.
5141 If any input operand uses the @code{f} constraint, all output reg
5142 constraints must use the @code{&} earlyclobber.
5144 The asm above would be written as
5147 asm ("foo" : "=&t" (a) : "f" (b));
5151 Some operands need to be in particular places on the stack. All
5152 output operands fall in this category---there is no other way to
5153 know which regs the outputs appear in unless the user indicates
5154 this in the constraints.
5156 Output operands must specifically indicate which reg an output
5157 appears in after an asm. @code{=f} is not allowed: the operand
5158 constraints must select a class with a single reg.
5161 Output operands may not be ``inserted'' between existing stack regs.
5162 Since no 387 opcode uses a read/write operand, all output operands
5163 are dead before the asm_operands, and are pushed by the asm_operands.
5164 It makes no sense to push anywhere but the top of the reg-stack.
5166 Output operands must start at the top of the reg-stack: output
5167 operands may not ``skip'' a reg.
5170 Some asm statements may need extra stack space for internal
5171 calculations. This can be guaranteed by clobbering stack registers
5172 unrelated to the inputs and outputs.
5176 Here are a couple of reasonable asms to want to write. This asm
5177 takes one input, which is internally popped, and produces two outputs.
5180 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
5183 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
5184 and replaces them with one output. The user must code the @code{st(1)}
5185 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
5188 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
5194 @section Controlling Names Used in Assembler Code
5195 @cindex assembler names for identifiers
5196 @cindex names used in assembler code
5197 @cindex identifiers, names in assembler code
5199 You can specify the name to be used in the assembler code for a C
5200 function or variable by writing the @code{asm} (or @code{__asm__})
5201 keyword after the declarator as follows:
5204 int foo asm ("myfoo") = 2;
5208 This specifies that the name to be used for the variable @code{foo} in
5209 the assembler code should be @samp{myfoo} rather than the usual
5212 On systems where an underscore is normally prepended to the name of a C
5213 function or variable, this feature allows you to define names for the
5214 linker that do not start with an underscore.
5216 It does not make sense to use this feature with a non-static local
5217 variable since such variables do not have assembler names. If you are
5218 trying to put the variable in a particular register, see @ref{Explicit
5219 Reg Vars}. GCC presently accepts such code with a warning, but will
5220 probably be changed to issue an error, rather than a warning, in the
5223 You cannot use @code{asm} in this way in a function @emph{definition}; but
5224 you can get the same effect by writing a declaration for the function
5225 before its definition and putting @code{asm} there, like this:
5228 extern func () asm ("FUNC");
5235 It is up to you to make sure that the assembler names you choose do not
5236 conflict with any other assembler symbols. Also, you must not use a
5237 register name; that would produce completely invalid assembler code. GCC
5238 does not as yet have the ability to store static variables in registers.
5239 Perhaps that will be added.
5241 @node Explicit Reg Vars
5242 @section Variables in Specified Registers
5243 @cindex explicit register variables
5244 @cindex variables in specified registers
5245 @cindex specified registers
5246 @cindex registers, global allocation
5248 GNU C allows you to put a few global variables into specified hardware
5249 registers. You can also specify the register in which an ordinary
5250 register variable should be allocated.
5254 Global register variables reserve registers throughout the program.
5255 This may be useful in programs such as programming language
5256 interpreters which have a couple of global variables that are accessed
5260 Local register variables in specific registers do not reserve the
5261 registers, except at the point where they are used as input or output
5262 operands in an @code{asm} statement and the @code{asm} statement itself is
5263 not deleted. The compiler's data flow analysis is capable of determining
5264 where the specified registers contain live values, and where they are
5265 available for other uses. Stores into local register variables may be deleted
5266 when they appear to be dead according to dataflow analysis. References
5267 to local register variables may be deleted or moved or simplified.
5269 These local variables are sometimes convenient for use with the extended
5270 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
5271 output of the assembler instruction directly into a particular register.
5272 (This will work provided the register you specify fits the constraints
5273 specified for that operand in the @code{asm}.)
5281 @node Global Reg Vars
5282 @subsection Defining Global Register Variables
5283 @cindex global register variables
5284 @cindex registers, global variables in
5286 You can define a global register variable in GNU C like this:
5289 register int *foo asm ("a5");
5293 Here @code{a5} is the name of the register which should be used. Choose a
5294 register which is normally saved and restored by function calls on your
5295 machine, so that library routines will not clobber it.
5297 Naturally the register name is cpu-dependent, so you would need to
5298 conditionalize your program according to cpu type. The register
5299 @code{a5} would be a good choice on a 68000 for a variable of pointer
5300 type. On machines with register windows, be sure to choose a ``global''
5301 register that is not affected magically by the function call mechanism.
5303 In addition, operating systems on one type of cpu may differ in how they
5304 name the registers; then you would need additional conditionals. For
5305 example, some 68000 operating systems call this register @code{%a5}.
5307 Eventually there may be a way of asking the compiler to choose a register
5308 automatically, but first we need to figure out how it should choose and
5309 how to enable you to guide the choice. No solution is evident.
5311 Defining a global register variable in a certain register reserves that
5312 register entirely for this use, at least within the current compilation.
5313 The register will not be allocated for any other purpose in the functions
5314 in the current compilation. The register will not be saved and restored by
5315 these functions. Stores into this register are never deleted even if they
5316 would appear to be dead, but references may be deleted or moved or
5319 It is not safe to access the global register variables from signal
5320 handlers, or from more than one thread of control, because the system
5321 library routines may temporarily use the register for other things (unless
5322 you recompile them specially for the task at hand).
5324 @cindex @code{qsort}, and global register variables
5325 It is not safe for one function that uses a global register variable to
5326 call another such function @code{foo} by way of a third function
5327 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
5328 different source file in which the variable wasn't declared). This is
5329 because @code{lose} might save the register and put some other value there.
5330 For example, you can't expect a global register variable to be available in
5331 the comparison-function that you pass to @code{qsort}, since @code{qsort}
5332 might have put something else in that register. (If you are prepared to
5333 recompile @code{qsort} with the same global register variable, you can
5334 solve this problem.)
5336 If you want to recompile @code{qsort} or other source files which do not
5337 actually use your global register variable, so that they will not use that
5338 register for any other purpose, then it suffices to specify the compiler
5339 option @option{-ffixed-@var{reg}}. You need not actually add a global
5340 register declaration to their source code.
5342 A function which can alter the value of a global register variable cannot
5343 safely be called from a function compiled without this variable, because it
5344 could clobber the value the caller expects to find there on return.
5345 Therefore, the function which is the entry point into the part of the
5346 program that uses the global register variable must explicitly save and
5347 restore the value which belongs to its caller.
5349 @cindex register variable after @code{longjmp}
5350 @cindex global register after @code{longjmp}
5351 @cindex value after @code{longjmp}
5354 On most machines, @code{longjmp} will restore to each global register
5355 variable the value it had at the time of the @code{setjmp}. On some
5356 machines, however, @code{longjmp} will not change the value of global
5357 register variables. To be portable, the function that called @code{setjmp}
5358 should make other arrangements to save the values of the global register
5359 variables, and to restore them in a @code{longjmp}. This way, the same
5360 thing will happen regardless of what @code{longjmp} does.
5362 All global register variable declarations must precede all function
5363 definitions. If such a declaration could appear after function
5364 definitions, the declaration would be too late to prevent the register from
5365 being used for other purposes in the preceding functions.
5367 Global register variables may not have initial values, because an
5368 executable file has no means to supply initial contents for a register.
5370 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5371 registers, but certain library functions, such as @code{getwd}, as well
5372 as the subroutines for division and remainder, modify g3 and g4. g1 and
5373 g2 are local temporaries.
5375 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5376 Of course, it will not do to use more than a few of those.
5378 @node Local Reg Vars
5379 @subsection Specifying Registers for Local Variables
5380 @cindex local variables, specifying registers
5381 @cindex specifying registers for local variables
5382 @cindex registers for local variables
5384 You can define a local register variable with a specified register
5388 register int *foo asm ("a5");
5392 Here @code{a5} is the name of the register which should be used. Note
5393 that this is the same syntax used for defining global register
5394 variables, but for a local variable it would appear within a function.
5396 Naturally the register name is cpu-dependent, but this is not a
5397 problem, since specific registers are most often useful with explicit
5398 assembler instructions (@pxref{Extended Asm}). Both of these things
5399 generally require that you conditionalize your program according to
5402 In addition, operating systems on one type of cpu may differ in how they
5403 name the registers; then you would need additional conditionals. For
5404 example, some 68000 operating systems call this register @code{%a5}.
5406 Defining such a register variable does not reserve the register; it
5407 remains available for other uses in places where flow control determines
5408 the variable's value is not live.
5410 This option does not guarantee that GCC will generate code that has
5411 this variable in the register you specify at all times. You may not
5412 code an explicit reference to this register in the @emph{assembler
5413 instruction template} part of an @code{asm} statement and assume it will
5414 always refer to this variable. However, using the variable as an
5415 @code{asm} @emph{operand} guarantees that the specified register is used
5418 Stores into local register variables may be deleted when they appear to be dead
5419 according to dataflow analysis. References to local register variables may
5420 be deleted or moved or simplified.
5422 As for global register variables, it's recommended that you choose a
5423 register which is normally saved and restored by function calls on
5424 your machine, so that library routines will not clobber it. A common
5425 pitfall is to initialize multiple call-clobbered registers with
5426 arbitrary expressions, where a function call or library call for an
5427 arithmetic operator will overwrite a register value from a previous
5428 assignment, for example @code{r0} below:
5430 register int *p1 asm ("r0") = @dots{};
5431 register int *p2 asm ("r1") = @dots{};
5433 In those cases, a solution is to use a temporary variable for
5434 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
5436 @node Alternate Keywords
5437 @section Alternate Keywords
5438 @cindex alternate keywords
5439 @cindex keywords, alternate
5441 @option{-ansi} and the various @option{-std} options disable certain
5442 keywords. This causes trouble when you want to use GNU C extensions, or
5443 a general-purpose header file that should be usable by all programs,
5444 including ISO C programs. The keywords @code{asm}, @code{typeof} and
5445 @code{inline} are not available in programs compiled with
5446 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
5447 program compiled with @option{-std=c99}). The ISO C99 keyword
5448 @code{restrict} is only available when @option{-std=gnu99} (which will
5449 eventually be the default) or @option{-std=c99} (or the equivalent
5450 @option{-std=iso9899:1999}) is used.
5452 The way to solve these problems is to put @samp{__} at the beginning and
5453 end of each problematical keyword. For example, use @code{__asm__}
5454 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
5456 Other C compilers won't accept these alternative keywords; if you want to
5457 compile with another compiler, you can define the alternate keywords as
5458 macros to replace them with the customary keywords. It looks like this:
5466 @findex __extension__
5468 @option{-pedantic} and other options cause warnings for many GNU C extensions.
5470 prevent such warnings within one expression by writing
5471 @code{__extension__} before the expression. @code{__extension__} has no
5472 effect aside from this.
5474 @node Incomplete Enums
5475 @section Incomplete @code{enum} Types
5477 You can define an @code{enum} tag without specifying its possible values.
5478 This results in an incomplete type, much like what you get if you write
5479 @code{struct foo} without describing the elements. A later declaration
5480 which does specify the possible values completes the type.
5482 You can't allocate variables or storage using the type while it is
5483 incomplete. However, you can work with pointers to that type.
5485 This extension may not be very useful, but it makes the handling of
5486 @code{enum} more consistent with the way @code{struct} and @code{union}
5489 This extension is not supported by GNU C++.
5491 @node Function Names
5492 @section Function Names as Strings
5493 @cindex @code{__func__} identifier
5494 @cindex @code{__FUNCTION__} identifier
5495 @cindex @code{__PRETTY_FUNCTION__} identifier
5497 GCC provides three magic variables which hold the name of the current
5498 function, as a string. The first of these is @code{__func__}, which
5499 is part of the C99 standard:
5501 The identifier @code{__func__} is implicitly declared by the translator
5502 as if, immediately following the opening brace of each function
5503 definition, the declaration
5506 static const char __func__[] = "function-name";
5510 appeared, where function-name is the name of the lexically-enclosing
5511 function. This name is the unadorned name of the function.
5513 @code{__FUNCTION__} is another name for @code{__func__}. Older
5514 versions of GCC recognize only this name. However, it is not
5515 standardized. For maximum portability, we recommend you use
5516 @code{__func__}, but provide a fallback definition with the
5520 #if __STDC_VERSION__ < 199901L
5522 # define __func__ __FUNCTION__
5524 # define __func__ "<unknown>"
5529 In C, @code{__PRETTY_FUNCTION__} is yet another name for
5530 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
5531 the type signature of the function as well as its bare name. For
5532 example, this program:
5536 extern int printf (char *, ...);
5543 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
5544 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
5562 __PRETTY_FUNCTION__ = void a::sub(int)
5565 These identifiers are not preprocessor macros. In GCC 3.3 and
5566 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
5567 were treated as string literals; they could be used to initialize
5568 @code{char} arrays, and they could be concatenated with other string
5569 literals. GCC 3.4 and later treat them as variables, like
5570 @code{__func__}. In C++, @code{__FUNCTION__} and
5571 @code{__PRETTY_FUNCTION__} have always been variables.
5573 @node Return Address
5574 @section Getting the Return or Frame Address of a Function
5576 These functions may be used to get information about the callers of a
5579 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
5580 This function returns the return address of the current function, or of
5581 one of its callers. The @var{level} argument is number of frames to
5582 scan up the call stack. A value of @code{0} yields the return address
5583 of the current function, a value of @code{1} yields the return address
5584 of the caller of the current function, and so forth. When inlining
5585 the expected behavior is that the function will return the address of
5586 the function that will be returned to. To work around this behavior use
5587 the @code{noinline} function attribute.
5589 The @var{level} argument must be a constant integer.
5591 On some machines it may be impossible to determine the return address of
5592 any function other than the current one; in such cases, or when the top
5593 of the stack has been reached, this function will return @code{0} or a
5594 random value. In addition, @code{__builtin_frame_address} may be used
5595 to determine if the top of the stack has been reached.
5597 This function should only be used with a nonzero argument for debugging
5601 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
5602 This function is similar to @code{__builtin_return_address}, but it
5603 returns the address of the function frame rather than the return address
5604 of the function. Calling @code{__builtin_frame_address} with a value of
5605 @code{0} yields the frame address of the current function, a value of
5606 @code{1} yields the frame address of the caller of the current function,
5609 The frame is the area on the stack which holds local variables and saved
5610 registers. The frame address is normally the address of the first word
5611 pushed on to the stack by the function. However, the exact definition
5612 depends upon the processor and the calling convention. If the processor
5613 has a dedicated frame pointer register, and the function has a frame,
5614 then @code{__builtin_frame_address} will return the value of the frame
5617 On some machines it may be impossible to determine the frame address of
5618 any function other than the current one; in such cases, or when the top
5619 of the stack has been reached, this function will return @code{0} if
5620 the first frame pointer is properly initialized by the startup code.
5622 This function should only be used with a nonzero argument for debugging
5626 @node Vector Extensions
5627 @section Using vector instructions through built-in functions
5629 On some targets, the instruction set contains SIMD vector instructions that
5630 operate on multiple values contained in one large register at the same time.
5631 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5634 The first step in using these extensions is to provide the necessary data
5635 types. This should be done using an appropriate @code{typedef}:
5638 typedef int v4si __attribute__ ((vector_size (16)));
5641 The @code{int} type specifies the base type, while the attribute specifies
5642 the vector size for the variable, measured in bytes. For example, the
5643 declaration above causes the compiler to set the mode for the @code{v4si}
5644 type to be 16 bytes wide and divided into @code{int} sized units. For
5645 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5646 corresponding mode of @code{foo} will be @acronym{V4SI}.
5648 The @code{vector_size} attribute is only applicable to integral and
5649 float scalars, although arrays, pointers, and function return values
5650 are allowed in conjunction with this construct.
5652 All the basic integer types can be used as base types, both as signed
5653 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5654 @code{long long}. In addition, @code{float} and @code{double} can be
5655 used to build floating-point vector types.
5657 Specifying a combination that is not valid for the current architecture
5658 will cause GCC to synthesize the instructions using a narrower mode.
5659 For example, if you specify a variable of type @code{V4SI} and your
5660 architecture does not allow for this specific SIMD type, GCC will
5661 produce code that uses 4 @code{SIs}.
5663 The types defined in this manner can be used with a subset of normal C
5664 operations. Currently, GCC will allow using the following operators
5665 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5667 The operations behave like C++ @code{valarrays}. Addition is defined as
5668 the addition of the corresponding elements of the operands. For
5669 example, in the code below, each of the 4 elements in @var{a} will be
5670 added to the corresponding 4 elements in @var{b} and the resulting
5671 vector will be stored in @var{c}.
5674 typedef int v4si __attribute__ ((vector_size (16)));
5681 Subtraction, multiplication, division, and the logical operations
5682 operate in a similar manner. Likewise, the result of using the unary
5683 minus or complement operators on a vector type is a vector whose
5684 elements are the negative or complemented values of the corresponding
5685 elements in the operand.
5687 You can declare variables and use them in function calls and returns, as
5688 well as in assignments and some casts. You can specify a vector type as
5689 a return type for a function. Vector types can also be used as function
5690 arguments. It is possible to cast from one vector type to another,
5691 provided they are of the same size (in fact, you can also cast vectors
5692 to and from other datatypes of the same size).
5694 You cannot operate between vectors of different lengths or different
5695 signedness without a cast.
5697 A port that supports hardware vector operations, usually provides a set
5698 of built-in functions that can be used to operate on vectors. For
5699 example, a function to add two vectors and multiply the result by a
5700 third could look like this:
5703 v4si f (v4si a, v4si b, v4si c)
5705 v4si tmp = __builtin_addv4si (a, b);
5706 return __builtin_mulv4si (tmp, c);
5713 @findex __builtin_offsetof
5715 GCC implements for both C and C++ a syntactic extension to implement
5716 the @code{offsetof} macro.
5720 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5722 offsetof_member_designator:
5724 | offsetof_member_designator "." @code{identifier}
5725 | offsetof_member_designator "[" @code{expr} "]"
5728 This extension is sufficient such that
5731 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5734 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5735 may be dependent. In either case, @var{member} may consist of a single
5736 identifier, or a sequence of member accesses and array references.
5738 @node Atomic Builtins
5739 @section Built-in functions for atomic memory access
5741 The following builtins are intended to be compatible with those described
5742 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5743 section 7.4. As such, they depart from the normal GCC practice of using
5744 the ``__builtin_'' prefix, and further that they are overloaded such that
5745 they work on multiple types.
5747 The definition given in the Intel documentation allows only for the use of
5748 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5749 counterparts. GCC will allow any integral scalar or pointer type that is
5750 1, 2, 4 or 8 bytes in length.
5752 Not all operations are supported by all target processors. If a particular
5753 operation cannot be implemented on the target processor, a warning will be
5754 generated and a call an external function will be generated. The external
5755 function will carry the same name as the builtin, with an additional suffix
5756 @samp{_@var{n}} where @var{n} is the size of the data type.
5758 @c ??? Should we have a mechanism to suppress this warning? This is almost
5759 @c useful for implementing the operation under the control of an external
5762 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5763 no memory operand will be moved across the operation, either forward or
5764 backward. Further, instructions will be issued as necessary to prevent the
5765 processor from speculating loads across the operation and from queuing stores
5766 after the operation.
5768 All of the routines are are described in the Intel documentation to take
5769 ``an optional list of variables protected by the memory barrier''. It's
5770 not clear what is meant by that; it could mean that @emph{only} the
5771 following variables are protected, or it could mean that these variables
5772 should in addition be protected. At present GCC ignores this list and
5773 protects all variables which are globally accessible. If in the future
5774 we make some use of this list, an empty list will continue to mean all
5775 globally accessible variables.
5778 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5779 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5780 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5781 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5782 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5783 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5784 @findex __sync_fetch_and_add
5785 @findex __sync_fetch_and_sub
5786 @findex __sync_fetch_and_or
5787 @findex __sync_fetch_and_and
5788 @findex __sync_fetch_and_xor
5789 @findex __sync_fetch_and_nand
5790 These builtins perform the operation suggested by the name, and
5791 returns the value that had previously been in memory. That is,
5794 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5795 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5798 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5799 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5800 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5801 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5802 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5803 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5804 @findex __sync_add_and_fetch
5805 @findex __sync_sub_and_fetch
5806 @findex __sync_or_and_fetch
5807 @findex __sync_and_and_fetch
5808 @findex __sync_xor_and_fetch
5809 @findex __sync_nand_and_fetch
5810 These builtins perform the operation suggested by the name, and
5811 return the new value. That is,
5814 @{ *ptr @var{op}= value; return *ptr; @}
5815 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5818 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5819 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5820 @findex __sync_bool_compare_and_swap
5821 @findex __sync_val_compare_and_swap
5822 These builtins perform an atomic compare and swap. That is, if the current
5823 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5826 The ``bool'' version returns true if the comparison is successful and
5827 @var{newval} was written. The ``val'' version returns the contents
5828 of @code{*@var{ptr}} before the operation.
5830 @item __sync_synchronize (...)
5831 @findex __sync_synchronize
5832 This builtin issues a full memory barrier.
5834 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5835 @findex __sync_lock_test_and_set
5836 This builtin, as described by Intel, is not a traditional test-and-set
5837 operation, but rather an atomic exchange operation. It writes @var{value}
5838 into @code{*@var{ptr}}, and returns the previous contents of
5841 Many targets have only minimal support for such locks, and do not support
5842 a full exchange operation. In this case, a target may support reduced
5843 functionality here by which the @emph{only} valid value to store is the
5844 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5845 is implementation defined.
5847 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5848 This means that references after the builtin cannot move to (or be
5849 speculated to) before the builtin, but previous memory stores may not
5850 be globally visible yet, and previous memory loads may not yet be
5853 @item void __sync_lock_release (@var{type} *ptr, ...)
5854 @findex __sync_lock_release
5855 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5856 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5858 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5859 This means that all previous memory stores are globally visible, and all
5860 previous memory loads have been satisfied, but following memory reads
5861 are not prevented from being speculated to before the barrier.
5864 @node Object Size Checking
5865 @section Object Size Checking Builtins
5866 @findex __builtin_object_size
5867 @findex __builtin___memcpy_chk
5868 @findex __builtin___mempcpy_chk
5869 @findex __builtin___memmove_chk
5870 @findex __builtin___memset_chk
5871 @findex __builtin___strcpy_chk
5872 @findex __builtin___stpcpy_chk
5873 @findex __builtin___strncpy_chk
5874 @findex __builtin___strcat_chk
5875 @findex __builtin___strncat_chk
5876 @findex __builtin___sprintf_chk
5877 @findex __builtin___snprintf_chk
5878 @findex __builtin___vsprintf_chk
5879 @findex __builtin___vsnprintf_chk
5880 @findex __builtin___printf_chk
5881 @findex __builtin___vprintf_chk
5882 @findex __builtin___fprintf_chk
5883 @findex __builtin___vfprintf_chk
5885 GCC implements a limited buffer overflow protection mechanism
5886 that can prevent some buffer overflow attacks.
5888 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5889 is a built-in construct that returns a constant number of bytes from
5890 @var{ptr} to the end of the object @var{ptr} pointer points to
5891 (if known at compile time). @code{__builtin_object_size} never evaluates
5892 its arguments for side-effects. If there are any side-effects in them, it
5893 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5894 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5895 point to and all of them are known at compile time, the returned number
5896 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5897 0 and minimum if nonzero. If it is not possible to determine which objects
5898 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5899 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5900 for @var{type} 2 or 3.
5902 @var{type} is an integer constant from 0 to 3. If the least significant
5903 bit is clear, objects are whole variables, if it is set, a closest
5904 surrounding subobject is considered the object a pointer points to.
5905 The second bit determines if maximum or minimum of remaining bytes
5909 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5910 char *p = &var.buf1[1], *q = &var.b;
5912 /* Here the object p points to is var. */
5913 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5914 /* The subobject p points to is var.buf1. */
5915 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5916 /* The object q points to is var. */
5917 assert (__builtin_object_size (q, 0)
5918 == (char *) (&var + 1) - (char *) &var.b);
5919 /* The subobject q points to is var.b. */
5920 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5924 There are built-in functions added for many common string operation
5925 functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
5926 built-in is provided. This built-in has an additional last argument,
5927 which is the number of bytes remaining in object the @var{dest}
5928 argument points to or @code{(size_t) -1} if the size is not known.
5930 The built-in functions are optimized into the normal string functions
5931 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5932 it is known at compile time that the destination object will not
5933 be overflown. If the compiler can determine at compile time the
5934 object will be always overflown, it issues a warning.
5936 The intended use can be e.g.
5940 #define bos0(dest) __builtin_object_size (dest, 0)
5941 #define memcpy(dest, src, n) \
5942 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5946 /* It is unknown what object p points to, so this is optimized
5947 into plain memcpy - no checking is possible. */
5948 memcpy (p, "abcde", n);
5949 /* Destination is known and length too. It is known at compile
5950 time there will be no overflow. */
5951 memcpy (&buf[5], "abcde", 5);
5952 /* Destination is known, but the length is not known at compile time.
5953 This will result in __memcpy_chk call that can check for overflow
5955 memcpy (&buf[5], "abcde", n);
5956 /* Destination is known and it is known at compile time there will
5957 be overflow. There will be a warning and __memcpy_chk call that
5958 will abort the program at runtime. */
5959 memcpy (&buf[6], "abcde", 5);
5962 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5963 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5964 @code{strcat} and @code{strncat}.
5966 There are also checking built-in functions for formatted output functions.
5968 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5969 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5970 const char *fmt, ...);
5971 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5973 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5974 const char *fmt, va_list ap);
5977 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5978 etc.@: functions and can contain implementation specific flags on what
5979 additional security measures the checking function might take, such as
5980 handling @code{%n} differently.
5982 The @var{os} argument is the object size @var{s} points to, like in the
5983 other built-in functions. There is a small difference in the behavior
5984 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5985 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5986 the checking function is called with @var{os} argument set to
5989 In addition to this, there are checking built-in functions
5990 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5991 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5992 These have just one additional argument, @var{flag}, right before
5993 format string @var{fmt}. If the compiler is able to optimize them to
5994 @code{fputc} etc.@: functions, it will, otherwise the checking function
5995 should be called and the @var{flag} argument passed to it.
5997 @node Other Builtins
5998 @section Other built-in functions provided by GCC
5999 @cindex built-in functions
6000 @findex __builtin_fpclassify
6001 @findex __builtin_isfinite
6002 @findex __builtin_isnormal
6003 @findex __builtin_isgreater
6004 @findex __builtin_isgreaterequal
6005 @findex __builtin_isinf_sign
6006 @findex __builtin_isless
6007 @findex __builtin_islessequal
6008 @findex __builtin_islessgreater
6009 @findex __builtin_isunordered
6010 @findex __builtin_powi
6011 @findex __builtin_powif
6012 @findex __builtin_powil
6170 @findex fprintf_unlocked
6172 @findex fputs_unlocked
6289 @findex printf_unlocked
6321 @findex significandf
6322 @findex significandl
6393 GCC provides a large number of built-in functions other than the ones
6394 mentioned above. Some of these are for internal use in the processing
6395 of exceptions or variable-length argument lists and will not be
6396 documented here because they may change from time to time; we do not
6397 recommend general use of these functions.
6399 The remaining functions are provided for optimization purposes.
6401 @opindex fno-builtin
6402 GCC includes built-in versions of many of the functions in the standard
6403 C library. The versions prefixed with @code{__builtin_} will always be
6404 treated as having the same meaning as the C library function even if you
6405 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
6406 Many of these functions are only optimized in certain cases; if they are
6407 not optimized in a particular case, a call to the library function will
6412 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
6413 @option{-std=c99}), the functions
6414 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
6415 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
6416 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
6417 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
6418 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
6419 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
6420 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
6421 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
6422 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
6423 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
6424 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
6425 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
6426 @code{signbitd64}, @code{signbitd128}, @code{significandf},
6427 @code{significandl}, @code{significand}, @code{sincosf},
6428 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
6429 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
6430 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
6431 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
6433 may be handled as built-in functions.
6434 All these functions have corresponding versions
6435 prefixed with @code{__builtin_}, which may be used even in strict C89
6438 The ISO C99 functions
6439 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
6440 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
6441 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
6442 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
6443 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
6444 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
6445 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
6446 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
6447 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
6448 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
6449 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
6450 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
6451 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
6452 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
6453 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
6454 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
6455 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
6456 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
6457 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
6458 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
6459 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
6460 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
6461 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
6462 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
6463 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
6464 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
6465 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
6466 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
6467 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
6468 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
6469 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
6470 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
6471 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
6472 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
6473 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
6474 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
6475 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
6476 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
6477 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
6478 are handled as built-in functions
6479 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6481 There are also built-in versions of the ISO C99 functions
6482 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
6483 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
6484 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
6485 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
6486 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
6487 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
6488 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
6489 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
6490 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
6491 that are recognized in any mode since ISO C90 reserves these names for
6492 the purpose to which ISO C99 puts them. All these functions have
6493 corresponding versions prefixed with @code{__builtin_}.
6495 The ISO C94 functions
6496 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
6497 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
6498 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
6500 are handled as built-in functions
6501 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6503 The ISO C90 functions
6504 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
6505 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
6506 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
6507 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
6508 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
6509 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
6510 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
6511 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
6512 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
6513 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
6514 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
6515 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
6516 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
6517 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
6518 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
6519 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
6520 are all recognized as built-in functions unless
6521 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
6522 is specified for an individual function). All of these functions have
6523 corresponding versions prefixed with @code{__builtin_}.
6525 GCC provides built-in versions of the ISO C99 floating point comparison
6526 macros that avoid raising exceptions for unordered operands. They have
6527 the same names as the standard macros ( @code{isgreater},
6528 @code{isgreaterequal}, @code{isless}, @code{islessequal},
6529 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
6530 prefixed. We intend for a library implementor to be able to simply
6531 @code{#define} each standard macro to its built-in equivalent.
6532 In the same fashion, GCC provides @code{fpclassify}, @code{isfinite},
6533 @code{isinf_sign} and @code{isnormal} built-ins used with
6534 @code{__builtin_} prefixed. The @code{isinf} and @code{isnan}
6535 builtins appear both with and without the @code{__builtin_} prefix.
6537 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
6539 You can use the built-in function @code{__builtin_types_compatible_p} to
6540 determine whether two types are the same.
6542 This built-in function returns 1 if the unqualified versions of the
6543 types @var{type1} and @var{type2} (which are types, not expressions) are
6544 compatible, 0 otherwise. The result of this built-in function can be
6545 used in integer constant expressions.
6547 This built-in function ignores top level qualifiers (e.g., @code{const},
6548 @code{volatile}). For example, @code{int} is equivalent to @code{const
6551 The type @code{int[]} and @code{int[5]} are compatible. On the other
6552 hand, @code{int} and @code{char *} are not compatible, even if the size
6553 of their types, on the particular architecture are the same. Also, the
6554 amount of pointer indirection is taken into account when determining
6555 similarity. Consequently, @code{short *} is not similar to
6556 @code{short **}. Furthermore, two types that are typedefed are
6557 considered compatible if their underlying types are compatible.
6559 An @code{enum} type is not considered to be compatible with another
6560 @code{enum} type even if both are compatible with the same integer
6561 type; this is what the C standard specifies.
6562 For example, @code{enum @{foo, bar@}} is not similar to
6563 @code{enum @{hot, dog@}}.
6565 You would typically use this function in code whose execution varies
6566 depending on the arguments' types. For example:
6571 typeof (x) tmp = (x); \
6572 if (__builtin_types_compatible_p (typeof (x), long double)) \
6573 tmp = foo_long_double (tmp); \
6574 else if (__builtin_types_compatible_p (typeof (x), double)) \
6575 tmp = foo_double (tmp); \
6576 else if (__builtin_types_compatible_p (typeof (x), float)) \
6577 tmp = foo_float (tmp); \
6584 @emph{Note:} This construct is only available for C@.
6588 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
6590 You can use the built-in function @code{__builtin_choose_expr} to
6591 evaluate code depending on the value of a constant expression. This
6592 built-in function returns @var{exp1} if @var{const_exp}, which is a
6593 constant expression that must be able to be determined at compile time,
6594 is nonzero. Otherwise it returns 0.
6596 This built-in function is analogous to the @samp{? :} operator in C,
6597 except that the expression returned has its type unaltered by promotion
6598 rules. Also, the built-in function does not evaluate the expression
6599 that was not chosen. For example, if @var{const_exp} evaluates to true,
6600 @var{exp2} is not evaluated even if it has side-effects.
6602 This built-in function can return an lvalue if the chosen argument is an
6605 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
6606 type. Similarly, if @var{exp2} is returned, its return type is the same
6613 __builtin_choose_expr ( \
6614 __builtin_types_compatible_p (typeof (x), double), \
6616 __builtin_choose_expr ( \
6617 __builtin_types_compatible_p (typeof (x), float), \
6619 /* @r{The void expression results in a compile-time error} \
6620 @r{when assigning the result to something.} */ \
6624 @emph{Note:} This construct is only available for C@. Furthermore, the
6625 unused expression (@var{exp1} or @var{exp2} depending on the value of
6626 @var{const_exp}) may still generate syntax errors. This may change in
6631 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6632 You can use the built-in function @code{__builtin_constant_p} to
6633 determine if a value is known to be constant at compile-time and hence
6634 that GCC can perform constant-folding on expressions involving that
6635 value. The argument of the function is the value to test. The function
6636 returns the integer 1 if the argument is known to be a compile-time
6637 constant and 0 if it is not known to be a compile-time constant. A
6638 return of 0 does not indicate that the value is @emph{not} a constant,
6639 but merely that GCC cannot prove it is a constant with the specified
6640 value of the @option{-O} option.
6642 You would typically use this function in an embedded application where
6643 memory was a critical resource. If you have some complex calculation,
6644 you may want it to be folded if it involves constants, but need to call
6645 a function if it does not. For example:
6648 #define Scale_Value(X) \
6649 (__builtin_constant_p (X) \
6650 ? ((X) * SCALE + OFFSET) : Scale (X))
6653 You may use this built-in function in either a macro or an inline
6654 function. However, if you use it in an inlined function and pass an
6655 argument of the function as the argument to the built-in, GCC will
6656 never return 1 when you call the inline function with a string constant
6657 or compound literal (@pxref{Compound Literals}) and will not return 1
6658 when you pass a constant numeric value to the inline function unless you
6659 specify the @option{-O} option.
6661 You may also use @code{__builtin_constant_p} in initializers for static
6662 data. For instance, you can write
6665 static const int table[] = @{
6666 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6672 This is an acceptable initializer even if @var{EXPRESSION} is not a
6673 constant expression. GCC must be more conservative about evaluating the
6674 built-in in this case, because it has no opportunity to perform
6677 Previous versions of GCC did not accept this built-in in data
6678 initializers. The earliest version where it is completely safe is
6682 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6683 @opindex fprofile-arcs
6684 You may use @code{__builtin_expect} to provide the compiler with
6685 branch prediction information. In general, you should prefer to
6686 use actual profile feedback for this (@option{-fprofile-arcs}), as
6687 programmers are notoriously bad at predicting how their programs
6688 actually perform. However, there are applications in which this
6689 data is hard to collect.
6691 The return value is the value of @var{exp}, which should be an integral
6692 expression. The semantics of the built-in are that it is expected that
6693 @var{exp} == @var{c}. For example:
6696 if (__builtin_expect (x, 0))
6701 would indicate that we do not expect to call @code{foo}, since
6702 we expect @code{x} to be zero. Since you are limited to integral
6703 expressions for @var{exp}, you should use constructions such as
6706 if (__builtin_expect (ptr != NULL, 1))
6711 when testing pointer or floating-point values.
6714 @deftypefn {Built-in Function} void __builtin_trap (void)
6715 This function causes the program to exit abnormally. GCC implements
6716 this function by using a target-dependent mechanism (such as
6717 intentionally executing an illegal instruction) or by calling
6718 @code{abort}. The mechanism used may vary from release to release so
6719 you should not rely on any particular implementation.
6722 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
6723 This function is used to flush the processor's instruction cache for
6724 the region of memory between @var{begin} inclusive and @var{end}
6725 exclusive. Some targets require that the instruction cache be
6726 flushed, after modifying memory containing code, in order to obtain
6727 deterministic behavior.
6729 If the target does not require instruction cache flushes,
6730 @code{__builtin___clear_cache} has no effect. Otherwise either
6731 instructions are emitted in-line to clear the instruction cache or a
6732 call to the @code{__clear_cache} function in libgcc is made.
6735 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6736 This function is used to minimize cache-miss latency by moving data into
6737 a cache before it is accessed.
6738 You can insert calls to @code{__builtin_prefetch} into code for which
6739 you know addresses of data in memory that is likely to be accessed soon.
6740 If the target supports them, data prefetch instructions will be generated.
6741 If the prefetch is done early enough before the access then the data will
6742 be in the cache by the time it is accessed.
6744 The value of @var{addr} is the address of the memory to prefetch.
6745 There are two optional arguments, @var{rw} and @var{locality}.
6746 The value of @var{rw} is a compile-time constant one or zero; one
6747 means that the prefetch is preparing for a write to the memory address
6748 and zero, the default, means that the prefetch is preparing for a read.
6749 The value @var{locality} must be a compile-time constant integer between
6750 zero and three. A value of zero means that the data has no temporal
6751 locality, so it need not be left in the cache after the access. A value
6752 of three means that the data has a high degree of temporal locality and
6753 should be left in all levels of cache possible. Values of one and two
6754 mean, respectively, a low or moderate degree of temporal locality. The
6758 for (i = 0; i < n; i++)
6761 __builtin_prefetch (&a[i+j], 1, 1);
6762 __builtin_prefetch (&b[i+j], 0, 1);
6767 Data prefetch does not generate faults if @var{addr} is invalid, but
6768 the address expression itself must be valid. For example, a prefetch
6769 of @code{p->next} will not fault if @code{p->next} is not a valid
6770 address, but evaluation will fault if @code{p} is not a valid address.
6772 If the target does not support data prefetch, the address expression
6773 is evaluated if it includes side effects but no other code is generated
6774 and GCC does not issue a warning.
6777 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6778 Returns a positive infinity, if supported by the floating-point format,
6779 else @code{DBL_MAX}. This function is suitable for implementing the
6780 ISO C macro @code{HUGE_VAL}.
6783 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6784 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6787 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6788 Similar to @code{__builtin_huge_val}, except the return
6789 type is @code{long double}.
6792 @deftypefn {Built-in Function} int __builtin_fpclassify (int, int, int, int, int, ...)
6793 This built-in implements the C99 fpclassify functionality. The first
6794 five int arguments should be the target library's notion of the
6795 possible FP classes and are used for return values. They must be
6796 constant values and they must appear in this order: @code{FP_NAN},
6797 @code{FP_INFINITE}, @code{FP_NORMAL}, @code{FP_SUBNORMAL} and
6798 @code{FP_ZERO}. The ellipsis is for exactly one floating point value
6799 to classify. GCC treats the last argument as type-generic, which
6800 means it does not do default promotion from float to double.
6803 @deftypefn {Built-in Function} double __builtin_inf (void)
6804 Similar to @code{__builtin_huge_val}, except a warning is generated
6805 if the target floating-point format does not support infinities.
6808 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6809 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6812 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6813 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6816 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6817 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6820 @deftypefn {Built-in Function} float __builtin_inff (void)
6821 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6822 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6825 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6826 Similar to @code{__builtin_inf}, except the return
6827 type is @code{long double}.
6830 @deftypefn {Built-in Function} int __builtin_isinf_sign (...)
6831 Similar to @code{isinf}, except the return value will be negative for
6832 an argument of @code{-Inf}. Note while the parameter list is an
6833 ellipsis, this function only accepts exactly one floating point
6834 argument. GCC treats this parameter as type-generic, which means it
6835 does not do default promotion from float to double.
6838 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6839 This is an implementation of the ISO C99 function @code{nan}.
6841 Since ISO C99 defines this function in terms of @code{strtod}, which we
6842 do not implement, a description of the parsing is in order. The string
6843 is parsed as by @code{strtol}; that is, the base is recognized by
6844 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6845 in the significand such that the least significant bit of the number
6846 is at the least significant bit of the significand. The number is
6847 truncated to fit the significand field provided. The significand is
6848 forced to be a quiet NaN@.
6850 This function, if given a string literal all of which would have been
6851 consumed by strtol, is evaluated early enough that it is considered a
6852 compile-time constant.
6855 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6856 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6859 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6860 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6863 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6864 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6867 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6868 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6871 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6872 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6875 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6876 Similar to @code{__builtin_nan}, except the significand is forced
6877 to be a signaling NaN@. The @code{nans} function is proposed by
6878 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6881 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6882 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6885 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6886 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6889 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6890 Returns one plus the index of the least significant 1-bit of @var{x}, or
6891 if @var{x} is zero, returns zero.
6894 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6895 Returns the number of leading 0-bits in @var{x}, starting at the most
6896 significant bit position. If @var{x} is 0, the result is undefined.
6899 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6900 Returns the number of trailing 0-bits in @var{x}, starting at the least
6901 significant bit position. If @var{x} is 0, the result is undefined.
6904 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6905 Returns the number of 1-bits in @var{x}.
6908 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6909 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6913 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6914 Similar to @code{__builtin_ffs}, except the argument type is
6915 @code{unsigned long}.
6918 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6919 Similar to @code{__builtin_clz}, except the argument type is
6920 @code{unsigned long}.
6923 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6924 Similar to @code{__builtin_ctz}, except the argument type is
6925 @code{unsigned long}.
6928 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6929 Similar to @code{__builtin_popcount}, except the argument type is
6930 @code{unsigned long}.
6933 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6934 Similar to @code{__builtin_parity}, except the argument type is
6935 @code{unsigned long}.
6938 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6939 Similar to @code{__builtin_ffs}, except the argument type is
6940 @code{unsigned long long}.
6943 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6944 Similar to @code{__builtin_clz}, except the argument type is
6945 @code{unsigned long long}.
6948 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6949 Similar to @code{__builtin_ctz}, except the argument type is
6950 @code{unsigned long long}.
6953 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6954 Similar to @code{__builtin_popcount}, except the argument type is
6955 @code{unsigned long long}.
6958 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6959 Similar to @code{__builtin_parity}, except the argument type is
6960 @code{unsigned long long}.
6963 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6964 Returns the first argument raised to the power of the second. Unlike the
6965 @code{pow} function no guarantees about precision and rounding are made.
6968 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6969 Similar to @code{__builtin_powi}, except the argument and return types
6973 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6974 Similar to @code{__builtin_powi}, except the argument and return types
6975 are @code{long double}.
6978 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6979 Returns @var{x} with the order of the bytes reversed; for example,
6980 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6984 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6985 Similar to @code{__builtin_bswap32}, except the argument and return types
6989 @node Target Builtins
6990 @section Built-in Functions Specific to Particular Target Machines
6992 On some target machines, GCC supports many built-in functions specific
6993 to those machines. Generally these generate calls to specific machine
6994 instructions, but allow the compiler to schedule those calls.
6997 * Alpha Built-in Functions::
6998 * ARM iWMMXt Built-in Functions::
6999 * ARM NEON Intrinsics::
7000 * Blackfin Built-in Functions::
7001 * FR-V Built-in Functions::
7002 * X86 Built-in Functions::
7003 * MIPS DSP Built-in Functions::
7004 * MIPS Paired-Single Support::
7005 * MIPS Loongson Built-in Functions::
7006 * PowerPC AltiVec Built-in Functions::
7007 * SPARC VIS Built-in Functions::
7008 * SPU Built-in Functions::
7011 @node Alpha Built-in Functions
7012 @subsection Alpha Built-in Functions
7014 These built-in functions are available for the Alpha family of
7015 processors, depending on the command-line switches used.
7017 The following built-in functions are always available. They
7018 all generate the machine instruction that is part of the name.
7021 long __builtin_alpha_implver (void)
7022 long __builtin_alpha_rpcc (void)
7023 long __builtin_alpha_amask (long)
7024 long __builtin_alpha_cmpbge (long, long)
7025 long __builtin_alpha_extbl (long, long)
7026 long __builtin_alpha_extwl (long, long)
7027 long __builtin_alpha_extll (long, long)
7028 long __builtin_alpha_extql (long, long)
7029 long __builtin_alpha_extwh (long, long)
7030 long __builtin_alpha_extlh (long, long)
7031 long __builtin_alpha_extqh (long, long)
7032 long __builtin_alpha_insbl (long, long)
7033 long __builtin_alpha_inswl (long, long)
7034 long __builtin_alpha_insll (long, long)
7035 long __builtin_alpha_insql (long, long)
7036 long __builtin_alpha_inswh (long, long)
7037 long __builtin_alpha_inslh (long, long)
7038 long __builtin_alpha_insqh (long, long)
7039 long __builtin_alpha_mskbl (long, long)
7040 long __builtin_alpha_mskwl (long, long)
7041 long __builtin_alpha_mskll (long, long)
7042 long __builtin_alpha_mskql (long, long)
7043 long __builtin_alpha_mskwh (long, long)
7044 long __builtin_alpha_msklh (long, long)
7045 long __builtin_alpha_mskqh (long, long)
7046 long __builtin_alpha_umulh (long, long)
7047 long __builtin_alpha_zap (long, long)
7048 long __builtin_alpha_zapnot (long, long)
7051 The following built-in functions are always with @option{-mmax}
7052 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
7053 later. They all generate the machine instruction that is part
7057 long __builtin_alpha_pklb (long)
7058 long __builtin_alpha_pkwb (long)
7059 long __builtin_alpha_unpkbl (long)
7060 long __builtin_alpha_unpkbw (long)
7061 long __builtin_alpha_minub8 (long, long)
7062 long __builtin_alpha_minsb8 (long, long)
7063 long __builtin_alpha_minuw4 (long, long)
7064 long __builtin_alpha_minsw4 (long, long)
7065 long __builtin_alpha_maxub8 (long, long)
7066 long __builtin_alpha_maxsb8 (long, long)
7067 long __builtin_alpha_maxuw4 (long, long)
7068 long __builtin_alpha_maxsw4 (long, long)
7069 long __builtin_alpha_perr (long, long)
7072 The following built-in functions are always with @option{-mcix}
7073 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
7074 later. They all generate the machine instruction that is part
7078 long __builtin_alpha_cttz (long)
7079 long __builtin_alpha_ctlz (long)
7080 long __builtin_alpha_ctpop (long)
7083 The following builtins are available on systems that use the OSF/1
7084 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
7085 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
7086 @code{rdval} and @code{wrval}.
7089 void *__builtin_thread_pointer (void)
7090 void __builtin_set_thread_pointer (void *)
7093 @node ARM iWMMXt Built-in Functions
7094 @subsection ARM iWMMXt Built-in Functions
7096 These built-in functions are available for the ARM family of
7097 processors when the @option{-mcpu=iwmmxt} switch is used:
7100 typedef int v2si __attribute__ ((vector_size (8)));
7101 typedef short v4hi __attribute__ ((vector_size (8)));
7102 typedef char v8qi __attribute__ ((vector_size (8)));
7104 int __builtin_arm_getwcx (int)
7105 void __builtin_arm_setwcx (int, int)
7106 int __builtin_arm_textrmsb (v8qi, int)
7107 int __builtin_arm_textrmsh (v4hi, int)
7108 int __builtin_arm_textrmsw (v2si, int)
7109 int __builtin_arm_textrmub (v8qi, int)
7110 int __builtin_arm_textrmuh (v4hi, int)
7111 int __builtin_arm_textrmuw (v2si, int)
7112 v8qi __builtin_arm_tinsrb (v8qi, int)
7113 v4hi __builtin_arm_tinsrh (v4hi, int)
7114 v2si __builtin_arm_tinsrw (v2si, int)
7115 long long __builtin_arm_tmia (long long, int, int)
7116 long long __builtin_arm_tmiabb (long long, int, int)
7117 long long __builtin_arm_tmiabt (long long, int, int)
7118 long long __builtin_arm_tmiaph (long long, int, int)
7119 long long __builtin_arm_tmiatb (long long, int, int)
7120 long long __builtin_arm_tmiatt (long long, int, int)
7121 int __builtin_arm_tmovmskb (v8qi)
7122 int __builtin_arm_tmovmskh (v4hi)
7123 int __builtin_arm_tmovmskw (v2si)
7124 long long __builtin_arm_waccb (v8qi)
7125 long long __builtin_arm_wacch (v4hi)
7126 long long __builtin_arm_waccw (v2si)
7127 v8qi __builtin_arm_waddb (v8qi, v8qi)
7128 v8qi __builtin_arm_waddbss (v8qi, v8qi)
7129 v8qi __builtin_arm_waddbus (v8qi, v8qi)
7130 v4hi __builtin_arm_waddh (v4hi, v4hi)
7131 v4hi __builtin_arm_waddhss (v4hi, v4hi)
7132 v4hi __builtin_arm_waddhus (v4hi, v4hi)
7133 v2si __builtin_arm_waddw (v2si, v2si)
7134 v2si __builtin_arm_waddwss (v2si, v2si)
7135 v2si __builtin_arm_waddwus (v2si, v2si)
7136 v8qi __builtin_arm_walign (v8qi, v8qi, int)
7137 long long __builtin_arm_wand(long long, long long)
7138 long long __builtin_arm_wandn (long long, long long)
7139 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
7140 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
7141 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
7142 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
7143 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
7144 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
7145 v2si __builtin_arm_wcmpeqw (v2si, v2si)
7146 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
7147 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
7148 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
7149 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
7150 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
7151 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
7152 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
7153 long long __builtin_arm_wmacsz (v4hi, v4hi)
7154 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
7155 long long __builtin_arm_wmacuz (v4hi, v4hi)
7156 v4hi __builtin_arm_wmadds (v4hi, v4hi)
7157 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
7158 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
7159 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
7160 v2si __builtin_arm_wmaxsw (v2si, v2si)
7161 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
7162 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
7163 v2si __builtin_arm_wmaxuw (v2si, v2si)
7164 v8qi __builtin_arm_wminsb (v8qi, v8qi)
7165 v4hi __builtin_arm_wminsh (v4hi, v4hi)
7166 v2si __builtin_arm_wminsw (v2si, v2si)
7167 v8qi __builtin_arm_wminub (v8qi, v8qi)
7168 v4hi __builtin_arm_wminuh (v4hi, v4hi)
7169 v2si __builtin_arm_wminuw (v2si, v2si)
7170 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
7171 v4hi __builtin_arm_wmulul (v4hi, v4hi)
7172 v4hi __builtin_arm_wmulum (v4hi, v4hi)
7173 long long __builtin_arm_wor (long long, long long)
7174 v2si __builtin_arm_wpackdss (long long, long long)
7175 v2si __builtin_arm_wpackdus (long long, long long)
7176 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
7177 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
7178 v4hi __builtin_arm_wpackwss (v2si, v2si)
7179 v4hi __builtin_arm_wpackwus (v2si, v2si)
7180 long long __builtin_arm_wrord (long long, long long)
7181 long long __builtin_arm_wrordi (long long, int)
7182 v4hi __builtin_arm_wrorh (v4hi, long long)
7183 v4hi __builtin_arm_wrorhi (v4hi, int)
7184 v2si __builtin_arm_wrorw (v2si, long long)
7185 v2si __builtin_arm_wrorwi (v2si, int)
7186 v2si __builtin_arm_wsadb (v8qi, v8qi)
7187 v2si __builtin_arm_wsadbz (v8qi, v8qi)
7188 v2si __builtin_arm_wsadh (v4hi, v4hi)
7189 v2si __builtin_arm_wsadhz (v4hi, v4hi)
7190 v4hi __builtin_arm_wshufh (v4hi, int)
7191 long long __builtin_arm_wslld (long long, long long)
7192 long long __builtin_arm_wslldi (long long, int)
7193 v4hi __builtin_arm_wsllh (v4hi, long long)
7194 v4hi __builtin_arm_wsllhi (v4hi, int)
7195 v2si __builtin_arm_wsllw (v2si, long long)
7196 v2si __builtin_arm_wsllwi (v2si, int)
7197 long long __builtin_arm_wsrad (long long, long long)
7198 long long __builtin_arm_wsradi (long long, int)
7199 v4hi __builtin_arm_wsrah (v4hi, long long)
7200 v4hi __builtin_arm_wsrahi (v4hi, int)
7201 v2si __builtin_arm_wsraw (v2si, long long)
7202 v2si __builtin_arm_wsrawi (v2si, int)
7203 long long __builtin_arm_wsrld (long long, long long)
7204 long long __builtin_arm_wsrldi (long long, int)
7205 v4hi __builtin_arm_wsrlh (v4hi, long long)
7206 v4hi __builtin_arm_wsrlhi (v4hi, int)
7207 v2si __builtin_arm_wsrlw (v2si, long long)
7208 v2si __builtin_arm_wsrlwi (v2si, int)
7209 v8qi __builtin_arm_wsubb (v8qi, v8qi)
7210 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
7211 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
7212 v4hi __builtin_arm_wsubh (v4hi, v4hi)
7213 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
7214 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
7215 v2si __builtin_arm_wsubw (v2si, v2si)
7216 v2si __builtin_arm_wsubwss (v2si, v2si)
7217 v2si __builtin_arm_wsubwus (v2si, v2si)
7218 v4hi __builtin_arm_wunpckehsb (v8qi)
7219 v2si __builtin_arm_wunpckehsh (v4hi)
7220 long long __builtin_arm_wunpckehsw (v2si)
7221 v4hi __builtin_arm_wunpckehub (v8qi)
7222 v2si __builtin_arm_wunpckehuh (v4hi)
7223 long long __builtin_arm_wunpckehuw (v2si)
7224 v4hi __builtin_arm_wunpckelsb (v8qi)
7225 v2si __builtin_arm_wunpckelsh (v4hi)
7226 long long __builtin_arm_wunpckelsw (v2si)
7227 v4hi __builtin_arm_wunpckelub (v8qi)
7228 v2si __builtin_arm_wunpckeluh (v4hi)
7229 long long __builtin_arm_wunpckeluw (v2si)
7230 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
7231 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
7232 v2si __builtin_arm_wunpckihw (v2si, v2si)
7233 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
7234 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
7235 v2si __builtin_arm_wunpckilw (v2si, v2si)
7236 long long __builtin_arm_wxor (long long, long long)
7237 long long __builtin_arm_wzero ()
7240 @node ARM NEON Intrinsics
7241 @subsection ARM NEON Intrinsics
7243 These built-in intrinsics for the ARM Advanced SIMD extension are available
7244 when the @option{-mfpu=neon} switch is used:
7246 @include arm-neon-intrinsics.texi
7248 @node Blackfin Built-in Functions
7249 @subsection Blackfin Built-in Functions
7251 Currently, there are two Blackfin-specific built-in functions. These are
7252 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
7253 using inline assembly; by using these built-in functions the compiler can
7254 automatically add workarounds for hardware errata involving these
7255 instructions. These functions are named as follows:
7258 void __builtin_bfin_csync (void)
7259 void __builtin_bfin_ssync (void)
7262 @node FR-V Built-in Functions
7263 @subsection FR-V Built-in Functions
7265 GCC provides many FR-V-specific built-in functions. In general,
7266 these functions are intended to be compatible with those described
7267 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
7268 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
7269 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
7270 pointer rather than by value.
7272 Most of the functions are named after specific FR-V instructions.
7273 Such functions are said to be ``directly mapped'' and are summarized
7274 here in tabular form.
7278 * Directly-mapped Integer Functions::
7279 * Directly-mapped Media Functions::
7280 * Raw read/write Functions::
7281 * Other Built-in Functions::
7284 @node Argument Types
7285 @subsubsection Argument Types
7287 The arguments to the built-in functions can be divided into three groups:
7288 register numbers, compile-time constants and run-time values. In order
7289 to make this classification clear at a glance, the arguments and return
7290 values are given the following pseudo types:
7292 @multitable @columnfractions .20 .30 .15 .35
7293 @item Pseudo type @tab Real C type @tab Constant? @tab Description
7294 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
7295 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
7296 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
7297 @item @code{uw2} @tab @code{unsigned long long} @tab No
7298 @tab an unsigned doubleword
7299 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
7300 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
7301 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
7302 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
7305 These pseudo types are not defined by GCC, they are simply a notational
7306 convenience used in this manual.
7308 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
7309 and @code{sw2} are evaluated at run time. They correspond to
7310 register operands in the underlying FR-V instructions.
7312 @code{const} arguments represent immediate operands in the underlying
7313 FR-V instructions. They must be compile-time constants.
7315 @code{acc} arguments are evaluated at compile time and specify the number
7316 of an accumulator register. For example, an @code{acc} argument of 2
7317 will select the ACC2 register.
7319 @code{iacc} arguments are similar to @code{acc} arguments but specify the
7320 number of an IACC register. See @pxref{Other Built-in Functions}
7323 @node Directly-mapped Integer Functions
7324 @subsubsection Directly-mapped Integer Functions
7326 The functions listed below map directly to FR-V I-type instructions.
7328 @multitable @columnfractions .45 .32 .23
7329 @item Function prototype @tab Example usage @tab Assembly output
7330 @item @code{sw1 __ADDSS (sw1, sw1)}
7331 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
7332 @tab @code{ADDSS @var{a},@var{b},@var{c}}
7333 @item @code{sw1 __SCAN (sw1, sw1)}
7334 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
7335 @tab @code{SCAN @var{a},@var{b},@var{c}}
7336 @item @code{sw1 __SCUTSS (sw1)}
7337 @tab @code{@var{b} = __SCUTSS (@var{a})}
7338 @tab @code{SCUTSS @var{a},@var{b}}
7339 @item @code{sw1 __SLASS (sw1, sw1)}
7340 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
7341 @tab @code{SLASS @var{a},@var{b},@var{c}}
7342 @item @code{void __SMASS (sw1, sw1)}
7343 @tab @code{__SMASS (@var{a}, @var{b})}
7344 @tab @code{SMASS @var{a},@var{b}}
7345 @item @code{void __SMSSS (sw1, sw1)}
7346 @tab @code{__SMSSS (@var{a}, @var{b})}
7347 @tab @code{SMSSS @var{a},@var{b}}
7348 @item @code{void __SMU (sw1, sw1)}
7349 @tab @code{__SMU (@var{a}, @var{b})}
7350 @tab @code{SMU @var{a},@var{b}}
7351 @item @code{sw2 __SMUL (sw1, sw1)}
7352 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
7353 @tab @code{SMUL @var{a},@var{b},@var{c}}
7354 @item @code{sw1 __SUBSS (sw1, sw1)}
7355 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
7356 @tab @code{SUBSS @var{a},@var{b},@var{c}}
7357 @item @code{uw2 __UMUL (uw1, uw1)}
7358 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
7359 @tab @code{UMUL @var{a},@var{b},@var{c}}
7362 @node Directly-mapped Media Functions
7363 @subsubsection Directly-mapped Media Functions
7365 The functions listed below map directly to FR-V M-type instructions.
7367 @multitable @columnfractions .45 .32 .23
7368 @item Function prototype @tab Example usage @tab Assembly output
7369 @item @code{uw1 __MABSHS (sw1)}
7370 @tab @code{@var{b} = __MABSHS (@var{a})}
7371 @tab @code{MABSHS @var{a},@var{b}}
7372 @item @code{void __MADDACCS (acc, acc)}
7373 @tab @code{__MADDACCS (@var{b}, @var{a})}
7374 @tab @code{MADDACCS @var{a},@var{b}}
7375 @item @code{sw1 __MADDHSS (sw1, sw1)}
7376 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
7377 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
7378 @item @code{uw1 __MADDHUS (uw1, uw1)}
7379 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
7380 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
7381 @item @code{uw1 __MAND (uw1, uw1)}
7382 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
7383 @tab @code{MAND @var{a},@var{b},@var{c}}
7384 @item @code{void __MASACCS (acc, acc)}
7385 @tab @code{__MASACCS (@var{b}, @var{a})}
7386 @tab @code{MASACCS @var{a},@var{b}}
7387 @item @code{uw1 __MAVEH (uw1, uw1)}
7388 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
7389 @tab @code{MAVEH @var{a},@var{b},@var{c}}
7390 @item @code{uw2 __MBTOH (uw1)}
7391 @tab @code{@var{b} = __MBTOH (@var{a})}
7392 @tab @code{MBTOH @var{a},@var{b}}
7393 @item @code{void __MBTOHE (uw1 *, uw1)}
7394 @tab @code{__MBTOHE (&@var{b}, @var{a})}
7395 @tab @code{MBTOHE @var{a},@var{b}}
7396 @item @code{void __MCLRACC (acc)}
7397 @tab @code{__MCLRACC (@var{a})}
7398 @tab @code{MCLRACC @var{a}}
7399 @item @code{void __MCLRACCA (void)}
7400 @tab @code{__MCLRACCA ()}
7401 @tab @code{MCLRACCA}
7402 @item @code{uw1 __Mcop1 (uw1, uw1)}
7403 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
7404 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
7405 @item @code{uw1 __Mcop2 (uw1, uw1)}
7406 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
7407 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
7408 @item @code{uw1 __MCPLHI (uw2, const)}
7409 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
7410 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
7411 @item @code{uw1 __MCPLI (uw2, const)}
7412 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
7413 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
7414 @item @code{void __MCPXIS (acc, sw1, sw1)}
7415 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
7416 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
7417 @item @code{void __MCPXIU (acc, uw1, uw1)}
7418 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
7419 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
7420 @item @code{void __MCPXRS (acc, sw1, sw1)}
7421 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
7422 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
7423 @item @code{void __MCPXRU (acc, uw1, uw1)}
7424 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
7425 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
7426 @item @code{uw1 __MCUT (acc, uw1)}
7427 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
7428 @tab @code{MCUT @var{a},@var{b},@var{c}}
7429 @item @code{uw1 __MCUTSS (acc, sw1)}
7430 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
7431 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
7432 @item @code{void __MDADDACCS (acc, acc)}
7433 @tab @code{__MDADDACCS (@var{b}, @var{a})}
7434 @tab @code{MDADDACCS @var{a},@var{b}}
7435 @item @code{void __MDASACCS (acc, acc)}
7436 @tab @code{__MDASACCS (@var{b}, @var{a})}
7437 @tab @code{MDASACCS @var{a},@var{b}}
7438 @item @code{uw2 __MDCUTSSI (acc, const)}
7439 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
7440 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
7441 @item @code{uw2 __MDPACKH (uw2, uw2)}
7442 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
7443 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
7444 @item @code{uw2 __MDROTLI (uw2, const)}
7445 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
7446 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
7447 @item @code{void __MDSUBACCS (acc, acc)}
7448 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
7449 @tab @code{MDSUBACCS @var{a},@var{b}}
7450 @item @code{void __MDUNPACKH (uw1 *, uw2)}
7451 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
7452 @tab @code{MDUNPACKH @var{a},@var{b}}
7453 @item @code{uw2 __MEXPDHD (uw1, const)}
7454 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
7455 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
7456 @item @code{uw1 __MEXPDHW (uw1, const)}
7457 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
7458 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
7459 @item @code{uw1 __MHDSETH (uw1, const)}
7460 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
7461 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
7462 @item @code{sw1 __MHDSETS (const)}
7463 @tab @code{@var{b} = __MHDSETS (@var{a})}
7464 @tab @code{MHDSETS #@var{a},@var{b}}
7465 @item @code{uw1 __MHSETHIH (uw1, const)}
7466 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
7467 @tab @code{MHSETHIH #@var{a},@var{b}}
7468 @item @code{sw1 __MHSETHIS (sw1, const)}
7469 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
7470 @tab @code{MHSETHIS #@var{a},@var{b}}
7471 @item @code{uw1 __MHSETLOH (uw1, const)}
7472 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
7473 @tab @code{MHSETLOH #@var{a},@var{b}}
7474 @item @code{sw1 __MHSETLOS (sw1, const)}
7475 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
7476 @tab @code{MHSETLOS #@var{a},@var{b}}
7477 @item @code{uw1 __MHTOB (uw2)}
7478 @tab @code{@var{b} = __MHTOB (@var{a})}
7479 @tab @code{MHTOB @var{a},@var{b}}
7480 @item @code{void __MMACHS (acc, sw1, sw1)}
7481 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
7482 @tab @code{MMACHS @var{a},@var{b},@var{c}}
7483 @item @code{void __MMACHU (acc, uw1, uw1)}
7484 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
7485 @tab @code{MMACHU @var{a},@var{b},@var{c}}
7486 @item @code{void __MMRDHS (acc, sw1, sw1)}
7487 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
7488 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
7489 @item @code{void __MMRDHU (acc, uw1, uw1)}
7490 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
7491 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
7492 @item @code{void __MMULHS (acc, sw1, sw1)}
7493 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
7494 @tab @code{MMULHS @var{a},@var{b},@var{c}}
7495 @item @code{void __MMULHU (acc, uw1, uw1)}
7496 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
7497 @tab @code{MMULHU @var{a},@var{b},@var{c}}
7498 @item @code{void __MMULXHS (acc, sw1, sw1)}
7499 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
7500 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
7501 @item @code{void __MMULXHU (acc, uw1, uw1)}
7502 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
7503 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
7504 @item @code{uw1 __MNOT (uw1)}
7505 @tab @code{@var{b} = __MNOT (@var{a})}
7506 @tab @code{MNOT @var{a},@var{b}}
7507 @item @code{uw1 __MOR (uw1, uw1)}
7508 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
7509 @tab @code{MOR @var{a},@var{b},@var{c}}
7510 @item @code{uw1 __MPACKH (uh, uh)}
7511 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
7512 @tab @code{MPACKH @var{a},@var{b},@var{c}}
7513 @item @code{sw2 __MQADDHSS (sw2, sw2)}
7514 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
7515 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
7516 @item @code{uw2 __MQADDHUS (uw2, uw2)}
7517 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
7518 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
7519 @item @code{void __MQCPXIS (acc, sw2, sw2)}
7520 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
7521 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
7522 @item @code{void __MQCPXIU (acc, uw2, uw2)}
7523 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
7524 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
7525 @item @code{void __MQCPXRS (acc, sw2, sw2)}
7526 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
7527 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
7528 @item @code{void __MQCPXRU (acc, uw2, uw2)}
7529 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
7530 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
7531 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
7532 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
7533 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
7534 @item @code{sw2 __MQLMTHS (sw2, sw2)}
7535 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
7536 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
7537 @item @code{void __MQMACHS (acc, sw2, sw2)}
7538 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
7539 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
7540 @item @code{void __MQMACHU (acc, uw2, uw2)}
7541 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
7542 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
7543 @item @code{void __MQMACXHS (acc, sw2, sw2)}
7544 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
7545 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
7546 @item @code{void __MQMULHS (acc, sw2, sw2)}
7547 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
7548 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
7549 @item @code{void __MQMULHU (acc, uw2, uw2)}
7550 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
7551 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
7552 @item @code{void __MQMULXHS (acc, sw2, sw2)}
7553 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
7554 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
7555 @item @code{void __MQMULXHU (acc, uw2, uw2)}
7556 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
7557 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
7558 @item @code{sw2 __MQSATHS (sw2, sw2)}
7559 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
7560 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
7561 @item @code{uw2 __MQSLLHI (uw2, int)}
7562 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
7563 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
7564 @item @code{sw2 __MQSRAHI (sw2, int)}
7565 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
7566 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
7567 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
7568 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
7569 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
7570 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
7571 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
7572 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
7573 @item @code{void __MQXMACHS (acc, sw2, sw2)}
7574 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
7575 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
7576 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
7577 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
7578 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
7579 @item @code{uw1 __MRDACC (acc)}
7580 @tab @code{@var{b} = __MRDACC (@var{a})}
7581 @tab @code{MRDACC @var{a},@var{b}}
7582 @item @code{uw1 __MRDACCG (acc)}
7583 @tab @code{@var{b} = __MRDACCG (@var{a})}
7584 @tab @code{MRDACCG @var{a},@var{b}}
7585 @item @code{uw1 __MROTLI (uw1, const)}
7586 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
7587 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
7588 @item @code{uw1 __MROTRI (uw1, const)}
7589 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
7590 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
7591 @item @code{sw1 __MSATHS (sw1, sw1)}
7592 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
7593 @tab @code{MSATHS @var{a},@var{b},@var{c}}
7594 @item @code{uw1 __MSATHU (uw1, uw1)}
7595 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
7596 @tab @code{MSATHU @var{a},@var{b},@var{c}}
7597 @item @code{uw1 __MSLLHI (uw1, const)}
7598 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
7599 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
7600 @item @code{sw1 __MSRAHI (sw1, const)}
7601 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
7602 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
7603 @item @code{uw1 __MSRLHI (uw1, const)}
7604 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
7605 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
7606 @item @code{void __MSUBACCS (acc, acc)}
7607 @tab @code{__MSUBACCS (@var{b}, @var{a})}
7608 @tab @code{MSUBACCS @var{a},@var{b}}
7609 @item @code{sw1 __MSUBHSS (sw1, sw1)}
7610 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
7611 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
7612 @item @code{uw1 __MSUBHUS (uw1, uw1)}
7613 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
7614 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
7615 @item @code{void __MTRAP (void)}
7616 @tab @code{__MTRAP ()}
7618 @item @code{uw2 __MUNPACKH (uw1)}
7619 @tab @code{@var{b} = __MUNPACKH (@var{a})}
7620 @tab @code{MUNPACKH @var{a},@var{b}}
7621 @item @code{uw1 __MWCUT (uw2, uw1)}
7622 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
7623 @tab @code{MWCUT @var{a},@var{b},@var{c}}
7624 @item @code{void __MWTACC (acc, uw1)}
7625 @tab @code{__MWTACC (@var{b}, @var{a})}
7626 @tab @code{MWTACC @var{a},@var{b}}
7627 @item @code{void __MWTACCG (acc, uw1)}
7628 @tab @code{__MWTACCG (@var{b}, @var{a})}
7629 @tab @code{MWTACCG @var{a},@var{b}}
7630 @item @code{uw1 __MXOR (uw1, uw1)}
7631 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
7632 @tab @code{MXOR @var{a},@var{b},@var{c}}
7635 @node Raw read/write Functions
7636 @subsubsection Raw read/write Functions
7638 This sections describes built-in functions related to read and write
7639 instructions to access memory. These functions generate
7640 @code{membar} instructions to flush the I/O load and stores where
7641 appropriate, as described in Fujitsu's manual described above.
7645 @item unsigned char __builtin_read8 (void *@var{data})
7646 @item unsigned short __builtin_read16 (void *@var{data})
7647 @item unsigned long __builtin_read32 (void *@var{data})
7648 @item unsigned long long __builtin_read64 (void *@var{data})
7650 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
7651 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
7652 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
7653 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
7656 @node Other Built-in Functions
7657 @subsubsection Other Built-in Functions
7659 This section describes built-in functions that are not named after
7660 a specific FR-V instruction.
7663 @item sw2 __IACCreadll (iacc @var{reg})
7664 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
7665 for future expansion and must be 0.
7667 @item sw1 __IACCreadl (iacc @var{reg})
7668 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
7669 Other values of @var{reg} are rejected as invalid.
7671 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
7672 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
7673 is reserved for future expansion and must be 0.
7675 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7676 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7677 is 1. Other values of @var{reg} are rejected as invalid.
7679 @item void __data_prefetch0 (const void *@var{x})
7680 Use the @code{dcpl} instruction to load the contents of address @var{x}
7681 into the data cache.
7683 @item void __data_prefetch (const void *@var{x})
7684 Use the @code{nldub} instruction to load the contents of address @var{x}
7685 into the data cache. The instruction will be issued in slot I1@.
7688 @node X86 Built-in Functions
7689 @subsection X86 Built-in Functions
7691 These built-in functions are available for the i386 and x86-64 family
7692 of computers, depending on the command-line switches used.
7694 Note that, if you specify command-line switches such as @option{-msse},
7695 the compiler could use the extended instruction sets even if the built-ins
7696 are not used explicitly in the program. For this reason, applications
7697 which perform runtime CPU detection must compile separate files for each
7698 supported architecture, using the appropriate flags. In particular,
7699 the file containing the CPU detection code should be compiled without
7702 The following machine modes are available for use with MMX built-in functions
7703 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7704 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7705 vector of eight 8-bit integers. Some of the built-in functions operate on
7706 MMX registers as a whole 64-bit entity, these use @code{V1DI} as their mode.
7708 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7709 of two 32-bit floating point values.
7711 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7712 floating point values. Some instructions use a vector of four 32-bit
7713 integers, these use @code{V4SI}. Finally, some instructions operate on an
7714 entire vector register, interpreting it as a 128-bit integer, these use mode
7717 In 64-bit mode, the x86-64 family of processors uses additional built-in
7718 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7719 floating point and @code{TC} 128-bit complex floating point values.
7721 The following floating point built-in functions are available in 64-bit
7722 mode. All of them implement the function that is part of the name.
7725 __float128 __builtin_fabsq (__float128)
7726 __float128 __builtin_copysignq (__float128, __float128)
7729 The following floating point built-in functions are made available in the
7733 @item __float128 __builtin_infq (void)
7734 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7737 The following built-in functions are made available by @option{-mmmx}.
7738 All of them generate the machine instruction that is part of the name.
7741 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7742 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7743 v2si __builtin_ia32_paddd (v2si, v2si)
7744 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7745 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7746 v2si __builtin_ia32_psubd (v2si, v2si)
7747 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7748 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7749 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7750 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7751 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7752 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7753 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7754 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7755 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7756 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7757 di __builtin_ia32_pand (di, di)
7758 di __builtin_ia32_pandn (di,di)
7759 di __builtin_ia32_por (di, di)
7760 di __builtin_ia32_pxor (di, di)
7761 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7762 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7763 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7764 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7765 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7766 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7767 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7768 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7769 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7770 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7771 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7772 v2si __builtin_ia32_punpckldq (v2si, v2si)
7773 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7774 v4hi __builtin_ia32_packssdw (v2si, v2si)
7775 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7777 v4hi __builtin_ia32_psllw (v4hi, v4hi)
7778 v2si __builtin_ia32_pslld (v2si, v2si)
7779 v1di __builtin_ia32_psllq (v1di, v1di)
7780 v4hi __builtin_ia32_psrlw (v4hi, v4hi)
7781 v2si __builtin_ia32_psrld (v2si, v2si)
7782 v1di __builtin_ia32_psrlq (v1di, v1di)
7783 v4hi __builtin_ia32_psraw (v4hi, v4hi)
7784 v2si __builtin_ia32_psrad (v2si, v2si)
7785 v4hi __builtin_ia32_psllwi (v4hi, int)
7786 v2si __builtin_ia32_pslldi (v2si, int)
7787 v1di __builtin_ia32_psllqi (v1di, int)
7788 v4hi __builtin_ia32_psrlwi (v4hi, int)
7789 v2si __builtin_ia32_psrldi (v2si, int)
7790 v1di __builtin_ia32_psrlqi (v1di, int)
7791 v4hi __builtin_ia32_psrawi (v4hi, int)
7792 v2si __builtin_ia32_psradi (v2si, int)
7796 The following built-in functions are made available either with
7797 @option{-msse}, or with a combination of @option{-m3dnow} and
7798 @option{-march=athlon}. All of them generate the machine
7799 instruction that is part of the name.
7802 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7803 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7804 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7805 v1di __builtin_ia32_psadbw (v8qi, v8qi)
7806 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7807 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7808 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7809 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7810 int __builtin_ia32_pextrw (v4hi, int)
7811 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7812 int __builtin_ia32_pmovmskb (v8qi)
7813 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7814 void __builtin_ia32_movntq (di *, di)
7815 void __builtin_ia32_sfence (void)
7818 The following built-in functions are available when @option{-msse} is used.
7819 All of them generate the machine instruction that is part of the name.
7822 int __builtin_ia32_comieq (v4sf, v4sf)
7823 int __builtin_ia32_comineq (v4sf, v4sf)
7824 int __builtin_ia32_comilt (v4sf, v4sf)
7825 int __builtin_ia32_comile (v4sf, v4sf)
7826 int __builtin_ia32_comigt (v4sf, v4sf)
7827 int __builtin_ia32_comige (v4sf, v4sf)
7828 int __builtin_ia32_ucomieq (v4sf, v4sf)
7829 int __builtin_ia32_ucomineq (v4sf, v4sf)
7830 int __builtin_ia32_ucomilt (v4sf, v4sf)
7831 int __builtin_ia32_ucomile (v4sf, v4sf)
7832 int __builtin_ia32_ucomigt (v4sf, v4sf)
7833 int __builtin_ia32_ucomige (v4sf, v4sf)
7834 v4sf __builtin_ia32_addps (v4sf, v4sf)
7835 v4sf __builtin_ia32_subps (v4sf, v4sf)
7836 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7837 v4sf __builtin_ia32_divps (v4sf, v4sf)
7838 v4sf __builtin_ia32_addss (v4sf, v4sf)
7839 v4sf __builtin_ia32_subss (v4sf, v4sf)
7840 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7841 v4sf __builtin_ia32_divss (v4sf, v4sf)
7842 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7843 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7844 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7845 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7846 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7847 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7848 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7849 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7850 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7851 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7852 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7853 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7854 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7855 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7856 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7857 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7858 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7859 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7860 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7861 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7862 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7863 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7864 v4sf __builtin_ia32_minps (v4sf, v4sf)
7865 v4sf __builtin_ia32_minss (v4sf, v4sf)
7866 v4sf __builtin_ia32_andps (v4sf, v4sf)
7867 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7868 v4sf __builtin_ia32_orps (v4sf, v4sf)
7869 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7870 v4sf __builtin_ia32_movss (v4sf, v4sf)
7871 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7872 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7873 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7874 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7875 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7876 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7877 v2si __builtin_ia32_cvtps2pi (v4sf)
7878 int __builtin_ia32_cvtss2si (v4sf)
7879 v2si __builtin_ia32_cvttps2pi (v4sf)
7880 int __builtin_ia32_cvttss2si (v4sf)
7881 v4sf __builtin_ia32_rcpps (v4sf)
7882 v4sf __builtin_ia32_rsqrtps (v4sf)
7883 v4sf __builtin_ia32_sqrtps (v4sf)
7884 v4sf __builtin_ia32_rcpss (v4sf)
7885 v4sf __builtin_ia32_rsqrtss (v4sf)
7886 v4sf __builtin_ia32_sqrtss (v4sf)
7887 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7888 void __builtin_ia32_movntps (float *, v4sf)
7889 int __builtin_ia32_movmskps (v4sf)
7892 The following built-in functions are available when @option{-msse} is used.
7895 @item v4sf __builtin_ia32_loadaps (float *)
7896 Generates the @code{movaps} machine instruction as a load from memory.
7897 @item void __builtin_ia32_storeaps (float *, v4sf)
7898 Generates the @code{movaps} machine instruction as a store to memory.
7899 @item v4sf __builtin_ia32_loadups (float *)
7900 Generates the @code{movups} machine instruction as a load from memory.
7901 @item void __builtin_ia32_storeups (float *, v4sf)
7902 Generates the @code{movups} machine instruction as a store to memory.
7903 @item v4sf __builtin_ia32_loadsss (float *)
7904 Generates the @code{movss} machine instruction as a load from memory.
7905 @item void __builtin_ia32_storess (float *, v4sf)
7906 Generates the @code{movss} machine instruction as a store to memory.
7907 @item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
7908 Generates the @code{movhps} machine instruction as a load from memory.
7909 @item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
7910 Generates the @code{movlps} machine instruction as a load from memory
7911 @item void __builtin_ia32_storehps (v2sf *, v4sf)
7912 Generates the @code{movhps} machine instruction as a store to memory.
7913 @item void __builtin_ia32_storelps (v2sf *, v4sf)
7914 Generates the @code{movlps} machine instruction as a store to memory.
7917 The following built-in functions are available when @option{-msse2} is used.
7918 All of them generate the machine instruction that is part of the name.
7921 int __builtin_ia32_comisdeq (v2df, v2df)
7922 int __builtin_ia32_comisdlt (v2df, v2df)
7923 int __builtin_ia32_comisdle (v2df, v2df)
7924 int __builtin_ia32_comisdgt (v2df, v2df)
7925 int __builtin_ia32_comisdge (v2df, v2df)
7926 int __builtin_ia32_comisdneq (v2df, v2df)
7927 int __builtin_ia32_ucomisdeq (v2df, v2df)
7928 int __builtin_ia32_ucomisdlt (v2df, v2df)
7929 int __builtin_ia32_ucomisdle (v2df, v2df)
7930 int __builtin_ia32_ucomisdgt (v2df, v2df)
7931 int __builtin_ia32_ucomisdge (v2df, v2df)
7932 int __builtin_ia32_ucomisdneq (v2df, v2df)
7933 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7934 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7935 v2df __builtin_ia32_cmplepd (v2df, v2df)
7936 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7937 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7938 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7939 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7940 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7941 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7942 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7943 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7944 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7945 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7946 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7947 v2df __builtin_ia32_cmplesd (v2df, v2df)
7948 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7949 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7950 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7951 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7952 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7953 v2di __builtin_ia32_paddq (v2di, v2di)
7954 v2di __builtin_ia32_psubq (v2di, v2di)
7955 v2df __builtin_ia32_addpd (v2df, v2df)
7956 v2df __builtin_ia32_subpd (v2df, v2df)
7957 v2df __builtin_ia32_mulpd (v2df, v2df)
7958 v2df __builtin_ia32_divpd (v2df, v2df)
7959 v2df __builtin_ia32_addsd (v2df, v2df)
7960 v2df __builtin_ia32_subsd (v2df, v2df)
7961 v2df __builtin_ia32_mulsd (v2df, v2df)
7962 v2df __builtin_ia32_divsd (v2df, v2df)
7963 v2df __builtin_ia32_minpd (v2df, v2df)
7964 v2df __builtin_ia32_maxpd (v2df, v2df)
7965 v2df __builtin_ia32_minsd (v2df, v2df)
7966 v2df __builtin_ia32_maxsd (v2df, v2df)
7967 v2df __builtin_ia32_andpd (v2df, v2df)
7968 v2df __builtin_ia32_andnpd (v2df, v2df)
7969 v2df __builtin_ia32_orpd (v2df, v2df)
7970 v2df __builtin_ia32_xorpd (v2df, v2df)
7971 v2df __builtin_ia32_movsd (v2df, v2df)
7972 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7973 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7974 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7975 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7976 v4si __builtin_ia32_paddd128 (v4si, v4si)
7977 v2di __builtin_ia32_paddq128 (v2di, v2di)
7978 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7979 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7980 v4si __builtin_ia32_psubd128 (v4si, v4si)
7981 v2di __builtin_ia32_psubq128 (v2di, v2di)
7982 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7983 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7984 v2di __builtin_ia32_pand128 (v2di, v2di)
7985 v2di __builtin_ia32_pandn128 (v2di, v2di)
7986 v2di __builtin_ia32_por128 (v2di, v2di)
7987 v2di __builtin_ia32_pxor128 (v2di, v2di)
7988 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7989 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7990 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7991 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7992 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7993 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7994 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7995 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7996 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7997 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7998 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7999 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
8000 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
8001 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
8002 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
8003 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
8004 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
8005 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
8006 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
8007 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
8008 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi)
8009 v8hi __builtin_ia32_packssdw128 (v4si, v4si)
8010 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi)
8011 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
8012 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
8013 v2df __builtin_ia32_loadupd (double *)
8014 void __builtin_ia32_storeupd (double *, v2df)
8015 v2df __builtin_ia32_loadhpd (v2df, double const *)
8016 v2df __builtin_ia32_loadlpd (v2df, double const *)
8017 int __builtin_ia32_movmskpd (v2df)
8018 int __builtin_ia32_pmovmskb128 (v16qi)
8019 void __builtin_ia32_movnti (int *, int)
8020 void __builtin_ia32_movntpd (double *, v2df)
8021 void __builtin_ia32_movntdq (v2df *, v2df)
8022 v4si __builtin_ia32_pshufd (v4si, int)
8023 v8hi __builtin_ia32_pshuflw (v8hi, int)
8024 v8hi __builtin_ia32_pshufhw (v8hi, int)
8025 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
8026 v2df __builtin_ia32_sqrtpd (v2df)
8027 v2df __builtin_ia32_sqrtsd (v2df)
8028 v2df __builtin_ia32_shufpd (v2df, v2df, int)
8029 v2df __builtin_ia32_cvtdq2pd (v4si)
8030 v4sf __builtin_ia32_cvtdq2ps (v4si)
8031 v4si __builtin_ia32_cvtpd2dq (v2df)
8032 v2si __builtin_ia32_cvtpd2pi (v2df)
8033 v4sf __builtin_ia32_cvtpd2ps (v2df)
8034 v4si __builtin_ia32_cvttpd2dq (v2df)
8035 v2si __builtin_ia32_cvttpd2pi (v2df)
8036 v2df __builtin_ia32_cvtpi2pd (v2si)
8037 int __builtin_ia32_cvtsd2si (v2df)
8038 int __builtin_ia32_cvttsd2si (v2df)
8039 long long __builtin_ia32_cvtsd2si64 (v2df)
8040 long long __builtin_ia32_cvttsd2si64 (v2df)
8041 v4si __builtin_ia32_cvtps2dq (v4sf)
8042 v2df __builtin_ia32_cvtps2pd (v4sf)
8043 v4si __builtin_ia32_cvttps2dq (v4sf)
8044 v2df __builtin_ia32_cvtsi2sd (v2df, int)
8045 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
8046 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
8047 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
8048 void __builtin_ia32_clflush (const void *)
8049 void __builtin_ia32_lfence (void)
8050 void __builtin_ia32_mfence (void)
8051 v16qi __builtin_ia32_loaddqu (const char *)
8052 void __builtin_ia32_storedqu (char *, v16qi)
8053 v1di __builtin_ia32_pmuludq (v2si, v2si)
8054 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
8055 v8hi __builtin_ia32_psllw128 (v8hi, v8hi)
8056 v4si __builtin_ia32_pslld128 (v4si, v4si)
8057 v2di __builtin_ia32_psllq128 (v2di, v2di)
8058 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi)
8059 v4si __builtin_ia32_psrld128 (v4si, v4si)
8060 v2di __builtin_ia32_psrlq128 (v2di, v2di)
8061 v8hi __builtin_ia32_psraw128 (v8hi, v8hi)
8062 v4si __builtin_ia32_psrad128 (v4si, v4si)
8063 v2di __builtin_ia32_pslldqi128 (v2di, int)
8064 v8hi __builtin_ia32_psllwi128 (v8hi, int)
8065 v4si __builtin_ia32_pslldi128 (v4si, int)
8066 v2di __builtin_ia32_psllqi128 (v2di, int)
8067 v2di __builtin_ia32_psrldqi128 (v2di, int)
8068 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
8069 v4si __builtin_ia32_psrldi128 (v4si, int)
8070 v2di __builtin_ia32_psrlqi128 (v2di, int)
8071 v8hi __builtin_ia32_psrawi128 (v8hi, int)
8072 v4si __builtin_ia32_psradi128 (v4si, int)
8073 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
8076 The following built-in functions are available when @option{-msse3} is used.
8077 All of them generate the machine instruction that is part of the name.
8080 v2df __builtin_ia32_addsubpd (v2df, v2df)
8081 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
8082 v2df __builtin_ia32_haddpd (v2df, v2df)
8083 v4sf __builtin_ia32_haddps (v4sf, v4sf)
8084 v2df __builtin_ia32_hsubpd (v2df, v2df)
8085 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
8086 v16qi __builtin_ia32_lddqu (char const *)
8087 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
8088 v2df __builtin_ia32_movddup (v2df)
8089 v4sf __builtin_ia32_movshdup (v4sf)
8090 v4sf __builtin_ia32_movsldup (v4sf)
8091 void __builtin_ia32_mwait (unsigned int, unsigned int)
8094 The following built-in functions are available when @option{-msse3} is used.
8097 @item v2df __builtin_ia32_loadddup (double const *)
8098 Generates the @code{movddup} machine instruction as a load from memory.
8101 The following built-in functions are available when @option{-mssse3} is used.
8102 All of them generate the machine instruction that is part of the name
8106 v2si __builtin_ia32_phaddd (v2si, v2si)
8107 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
8108 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
8109 v2si __builtin_ia32_phsubd (v2si, v2si)
8110 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
8111 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
8112 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi)
8113 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
8114 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
8115 v8qi __builtin_ia32_psignb (v8qi, v8qi)
8116 v2si __builtin_ia32_psignd (v2si, v2si)
8117 v4hi __builtin_ia32_psignw (v4hi, v4hi)
8118 v1di __builtin_ia32_palignr (v1di, v1di, int)
8119 v8qi __builtin_ia32_pabsb (v8qi)
8120 v2si __builtin_ia32_pabsd (v2si)
8121 v4hi __builtin_ia32_pabsw (v4hi)
8124 The following built-in functions are available when @option{-mssse3} is used.
8125 All of them generate the machine instruction that is part of the name
8129 v4si __builtin_ia32_phaddd128 (v4si, v4si)
8130 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
8131 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
8132 v4si __builtin_ia32_phsubd128 (v4si, v4si)
8133 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
8134 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
8135 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
8136 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
8137 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
8138 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
8139 v4si __builtin_ia32_psignd128 (v4si, v4si)
8140 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
8141 v2di __builtin_ia32_palignr128 (v2di, v2di, int)
8142 v16qi __builtin_ia32_pabsb128 (v16qi)
8143 v4si __builtin_ia32_pabsd128 (v4si)
8144 v8hi __builtin_ia32_pabsw128 (v8hi)
8147 The following built-in functions are available when @option{-msse4.1} is
8148 used. All of them generate the machine instruction that is part of the
8152 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
8153 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
8154 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
8155 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
8156 v2df __builtin_ia32_dppd (v2df, v2df, const int)
8157 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
8158 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
8159 v2di __builtin_ia32_movntdqa (v2di *);
8160 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
8161 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
8162 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
8163 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
8164 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
8165 v8hi __builtin_ia32_phminposuw128 (v8hi)
8166 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
8167 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
8168 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
8169 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
8170 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
8171 v4si __builtin_ia32_pminsd128 (v4si, v4si)
8172 v4si __builtin_ia32_pminud128 (v4si, v4si)
8173 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
8174 v4si __builtin_ia32_pmovsxbd128 (v16qi)
8175 v2di __builtin_ia32_pmovsxbq128 (v16qi)
8176 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
8177 v2di __builtin_ia32_pmovsxdq128 (v4si)
8178 v4si __builtin_ia32_pmovsxwd128 (v8hi)
8179 v2di __builtin_ia32_pmovsxwq128 (v8hi)
8180 v4si __builtin_ia32_pmovzxbd128 (v16qi)
8181 v2di __builtin_ia32_pmovzxbq128 (v16qi)
8182 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
8183 v2di __builtin_ia32_pmovzxdq128 (v4si)
8184 v4si __builtin_ia32_pmovzxwd128 (v8hi)
8185 v2di __builtin_ia32_pmovzxwq128 (v8hi)
8186 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
8187 v4si __builtin_ia32_pmulld128 (v4si, v4si)
8188 int __builtin_ia32_ptestc128 (v2di, v2di)
8189 int __builtin_ia32_ptestnzc128 (v2di, v2di)
8190 int __builtin_ia32_ptestz128 (v2di, v2di)
8191 v2df __builtin_ia32_roundpd (v2df, const int)
8192 v4sf __builtin_ia32_roundps (v4sf, const int)
8193 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
8194 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
8197 The following built-in functions are available when @option{-msse4.1} is
8201 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
8202 Generates the @code{insertps} machine instruction.
8203 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
8204 Generates the @code{pextrb} machine instruction.
8205 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
8206 Generates the @code{pinsrb} machine instruction.
8207 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
8208 Generates the @code{pinsrd} machine instruction.
8209 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
8210 Generates the @code{pinsrq} machine instruction in 64bit mode.
8213 The following built-in functions are changed to generate new SSE4.1
8214 instructions when @option{-msse4.1} is used.
8217 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
8218 Generates the @code{extractps} machine instruction.
8219 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
8220 Generates the @code{pextrd} machine instruction.
8221 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
8222 Generates the @code{pextrq} machine instruction in 64bit mode.
8225 The following built-in functions are available when @option{-msse4.2} is
8226 used. All of them generate the machine instruction that is part of the
8230 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
8231 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
8232 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
8233 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
8234 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
8235 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
8236 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
8237 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
8238 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
8239 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
8240 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
8241 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
8242 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
8243 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
8244 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
8247 The following built-in functions are available when @option{-msse4.2} is
8251 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
8252 Generates the @code{crc32b} machine instruction.
8253 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
8254 Generates the @code{crc32w} machine instruction.
8255 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
8256 Generates the @code{crc32l} machine instruction.
8257 @item unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)
8260 The following built-in functions are changed to generate new SSE4.2
8261 instructions when @option{-msse4.2} is used.
8264 @item int __builtin_popcount (unsigned int)
8265 Generates the @code{popcntl} machine instruction.
8266 @item int __builtin_popcountl (unsigned long)
8267 Generates the @code{popcntl} or @code{popcntq} machine instruction,
8268 depending on the size of @code{unsigned long}.
8269 @item int __builtin_popcountll (unsigned long long)
8270 Generates the @code{popcntq} machine instruction.
8273 The following built-in functions are available when @option{-maes} is
8274 used. All of them generate the machine instruction that is part of the
8278 v2di __builtin_ia32_aesenc128 (v2di, v2di)
8279 v2di __builtin_ia32_aesenclast128 (v2di, v2di)
8280 v2di __builtin_ia32_aesdec128 (v2di, v2di)
8281 v2di __builtin_ia32_aesdeclast128 (v2di, v2di)
8282 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int)
8283 v2di __builtin_ia32_aesimc128 (v2di)
8286 The following built-in function is available when @option{-mpclmul} is
8290 @item v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)
8291 Generates the @code{pclmulqdq} machine instruction.
8294 The following built-in functions are available when @option{-msse4a} is used.
8295 All of them generate the machine instruction that is part of the name.
8298 void __builtin_ia32_movntsd (double *, v2df)
8299 void __builtin_ia32_movntss (float *, v4sf)
8300 v2di __builtin_ia32_extrq (v2di, v16qi)
8301 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
8302 v2di __builtin_ia32_insertq (v2di, v2di)
8303 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
8306 The following built-in functions are available when @option{-msse5} is used.
8307 All of them generate the machine instruction that is part of the name
8311 v2df __builtin_ia32_comeqpd (v2df, v2df)
8312 v2df __builtin_ia32_comeqps (v2df, v2df)
8313 v4sf __builtin_ia32_comeqsd (v4sf, v4sf)
8314 v4sf __builtin_ia32_comeqss (v4sf, v4sf)
8315 v2df __builtin_ia32_comfalsepd (v2df, v2df)
8316 v2df __builtin_ia32_comfalseps (v2df, v2df)
8317 v4sf __builtin_ia32_comfalsesd (v4sf, v4sf)
8318 v4sf __builtin_ia32_comfalsess (v4sf, v4sf)
8319 v2df __builtin_ia32_comgepd (v2df, v2df)
8320 v2df __builtin_ia32_comgeps (v2df, v2df)
8321 v4sf __builtin_ia32_comgesd (v4sf, v4sf)
8322 v4sf __builtin_ia32_comgess (v4sf, v4sf)
8323 v2df __builtin_ia32_comgtpd (v2df, v2df)
8324 v2df __builtin_ia32_comgtps (v2df, v2df)
8325 v4sf __builtin_ia32_comgtsd (v4sf, v4sf)
8326 v4sf __builtin_ia32_comgtss (v4sf, v4sf)
8327 v2df __builtin_ia32_comlepd (v2df, v2df)
8328 v2df __builtin_ia32_comleps (v2df, v2df)
8329 v4sf __builtin_ia32_comlesd (v4sf, v4sf)
8330 v4sf __builtin_ia32_comless (v4sf, v4sf)
8331 v2df __builtin_ia32_comltpd (v2df, v2df)
8332 v2df __builtin_ia32_comltps (v2df, v2df)
8333 v4sf __builtin_ia32_comltsd (v4sf, v4sf)
8334 v4sf __builtin_ia32_comltss (v4sf, v4sf)
8335 v2df __builtin_ia32_comnepd (v2df, v2df)
8336 v2df __builtin_ia32_comneps (v2df, v2df)
8337 v4sf __builtin_ia32_comnesd (v4sf, v4sf)
8338 v4sf __builtin_ia32_comness (v4sf, v4sf)
8339 v2df __builtin_ia32_comordpd (v2df, v2df)
8340 v2df __builtin_ia32_comordps (v2df, v2df)
8341 v4sf __builtin_ia32_comordsd (v4sf, v4sf)
8342 v4sf __builtin_ia32_comordss (v4sf, v4sf)
8343 v2df __builtin_ia32_comtruepd (v2df, v2df)
8344 v2df __builtin_ia32_comtrueps (v2df, v2df)
8345 v4sf __builtin_ia32_comtruesd (v4sf, v4sf)
8346 v4sf __builtin_ia32_comtruess (v4sf, v4sf)
8347 v2df __builtin_ia32_comueqpd (v2df, v2df)
8348 v2df __builtin_ia32_comueqps (v2df, v2df)
8349 v4sf __builtin_ia32_comueqsd (v4sf, v4sf)
8350 v4sf __builtin_ia32_comueqss (v4sf, v4sf)
8351 v2df __builtin_ia32_comugepd (v2df, v2df)
8352 v2df __builtin_ia32_comugeps (v2df, v2df)
8353 v4sf __builtin_ia32_comugesd (v4sf, v4sf)
8354 v4sf __builtin_ia32_comugess (v4sf, v4sf)
8355 v2df __builtin_ia32_comugtpd (v2df, v2df)
8356 v2df __builtin_ia32_comugtps (v2df, v2df)
8357 v4sf __builtin_ia32_comugtsd (v4sf, v4sf)
8358 v4sf __builtin_ia32_comugtss (v4sf, v4sf)
8359 v2df __builtin_ia32_comulepd (v2df, v2df)
8360 v2df __builtin_ia32_comuleps (v2df, v2df)
8361 v4sf __builtin_ia32_comulesd (v4sf, v4sf)
8362 v4sf __builtin_ia32_comuless (v4sf, v4sf)
8363 v2df __builtin_ia32_comultpd (v2df, v2df)
8364 v2df __builtin_ia32_comultps (v2df, v2df)
8365 v4sf __builtin_ia32_comultsd (v4sf, v4sf)
8366 v4sf __builtin_ia32_comultss (v4sf, v4sf)
8367 v2df __builtin_ia32_comunepd (v2df, v2df)
8368 v2df __builtin_ia32_comuneps (v2df, v2df)
8369 v4sf __builtin_ia32_comunesd (v4sf, v4sf)
8370 v4sf __builtin_ia32_comuness (v4sf, v4sf)
8371 v2df __builtin_ia32_comunordpd (v2df, v2df)
8372 v2df __builtin_ia32_comunordps (v2df, v2df)
8373 v4sf __builtin_ia32_comunordsd (v4sf, v4sf)
8374 v4sf __builtin_ia32_comunordss (v4sf, v4sf)
8375 v2df __builtin_ia32_fmaddpd (v2df, v2df, v2df)
8376 v4sf __builtin_ia32_fmaddps (v4sf, v4sf, v4sf)
8377 v2df __builtin_ia32_fmaddsd (v2df, v2df, v2df)
8378 v4sf __builtin_ia32_fmaddss (v4sf, v4sf, v4sf)
8379 v2df __builtin_ia32_fmsubpd (v2df, v2df, v2df)
8380 v4sf __builtin_ia32_fmsubps (v4sf, v4sf, v4sf)
8381 v2df __builtin_ia32_fmsubsd (v2df, v2df, v2df)
8382 v4sf __builtin_ia32_fmsubss (v4sf, v4sf, v4sf)
8383 v2df __builtin_ia32_fnmaddpd (v2df, v2df, v2df)
8384 v4sf __builtin_ia32_fnmaddps (v4sf, v4sf, v4sf)
8385 v2df __builtin_ia32_fnmaddsd (v2df, v2df, v2df)
8386 v4sf __builtin_ia32_fnmaddss (v4sf, v4sf, v4sf)
8387 v2df __builtin_ia32_fnmsubpd (v2df, v2df, v2df)
8388 v4sf __builtin_ia32_fnmsubps (v4sf, v4sf, v4sf)
8389 v2df __builtin_ia32_fnmsubsd (v2df, v2df, v2df)
8390 v4sf __builtin_ia32_fnmsubss (v4sf, v4sf, v4sf)
8391 v2df __builtin_ia32_frczpd (v2df)
8392 v4sf __builtin_ia32_frczps (v4sf)
8393 v2df __builtin_ia32_frczsd (v2df, v2df)
8394 v4sf __builtin_ia32_frczss (v4sf, v4sf)
8395 v2di __builtin_ia32_pcmov (v2di, v2di, v2di)
8396 v2di __builtin_ia32_pcmov_v2di (v2di, v2di, v2di)
8397 v4si __builtin_ia32_pcmov_v4si (v4si, v4si, v4si)
8398 v8hi __builtin_ia32_pcmov_v8hi (v8hi, v8hi, v8hi)
8399 v16qi __builtin_ia32_pcmov_v16qi (v16qi, v16qi, v16qi)
8400 v2df __builtin_ia32_pcmov_v2df (v2df, v2df, v2df)
8401 v4sf __builtin_ia32_pcmov_v4sf (v4sf, v4sf, v4sf)
8402 v16qi __builtin_ia32_pcomeqb (v16qi, v16qi)
8403 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8404 v4si __builtin_ia32_pcomeqd (v4si, v4si)
8405 v2di __builtin_ia32_pcomeqq (v2di, v2di)
8406 v16qi __builtin_ia32_pcomequb (v16qi, v16qi)
8407 v4si __builtin_ia32_pcomequd (v4si, v4si)
8408 v2di __builtin_ia32_pcomequq (v2di, v2di)
8409 v8hi __builtin_ia32_pcomequw (v8hi, v8hi)
8410 v8hi __builtin_ia32_pcomeqw (v8hi, v8hi)
8411 v16qi __builtin_ia32_pcomfalseb (v16qi, v16qi)
8412 v4si __builtin_ia32_pcomfalsed (v4si, v4si)
8413 v2di __builtin_ia32_pcomfalseq (v2di, v2di)
8414 v16qi __builtin_ia32_pcomfalseub (v16qi, v16qi)
8415 v4si __builtin_ia32_pcomfalseud (v4si, v4si)
8416 v2di __builtin_ia32_pcomfalseuq (v2di, v2di)
8417 v8hi __builtin_ia32_pcomfalseuw (v8hi, v8hi)
8418 v8hi __builtin_ia32_pcomfalsew (v8hi, v8hi)
8419 v16qi __builtin_ia32_pcomgeb (v16qi, v16qi)
8420 v4si __builtin_ia32_pcomged (v4si, v4si)
8421 v2di __builtin_ia32_pcomgeq (v2di, v2di)
8422 v16qi __builtin_ia32_pcomgeub (v16qi, v16qi)
8423 v4si __builtin_ia32_pcomgeud (v4si, v4si)
8424 v2di __builtin_ia32_pcomgeuq (v2di, v2di)
8425 v8hi __builtin_ia32_pcomgeuw (v8hi, v8hi)
8426 v8hi __builtin_ia32_pcomgew (v8hi, v8hi)
8427 v16qi __builtin_ia32_pcomgtb (v16qi, v16qi)
8428 v4si __builtin_ia32_pcomgtd (v4si, v4si)
8429 v2di __builtin_ia32_pcomgtq (v2di, v2di)
8430 v16qi __builtin_ia32_pcomgtub (v16qi, v16qi)
8431 v4si __builtin_ia32_pcomgtud (v4si, v4si)
8432 v2di __builtin_ia32_pcomgtuq (v2di, v2di)
8433 v8hi __builtin_ia32_pcomgtuw (v8hi, v8hi)
8434 v8hi __builtin_ia32_pcomgtw (v8hi, v8hi)
8435 v16qi __builtin_ia32_pcomleb (v16qi, v16qi)
8436 v4si __builtin_ia32_pcomled (v4si, v4si)
8437 v2di __builtin_ia32_pcomleq (v2di, v2di)
8438 v16qi __builtin_ia32_pcomleub (v16qi, v16qi)
8439 v4si __builtin_ia32_pcomleud (v4si, v4si)
8440 v2di __builtin_ia32_pcomleuq (v2di, v2di)
8441 v8hi __builtin_ia32_pcomleuw (v8hi, v8hi)
8442 v8hi __builtin_ia32_pcomlew (v8hi, v8hi)
8443 v16qi __builtin_ia32_pcomltb (v16qi, v16qi)
8444 v4si __builtin_ia32_pcomltd (v4si, v4si)
8445 v2di __builtin_ia32_pcomltq (v2di, v2di)
8446 v16qi __builtin_ia32_pcomltub (v16qi, v16qi)
8447 v4si __builtin_ia32_pcomltud (v4si, v4si)
8448 v2di __builtin_ia32_pcomltuq (v2di, v2di)
8449 v8hi __builtin_ia32_pcomltuw (v8hi, v8hi)
8450 v8hi __builtin_ia32_pcomltw (v8hi, v8hi)
8451 v16qi __builtin_ia32_pcomneb (v16qi, v16qi)
8452 v4si __builtin_ia32_pcomned (v4si, v4si)
8453 v2di __builtin_ia32_pcomneq (v2di, v2di)
8454 v16qi __builtin_ia32_pcomneub (v16qi, v16qi)
8455 v4si __builtin_ia32_pcomneud (v4si, v4si)
8456 v2di __builtin_ia32_pcomneuq (v2di, v2di)
8457 v8hi __builtin_ia32_pcomneuw (v8hi, v8hi)
8458 v8hi __builtin_ia32_pcomnew (v8hi, v8hi)
8459 v16qi __builtin_ia32_pcomtrueb (v16qi, v16qi)
8460 v4si __builtin_ia32_pcomtrued (v4si, v4si)
8461 v2di __builtin_ia32_pcomtrueq (v2di, v2di)
8462 v16qi __builtin_ia32_pcomtrueub (v16qi, v16qi)
8463 v4si __builtin_ia32_pcomtrueud (v4si, v4si)
8464 v2di __builtin_ia32_pcomtrueuq (v2di, v2di)
8465 v8hi __builtin_ia32_pcomtrueuw (v8hi, v8hi)
8466 v8hi __builtin_ia32_pcomtruew (v8hi, v8hi)
8467 v4df __builtin_ia32_permpd (v2df, v2df, v16qi)
8468 v4sf __builtin_ia32_permps (v4sf, v4sf, v16qi)
8469 v4si __builtin_ia32_phaddbd (v16qi)
8470 v2di __builtin_ia32_phaddbq (v16qi)
8471 v8hi __builtin_ia32_phaddbw (v16qi)
8472 v2di __builtin_ia32_phadddq (v4si)
8473 v4si __builtin_ia32_phaddubd (v16qi)
8474 v2di __builtin_ia32_phaddubq (v16qi)
8475 v8hi __builtin_ia32_phaddubw (v16qi)
8476 v2di __builtin_ia32_phaddudq (v4si)
8477 v4si __builtin_ia32_phadduwd (v8hi)
8478 v2di __builtin_ia32_phadduwq (v8hi)
8479 v4si __builtin_ia32_phaddwd (v8hi)
8480 v2di __builtin_ia32_phaddwq (v8hi)
8481 v8hi __builtin_ia32_phsubbw (v16qi)
8482 v2di __builtin_ia32_phsubdq (v4si)
8483 v4si __builtin_ia32_phsubwd (v8hi)
8484 v4si __builtin_ia32_pmacsdd (v4si, v4si, v4si)
8485 v2di __builtin_ia32_pmacsdqh (v4si, v4si, v2di)
8486 v2di __builtin_ia32_pmacsdql (v4si, v4si, v2di)
8487 v4si __builtin_ia32_pmacssdd (v4si, v4si, v4si)
8488 v2di __builtin_ia32_pmacssdqh (v4si, v4si, v2di)
8489 v2di __builtin_ia32_pmacssdql (v4si, v4si, v2di)
8490 v4si __builtin_ia32_pmacsswd (v8hi, v8hi, v4si)
8491 v8hi __builtin_ia32_pmacssww (v8hi, v8hi, v8hi)
8492 v4si __builtin_ia32_pmacswd (v8hi, v8hi, v4si)
8493 v8hi __builtin_ia32_pmacsww (v8hi, v8hi, v8hi)
8494 v4si __builtin_ia32_pmadcsswd (v8hi, v8hi, v4si)
8495 v4si __builtin_ia32_pmadcswd (v8hi, v8hi, v4si)
8496 v16qi __builtin_ia32_pperm (v16qi, v16qi, v16qi)
8497 v16qi __builtin_ia32_protb (v16qi, v16qi)
8498 v4si __builtin_ia32_protd (v4si, v4si)
8499 v2di __builtin_ia32_protq (v2di, v2di)
8500 v8hi __builtin_ia32_protw (v8hi, v8hi)
8501 v16qi __builtin_ia32_pshab (v16qi, v16qi)
8502 v4si __builtin_ia32_pshad (v4si, v4si)
8503 v2di __builtin_ia32_pshaq (v2di, v2di)
8504 v8hi __builtin_ia32_pshaw (v8hi, v8hi)
8505 v16qi __builtin_ia32_pshlb (v16qi, v16qi)
8506 v4si __builtin_ia32_pshld (v4si, v4si)
8507 v2di __builtin_ia32_pshlq (v2di, v2di)
8508 v8hi __builtin_ia32_pshlw (v8hi, v8hi)
8511 The following builtin-in functions are available when @option{-msse5}
8512 is used. The second argument must be an integer constant and generate
8513 the machine instruction that is part of the name with the @samp{_imm}
8517 v16qi __builtin_ia32_protb_imm (v16qi, int)
8518 v4si __builtin_ia32_protd_imm (v4si, int)
8519 v2di __builtin_ia32_protq_imm (v2di, int)
8520 v8hi __builtin_ia32_protw_imm (v8hi, int)
8523 The following built-in functions are available when @option{-m3dnow} is used.
8524 All of them generate the machine instruction that is part of the name.
8527 void __builtin_ia32_femms (void)
8528 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
8529 v2si __builtin_ia32_pf2id (v2sf)
8530 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
8531 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
8532 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
8533 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
8534 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
8535 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
8536 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
8537 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
8538 v2sf __builtin_ia32_pfrcp (v2sf)
8539 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
8540 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
8541 v2sf __builtin_ia32_pfrsqrt (v2sf)
8542 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
8543 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
8544 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
8545 v2sf __builtin_ia32_pi2fd (v2si)
8546 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
8549 The following built-in functions are available when both @option{-m3dnow}
8550 and @option{-march=athlon} are used. All of them generate the machine
8551 instruction that is part of the name.
8554 v2si __builtin_ia32_pf2iw (v2sf)
8555 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
8556 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
8557 v2sf __builtin_ia32_pi2fw (v2si)
8558 v2sf __builtin_ia32_pswapdsf (v2sf)
8559 v2si __builtin_ia32_pswapdsi (v2si)
8562 @node MIPS DSP Built-in Functions
8563 @subsection MIPS DSP Built-in Functions
8565 The MIPS DSP Application-Specific Extension (ASE) includes new
8566 instructions that are designed to improve the performance of DSP and
8567 media applications. It provides instructions that operate on packed
8568 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
8570 GCC supports MIPS DSP operations using both the generic
8571 vector extensions (@pxref{Vector Extensions}) and a collection of
8572 MIPS-specific built-in functions. Both kinds of support are
8573 enabled by the @option{-mdsp} command-line option.
8575 Revision 2 of the ASE was introduced in the second half of 2006.
8576 This revision adds extra instructions to the original ASE, but is
8577 otherwise backwards-compatible with it. You can select revision 2
8578 using the command-line option @option{-mdspr2}; this option implies
8581 At present, GCC only provides support for operations on 32-bit
8582 vectors. The vector type associated with 8-bit integer data is
8583 usually called @code{v4i8}, the vector type associated with Q7
8584 is usually called @code{v4q7}, the vector type associated with 16-bit
8585 integer data is usually called @code{v2i16}, and the vector type
8586 associated with Q15 is usually called @code{v2q15}. They can be
8587 defined in C as follows:
8590 typedef signed char v4i8 __attribute__ ((vector_size(4)));
8591 typedef signed char v4q7 __attribute__ ((vector_size(4)));
8592 typedef short v2i16 __attribute__ ((vector_size(4)));
8593 typedef short v2q15 __attribute__ ((vector_size(4)));
8596 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
8597 initialized in the same way as aggregates. For example:
8600 v4i8 a = @{1, 2, 3, 4@};
8602 b = (v4i8) @{5, 6, 7, 8@};
8604 v2q15 c = @{0x0fcb, 0x3a75@};
8606 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
8609 @emph{Note:} The CPU's endianness determines the order in which values
8610 are packed. On little-endian targets, the first value is the least
8611 significant and the last value is the most significant. The opposite
8612 order applies to big-endian targets. For example, the code above will
8613 set the lowest byte of @code{a} to @code{1} on little-endian targets
8614 and @code{4} on big-endian targets.
8616 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
8617 representation. As shown in this example, the integer representation
8618 of a Q7 value can be obtained by multiplying the fractional value by
8619 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
8620 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
8623 The table below lists the @code{v4i8} and @code{v2q15} operations for which
8624 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
8625 and @code{c} and @code{d} are @code{v2q15} values.
8627 @multitable @columnfractions .50 .50
8628 @item C code @tab MIPS instruction
8629 @item @code{a + b} @tab @code{addu.qb}
8630 @item @code{c + d} @tab @code{addq.ph}
8631 @item @code{a - b} @tab @code{subu.qb}
8632 @item @code{c - d} @tab @code{subq.ph}
8635 The table below lists the @code{v2i16} operation for which
8636 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
8637 @code{v2i16} values.
8639 @multitable @columnfractions .50 .50
8640 @item C code @tab MIPS instruction
8641 @item @code{e * f} @tab @code{mul.ph}
8644 It is easier to describe the DSP built-in functions if we first define
8645 the following types:
8650 typedef unsigned int ui32;
8651 typedef long long a64;
8654 @code{q31} and @code{i32} are actually the same as @code{int}, but we
8655 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
8656 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
8657 @code{long long}, but we use @code{a64} to indicate values that will
8658 be placed in one of the four DSP accumulators (@code{$ac0},
8659 @code{$ac1}, @code{$ac2} or @code{$ac3}).
8661 Also, some built-in functions prefer or require immediate numbers as
8662 parameters, because the corresponding DSP instructions accept both immediate
8663 numbers and register operands, or accept immediate numbers only. The
8664 immediate parameters are listed as follows.
8673 imm_n32_31: -32 to 31.
8674 imm_n512_511: -512 to 511.
8677 The following built-in functions map directly to a particular MIPS DSP
8678 instruction. Please refer to the architecture specification
8679 for details on what each instruction does.
8682 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
8683 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
8684 q31 __builtin_mips_addq_s_w (q31, q31)
8685 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
8686 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
8687 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
8688 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
8689 q31 __builtin_mips_subq_s_w (q31, q31)
8690 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
8691 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
8692 i32 __builtin_mips_addsc (i32, i32)
8693 i32 __builtin_mips_addwc (i32, i32)
8694 i32 __builtin_mips_modsub (i32, i32)
8695 i32 __builtin_mips_raddu_w_qb (v4i8)
8696 v2q15 __builtin_mips_absq_s_ph (v2q15)
8697 q31 __builtin_mips_absq_s_w (q31)
8698 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
8699 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
8700 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
8701 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
8702 q31 __builtin_mips_preceq_w_phl (v2q15)
8703 q31 __builtin_mips_preceq_w_phr (v2q15)
8704 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
8705 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
8706 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
8707 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
8708 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
8709 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
8710 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
8711 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
8712 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
8713 v4i8 __builtin_mips_shll_qb (v4i8, i32)
8714 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
8715 v2q15 __builtin_mips_shll_ph (v2q15, i32)
8716 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
8717 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
8718 q31 __builtin_mips_shll_s_w (q31, imm0_31)
8719 q31 __builtin_mips_shll_s_w (q31, i32)
8720 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
8721 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
8722 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
8723 v2q15 __builtin_mips_shra_ph (v2q15, i32)
8724 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
8725 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
8726 q31 __builtin_mips_shra_r_w (q31, imm0_31)
8727 q31 __builtin_mips_shra_r_w (q31, i32)
8728 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
8729 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
8730 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
8731 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
8732 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
8733 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
8734 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
8735 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
8736 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
8737 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
8738 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
8739 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
8740 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
8741 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
8742 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
8743 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
8744 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
8745 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
8746 i32 __builtin_mips_bitrev (i32)
8747 i32 __builtin_mips_insv (i32, i32)
8748 v4i8 __builtin_mips_repl_qb (imm0_255)
8749 v4i8 __builtin_mips_repl_qb (i32)
8750 v2q15 __builtin_mips_repl_ph (imm_n512_511)
8751 v2q15 __builtin_mips_repl_ph (i32)
8752 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
8753 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
8754 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
8755 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
8756 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
8757 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
8758 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
8759 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
8760 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
8761 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
8762 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
8763 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
8764 i32 __builtin_mips_extr_w (a64, imm0_31)
8765 i32 __builtin_mips_extr_w (a64, i32)
8766 i32 __builtin_mips_extr_r_w (a64, imm0_31)
8767 i32 __builtin_mips_extr_s_h (a64, i32)
8768 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
8769 i32 __builtin_mips_extr_rs_w (a64, i32)
8770 i32 __builtin_mips_extr_s_h (a64, imm0_31)
8771 i32 __builtin_mips_extr_r_w (a64, i32)
8772 i32 __builtin_mips_extp (a64, imm0_31)
8773 i32 __builtin_mips_extp (a64, i32)
8774 i32 __builtin_mips_extpdp (a64, imm0_31)
8775 i32 __builtin_mips_extpdp (a64, i32)
8776 a64 __builtin_mips_shilo (a64, imm_n32_31)
8777 a64 __builtin_mips_shilo (a64, i32)
8778 a64 __builtin_mips_mthlip (a64, i32)
8779 void __builtin_mips_wrdsp (i32, imm0_63)
8780 i32 __builtin_mips_rddsp (imm0_63)
8781 i32 __builtin_mips_lbux (void *, i32)
8782 i32 __builtin_mips_lhx (void *, i32)
8783 i32 __builtin_mips_lwx (void *, i32)
8784 i32 __builtin_mips_bposge32 (void)
8787 The following built-in functions map directly to a particular MIPS DSP REV 2
8788 instruction. Please refer to the architecture specification
8789 for details on what each instruction does.
8792 v4q7 __builtin_mips_absq_s_qb (v4q7);
8793 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
8794 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
8795 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
8796 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
8797 i32 __builtin_mips_append (i32, i32, imm0_31);
8798 i32 __builtin_mips_balign (i32, i32, imm0_3);
8799 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
8800 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
8801 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
8802 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
8803 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
8804 a64 __builtin_mips_madd (a64, i32, i32);
8805 a64 __builtin_mips_maddu (a64, ui32, ui32);
8806 a64 __builtin_mips_msub (a64, i32, i32);
8807 a64 __builtin_mips_msubu (a64, ui32, ui32);
8808 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
8809 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
8810 q31 __builtin_mips_mulq_rs_w (q31, q31);
8811 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
8812 q31 __builtin_mips_mulq_s_w (q31, q31);
8813 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
8814 a64 __builtin_mips_mult (i32, i32);
8815 a64 __builtin_mips_multu (ui32, ui32);
8816 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
8817 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
8818 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
8819 i32 __builtin_mips_prepend (i32, i32, imm0_31);
8820 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
8821 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
8822 v4i8 __builtin_mips_shra_qb (v4i8, i32);
8823 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
8824 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
8825 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
8826 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
8827 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
8828 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
8829 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
8830 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
8831 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
8832 q31 __builtin_mips_addqh_w (q31, q31);
8833 q31 __builtin_mips_addqh_r_w (q31, q31);
8834 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
8835 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
8836 q31 __builtin_mips_subqh_w (q31, q31);
8837 q31 __builtin_mips_subqh_r_w (q31, q31);
8838 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
8839 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
8840 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
8841 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
8842 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
8843 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
8847 @node MIPS Paired-Single Support
8848 @subsection MIPS Paired-Single Support
8850 The MIPS64 architecture includes a number of instructions that
8851 operate on pairs of single-precision floating-point values.
8852 Each pair is packed into a 64-bit floating-point register,
8853 with one element being designated the ``upper half'' and
8854 the other being designated the ``lower half''.
8856 GCC supports paired-single operations using both the generic
8857 vector extensions (@pxref{Vector Extensions}) and a collection of
8858 MIPS-specific built-in functions. Both kinds of support are
8859 enabled by the @option{-mpaired-single} command-line option.
8861 The vector type associated with paired-single values is usually
8862 called @code{v2sf}. It can be defined in C as follows:
8865 typedef float v2sf __attribute__ ((vector_size (8)));
8868 @code{v2sf} values are initialized in the same way as aggregates.
8872 v2sf a = @{1.5, 9.1@};
8875 b = (v2sf) @{e, f@};
8878 @emph{Note:} The CPU's endianness determines which value is stored in
8879 the upper half of a register and which value is stored in the lower half.
8880 On little-endian targets, the first value is the lower one and the second
8881 value is the upper one. The opposite order applies to big-endian targets.
8882 For example, the code above will set the lower half of @code{a} to
8883 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
8885 @node MIPS Loongson Built-in Functions
8886 @subsection MIPS Loongson Built-in Functions
8888 GCC provides intrinsics to access the SIMD instructions provided by the
8889 ST Microelectronics Loongson-2E and -2F processors. These intrinsics,
8890 available after inclusion of the @code{loongson.h} header file,
8891 operate on the following 64-bit vector types:
8894 @item @code{uint8x8_t}, a vector of eight unsigned 8-bit integers;
8895 @item @code{uint16x4_t}, a vector of four unsigned 16-bit integers;
8896 @item @code{uint32x2_t}, a vector of two unsigned 32-bit integers;
8897 @item @code{int8x8_t}, a vector of eight signed 8-bit integers;
8898 @item @code{int16x4_t}, a vector of four signed 16-bit integers;
8899 @item @code{int32x2_t}, a vector of two signed 32-bit integers.
8902 The intrinsics provided are listed below; each is named after the
8903 machine instruction to which it corresponds, with suffixes added as
8904 appropriate to distinguish intrinsics that expand to the same machine
8905 instruction yet have different argument types. Refer to the architecture
8906 documentation for a description of the functionality of each
8910 int16x4_t packsswh (int32x2_t s, int32x2_t t);
8911 int8x8_t packsshb (int16x4_t s, int16x4_t t);
8912 uint8x8_t packushb (uint16x4_t s, uint16x4_t t);
8913 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t);
8914 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t);
8915 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t);
8916 int32x2_t paddw_s (int32x2_t s, int32x2_t t);
8917 int16x4_t paddh_s (int16x4_t s, int16x4_t t);
8918 int8x8_t paddb_s (int8x8_t s, int8x8_t t);
8919 uint64_t paddd_u (uint64_t s, uint64_t t);
8920 int64_t paddd_s (int64_t s, int64_t t);
8921 int16x4_t paddsh (int16x4_t s, int16x4_t t);
8922 int8x8_t paddsb (int8x8_t s, int8x8_t t);
8923 uint16x4_t paddush (uint16x4_t s, uint16x4_t t);
8924 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t);
8925 uint64_t pandn_ud (uint64_t s, uint64_t t);
8926 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t);
8927 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t);
8928 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t);
8929 int64_t pandn_sd (int64_t s, int64_t t);
8930 int32x2_t pandn_sw (int32x2_t s, int32x2_t t);
8931 int16x4_t pandn_sh (int16x4_t s, int16x4_t t);
8932 int8x8_t pandn_sb (int8x8_t s, int8x8_t t);
8933 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t);
8934 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t);
8935 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t);
8936 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t);
8937 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t);
8938 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t);
8939 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t);
8940 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t);
8941 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t);
8942 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t);
8943 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t);
8944 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t);
8945 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t);
8946 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t);
8947 uint16x4_t pextrh_u (uint16x4_t s, int field);
8948 int16x4_t pextrh_s (int16x4_t s, int field);
8949 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t);
8950 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t);
8951 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t);
8952 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t);
8953 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t);
8954 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t);
8955 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t);
8956 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t);
8957 int32x2_t pmaddhw (int16x4_t s, int16x4_t t);
8958 int16x4_t pmaxsh (int16x4_t s, int16x4_t t);
8959 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t);
8960 int16x4_t pminsh (int16x4_t s, int16x4_t t);
8961 uint8x8_t pminub (uint8x8_t s, uint8x8_t t);
8962 uint8x8_t pmovmskb_u (uint8x8_t s);
8963 int8x8_t pmovmskb_s (int8x8_t s);
8964 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t);
8965 int16x4_t pmulhh (int16x4_t s, int16x4_t t);
8966 int16x4_t pmullh (int16x4_t s, int16x4_t t);
8967 int64_t pmuluw (uint32x2_t s, uint32x2_t t);
8968 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t);
8969 uint16x4_t biadd (uint8x8_t s);
8970 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t);
8971 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order);
8972 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order);
8973 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount);
8974 int16x4_t psllh_s (int16x4_t s, uint8_t amount);
8975 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount);
8976 int32x2_t psllw_s (int32x2_t s, uint8_t amount);
8977 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount);
8978 int16x4_t psrlh_s (int16x4_t s, uint8_t amount);
8979 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount);
8980 int32x2_t psrlw_s (int32x2_t s, uint8_t amount);
8981 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount);
8982 int16x4_t psrah_s (int16x4_t s, uint8_t amount);
8983 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount);
8984 int32x2_t psraw_s (int32x2_t s, uint8_t amount);
8985 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t);
8986 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t);
8987 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t);
8988 int32x2_t psubw_s (int32x2_t s, int32x2_t t);
8989 int16x4_t psubh_s (int16x4_t s, int16x4_t t);
8990 int8x8_t psubb_s (int8x8_t s, int8x8_t t);
8991 uint64_t psubd_u (uint64_t s, uint64_t t);
8992 int64_t psubd_s (int64_t s, int64_t t);
8993 int16x4_t psubsh (int16x4_t s, int16x4_t t);
8994 int8x8_t psubsb (int8x8_t s, int8x8_t t);
8995 uint16x4_t psubush (uint16x4_t s, uint16x4_t t);
8996 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t);
8997 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t);
8998 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t);
8999 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t);
9000 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t);
9001 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t);
9002 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t);
9003 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t);
9004 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t);
9005 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t);
9006 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t);
9007 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t);
9008 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t);
9012 * Paired-Single Arithmetic::
9013 * Paired-Single Built-in Functions::
9014 * MIPS-3D Built-in Functions::
9017 @node Paired-Single Arithmetic
9018 @subsubsection Paired-Single Arithmetic
9020 The table below lists the @code{v2sf} operations for which hardware
9021 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
9022 values and @code{x} is an integral value.
9024 @multitable @columnfractions .50 .50
9025 @item C code @tab MIPS instruction
9026 @item @code{a + b} @tab @code{add.ps}
9027 @item @code{a - b} @tab @code{sub.ps}
9028 @item @code{-a} @tab @code{neg.ps}
9029 @item @code{a * b} @tab @code{mul.ps}
9030 @item @code{a * b + c} @tab @code{madd.ps}
9031 @item @code{a * b - c} @tab @code{msub.ps}
9032 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
9033 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
9034 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
9037 Note that the multiply-accumulate instructions can be disabled
9038 using the command-line option @code{-mno-fused-madd}.
9040 @node Paired-Single Built-in Functions
9041 @subsubsection Paired-Single Built-in Functions
9043 The following paired-single functions map directly to a particular
9044 MIPS instruction. Please refer to the architecture specification
9045 for details on what each instruction does.
9048 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
9049 Pair lower lower (@code{pll.ps}).
9051 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
9052 Pair upper lower (@code{pul.ps}).
9054 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
9055 Pair lower upper (@code{plu.ps}).
9057 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
9058 Pair upper upper (@code{puu.ps}).
9060 @item v2sf __builtin_mips_cvt_ps_s (float, float)
9061 Convert pair to paired single (@code{cvt.ps.s}).
9063 @item float __builtin_mips_cvt_s_pl (v2sf)
9064 Convert pair lower to single (@code{cvt.s.pl}).
9066 @item float __builtin_mips_cvt_s_pu (v2sf)
9067 Convert pair upper to single (@code{cvt.s.pu}).
9069 @item v2sf __builtin_mips_abs_ps (v2sf)
9070 Absolute value (@code{abs.ps}).
9072 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
9073 Align variable (@code{alnv.ps}).
9075 @emph{Note:} The value of the third parameter must be 0 or 4
9076 modulo 8, otherwise the result will be unpredictable. Please read the
9077 instruction description for details.
9080 The following multi-instruction functions are also available.
9081 In each case, @var{cond} can be any of the 16 floating-point conditions:
9082 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9083 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
9084 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9087 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9088 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9089 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
9090 @code{movt.ps}/@code{movf.ps}).
9092 The @code{movt} functions return the value @var{x} computed by:
9095 c.@var{cond}.ps @var{cc},@var{a},@var{b}
9096 mov.ps @var{x},@var{c}
9097 movt.ps @var{x},@var{d},@var{cc}
9100 The @code{movf} functions are similar but use @code{movf.ps} instead
9103 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9104 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9105 Comparison of two paired-single values (@code{c.@var{cond}.ps},
9106 @code{bc1t}/@code{bc1f}).
9108 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9109 and return either the upper or lower half of the result. For example:
9113 if (__builtin_mips_upper_c_eq_ps (a, b))
9114 upper_halves_are_equal ();
9116 upper_halves_are_unequal ();
9118 if (__builtin_mips_lower_c_eq_ps (a, b))
9119 lower_halves_are_equal ();
9121 lower_halves_are_unequal ();
9125 @node MIPS-3D Built-in Functions
9126 @subsubsection MIPS-3D Built-in Functions
9128 The MIPS-3D Application-Specific Extension (ASE) includes additional
9129 paired-single instructions that are designed to improve the performance
9130 of 3D graphics operations. Support for these instructions is controlled
9131 by the @option{-mips3d} command-line option.
9133 The functions listed below map directly to a particular MIPS-3D
9134 instruction. Please refer to the architecture specification for
9135 more details on what each instruction does.
9138 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
9139 Reduction add (@code{addr.ps}).
9141 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
9142 Reduction multiply (@code{mulr.ps}).
9144 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
9145 Convert paired single to paired word (@code{cvt.pw.ps}).
9147 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
9148 Convert paired word to paired single (@code{cvt.ps.pw}).
9150 @item float __builtin_mips_recip1_s (float)
9151 @itemx double __builtin_mips_recip1_d (double)
9152 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
9153 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
9155 @item float __builtin_mips_recip2_s (float, float)
9156 @itemx double __builtin_mips_recip2_d (double, double)
9157 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
9158 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
9160 @item float __builtin_mips_rsqrt1_s (float)
9161 @itemx double __builtin_mips_rsqrt1_d (double)
9162 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
9163 Reduced precision reciprocal square root (sequence step 1)
9164 (@code{rsqrt1.@var{fmt}}).
9166 @item float __builtin_mips_rsqrt2_s (float, float)
9167 @itemx double __builtin_mips_rsqrt2_d (double, double)
9168 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
9169 Reduced precision reciprocal square root (sequence step 2)
9170 (@code{rsqrt2.@var{fmt}}).
9173 The following multi-instruction functions are also available.
9174 In each case, @var{cond} can be any of the 16 floating-point conditions:
9175 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
9176 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
9177 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
9180 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
9181 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
9182 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
9183 @code{bc1t}/@code{bc1f}).
9185 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
9186 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
9191 if (__builtin_mips_cabs_eq_s (a, b))
9197 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9198 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9199 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
9200 @code{bc1t}/@code{bc1f}).
9202 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
9203 and return either the upper or lower half of the result. For example:
9207 if (__builtin_mips_upper_cabs_eq_ps (a, b))
9208 upper_halves_are_equal ();
9210 upper_halves_are_unequal ();
9212 if (__builtin_mips_lower_cabs_eq_ps (a, b))
9213 lower_halves_are_equal ();
9215 lower_halves_are_unequal ();
9218 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9219 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9220 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
9221 @code{movt.ps}/@code{movf.ps}).
9223 The @code{movt} functions return the value @var{x} computed by:
9226 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
9227 mov.ps @var{x},@var{c}
9228 movt.ps @var{x},@var{d},@var{cc}
9231 The @code{movf} functions are similar but use @code{movf.ps} instead
9234 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9235 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9236 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9237 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
9238 Comparison of two paired-single values
9239 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9240 @code{bc1any2t}/@code{bc1any2f}).
9242 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
9243 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
9244 result is true and the @code{all} forms return true if both results are true.
9249 if (__builtin_mips_any_c_eq_ps (a, b))
9254 if (__builtin_mips_all_c_eq_ps (a, b))
9260 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9261 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9262 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9263 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
9264 Comparison of four paired-single values
9265 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
9266 @code{bc1any4t}/@code{bc1any4f}).
9268 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
9269 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
9270 The @code{any} forms return true if any of the four results are true
9271 and the @code{all} forms return true if all four results are true.
9276 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
9281 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
9288 @node PowerPC AltiVec Built-in Functions
9289 @subsection PowerPC AltiVec Built-in Functions
9291 GCC provides an interface for the PowerPC family of processors to access
9292 the AltiVec operations described in Motorola's AltiVec Programming
9293 Interface Manual. The interface is made available by including
9294 @code{<altivec.h>} and using @option{-maltivec} and
9295 @option{-mabi=altivec}. The interface supports the following vector
9299 vector unsigned char
9303 vector unsigned short
9314 GCC's implementation of the high-level language interface available from
9315 C and C++ code differs from Motorola's documentation in several ways.
9320 A vector constant is a list of constant expressions within curly braces.
9323 A vector initializer requires no cast if the vector constant is of the
9324 same type as the variable it is initializing.
9327 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9328 vector type is the default signedness of the base type. The default
9329 varies depending on the operating system, so a portable program should
9330 always specify the signedness.
9333 Compiling with @option{-maltivec} adds keywords @code{__vector},
9334 @code{vector}, @code{__pixel}, @code{pixel}, @code{__bool} and
9335 @code{bool}. When compiling ISO C, the context-sensitive substitution
9336 of the keywords @code{vector}, @code{pixel} and @code{bool} is
9337 disabled. To use them, you must include @code{<altivec.h>} instead.
9340 GCC allows using a @code{typedef} name as the type specifier for a
9344 For C, overloaded functions are implemented with macros so the following
9348 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9351 Since @code{vec_add} is a macro, the vector constant in the example
9352 is treated as four separate arguments. Wrap the entire argument in
9353 parentheses for this to work.
9356 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
9357 Internally, GCC uses built-in functions to achieve the functionality in
9358 the aforementioned header file, but they are not supported and are
9359 subject to change without notice.
9361 The following interfaces are supported for the generic and specific
9362 AltiVec operations and the AltiVec predicates. In cases where there
9363 is a direct mapping between generic and specific operations, only the
9364 generic names are shown here, although the specific operations can also
9367 Arguments that are documented as @code{const int} require literal
9368 integral values within the range required for that operation.
9371 vector signed char vec_abs (vector signed char);
9372 vector signed short vec_abs (vector signed short);
9373 vector signed int vec_abs (vector signed int);
9374 vector float vec_abs (vector float);
9376 vector signed char vec_abss (vector signed char);
9377 vector signed short vec_abss (vector signed short);
9378 vector signed int vec_abss (vector signed int);
9380 vector signed char vec_add (vector bool char, vector signed char);
9381 vector signed char vec_add (vector signed char, vector bool char);
9382 vector signed char vec_add (vector signed char, vector signed char);
9383 vector unsigned char vec_add (vector bool char, vector unsigned char);
9384 vector unsigned char vec_add (vector unsigned char, vector bool char);
9385 vector unsigned char vec_add (vector unsigned char,
9386 vector unsigned char);
9387 vector signed short vec_add (vector bool short, vector signed short);
9388 vector signed short vec_add (vector signed short, vector bool short);
9389 vector signed short vec_add (vector signed short, vector signed short);
9390 vector unsigned short vec_add (vector bool short,
9391 vector unsigned short);
9392 vector unsigned short vec_add (vector unsigned short,
9394 vector unsigned short vec_add (vector unsigned short,
9395 vector unsigned short);
9396 vector signed int vec_add (vector bool int, vector signed int);
9397 vector signed int vec_add (vector signed int, vector bool int);
9398 vector signed int vec_add (vector signed int, vector signed int);
9399 vector unsigned int vec_add (vector bool int, vector unsigned int);
9400 vector unsigned int vec_add (vector unsigned int, vector bool int);
9401 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
9402 vector float vec_add (vector float, vector float);
9404 vector float vec_vaddfp (vector float, vector float);
9406 vector signed int vec_vadduwm (vector bool int, vector signed int);
9407 vector signed int vec_vadduwm (vector signed int, vector bool int);
9408 vector signed int vec_vadduwm (vector signed int, vector signed int);
9409 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
9410 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
9411 vector unsigned int vec_vadduwm (vector unsigned int,
9412 vector unsigned int);
9414 vector signed short vec_vadduhm (vector bool short,
9415 vector signed short);
9416 vector signed short vec_vadduhm (vector signed short,
9418 vector signed short vec_vadduhm (vector signed short,
9419 vector signed short);
9420 vector unsigned short vec_vadduhm (vector bool short,
9421 vector unsigned short);
9422 vector unsigned short vec_vadduhm (vector unsigned short,
9424 vector unsigned short vec_vadduhm (vector unsigned short,
9425 vector unsigned short);
9427 vector signed char vec_vaddubm (vector bool char, vector signed char);
9428 vector signed char vec_vaddubm (vector signed char, vector bool char);
9429 vector signed char vec_vaddubm (vector signed char, vector signed char);
9430 vector unsigned char vec_vaddubm (vector bool char,
9431 vector unsigned char);
9432 vector unsigned char vec_vaddubm (vector unsigned char,
9434 vector unsigned char vec_vaddubm (vector unsigned char,
9435 vector unsigned char);
9437 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
9439 vector unsigned char vec_adds (vector bool char, vector unsigned char);
9440 vector unsigned char vec_adds (vector unsigned char, vector bool char);
9441 vector unsigned char vec_adds (vector unsigned char,
9442 vector unsigned char);
9443 vector signed char vec_adds (vector bool char, vector signed char);
9444 vector signed char vec_adds (vector signed char, vector bool char);
9445 vector signed char vec_adds (vector signed char, vector signed char);
9446 vector unsigned short vec_adds (vector bool short,
9447 vector unsigned short);
9448 vector unsigned short vec_adds (vector unsigned short,
9450 vector unsigned short vec_adds (vector unsigned short,
9451 vector unsigned short);
9452 vector signed short vec_adds (vector bool short, vector signed short);
9453 vector signed short vec_adds (vector signed short, vector bool short);
9454 vector signed short vec_adds (vector signed short, vector signed short);
9455 vector unsigned int vec_adds (vector bool int, vector unsigned int);
9456 vector unsigned int vec_adds (vector unsigned int, vector bool int);
9457 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
9458 vector signed int vec_adds (vector bool int, vector signed int);
9459 vector signed int vec_adds (vector signed int, vector bool int);
9460 vector signed int vec_adds (vector signed int, vector signed int);
9462 vector signed int vec_vaddsws (vector bool int, vector signed int);
9463 vector signed int vec_vaddsws (vector signed int, vector bool int);
9464 vector signed int vec_vaddsws (vector signed int, vector signed int);
9466 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
9467 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
9468 vector unsigned int vec_vadduws (vector unsigned int,
9469 vector unsigned int);
9471 vector signed short vec_vaddshs (vector bool short,
9472 vector signed short);
9473 vector signed short vec_vaddshs (vector signed short,
9475 vector signed short vec_vaddshs (vector signed short,
9476 vector signed short);
9478 vector unsigned short vec_vadduhs (vector bool short,
9479 vector unsigned short);
9480 vector unsigned short vec_vadduhs (vector unsigned short,
9482 vector unsigned short vec_vadduhs (vector unsigned short,
9483 vector unsigned short);
9485 vector signed char vec_vaddsbs (vector bool char, vector signed char);
9486 vector signed char vec_vaddsbs (vector signed char, vector bool char);
9487 vector signed char vec_vaddsbs (vector signed char, vector signed char);
9489 vector unsigned char vec_vaddubs (vector bool char,
9490 vector unsigned char);
9491 vector unsigned char vec_vaddubs (vector unsigned char,
9493 vector unsigned char vec_vaddubs (vector unsigned char,
9494 vector unsigned char);
9496 vector float vec_and (vector float, vector float);
9497 vector float vec_and (vector float, vector bool int);
9498 vector float vec_and (vector bool int, vector float);
9499 vector bool int vec_and (vector bool int, vector bool int);
9500 vector signed int vec_and (vector bool int, vector signed int);
9501 vector signed int vec_and (vector signed int, vector bool int);
9502 vector signed int vec_and (vector signed int, vector signed int);
9503 vector unsigned int vec_and (vector bool int, vector unsigned int);
9504 vector unsigned int vec_and (vector unsigned int, vector bool int);
9505 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
9506 vector bool short vec_and (vector bool short, vector bool short);
9507 vector signed short vec_and (vector bool short, vector signed short);
9508 vector signed short vec_and (vector signed short, vector bool short);
9509 vector signed short vec_and (vector signed short, vector signed short);
9510 vector unsigned short vec_and (vector bool short,
9511 vector unsigned short);
9512 vector unsigned short vec_and (vector unsigned short,
9514 vector unsigned short vec_and (vector unsigned short,
9515 vector unsigned short);
9516 vector signed char vec_and (vector bool char, vector signed char);
9517 vector bool char vec_and (vector bool char, vector bool char);
9518 vector signed char vec_and (vector signed char, vector bool char);
9519 vector signed char vec_and (vector signed char, vector signed char);
9520 vector unsigned char vec_and (vector bool char, vector unsigned char);
9521 vector unsigned char vec_and (vector unsigned char, vector bool char);
9522 vector unsigned char vec_and (vector unsigned char,
9523 vector unsigned char);
9525 vector float vec_andc (vector float, vector float);
9526 vector float vec_andc (vector float, vector bool int);
9527 vector float vec_andc (vector bool int, vector float);
9528 vector bool int vec_andc (vector bool int, vector bool int);
9529 vector signed int vec_andc (vector bool int, vector signed int);
9530 vector signed int vec_andc (vector signed int, vector bool int);
9531 vector signed int vec_andc (vector signed int, vector signed int);
9532 vector unsigned int vec_andc (vector bool int, vector unsigned int);
9533 vector unsigned int vec_andc (vector unsigned int, vector bool int);
9534 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
9535 vector bool short vec_andc (vector bool short, vector bool short);
9536 vector signed short vec_andc (vector bool short, vector signed short);
9537 vector signed short vec_andc (vector signed short, vector bool short);
9538 vector signed short vec_andc (vector signed short, vector signed short);
9539 vector unsigned short vec_andc (vector bool short,
9540 vector unsigned short);
9541 vector unsigned short vec_andc (vector unsigned short,
9543 vector unsigned short vec_andc (vector unsigned short,
9544 vector unsigned short);
9545 vector signed char vec_andc (vector bool char, vector signed char);
9546 vector bool char vec_andc (vector bool char, vector bool char);
9547 vector signed char vec_andc (vector signed char, vector bool char);
9548 vector signed char vec_andc (vector signed char, vector signed char);
9549 vector unsigned char vec_andc (vector bool char, vector unsigned char);
9550 vector unsigned char vec_andc (vector unsigned char, vector bool char);
9551 vector unsigned char vec_andc (vector unsigned char,
9552 vector unsigned char);
9554 vector unsigned char vec_avg (vector unsigned char,
9555 vector unsigned char);
9556 vector signed char vec_avg (vector signed char, vector signed char);
9557 vector unsigned short vec_avg (vector unsigned short,
9558 vector unsigned short);
9559 vector signed short vec_avg (vector signed short, vector signed short);
9560 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
9561 vector signed int vec_avg (vector signed int, vector signed int);
9563 vector signed int vec_vavgsw (vector signed int, vector signed int);
9565 vector unsigned int vec_vavguw (vector unsigned int,
9566 vector unsigned int);
9568 vector signed short vec_vavgsh (vector signed short,
9569 vector signed short);
9571 vector unsigned short vec_vavguh (vector unsigned short,
9572 vector unsigned short);
9574 vector signed char vec_vavgsb (vector signed char, vector signed char);
9576 vector unsigned char vec_vavgub (vector unsigned char,
9577 vector unsigned char);
9579 vector float vec_ceil (vector float);
9581 vector signed int vec_cmpb (vector float, vector float);
9583 vector bool char vec_cmpeq (vector signed char, vector signed char);
9584 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
9585 vector bool short vec_cmpeq (vector signed short, vector signed short);
9586 vector bool short vec_cmpeq (vector unsigned short,
9587 vector unsigned short);
9588 vector bool int vec_cmpeq (vector signed int, vector signed int);
9589 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
9590 vector bool int vec_cmpeq (vector float, vector float);
9592 vector bool int vec_vcmpeqfp (vector float, vector float);
9594 vector bool int vec_vcmpequw (vector signed int, vector signed int);
9595 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
9597 vector bool short vec_vcmpequh (vector signed short,
9598 vector signed short);
9599 vector bool short vec_vcmpequh (vector unsigned short,
9600 vector unsigned short);
9602 vector bool char vec_vcmpequb (vector signed char, vector signed char);
9603 vector bool char vec_vcmpequb (vector unsigned char,
9604 vector unsigned char);
9606 vector bool int vec_cmpge (vector float, vector float);
9608 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
9609 vector bool char vec_cmpgt (vector signed char, vector signed char);
9610 vector bool short vec_cmpgt (vector unsigned short,
9611 vector unsigned short);
9612 vector bool short vec_cmpgt (vector signed short, vector signed short);
9613 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
9614 vector bool int vec_cmpgt (vector signed int, vector signed int);
9615 vector bool int vec_cmpgt (vector float, vector float);
9617 vector bool int vec_vcmpgtfp (vector float, vector float);
9619 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
9621 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
9623 vector bool short vec_vcmpgtsh (vector signed short,
9624 vector signed short);
9626 vector bool short vec_vcmpgtuh (vector unsigned short,
9627 vector unsigned short);
9629 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
9631 vector bool char vec_vcmpgtub (vector unsigned char,
9632 vector unsigned char);
9634 vector bool int vec_cmple (vector float, vector float);
9636 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
9637 vector bool char vec_cmplt (vector signed char, vector signed char);
9638 vector bool short vec_cmplt (vector unsigned short,
9639 vector unsigned short);
9640 vector bool short vec_cmplt (vector signed short, vector signed short);
9641 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
9642 vector bool int vec_cmplt (vector signed int, vector signed int);
9643 vector bool int vec_cmplt (vector float, vector float);
9645 vector float vec_ctf (vector unsigned int, const int);
9646 vector float vec_ctf (vector signed int, const int);
9648 vector float vec_vcfsx (vector signed int, const int);
9650 vector float vec_vcfux (vector unsigned int, const int);
9652 vector signed int vec_cts (vector float, const int);
9654 vector unsigned int vec_ctu (vector float, const int);
9656 void vec_dss (const int);
9658 void vec_dssall (void);
9660 void vec_dst (const vector unsigned char *, int, const int);
9661 void vec_dst (const vector signed char *, int, const int);
9662 void vec_dst (const vector bool char *, int, const int);
9663 void vec_dst (const vector unsigned short *, int, const int);
9664 void vec_dst (const vector signed short *, int, const int);
9665 void vec_dst (const vector bool short *, int, const int);
9666 void vec_dst (const vector pixel *, int, const int);
9667 void vec_dst (const vector unsigned int *, int, const int);
9668 void vec_dst (const vector signed int *, int, const int);
9669 void vec_dst (const vector bool int *, int, const int);
9670 void vec_dst (const vector float *, int, const int);
9671 void vec_dst (const unsigned char *, int, const int);
9672 void vec_dst (const signed char *, int, const int);
9673 void vec_dst (const unsigned short *, int, const int);
9674 void vec_dst (const short *, int, const int);
9675 void vec_dst (const unsigned int *, int, const int);
9676 void vec_dst (const int *, int, const int);
9677 void vec_dst (const unsigned long *, int, const int);
9678 void vec_dst (const long *, int, const int);
9679 void vec_dst (const float *, int, const int);
9681 void vec_dstst (const vector unsigned char *, int, const int);
9682 void vec_dstst (const vector signed char *, int, const int);
9683 void vec_dstst (const vector bool char *, int, const int);
9684 void vec_dstst (const vector unsigned short *, int, const int);
9685 void vec_dstst (const vector signed short *, int, const int);
9686 void vec_dstst (const vector bool short *, int, const int);
9687 void vec_dstst (const vector pixel *, int, const int);
9688 void vec_dstst (const vector unsigned int *, int, const int);
9689 void vec_dstst (const vector signed int *, int, const int);
9690 void vec_dstst (const vector bool int *, int, const int);
9691 void vec_dstst (const vector float *, int, const int);
9692 void vec_dstst (const unsigned char *, int, const int);
9693 void vec_dstst (const signed char *, int, const int);
9694 void vec_dstst (const unsigned short *, int, const int);
9695 void vec_dstst (const short *, int, const int);
9696 void vec_dstst (const unsigned int *, int, const int);
9697 void vec_dstst (const int *, int, const int);
9698 void vec_dstst (const unsigned long *, int, const int);
9699 void vec_dstst (const long *, int, const int);
9700 void vec_dstst (const float *, int, const int);
9702 void vec_dststt (const vector unsigned char *, int, const int);
9703 void vec_dststt (const vector signed char *, int, const int);
9704 void vec_dststt (const vector bool char *, int, const int);
9705 void vec_dststt (const vector unsigned short *, int, const int);
9706 void vec_dststt (const vector signed short *, int, const int);
9707 void vec_dststt (const vector bool short *, int, const int);
9708 void vec_dststt (const vector pixel *, int, const int);
9709 void vec_dststt (const vector unsigned int *, int, const int);
9710 void vec_dststt (const vector signed int *, int, const int);
9711 void vec_dststt (const vector bool int *, int, const int);
9712 void vec_dststt (const vector float *, int, const int);
9713 void vec_dststt (const unsigned char *, int, const int);
9714 void vec_dststt (const signed char *, int, const int);
9715 void vec_dststt (const unsigned short *, int, const int);
9716 void vec_dststt (const short *, int, const int);
9717 void vec_dststt (const unsigned int *, int, const int);
9718 void vec_dststt (const int *, int, const int);
9719 void vec_dststt (const unsigned long *, int, const int);
9720 void vec_dststt (const long *, int, const int);
9721 void vec_dststt (const float *, int, const int);
9723 void vec_dstt (const vector unsigned char *, int, const int);
9724 void vec_dstt (const vector signed char *, int, const int);
9725 void vec_dstt (const vector bool char *, int, const int);
9726 void vec_dstt (const vector unsigned short *, int, const int);
9727 void vec_dstt (const vector signed short *, int, const int);
9728 void vec_dstt (const vector bool short *, int, const int);
9729 void vec_dstt (const vector pixel *, int, const int);
9730 void vec_dstt (const vector unsigned int *, int, const int);
9731 void vec_dstt (const vector signed int *, int, const int);
9732 void vec_dstt (const vector bool int *, int, const int);
9733 void vec_dstt (const vector float *, int, const int);
9734 void vec_dstt (const unsigned char *, int, const int);
9735 void vec_dstt (const signed char *, int, const int);
9736 void vec_dstt (const unsigned short *, int, const int);
9737 void vec_dstt (const short *, int, const int);
9738 void vec_dstt (const unsigned int *, int, const int);
9739 void vec_dstt (const int *, int, const int);
9740 void vec_dstt (const unsigned long *, int, const int);
9741 void vec_dstt (const long *, int, const int);
9742 void vec_dstt (const float *, int, const int);
9744 vector float vec_expte (vector float);
9746 vector float vec_floor (vector float);
9748 vector float vec_ld (int, const vector float *);
9749 vector float vec_ld (int, const float *);
9750 vector bool int vec_ld (int, const vector bool int *);
9751 vector signed int vec_ld (int, const vector signed int *);
9752 vector signed int vec_ld (int, const int *);
9753 vector signed int vec_ld (int, const long *);
9754 vector unsigned int vec_ld (int, const vector unsigned int *);
9755 vector unsigned int vec_ld (int, const unsigned int *);
9756 vector unsigned int vec_ld (int, const unsigned long *);
9757 vector bool short vec_ld (int, const vector bool short *);
9758 vector pixel vec_ld (int, const vector pixel *);
9759 vector signed short vec_ld (int, const vector signed short *);
9760 vector signed short vec_ld (int, const short *);
9761 vector unsigned short vec_ld (int, const vector unsigned short *);
9762 vector unsigned short vec_ld (int, const unsigned short *);
9763 vector bool char vec_ld (int, const vector bool char *);
9764 vector signed char vec_ld (int, const vector signed char *);
9765 vector signed char vec_ld (int, const signed char *);
9766 vector unsigned char vec_ld (int, const vector unsigned char *);
9767 vector unsigned char vec_ld (int, const unsigned char *);
9769 vector signed char vec_lde (int, const signed char *);
9770 vector unsigned char vec_lde (int, const unsigned char *);
9771 vector signed short vec_lde (int, const short *);
9772 vector unsigned short vec_lde (int, const unsigned short *);
9773 vector float vec_lde (int, const float *);
9774 vector signed int vec_lde (int, const int *);
9775 vector unsigned int vec_lde (int, const unsigned int *);
9776 vector signed int vec_lde (int, const long *);
9777 vector unsigned int vec_lde (int, const unsigned long *);
9779 vector float vec_lvewx (int, float *);
9780 vector signed int vec_lvewx (int, int *);
9781 vector unsigned int vec_lvewx (int, unsigned int *);
9782 vector signed int vec_lvewx (int, long *);
9783 vector unsigned int vec_lvewx (int, unsigned long *);
9785 vector signed short vec_lvehx (int, short *);
9786 vector unsigned short vec_lvehx (int, unsigned short *);
9788 vector signed char vec_lvebx (int, char *);
9789 vector unsigned char vec_lvebx (int, unsigned char *);
9791 vector float vec_ldl (int, const vector float *);
9792 vector float vec_ldl (int, const float *);
9793 vector bool int vec_ldl (int, const vector bool int *);
9794 vector signed int vec_ldl (int, const vector signed int *);
9795 vector signed int vec_ldl (int, const int *);
9796 vector signed int vec_ldl (int, const long *);
9797 vector unsigned int vec_ldl (int, const vector unsigned int *);
9798 vector unsigned int vec_ldl (int, const unsigned int *);
9799 vector unsigned int vec_ldl (int, const unsigned long *);
9800 vector bool short vec_ldl (int, const vector bool short *);
9801 vector pixel vec_ldl (int, const vector pixel *);
9802 vector signed short vec_ldl (int, const vector signed short *);
9803 vector signed short vec_ldl (int, const short *);
9804 vector unsigned short vec_ldl (int, const vector unsigned short *);
9805 vector unsigned short vec_ldl (int, const unsigned short *);
9806 vector bool char vec_ldl (int, const vector bool char *);
9807 vector signed char vec_ldl (int, const vector signed char *);
9808 vector signed char vec_ldl (int, const signed char *);
9809 vector unsigned char vec_ldl (int, const vector unsigned char *);
9810 vector unsigned char vec_ldl (int, const unsigned char *);
9812 vector float vec_loge (vector float);
9814 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
9815 vector unsigned char vec_lvsl (int, const volatile signed char *);
9816 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
9817 vector unsigned char vec_lvsl (int, const volatile short *);
9818 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
9819 vector unsigned char vec_lvsl (int, const volatile int *);
9820 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
9821 vector unsigned char vec_lvsl (int, const volatile long *);
9822 vector unsigned char vec_lvsl (int, const volatile float *);
9824 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
9825 vector unsigned char vec_lvsr (int, const volatile signed char *);
9826 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
9827 vector unsigned char vec_lvsr (int, const volatile short *);
9828 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
9829 vector unsigned char vec_lvsr (int, const volatile int *);
9830 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
9831 vector unsigned char vec_lvsr (int, const volatile long *);
9832 vector unsigned char vec_lvsr (int, const volatile float *);
9834 vector float vec_madd (vector float, vector float, vector float);
9836 vector signed short vec_madds (vector signed short,
9837 vector signed short,
9838 vector signed short);
9840 vector unsigned char vec_max (vector bool char, vector unsigned char);
9841 vector unsigned char vec_max (vector unsigned char, vector bool char);
9842 vector unsigned char vec_max (vector unsigned char,
9843 vector unsigned char);
9844 vector signed char vec_max (vector bool char, vector signed char);
9845 vector signed char vec_max (vector signed char, vector bool char);
9846 vector signed char vec_max (vector signed char, vector signed char);
9847 vector unsigned short vec_max (vector bool short,
9848 vector unsigned short);
9849 vector unsigned short vec_max (vector unsigned short,
9851 vector unsigned short vec_max (vector unsigned short,
9852 vector unsigned short);
9853 vector signed short vec_max (vector bool short, vector signed short);
9854 vector signed short vec_max (vector signed short, vector bool short);
9855 vector signed short vec_max (vector signed short, vector signed short);
9856 vector unsigned int vec_max (vector bool int, vector unsigned int);
9857 vector unsigned int vec_max (vector unsigned int, vector bool int);
9858 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
9859 vector signed int vec_max (vector bool int, vector signed int);
9860 vector signed int vec_max (vector signed int, vector bool int);
9861 vector signed int vec_max (vector signed int, vector signed int);
9862 vector float vec_max (vector float, vector float);
9864 vector float vec_vmaxfp (vector float, vector float);
9866 vector signed int vec_vmaxsw (vector bool int, vector signed int);
9867 vector signed int vec_vmaxsw (vector signed int, vector bool int);
9868 vector signed int vec_vmaxsw (vector signed int, vector signed int);
9870 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
9871 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
9872 vector unsigned int vec_vmaxuw (vector unsigned int,
9873 vector unsigned int);
9875 vector signed short vec_vmaxsh (vector bool short, vector signed short);
9876 vector signed short vec_vmaxsh (vector signed short, vector bool short);
9877 vector signed short vec_vmaxsh (vector signed short,
9878 vector signed short);
9880 vector unsigned short vec_vmaxuh (vector bool short,
9881 vector unsigned short);
9882 vector unsigned short vec_vmaxuh (vector unsigned short,
9884 vector unsigned short vec_vmaxuh (vector unsigned short,
9885 vector unsigned short);
9887 vector signed char vec_vmaxsb (vector bool char, vector signed char);
9888 vector signed char vec_vmaxsb (vector signed char, vector bool char);
9889 vector signed char vec_vmaxsb (vector signed char, vector signed char);
9891 vector unsigned char vec_vmaxub (vector bool char,
9892 vector unsigned char);
9893 vector unsigned char vec_vmaxub (vector unsigned char,
9895 vector unsigned char vec_vmaxub (vector unsigned char,
9896 vector unsigned char);
9898 vector bool char vec_mergeh (vector bool char, vector bool char);
9899 vector signed char vec_mergeh (vector signed char, vector signed char);
9900 vector unsigned char vec_mergeh (vector unsigned char,
9901 vector unsigned char);
9902 vector bool short vec_mergeh (vector bool short, vector bool short);
9903 vector pixel vec_mergeh (vector pixel, vector pixel);
9904 vector signed short vec_mergeh (vector signed short,
9905 vector signed short);
9906 vector unsigned short vec_mergeh (vector unsigned short,
9907 vector unsigned short);
9908 vector float vec_mergeh (vector float, vector float);
9909 vector bool int vec_mergeh (vector bool int, vector bool int);
9910 vector signed int vec_mergeh (vector signed int, vector signed int);
9911 vector unsigned int vec_mergeh (vector unsigned int,
9912 vector unsigned int);
9914 vector float vec_vmrghw (vector float, vector float);
9915 vector bool int vec_vmrghw (vector bool int, vector bool int);
9916 vector signed int vec_vmrghw (vector signed int, vector signed int);
9917 vector unsigned int vec_vmrghw (vector unsigned int,
9918 vector unsigned int);
9920 vector bool short vec_vmrghh (vector bool short, vector bool short);
9921 vector signed short vec_vmrghh (vector signed short,
9922 vector signed short);
9923 vector unsigned short vec_vmrghh (vector unsigned short,
9924 vector unsigned short);
9925 vector pixel vec_vmrghh (vector pixel, vector pixel);
9927 vector bool char vec_vmrghb (vector bool char, vector bool char);
9928 vector signed char vec_vmrghb (vector signed char, vector signed char);
9929 vector unsigned char vec_vmrghb (vector unsigned char,
9930 vector unsigned char);
9932 vector bool char vec_mergel (vector bool char, vector bool char);
9933 vector signed char vec_mergel (vector signed char, vector signed char);
9934 vector unsigned char vec_mergel (vector unsigned char,
9935 vector unsigned char);
9936 vector bool short vec_mergel (vector bool short, vector bool short);
9937 vector pixel vec_mergel (vector pixel, vector pixel);
9938 vector signed short vec_mergel (vector signed short,
9939 vector signed short);
9940 vector unsigned short vec_mergel (vector unsigned short,
9941 vector unsigned short);
9942 vector float vec_mergel (vector float, vector float);
9943 vector bool int vec_mergel (vector bool int, vector bool int);
9944 vector signed int vec_mergel (vector signed int, vector signed int);
9945 vector unsigned int vec_mergel (vector unsigned int,
9946 vector unsigned int);
9948 vector float vec_vmrglw (vector float, vector float);
9949 vector signed int vec_vmrglw (vector signed int, vector signed int);
9950 vector unsigned int vec_vmrglw (vector unsigned int,
9951 vector unsigned int);
9952 vector bool int vec_vmrglw (vector bool int, vector bool int);
9954 vector bool short vec_vmrglh (vector bool short, vector bool short);
9955 vector signed short vec_vmrglh (vector signed short,
9956 vector signed short);
9957 vector unsigned short vec_vmrglh (vector unsigned short,
9958 vector unsigned short);
9959 vector pixel vec_vmrglh (vector pixel, vector pixel);
9961 vector bool char vec_vmrglb (vector bool char, vector bool char);
9962 vector signed char vec_vmrglb (vector signed char, vector signed char);
9963 vector unsigned char vec_vmrglb (vector unsigned char,
9964 vector unsigned char);
9966 vector unsigned short vec_mfvscr (void);
9968 vector unsigned char vec_min (vector bool char, vector unsigned char);
9969 vector unsigned char vec_min (vector unsigned char, vector bool char);
9970 vector unsigned char vec_min (vector unsigned char,
9971 vector unsigned char);
9972 vector signed char vec_min (vector bool char, vector signed char);
9973 vector signed char vec_min (vector signed char, vector bool char);
9974 vector signed char vec_min (vector signed char, vector signed char);
9975 vector unsigned short vec_min (vector bool short,
9976 vector unsigned short);
9977 vector unsigned short vec_min (vector unsigned short,
9979 vector unsigned short vec_min (vector unsigned short,
9980 vector unsigned short);
9981 vector signed short vec_min (vector bool short, vector signed short);
9982 vector signed short vec_min (vector signed short, vector bool short);
9983 vector signed short vec_min (vector signed short, vector signed short);
9984 vector unsigned int vec_min (vector bool int, vector unsigned int);
9985 vector unsigned int vec_min (vector unsigned int, vector bool int);
9986 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
9987 vector signed int vec_min (vector bool int, vector signed int);
9988 vector signed int vec_min (vector signed int, vector bool int);
9989 vector signed int vec_min (vector signed int, vector signed int);
9990 vector float vec_min (vector float, vector float);
9992 vector float vec_vminfp (vector float, vector float);
9994 vector signed int vec_vminsw (vector bool int, vector signed int);
9995 vector signed int vec_vminsw (vector signed int, vector bool int);
9996 vector signed int vec_vminsw (vector signed int, vector signed int);
9998 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
9999 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
10000 vector unsigned int vec_vminuw (vector unsigned int,
10001 vector unsigned int);
10003 vector signed short vec_vminsh (vector bool short, vector signed short);
10004 vector signed short vec_vminsh (vector signed short, vector bool short);
10005 vector signed short vec_vminsh (vector signed short,
10006 vector signed short);
10008 vector unsigned short vec_vminuh (vector bool short,
10009 vector unsigned short);
10010 vector unsigned short vec_vminuh (vector unsigned short,
10011 vector bool short);
10012 vector unsigned short vec_vminuh (vector unsigned short,
10013 vector unsigned short);
10015 vector signed char vec_vminsb (vector bool char, vector signed char);
10016 vector signed char vec_vminsb (vector signed char, vector bool char);
10017 vector signed char vec_vminsb (vector signed char, vector signed char);
10019 vector unsigned char vec_vminub (vector bool char,
10020 vector unsigned char);
10021 vector unsigned char vec_vminub (vector unsigned char,
10023 vector unsigned char vec_vminub (vector unsigned char,
10024 vector unsigned char);
10026 vector signed short vec_mladd (vector signed short,
10027 vector signed short,
10028 vector signed short);
10029 vector signed short vec_mladd (vector signed short,
10030 vector unsigned short,
10031 vector unsigned short);
10032 vector signed short vec_mladd (vector unsigned short,
10033 vector signed short,
10034 vector signed short);
10035 vector unsigned short vec_mladd (vector unsigned short,
10036 vector unsigned short,
10037 vector unsigned short);
10039 vector signed short vec_mradds (vector signed short,
10040 vector signed short,
10041 vector signed short);
10043 vector unsigned int vec_msum (vector unsigned char,
10044 vector unsigned char,
10045 vector unsigned int);
10046 vector signed int vec_msum (vector signed char,
10047 vector unsigned char,
10048 vector signed int);
10049 vector unsigned int vec_msum (vector unsigned short,
10050 vector unsigned short,
10051 vector unsigned int);
10052 vector signed int vec_msum (vector signed short,
10053 vector signed short,
10054 vector signed int);
10056 vector signed int vec_vmsumshm (vector signed short,
10057 vector signed short,
10058 vector signed int);
10060 vector unsigned int vec_vmsumuhm (vector unsigned short,
10061 vector unsigned short,
10062 vector unsigned int);
10064 vector signed int vec_vmsummbm (vector signed char,
10065 vector unsigned char,
10066 vector signed int);
10068 vector unsigned int vec_vmsumubm (vector unsigned char,
10069 vector unsigned char,
10070 vector unsigned int);
10072 vector unsigned int vec_msums (vector unsigned short,
10073 vector unsigned short,
10074 vector unsigned int);
10075 vector signed int vec_msums (vector signed short,
10076 vector signed short,
10077 vector signed int);
10079 vector signed int vec_vmsumshs (vector signed short,
10080 vector signed short,
10081 vector signed int);
10083 vector unsigned int vec_vmsumuhs (vector unsigned short,
10084 vector unsigned short,
10085 vector unsigned int);
10087 void vec_mtvscr (vector signed int);
10088 void vec_mtvscr (vector unsigned int);
10089 void vec_mtvscr (vector bool int);
10090 void vec_mtvscr (vector signed short);
10091 void vec_mtvscr (vector unsigned short);
10092 void vec_mtvscr (vector bool short);
10093 void vec_mtvscr (vector pixel);
10094 void vec_mtvscr (vector signed char);
10095 void vec_mtvscr (vector unsigned char);
10096 void vec_mtvscr (vector bool char);
10098 vector unsigned short vec_mule (vector unsigned char,
10099 vector unsigned char);
10100 vector signed short vec_mule (vector signed char,
10101 vector signed char);
10102 vector unsigned int vec_mule (vector unsigned short,
10103 vector unsigned short);
10104 vector signed int vec_mule (vector signed short, vector signed short);
10106 vector signed int vec_vmulesh (vector signed short,
10107 vector signed short);
10109 vector unsigned int vec_vmuleuh (vector unsigned short,
10110 vector unsigned short);
10112 vector signed short vec_vmulesb (vector signed char,
10113 vector signed char);
10115 vector unsigned short vec_vmuleub (vector unsigned char,
10116 vector unsigned char);
10118 vector unsigned short vec_mulo (vector unsigned char,
10119 vector unsigned char);
10120 vector signed short vec_mulo (vector signed char, vector signed char);
10121 vector unsigned int vec_mulo (vector unsigned short,
10122 vector unsigned short);
10123 vector signed int vec_mulo (vector signed short, vector signed short);
10125 vector signed int vec_vmulosh (vector signed short,
10126 vector signed short);
10128 vector unsigned int vec_vmulouh (vector unsigned short,
10129 vector unsigned short);
10131 vector signed short vec_vmulosb (vector signed char,
10132 vector signed char);
10134 vector unsigned short vec_vmuloub (vector unsigned char,
10135 vector unsigned char);
10137 vector float vec_nmsub (vector float, vector float, vector float);
10139 vector float vec_nor (vector float, vector float);
10140 vector signed int vec_nor (vector signed int, vector signed int);
10141 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
10142 vector bool int vec_nor (vector bool int, vector bool int);
10143 vector signed short vec_nor (vector signed short, vector signed short);
10144 vector unsigned short vec_nor (vector unsigned short,
10145 vector unsigned short);
10146 vector bool short vec_nor (vector bool short, vector bool short);
10147 vector signed char vec_nor (vector signed char, vector signed char);
10148 vector unsigned char vec_nor (vector unsigned char,
10149 vector unsigned char);
10150 vector bool char vec_nor (vector bool char, vector bool char);
10152 vector float vec_or (vector float, vector float);
10153 vector float vec_or (vector float, vector bool int);
10154 vector float vec_or (vector bool int, vector float);
10155 vector bool int vec_or (vector bool int, vector bool int);
10156 vector signed int vec_or (vector bool int, vector signed int);
10157 vector signed int vec_or (vector signed int, vector bool int);
10158 vector signed int vec_or (vector signed int, vector signed int);
10159 vector unsigned int vec_or (vector bool int, vector unsigned int);
10160 vector unsigned int vec_or (vector unsigned int, vector bool int);
10161 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
10162 vector bool short vec_or (vector bool short, vector bool short);
10163 vector signed short vec_or (vector bool short, vector signed short);
10164 vector signed short vec_or (vector signed short, vector bool short);
10165 vector signed short vec_or (vector signed short, vector signed short);
10166 vector unsigned short vec_or (vector bool short, vector unsigned short);
10167 vector unsigned short vec_or (vector unsigned short, vector bool short);
10168 vector unsigned short vec_or (vector unsigned short,
10169 vector unsigned short);
10170 vector signed char vec_or (vector bool char, vector signed char);
10171 vector bool char vec_or (vector bool char, vector bool char);
10172 vector signed char vec_or (vector signed char, vector bool char);
10173 vector signed char vec_or (vector signed char, vector signed char);
10174 vector unsigned char vec_or (vector bool char, vector unsigned char);
10175 vector unsigned char vec_or (vector unsigned char, vector bool char);
10176 vector unsigned char vec_or (vector unsigned char,
10177 vector unsigned char);
10179 vector signed char vec_pack (vector signed short, vector signed short);
10180 vector unsigned char vec_pack (vector unsigned short,
10181 vector unsigned short);
10182 vector bool char vec_pack (vector bool short, vector bool short);
10183 vector signed short vec_pack (vector signed int, vector signed int);
10184 vector unsigned short vec_pack (vector unsigned int,
10185 vector unsigned int);
10186 vector bool short vec_pack (vector bool int, vector bool int);
10188 vector bool short vec_vpkuwum (vector bool int, vector bool int);
10189 vector signed short vec_vpkuwum (vector signed int, vector signed int);
10190 vector unsigned short vec_vpkuwum (vector unsigned int,
10191 vector unsigned int);
10193 vector bool char vec_vpkuhum (vector bool short, vector bool short);
10194 vector signed char vec_vpkuhum (vector signed short,
10195 vector signed short);
10196 vector unsigned char vec_vpkuhum (vector unsigned short,
10197 vector unsigned short);
10199 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
10201 vector unsigned char vec_packs (vector unsigned short,
10202 vector unsigned short);
10203 vector signed char vec_packs (vector signed short, vector signed short);
10204 vector unsigned short vec_packs (vector unsigned int,
10205 vector unsigned int);
10206 vector signed short vec_packs (vector signed int, vector signed int);
10208 vector signed short vec_vpkswss (vector signed int, vector signed int);
10210 vector unsigned short vec_vpkuwus (vector unsigned int,
10211 vector unsigned int);
10213 vector signed char vec_vpkshss (vector signed short,
10214 vector signed short);
10216 vector unsigned char vec_vpkuhus (vector unsigned short,
10217 vector unsigned short);
10219 vector unsigned char vec_packsu (vector unsigned short,
10220 vector unsigned short);
10221 vector unsigned char vec_packsu (vector signed short,
10222 vector signed short);
10223 vector unsigned short vec_packsu (vector unsigned int,
10224 vector unsigned int);
10225 vector unsigned short vec_packsu (vector signed int, vector signed int);
10227 vector unsigned short vec_vpkswus (vector signed int,
10228 vector signed int);
10230 vector unsigned char vec_vpkshus (vector signed short,
10231 vector signed short);
10233 vector float vec_perm (vector float,
10235 vector unsigned char);
10236 vector signed int vec_perm (vector signed int,
10238 vector unsigned char);
10239 vector unsigned int vec_perm (vector unsigned int,
10240 vector unsigned int,
10241 vector unsigned char);
10242 vector bool int vec_perm (vector bool int,
10244 vector unsigned char);
10245 vector signed short vec_perm (vector signed short,
10246 vector signed short,
10247 vector unsigned char);
10248 vector unsigned short vec_perm (vector unsigned short,
10249 vector unsigned short,
10250 vector unsigned char);
10251 vector bool short vec_perm (vector bool short,
10253 vector unsigned char);
10254 vector pixel vec_perm (vector pixel,
10256 vector unsigned char);
10257 vector signed char vec_perm (vector signed char,
10258 vector signed char,
10259 vector unsigned char);
10260 vector unsigned char vec_perm (vector unsigned char,
10261 vector unsigned char,
10262 vector unsigned char);
10263 vector bool char vec_perm (vector bool char,
10265 vector unsigned char);
10267 vector float vec_re (vector float);
10269 vector signed char vec_rl (vector signed char,
10270 vector unsigned char);
10271 vector unsigned char vec_rl (vector unsigned char,
10272 vector unsigned char);
10273 vector signed short vec_rl (vector signed short, vector unsigned short);
10274 vector unsigned short vec_rl (vector unsigned short,
10275 vector unsigned short);
10276 vector signed int vec_rl (vector signed int, vector unsigned int);
10277 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
10279 vector signed int vec_vrlw (vector signed int, vector unsigned int);
10280 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
10282 vector signed short vec_vrlh (vector signed short,
10283 vector unsigned short);
10284 vector unsigned short vec_vrlh (vector unsigned short,
10285 vector unsigned short);
10287 vector signed char vec_vrlb (vector signed char, vector unsigned char);
10288 vector unsigned char vec_vrlb (vector unsigned char,
10289 vector unsigned char);
10291 vector float vec_round (vector float);
10293 vector float vec_rsqrte (vector float);
10295 vector float vec_sel (vector float, vector float, vector bool int);
10296 vector float vec_sel (vector float, vector float, vector unsigned int);
10297 vector signed int vec_sel (vector signed int,
10300 vector signed int vec_sel (vector signed int,
10302 vector unsigned int);
10303 vector unsigned int vec_sel (vector unsigned int,
10304 vector unsigned int,
10306 vector unsigned int vec_sel (vector unsigned int,
10307 vector unsigned int,
10308 vector unsigned int);
10309 vector bool int vec_sel (vector bool int,
10312 vector bool int vec_sel (vector bool int,
10314 vector unsigned int);
10315 vector signed short vec_sel (vector signed short,
10316 vector signed short,
10317 vector bool short);
10318 vector signed short vec_sel (vector signed short,
10319 vector signed short,
10320 vector unsigned short);
10321 vector unsigned short vec_sel (vector unsigned short,
10322 vector unsigned short,
10323 vector bool short);
10324 vector unsigned short vec_sel (vector unsigned short,
10325 vector unsigned short,
10326 vector unsigned short);
10327 vector bool short vec_sel (vector bool short,
10329 vector bool short);
10330 vector bool short vec_sel (vector bool short,
10332 vector unsigned short);
10333 vector signed char vec_sel (vector signed char,
10334 vector signed char,
10336 vector signed char vec_sel (vector signed char,
10337 vector signed char,
10338 vector unsigned char);
10339 vector unsigned char vec_sel (vector unsigned char,
10340 vector unsigned char,
10342 vector unsigned char vec_sel (vector unsigned char,
10343 vector unsigned char,
10344 vector unsigned char);
10345 vector bool char vec_sel (vector bool char,
10348 vector bool char vec_sel (vector bool char,
10350 vector unsigned char);
10352 vector signed char vec_sl (vector signed char,
10353 vector unsigned char);
10354 vector unsigned char vec_sl (vector unsigned char,
10355 vector unsigned char);
10356 vector signed short vec_sl (vector signed short, vector unsigned short);
10357 vector unsigned short vec_sl (vector unsigned short,
10358 vector unsigned short);
10359 vector signed int vec_sl (vector signed int, vector unsigned int);
10360 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
10362 vector signed int vec_vslw (vector signed int, vector unsigned int);
10363 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
10365 vector signed short vec_vslh (vector signed short,
10366 vector unsigned short);
10367 vector unsigned short vec_vslh (vector unsigned short,
10368 vector unsigned short);
10370 vector signed char vec_vslb (vector signed char, vector unsigned char);
10371 vector unsigned char vec_vslb (vector unsigned char,
10372 vector unsigned char);
10374 vector float vec_sld (vector float, vector float, const int);
10375 vector signed int vec_sld (vector signed int,
10378 vector unsigned int vec_sld (vector unsigned int,
10379 vector unsigned int,
10381 vector bool int vec_sld (vector bool int,
10384 vector signed short vec_sld (vector signed short,
10385 vector signed short,
10387 vector unsigned short vec_sld (vector unsigned short,
10388 vector unsigned short,
10390 vector bool short vec_sld (vector bool short,
10393 vector pixel vec_sld (vector pixel,
10396 vector signed char vec_sld (vector signed char,
10397 vector signed char,
10399 vector unsigned char vec_sld (vector unsigned char,
10400 vector unsigned char,
10402 vector bool char vec_sld (vector bool char,
10406 vector signed int vec_sll (vector signed int,
10407 vector unsigned int);
10408 vector signed int vec_sll (vector signed int,
10409 vector unsigned short);
10410 vector signed int vec_sll (vector signed int,
10411 vector unsigned char);
10412 vector unsigned int vec_sll (vector unsigned int,
10413 vector unsigned int);
10414 vector unsigned int vec_sll (vector unsigned int,
10415 vector unsigned short);
10416 vector unsigned int vec_sll (vector unsigned int,
10417 vector unsigned char);
10418 vector bool int vec_sll (vector bool int,
10419 vector unsigned int);
10420 vector bool int vec_sll (vector bool int,
10421 vector unsigned short);
10422 vector bool int vec_sll (vector bool int,
10423 vector unsigned char);
10424 vector signed short vec_sll (vector signed short,
10425 vector unsigned int);
10426 vector signed short vec_sll (vector signed short,
10427 vector unsigned short);
10428 vector signed short vec_sll (vector signed short,
10429 vector unsigned char);
10430 vector unsigned short vec_sll (vector unsigned short,
10431 vector unsigned int);
10432 vector unsigned short vec_sll (vector unsigned short,
10433 vector unsigned short);
10434 vector unsigned short vec_sll (vector unsigned short,
10435 vector unsigned char);
10436 vector bool short vec_sll (vector bool short, vector unsigned int);
10437 vector bool short vec_sll (vector bool short, vector unsigned short);
10438 vector bool short vec_sll (vector bool short, vector unsigned char);
10439 vector pixel vec_sll (vector pixel, vector unsigned int);
10440 vector pixel vec_sll (vector pixel, vector unsigned short);
10441 vector pixel vec_sll (vector pixel, vector unsigned char);
10442 vector signed char vec_sll (vector signed char, vector unsigned int);
10443 vector signed char vec_sll (vector signed char, vector unsigned short);
10444 vector signed char vec_sll (vector signed char, vector unsigned char);
10445 vector unsigned char vec_sll (vector unsigned char,
10446 vector unsigned int);
10447 vector unsigned char vec_sll (vector unsigned char,
10448 vector unsigned short);
10449 vector unsigned char vec_sll (vector unsigned char,
10450 vector unsigned char);
10451 vector bool char vec_sll (vector bool char, vector unsigned int);
10452 vector bool char vec_sll (vector bool char, vector unsigned short);
10453 vector bool char vec_sll (vector bool char, vector unsigned char);
10455 vector float vec_slo (vector float, vector signed char);
10456 vector float vec_slo (vector float, vector unsigned char);
10457 vector signed int vec_slo (vector signed int, vector signed char);
10458 vector signed int vec_slo (vector signed int, vector unsigned char);
10459 vector unsigned int vec_slo (vector unsigned int, vector signed char);
10460 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
10461 vector signed short vec_slo (vector signed short, vector signed char);
10462 vector signed short vec_slo (vector signed short, vector unsigned char);
10463 vector unsigned short vec_slo (vector unsigned short,
10464 vector signed char);
10465 vector unsigned short vec_slo (vector unsigned short,
10466 vector unsigned char);
10467 vector pixel vec_slo (vector pixel, vector signed char);
10468 vector pixel vec_slo (vector pixel, vector unsigned char);
10469 vector signed char vec_slo (vector signed char, vector signed char);
10470 vector signed char vec_slo (vector signed char, vector unsigned char);
10471 vector unsigned char vec_slo (vector unsigned char, vector signed char);
10472 vector unsigned char vec_slo (vector unsigned char,
10473 vector unsigned char);
10475 vector signed char vec_splat (vector signed char, const int);
10476 vector unsigned char vec_splat (vector unsigned char, const int);
10477 vector bool char vec_splat (vector bool char, const int);
10478 vector signed short vec_splat (vector signed short, const int);
10479 vector unsigned short vec_splat (vector unsigned short, const int);
10480 vector bool short vec_splat (vector bool short, const int);
10481 vector pixel vec_splat (vector pixel, const int);
10482 vector float vec_splat (vector float, const int);
10483 vector signed int vec_splat (vector signed int, const int);
10484 vector unsigned int vec_splat (vector unsigned int, const int);
10485 vector bool int vec_splat (vector bool int, const int);
10487 vector float vec_vspltw (vector float, const int);
10488 vector signed int vec_vspltw (vector signed int, const int);
10489 vector unsigned int vec_vspltw (vector unsigned int, const int);
10490 vector bool int vec_vspltw (vector bool int, const int);
10492 vector bool short vec_vsplth (vector bool short, const int);
10493 vector signed short vec_vsplth (vector signed short, const int);
10494 vector unsigned short vec_vsplth (vector unsigned short, const int);
10495 vector pixel vec_vsplth (vector pixel, const int);
10497 vector signed char vec_vspltb (vector signed char, const int);
10498 vector unsigned char vec_vspltb (vector unsigned char, const int);
10499 vector bool char vec_vspltb (vector bool char, const int);
10501 vector signed char vec_splat_s8 (const int);
10503 vector signed short vec_splat_s16 (const int);
10505 vector signed int vec_splat_s32 (const int);
10507 vector unsigned char vec_splat_u8 (const int);
10509 vector unsigned short vec_splat_u16 (const int);
10511 vector unsigned int vec_splat_u32 (const int);
10513 vector signed char vec_sr (vector signed char, vector unsigned char);
10514 vector unsigned char vec_sr (vector unsigned char,
10515 vector unsigned char);
10516 vector signed short vec_sr (vector signed short,
10517 vector unsigned short);
10518 vector unsigned short vec_sr (vector unsigned short,
10519 vector unsigned short);
10520 vector signed int vec_sr (vector signed int, vector unsigned int);
10521 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
10523 vector signed int vec_vsrw (vector signed int, vector unsigned int);
10524 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
10526 vector signed short vec_vsrh (vector signed short,
10527 vector unsigned short);
10528 vector unsigned short vec_vsrh (vector unsigned short,
10529 vector unsigned short);
10531 vector signed char vec_vsrb (vector signed char, vector unsigned char);
10532 vector unsigned char vec_vsrb (vector unsigned char,
10533 vector unsigned char);
10535 vector signed char vec_sra (vector signed char, vector unsigned char);
10536 vector unsigned char vec_sra (vector unsigned char,
10537 vector unsigned char);
10538 vector signed short vec_sra (vector signed short,
10539 vector unsigned short);
10540 vector unsigned short vec_sra (vector unsigned short,
10541 vector unsigned short);
10542 vector signed int vec_sra (vector signed int, vector unsigned int);
10543 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
10545 vector signed int vec_vsraw (vector signed int, vector unsigned int);
10546 vector unsigned int vec_vsraw (vector unsigned int,
10547 vector unsigned int);
10549 vector signed short vec_vsrah (vector signed short,
10550 vector unsigned short);
10551 vector unsigned short vec_vsrah (vector unsigned short,
10552 vector unsigned short);
10554 vector signed char vec_vsrab (vector signed char, vector unsigned char);
10555 vector unsigned char vec_vsrab (vector unsigned char,
10556 vector unsigned char);
10558 vector signed int vec_srl (vector signed int, vector unsigned int);
10559 vector signed int vec_srl (vector signed int, vector unsigned short);
10560 vector signed int vec_srl (vector signed int, vector unsigned char);
10561 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
10562 vector unsigned int vec_srl (vector unsigned int,
10563 vector unsigned short);
10564 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
10565 vector bool int vec_srl (vector bool int, vector unsigned int);
10566 vector bool int vec_srl (vector bool int, vector unsigned short);
10567 vector bool int vec_srl (vector bool int, vector unsigned char);
10568 vector signed short vec_srl (vector signed short, vector unsigned int);
10569 vector signed short vec_srl (vector signed short,
10570 vector unsigned short);
10571 vector signed short vec_srl (vector signed short, vector unsigned char);
10572 vector unsigned short vec_srl (vector unsigned short,
10573 vector unsigned int);
10574 vector unsigned short vec_srl (vector unsigned short,
10575 vector unsigned short);
10576 vector unsigned short vec_srl (vector unsigned short,
10577 vector unsigned char);
10578 vector bool short vec_srl (vector bool short, vector unsigned int);
10579 vector bool short vec_srl (vector bool short, vector unsigned short);
10580 vector bool short vec_srl (vector bool short, vector unsigned char);
10581 vector pixel vec_srl (vector pixel, vector unsigned int);
10582 vector pixel vec_srl (vector pixel, vector unsigned short);
10583 vector pixel vec_srl (vector pixel, vector unsigned char);
10584 vector signed char vec_srl (vector signed char, vector unsigned int);
10585 vector signed char vec_srl (vector signed char, vector unsigned short);
10586 vector signed char vec_srl (vector signed char, vector unsigned char);
10587 vector unsigned char vec_srl (vector unsigned char,
10588 vector unsigned int);
10589 vector unsigned char vec_srl (vector unsigned char,
10590 vector unsigned short);
10591 vector unsigned char vec_srl (vector unsigned char,
10592 vector unsigned char);
10593 vector bool char vec_srl (vector bool char, vector unsigned int);
10594 vector bool char vec_srl (vector bool char, vector unsigned short);
10595 vector bool char vec_srl (vector bool char, vector unsigned char);
10597 vector float vec_sro (vector float, vector signed char);
10598 vector float vec_sro (vector float, vector unsigned char);
10599 vector signed int vec_sro (vector signed int, vector signed char);
10600 vector signed int vec_sro (vector signed int, vector unsigned char);
10601 vector unsigned int vec_sro (vector unsigned int, vector signed char);
10602 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
10603 vector signed short vec_sro (vector signed short, vector signed char);
10604 vector signed short vec_sro (vector signed short, vector unsigned char);
10605 vector unsigned short vec_sro (vector unsigned short,
10606 vector signed char);
10607 vector unsigned short vec_sro (vector unsigned short,
10608 vector unsigned char);
10609 vector pixel vec_sro (vector pixel, vector signed char);
10610 vector pixel vec_sro (vector pixel, vector unsigned char);
10611 vector signed char vec_sro (vector signed char, vector signed char);
10612 vector signed char vec_sro (vector signed char, vector unsigned char);
10613 vector unsigned char vec_sro (vector unsigned char, vector signed char);
10614 vector unsigned char vec_sro (vector unsigned char,
10615 vector unsigned char);
10617 void vec_st (vector float, int, vector float *);
10618 void vec_st (vector float, int, float *);
10619 void vec_st (vector signed int, int, vector signed int *);
10620 void vec_st (vector signed int, int, int *);
10621 void vec_st (vector unsigned int, int, vector unsigned int *);
10622 void vec_st (vector unsigned int, int, unsigned int *);
10623 void vec_st (vector bool int, int, vector bool int *);
10624 void vec_st (vector bool int, int, unsigned int *);
10625 void vec_st (vector bool int, int, int *);
10626 void vec_st (vector signed short, int, vector signed short *);
10627 void vec_st (vector signed short, int, short *);
10628 void vec_st (vector unsigned short, int, vector unsigned short *);
10629 void vec_st (vector unsigned short, int, unsigned short *);
10630 void vec_st (vector bool short, int, vector bool short *);
10631 void vec_st (vector bool short, int, unsigned short *);
10632 void vec_st (vector pixel, int, vector pixel *);
10633 void vec_st (vector pixel, int, unsigned short *);
10634 void vec_st (vector pixel, int, short *);
10635 void vec_st (vector bool short, int, short *);
10636 void vec_st (vector signed char, int, vector signed char *);
10637 void vec_st (vector signed char, int, signed char *);
10638 void vec_st (vector unsigned char, int, vector unsigned char *);
10639 void vec_st (vector unsigned char, int, unsigned char *);
10640 void vec_st (vector bool char, int, vector bool char *);
10641 void vec_st (vector bool char, int, unsigned char *);
10642 void vec_st (vector bool char, int, signed char *);
10644 void vec_ste (vector signed char, int, signed char *);
10645 void vec_ste (vector unsigned char, int, unsigned char *);
10646 void vec_ste (vector bool char, int, signed char *);
10647 void vec_ste (vector bool char, int, unsigned char *);
10648 void vec_ste (vector signed short, int, short *);
10649 void vec_ste (vector unsigned short, int, unsigned short *);
10650 void vec_ste (vector bool short, int, short *);
10651 void vec_ste (vector bool short, int, unsigned short *);
10652 void vec_ste (vector pixel, int, short *);
10653 void vec_ste (vector pixel, int, unsigned short *);
10654 void vec_ste (vector float, int, float *);
10655 void vec_ste (vector signed int, int, int *);
10656 void vec_ste (vector unsigned int, int, unsigned int *);
10657 void vec_ste (vector bool int, int, int *);
10658 void vec_ste (vector bool int, int, unsigned int *);
10660 void vec_stvewx (vector float, int, float *);
10661 void vec_stvewx (vector signed int, int, int *);
10662 void vec_stvewx (vector unsigned int, int, unsigned int *);
10663 void vec_stvewx (vector bool int, int, int *);
10664 void vec_stvewx (vector bool int, int, unsigned int *);
10666 void vec_stvehx (vector signed short, int, short *);
10667 void vec_stvehx (vector unsigned short, int, unsigned short *);
10668 void vec_stvehx (vector bool short, int, short *);
10669 void vec_stvehx (vector bool short, int, unsigned short *);
10670 void vec_stvehx (vector pixel, int, short *);
10671 void vec_stvehx (vector pixel, int, unsigned short *);
10673 void vec_stvebx (vector signed char, int, signed char *);
10674 void vec_stvebx (vector unsigned char, int, unsigned char *);
10675 void vec_stvebx (vector bool char, int, signed char *);
10676 void vec_stvebx (vector bool char, int, unsigned char *);
10678 void vec_stl (vector float, int, vector float *);
10679 void vec_stl (vector float, int, float *);
10680 void vec_stl (vector signed int, int, vector signed int *);
10681 void vec_stl (vector signed int, int, int *);
10682 void vec_stl (vector unsigned int, int, vector unsigned int *);
10683 void vec_stl (vector unsigned int, int, unsigned int *);
10684 void vec_stl (vector bool int, int, vector bool int *);
10685 void vec_stl (vector bool int, int, unsigned int *);
10686 void vec_stl (vector bool int, int, int *);
10687 void vec_stl (vector signed short, int, vector signed short *);
10688 void vec_stl (vector signed short, int, short *);
10689 void vec_stl (vector unsigned short, int, vector unsigned short *);
10690 void vec_stl (vector unsigned short, int, unsigned short *);
10691 void vec_stl (vector bool short, int, vector bool short *);
10692 void vec_stl (vector bool short, int, unsigned short *);
10693 void vec_stl (vector bool short, int, short *);
10694 void vec_stl (vector pixel, int, vector pixel *);
10695 void vec_stl (vector pixel, int, unsigned short *);
10696 void vec_stl (vector pixel, int, short *);
10697 void vec_stl (vector signed char, int, vector signed char *);
10698 void vec_stl (vector signed char, int, signed char *);
10699 void vec_stl (vector unsigned char, int, vector unsigned char *);
10700 void vec_stl (vector unsigned char, int, unsigned char *);
10701 void vec_stl (vector bool char, int, vector bool char *);
10702 void vec_stl (vector bool char, int, unsigned char *);
10703 void vec_stl (vector bool char, int, signed char *);
10705 vector signed char vec_sub (vector bool char, vector signed char);
10706 vector signed char vec_sub (vector signed char, vector bool char);
10707 vector signed char vec_sub (vector signed char, vector signed char);
10708 vector unsigned char vec_sub (vector bool char, vector unsigned char);
10709 vector unsigned char vec_sub (vector unsigned char, vector bool char);
10710 vector unsigned char vec_sub (vector unsigned char,
10711 vector unsigned char);
10712 vector signed short vec_sub (vector bool short, vector signed short);
10713 vector signed short vec_sub (vector signed short, vector bool short);
10714 vector signed short vec_sub (vector signed short, vector signed short);
10715 vector unsigned short vec_sub (vector bool short,
10716 vector unsigned short);
10717 vector unsigned short vec_sub (vector unsigned short,
10718 vector bool short);
10719 vector unsigned short vec_sub (vector unsigned short,
10720 vector unsigned short);
10721 vector signed int vec_sub (vector bool int, vector signed int);
10722 vector signed int vec_sub (vector signed int, vector bool int);
10723 vector signed int vec_sub (vector signed int, vector signed int);
10724 vector unsigned int vec_sub (vector bool int, vector unsigned int);
10725 vector unsigned int vec_sub (vector unsigned int, vector bool int);
10726 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
10727 vector float vec_sub (vector float, vector float);
10729 vector float vec_vsubfp (vector float, vector float);
10731 vector signed int vec_vsubuwm (vector bool int, vector signed int);
10732 vector signed int vec_vsubuwm (vector signed int, vector bool int);
10733 vector signed int vec_vsubuwm (vector signed int, vector signed int);
10734 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
10735 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
10736 vector unsigned int vec_vsubuwm (vector unsigned int,
10737 vector unsigned int);
10739 vector signed short vec_vsubuhm (vector bool short,
10740 vector signed short);
10741 vector signed short vec_vsubuhm (vector signed short,
10742 vector bool short);
10743 vector signed short vec_vsubuhm (vector signed short,
10744 vector signed short);
10745 vector unsigned short vec_vsubuhm (vector bool short,
10746 vector unsigned short);
10747 vector unsigned short vec_vsubuhm (vector unsigned short,
10748 vector bool short);
10749 vector unsigned short vec_vsubuhm (vector unsigned short,
10750 vector unsigned short);
10752 vector signed char vec_vsububm (vector bool char, vector signed char);
10753 vector signed char vec_vsububm (vector signed char, vector bool char);
10754 vector signed char vec_vsububm (vector signed char, vector signed char);
10755 vector unsigned char vec_vsububm (vector bool char,
10756 vector unsigned char);
10757 vector unsigned char vec_vsububm (vector unsigned char,
10759 vector unsigned char vec_vsububm (vector unsigned char,
10760 vector unsigned char);
10762 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
10764 vector unsigned char vec_subs (vector bool char, vector unsigned char);
10765 vector unsigned char vec_subs (vector unsigned char, vector bool char);
10766 vector unsigned char vec_subs (vector unsigned char,
10767 vector unsigned char);
10768 vector signed char vec_subs (vector bool char, vector signed char);
10769 vector signed char vec_subs (vector signed char, vector bool char);
10770 vector signed char vec_subs (vector signed char, vector signed char);
10771 vector unsigned short vec_subs (vector bool short,
10772 vector unsigned short);
10773 vector unsigned short vec_subs (vector unsigned short,
10774 vector bool short);
10775 vector unsigned short vec_subs (vector unsigned short,
10776 vector unsigned short);
10777 vector signed short vec_subs (vector bool short, vector signed short);
10778 vector signed short vec_subs (vector signed short, vector bool short);
10779 vector signed short vec_subs (vector signed short, vector signed short);
10780 vector unsigned int vec_subs (vector bool int, vector unsigned int);
10781 vector unsigned int vec_subs (vector unsigned int, vector bool int);
10782 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
10783 vector signed int vec_subs (vector bool int, vector signed int);
10784 vector signed int vec_subs (vector signed int, vector bool int);
10785 vector signed int vec_subs (vector signed int, vector signed int);
10787 vector signed int vec_vsubsws (vector bool int, vector signed int);
10788 vector signed int vec_vsubsws (vector signed int, vector bool int);
10789 vector signed int vec_vsubsws (vector signed int, vector signed int);
10791 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
10792 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
10793 vector unsigned int vec_vsubuws (vector unsigned int,
10794 vector unsigned int);
10796 vector signed short vec_vsubshs (vector bool short,
10797 vector signed short);
10798 vector signed short vec_vsubshs (vector signed short,
10799 vector bool short);
10800 vector signed short vec_vsubshs (vector signed short,
10801 vector signed short);
10803 vector unsigned short vec_vsubuhs (vector bool short,
10804 vector unsigned short);
10805 vector unsigned short vec_vsubuhs (vector unsigned short,
10806 vector bool short);
10807 vector unsigned short vec_vsubuhs (vector unsigned short,
10808 vector unsigned short);
10810 vector signed char vec_vsubsbs (vector bool char, vector signed char);
10811 vector signed char vec_vsubsbs (vector signed char, vector bool char);
10812 vector signed char vec_vsubsbs (vector signed char, vector signed char);
10814 vector unsigned char vec_vsububs (vector bool char,
10815 vector unsigned char);
10816 vector unsigned char vec_vsububs (vector unsigned char,
10818 vector unsigned char vec_vsububs (vector unsigned char,
10819 vector unsigned char);
10821 vector unsigned int vec_sum4s (vector unsigned char,
10822 vector unsigned int);
10823 vector signed int vec_sum4s (vector signed char, vector signed int);
10824 vector signed int vec_sum4s (vector signed short, vector signed int);
10826 vector signed int vec_vsum4shs (vector signed short, vector signed int);
10828 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
10830 vector unsigned int vec_vsum4ubs (vector unsigned char,
10831 vector unsigned int);
10833 vector signed int vec_sum2s (vector signed int, vector signed int);
10835 vector signed int vec_sums (vector signed int, vector signed int);
10837 vector float vec_trunc (vector float);
10839 vector signed short vec_unpackh (vector signed char);
10840 vector bool short vec_unpackh (vector bool char);
10841 vector signed int vec_unpackh (vector signed short);
10842 vector bool int vec_unpackh (vector bool short);
10843 vector unsigned int vec_unpackh (vector pixel);
10845 vector bool int vec_vupkhsh (vector bool short);
10846 vector signed int vec_vupkhsh (vector signed short);
10848 vector unsigned int vec_vupkhpx (vector pixel);
10850 vector bool short vec_vupkhsb (vector bool char);
10851 vector signed short vec_vupkhsb (vector signed char);
10853 vector signed short vec_unpackl (vector signed char);
10854 vector bool short vec_unpackl (vector bool char);
10855 vector unsigned int vec_unpackl (vector pixel);
10856 vector signed int vec_unpackl (vector signed short);
10857 vector bool int vec_unpackl (vector bool short);
10859 vector unsigned int vec_vupklpx (vector pixel);
10861 vector bool int vec_vupklsh (vector bool short);
10862 vector signed int vec_vupklsh (vector signed short);
10864 vector bool short vec_vupklsb (vector bool char);
10865 vector signed short vec_vupklsb (vector signed char);
10867 vector float vec_xor (vector float, vector float);
10868 vector float vec_xor (vector float, vector bool int);
10869 vector float vec_xor (vector bool int, vector float);
10870 vector bool int vec_xor (vector bool int, vector bool int);
10871 vector signed int vec_xor (vector bool int, vector signed int);
10872 vector signed int vec_xor (vector signed int, vector bool int);
10873 vector signed int vec_xor (vector signed int, vector signed int);
10874 vector unsigned int vec_xor (vector bool int, vector unsigned int);
10875 vector unsigned int vec_xor (vector unsigned int, vector bool int);
10876 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
10877 vector bool short vec_xor (vector bool short, vector bool short);
10878 vector signed short vec_xor (vector bool short, vector signed short);
10879 vector signed short vec_xor (vector signed short, vector bool short);
10880 vector signed short vec_xor (vector signed short, vector signed short);
10881 vector unsigned short vec_xor (vector bool short,
10882 vector unsigned short);
10883 vector unsigned short vec_xor (vector unsigned short,
10884 vector bool short);
10885 vector unsigned short vec_xor (vector unsigned short,
10886 vector unsigned short);
10887 vector signed char vec_xor (vector bool char, vector signed char);
10888 vector bool char vec_xor (vector bool char, vector bool char);
10889 vector signed char vec_xor (vector signed char, vector bool char);
10890 vector signed char vec_xor (vector signed char, vector signed char);
10891 vector unsigned char vec_xor (vector bool char, vector unsigned char);
10892 vector unsigned char vec_xor (vector unsigned char, vector bool char);
10893 vector unsigned char vec_xor (vector unsigned char,
10894 vector unsigned char);
10896 int vec_all_eq (vector signed char, vector bool char);
10897 int vec_all_eq (vector signed char, vector signed char);
10898 int vec_all_eq (vector unsigned char, vector bool char);
10899 int vec_all_eq (vector unsigned char, vector unsigned char);
10900 int vec_all_eq (vector bool char, vector bool char);
10901 int vec_all_eq (vector bool char, vector unsigned char);
10902 int vec_all_eq (vector bool char, vector signed char);
10903 int vec_all_eq (vector signed short, vector bool short);
10904 int vec_all_eq (vector signed short, vector signed short);
10905 int vec_all_eq (vector unsigned short, vector bool short);
10906 int vec_all_eq (vector unsigned short, vector unsigned short);
10907 int vec_all_eq (vector bool short, vector bool short);
10908 int vec_all_eq (vector bool short, vector unsigned short);
10909 int vec_all_eq (vector bool short, vector signed short);
10910 int vec_all_eq (vector pixel, vector pixel);
10911 int vec_all_eq (vector signed int, vector bool int);
10912 int vec_all_eq (vector signed int, vector signed int);
10913 int vec_all_eq (vector unsigned int, vector bool int);
10914 int vec_all_eq (vector unsigned int, vector unsigned int);
10915 int vec_all_eq (vector bool int, vector bool int);
10916 int vec_all_eq (vector bool int, vector unsigned int);
10917 int vec_all_eq (vector bool int, vector signed int);
10918 int vec_all_eq (vector float, vector float);
10920 int vec_all_ge (vector bool char, vector unsigned char);
10921 int vec_all_ge (vector unsigned char, vector bool char);
10922 int vec_all_ge (vector unsigned char, vector unsigned char);
10923 int vec_all_ge (vector bool char, vector signed char);
10924 int vec_all_ge (vector signed char, vector bool char);
10925 int vec_all_ge (vector signed char, vector signed char);
10926 int vec_all_ge (vector bool short, vector unsigned short);
10927 int vec_all_ge (vector unsigned short, vector bool short);
10928 int vec_all_ge (vector unsigned short, vector unsigned short);
10929 int vec_all_ge (vector signed short, vector signed short);
10930 int vec_all_ge (vector bool short, vector signed short);
10931 int vec_all_ge (vector signed short, vector bool short);
10932 int vec_all_ge (vector bool int, vector unsigned int);
10933 int vec_all_ge (vector unsigned int, vector bool int);
10934 int vec_all_ge (vector unsigned int, vector unsigned int);
10935 int vec_all_ge (vector bool int, vector signed int);
10936 int vec_all_ge (vector signed int, vector bool int);
10937 int vec_all_ge (vector signed int, vector signed int);
10938 int vec_all_ge (vector float, vector float);
10940 int vec_all_gt (vector bool char, vector unsigned char);
10941 int vec_all_gt (vector unsigned char, vector bool char);
10942 int vec_all_gt (vector unsigned char, vector unsigned char);
10943 int vec_all_gt (vector bool char, vector signed char);
10944 int vec_all_gt (vector signed char, vector bool char);
10945 int vec_all_gt (vector signed char, vector signed char);
10946 int vec_all_gt (vector bool short, vector unsigned short);
10947 int vec_all_gt (vector unsigned short, vector bool short);
10948 int vec_all_gt (vector unsigned short, vector unsigned short);
10949 int vec_all_gt (vector bool short, vector signed short);
10950 int vec_all_gt (vector signed short, vector bool short);
10951 int vec_all_gt (vector signed short, vector signed short);
10952 int vec_all_gt (vector bool int, vector unsigned int);
10953 int vec_all_gt (vector unsigned int, vector bool int);
10954 int vec_all_gt (vector unsigned int, vector unsigned int);
10955 int vec_all_gt (vector bool int, vector signed int);
10956 int vec_all_gt (vector signed int, vector bool int);
10957 int vec_all_gt (vector signed int, vector signed int);
10958 int vec_all_gt (vector float, vector float);
10960 int vec_all_in (vector float, vector float);
10962 int vec_all_le (vector bool char, vector unsigned char);
10963 int vec_all_le (vector unsigned char, vector bool char);
10964 int vec_all_le (vector unsigned char, vector unsigned char);
10965 int vec_all_le (vector bool char, vector signed char);
10966 int vec_all_le (vector signed char, vector bool char);
10967 int vec_all_le (vector signed char, vector signed char);
10968 int vec_all_le (vector bool short, vector unsigned short);
10969 int vec_all_le (vector unsigned short, vector bool short);
10970 int vec_all_le (vector unsigned short, vector unsigned short);
10971 int vec_all_le (vector bool short, vector signed short);
10972 int vec_all_le (vector signed short, vector bool short);
10973 int vec_all_le (vector signed short, vector signed short);
10974 int vec_all_le (vector bool int, vector unsigned int);
10975 int vec_all_le (vector unsigned int, vector bool int);
10976 int vec_all_le (vector unsigned int, vector unsigned int);
10977 int vec_all_le (vector bool int, vector signed int);
10978 int vec_all_le (vector signed int, vector bool int);
10979 int vec_all_le (vector signed int, vector signed int);
10980 int vec_all_le (vector float, vector float);
10982 int vec_all_lt (vector bool char, vector unsigned char);
10983 int vec_all_lt (vector unsigned char, vector bool char);
10984 int vec_all_lt (vector unsigned char, vector unsigned char);
10985 int vec_all_lt (vector bool char, vector signed char);
10986 int vec_all_lt (vector signed char, vector bool char);
10987 int vec_all_lt (vector signed char, vector signed char);
10988 int vec_all_lt (vector bool short, vector unsigned short);
10989 int vec_all_lt (vector unsigned short, vector bool short);
10990 int vec_all_lt (vector unsigned short, vector unsigned short);
10991 int vec_all_lt (vector bool short, vector signed short);
10992 int vec_all_lt (vector signed short, vector bool short);
10993 int vec_all_lt (vector signed short, vector signed short);
10994 int vec_all_lt (vector bool int, vector unsigned int);
10995 int vec_all_lt (vector unsigned int, vector bool int);
10996 int vec_all_lt (vector unsigned int, vector unsigned int);
10997 int vec_all_lt (vector bool int, vector signed int);
10998 int vec_all_lt (vector signed int, vector bool int);
10999 int vec_all_lt (vector signed int, vector signed int);
11000 int vec_all_lt (vector float, vector float);
11002 int vec_all_nan (vector float);
11004 int vec_all_ne (vector signed char, vector bool char);
11005 int vec_all_ne (vector signed char, vector signed char);
11006 int vec_all_ne (vector unsigned char, vector bool char);
11007 int vec_all_ne (vector unsigned char, vector unsigned char);
11008 int vec_all_ne (vector bool char, vector bool char);
11009 int vec_all_ne (vector bool char, vector unsigned char);
11010 int vec_all_ne (vector bool char, vector signed char);
11011 int vec_all_ne (vector signed short, vector bool short);
11012 int vec_all_ne (vector signed short, vector signed short);
11013 int vec_all_ne (vector unsigned short, vector bool short);
11014 int vec_all_ne (vector unsigned short, vector unsigned short);
11015 int vec_all_ne (vector bool short, vector bool short);
11016 int vec_all_ne (vector bool short, vector unsigned short);
11017 int vec_all_ne (vector bool short, vector signed short);
11018 int vec_all_ne (vector pixel, vector pixel);
11019 int vec_all_ne (vector signed int, vector bool int);
11020 int vec_all_ne (vector signed int, vector signed int);
11021 int vec_all_ne (vector unsigned int, vector bool int);
11022 int vec_all_ne (vector unsigned int, vector unsigned int);
11023 int vec_all_ne (vector bool int, vector bool int);
11024 int vec_all_ne (vector bool int, vector unsigned int);
11025 int vec_all_ne (vector bool int, vector signed int);
11026 int vec_all_ne (vector float, vector float);
11028 int vec_all_nge (vector float, vector float);
11030 int vec_all_ngt (vector float, vector float);
11032 int vec_all_nle (vector float, vector float);
11034 int vec_all_nlt (vector float, vector float);
11036 int vec_all_numeric (vector float);
11038 int vec_any_eq (vector signed char, vector bool char);
11039 int vec_any_eq (vector signed char, vector signed char);
11040 int vec_any_eq (vector unsigned char, vector bool char);
11041 int vec_any_eq (vector unsigned char, vector unsigned char);
11042 int vec_any_eq (vector bool char, vector bool char);
11043 int vec_any_eq (vector bool char, vector unsigned char);
11044 int vec_any_eq (vector bool char, vector signed char);
11045 int vec_any_eq (vector signed short, vector bool short);
11046 int vec_any_eq (vector signed short, vector signed short);
11047 int vec_any_eq (vector unsigned short, vector bool short);
11048 int vec_any_eq (vector unsigned short, vector unsigned short);
11049 int vec_any_eq (vector bool short, vector bool short);
11050 int vec_any_eq (vector bool short, vector unsigned short);
11051 int vec_any_eq (vector bool short, vector signed short);
11052 int vec_any_eq (vector pixel, vector pixel);
11053 int vec_any_eq (vector signed int, vector bool int);
11054 int vec_any_eq (vector signed int, vector signed int);
11055 int vec_any_eq (vector unsigned int, vector bool int);
11056 int vec_any_eq (vector unsigned int, vector unsigned int);
11057 int vec_any_eq (vector bool int, vector bool int);
11058 int vec_any_eq (vector bool int, vector unsigned int);
11059 int vec_any_eq (vector bool int, vector signed int);
11060 int vec_any_eq (vector float, vector float);
11062 int vec_any_ge (vector signed char, vector bool char);
11063 int vec_any_ge (vector unsigned char, vector bool char);
11064 int vec_any_ge (vector unsigned char, vector unsigned char);
11065 int vec_any_ge (vector signed char, vector signed char);
11066 int vec_any_ge (vector bool char, vector unsigned char);
11067 int vec_any_ge (vector bool char, vector signed char);
11068 int vec_any_ge (vector unsigned short, vector bool short);
11069 int vec_any_ge (vector unsigned short, vector unsigned short);
11070 int vec_any_ge (vector signed short, vector signed short);
11071 int vec_any_ge (vector signed short, vector bool short);
11072 int vec_any_ge (vector bool short, vector unsigned short);
11073 int vec_any_ge (vector bool short, vector signed short);
11074 int vec_any_ge (vector signed int, vector bool int);
11075 int vec_any_ge (vector unsigned int, vector bool int);
11076 int vec_any_ge (vector unsigned int, vector unsigned int);
11077 int vec_any_ge (vector signed int, vector signed int);
11078 int vec_any_ge (vector bool int, vector unsigned int);
11079 int vec_any_ge (vector bool int, vector signed int);
11080 int vec_any_ge (vector float, vector float);
11082 int vec_any_gt (vector bool char, vector unsigned char);
11083 int vec_any_gt (vector unsigned char, vector bool char);
11084 int vec_any_gt (vector unsigned char, vector unsigned char);
11085 int vec_any_gt (vector bool char, vector signed char);
11086 int vec_any_gt (vector signed char, vector bool char);
11087 int vec_any_gt (vector signed char, vector signed char);
11088 int vec_any_gt (vector bool short, vector unsigned short);
11089 int vec_any_gt (vector unsigned short, vector bool short);
11090 int vec_any_gt (vector unsigned short, vector unsigned short);
11091 int vec_any_gt (vector bool short, vector signed short);
11092 int vec_any_gt (vector signed short, vector bool short);
11093 int vec_any_gt (vector signed short, vector signed short);
11094 int vec_any_gt (vector bool int, vector unsigned int);
11095 int vec_any_gt (vector unsigned int, vector bool int);
11096 int vec_any_gt (vector unsigned int, vector unsigned int);
11097 int vec_any_gt (vector bool int, vector signed int);
11098 int vec_any_gt (vector signed int, vector bool int);
11099 int vec_any_gt (vector signed int, vector signed int);
11100 int vec_any_gt (vector float, vector float);
11102 int vec_any_le (vector bool char, vector unsigned char);
11103 int vec_any_le (vector unsigned char, vector bool char);
11104 int vec_any_le (vector unsigned char, vector unsigned char);
11105 int vec_any_le (vector bool char, vector signed char);
11106 int vec_any_le (vector signed char, vector bool char);
11107 int vec_any_le (vector signed char, vector signed char);
11108 int vec_any_le (vector bool short, vector unsigned short);
11109 int vec_any_le (vector unsigned short, vector bool short);
11110 int vec_any_le (vector unsigned short, vector unsigned short);
11111 int vec_any_le (vector bool short, vector signed short);
11112 int vec_any_le (vector signed short, vector bool short);
11113 int vec_any_le (vector signed short, vector signed short);
11114 int vec_any_le (vector bool int, vector unsigned int);
11115 int vec_any_le (vector unsigned int, vector bool int);
11116 int vec_any_le (vector unsigned int, vector unsigned int);
11117 int vec_any_le (vector bool int, vector signed int);
11118 int vec_any_le (vector signed int, vector bool int);
11119 int vec_any_le (vector signed int, vector signed int);
11120 int vec_any_le (vector float, vector float);
11122 int vec_any_lt (vector bool char, vector unsigned char);
11123 int vec_any_lt (vector unsigned char, vector bool char);
11124 int vec_any_lt (vector unsigned char, vector unsigned char);
11125 int vec_any_lt (vector bool char, vector signed char);
11126 int vec_any_lt (vector signed char, vector bool char);
11127 int vec_any_lt (vector signed char, vector signed char);
11128 int vec_any_lt (vector bool short, vector unsigned short);
11129 int vec_any_lt (vector unsigned short, vector bool short);
11130 int vec_any_lt (vector unsigned short, vector unsigned short);
11131 int vec_any_lt (vector bool short, vector signed short);
11132 int vec_any_lt (vector signed short, vector bool short);
11133 int vec_any_lt (vector signed short, vector signed short);
11134 int vec_any_lt (vector bool int, vector unsigned int);
11135 int vec_any_lt (vector unsigned int, vector bool int);
11136 int vec_any_lt (vector unsigned int, vector unsigned int);
11137 int vec_any_lt (vector bool int, vector signed int);
11138 int vec_any_lt (vector signed int, vector bool int);
11139 int vec_any_lt (vector signed int, vector signed int);
11140 int vec_any_lt (vector float, vector float);
11142 int vec_any_nan (vector float);
11144 int vec_any_ne (vector signed char, vector bool char);
11145 int vec_any_ne (vector signed char, vector signed char);
11146 int vec_any_ne (vector unsigned char, vector bool char);
11147 int vec_any_ne (vector unsigned char, vector unsigned char);
11148 int vec_any_ne (vector bool char, vector bool char);
11149 int vec_any_ne (vector bool char, vector unsigned char);
11150 int vec_any_ne (vector bool char, vector signed char);
11151 int vec_any_ne (vector signed short, vector bool short);
11152 int vec_any_ne (vector signed short, vector signed short);
11153 int vec_any_ne (vector unsigned short, vector bool short);
11154 int vec_any_ne (vector unsigned short, vector unsigned short);
11155 int vec_any_ne (vector bool short, vector bool short);
11156 int vec_any_ne (vector bool short, vector unsigned short);
11157 int vec_any_ne (vector bool short, vector signed short);
11158 int vec_any_ne (vector pixel, vector pixel);
11159 int vec_any_ne (vector signed int, vector bool int);
11160 int vec_any_ne (vector signed int, vector signed int);
11161 int vec_any_ne (vector unsigned int, vector bool int);
11162 int vec_any_ne (vector unsigned int, vector unsigned int);
11163 int vec_any_ne (vector bool int, vector bool int);
11164 int vec_any_ne (vector bool int, vector unsigned int);
11165 int vec_any_ne (vector bool int, vector signed int);
11166 int vec_any_ne (vector float, vector float);
11168 int vec_any_nge (vector float, vector float);
11170 int vec_any_ngt (vector float, vector float);
11172 int vec_any_nle (vector float, vector float);
11174 int vec_any_nlt (vector float, vector float);
11176 int vec_any_numeric (vector float);
11178 int vec_any_out (vector float, vector float);
11181 @node SPARC VIS Built-in Functions
11182 @subsection SPARC VIS Built-in Functions
11184 GCC supports SIMD operations on the SPARC using both the generic vector
11185 extensions (@pxref{Vector Extensions}) as well as built-in functions for
11186 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
11187 switch, the VIS extension is exposed as the following built-in functions:
11190 typedef int v2si __attribute__ ((vector_size (8)));
11191 typedef short v4hi __attribute__ ((vector_size (8)));
11192 typedef short v2hi __attribute__ ((vector_size (4)));
11193 typedef char v8qi __attribute__ ((vector_size (8)));
11194 typedef char v4qi __attribute__ ((vector_size (4)));
11196 void * __builtin_vis_alignaddr (void *, long);
11197 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
11198 v2si __builtin_vis_faligndatav2si (v2si, v2si);
11199 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
11200 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
11202 v4hi __builtin_vis_fexpand (v4qi);
11204 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
11205 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
11206 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
11207 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
11208 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
11209 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
11210 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
11212 v4qi __builtin_vis_fpack16 (v4hi);
11213 v8qi __builtin_vis_fpack32 (v2si, v2si);
11214 v2hi __builtin_vis_fpackfix (v2si);
11215 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
11217 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
11220 @node SPU Built-in Functions
11221 @subsection SPU Built-in Functions
11223 GCC provides extensions for the SPU processor as described in the
11224 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
11225 found at @uref{http://cell.scei.co.jp/} or
11226 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
11227 implementation differs in several ways.
11232 The optional extension of specifying vector constants in parentheses is
11236 A vector initializer requires no cast if the vector constant is of the
11237 same type as the variable it is initializing.
11240 If @code{signed} or @code{unsigned} is omitted, the signedness of the
11241 vector type is the default signedness of the base type. The default
11242 varies depending on the operating system, so a portable program should
11243 always specify the signedness.
11246 By default, the keyword @code{__vector} is added. The macro
11247 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
11251 GCC allows using a @code{typedef} name as the type specifier for a
11255 For C, overloaded functions are implemented with macros so the following
11259 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
11262 Since @code{spu_add} is a macro, the vector constant in the example
11263 is treated as four separate arguments. Wrap the entire argument in
11264 parentheses for this to work.
11267 The extended version of @code{__builtin_expect} is not supported.
11271 @emph{Note:} Only the interface described in the aforementioned
11272 specification is supported. Internally, GCC uses built-in functions to
11273 implement the required functionality, but these are not supported and
11274 are subject to change without notice.
11276 @node Target Format Checks
11277 @section Format Checks Specific to Particular Target Machines
11279 For some target machines, GCC supports additional options to the
11281 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
11284 * Solaris Format Checks::
11287 @node Solaris Format Checks
11288 @subsection Solaris Format Checks
11290 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
11291 check. @code{cmn_err} accepts a subset of the standard @code{printf}
11292 conversions, and the two-argument @code{%b} conversion for displaying
11293 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
11296 @section Pragmas Accepted by GCC
11300 GCC supports several types of pragmas, primarily in order to compile
11301 code originally written for other compilers. Note that in general
11302 we do not recommend the use of pragmas; @xref{Function Attributes},
11303 for further explanation.
11308 * RS/6000 and PowerPC Pragmas::
11310 * Solaris Pragmas::
11311 * Symbol-Renaming Pragmas::
11312 * Structure-Packing Pragmas::
11314 * Diagnostic Pragmas::
11315 * Visibility Pragmas::
11316 * Push/Pop Macro Pragmas::
11317 * Function Specific Option Pragmas::
11321 @subsection ARM Pragmas
11323 The ARM target defines pragmas for controlling the default addition of
11324 @code{long_call} and @code{short_call} attributes to functions.
11325 @xref{Function Attributes}, for information about the effects of these
11330 @cindex pragma, long_calls
11331 Set all subsequent functions to have the @code{long_call} attribute.
11333 @item no_long_calls
11334 @cindex pragma, no_long_calls
11335 Set all subsequent functions to have the @code{short_call} attribute.
11337 @item long_calls_off
11338 @cindex pragma, long_calls_off
11339 Do not affect the @code{long_call} or @code{short_call} attributes of
11340 subsequent functions.
11344 @subsection M32C Pragmas
11347 @item memregs @var{number}
11348 @cindex pragma, memregs
11349 Overrides the command line option @code{-memregs=} for the current
11350 file. Use with care! This pragma must be before any function in the
11351 file, and mixing different memregs values in different objects may
11352 make them incompatible. This pragma is useful when a
11353 performance-critical function uses a memreg for temporary values,
11354 as it may allow you to reduce the number of memregs used.
11358 @node RS/6000 and PowerPC Pragmas
11359 @subsection RS/6000 and PowerPC Pragmas
11361 The RS/6000 and PowerPC targets define one pragma for controlling
11362 whether or not the @code{longcall} attribute is added to function
11363 declarations by default. This pragma overrides the @option{-mlongcall}
11364 option, but not the @code{longcall} and @code{shortcall} attributes.
11365 @xref{RS/6000 and PowerPC Options}, for more information about when long
11366 calls are and are not necessary.
11370 @cindex pragma, longcall
11371 Apply the @code{longcall} attribute to all subsequent function
11375 Do not apply the @code{longcall} attribute to subsequent function
11379 @c Describe h8300 pragmas here.
11380 @c Describe sh pragmas here.
11381 @c Describe v850 pragmas here.
11383 @node Darwin Pragmas
11384 @subsection Darwin Pragmas
11386 The following pragmas are available for all architectures running the
11387 Darwin operating system. These are useful for compatibility with other
11391 @item mark @var{tokens}@dots{}
11392 @cindex pragma, mark
11393 This pragma is accepted, but has no effect.
11395 @item options align=@var{alignment}
11396 @cindex pragma, options align
11397 This pragma sets the alignment of fields in structures. The values of
11398 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
11399 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
11400 properly; to restore the previous setting, use @code{reset} for the
11403 @item segment @var{tokens}@dots{}
11404 @cindex pragma, segment
11405 This pragma is accepted, but has no effect.
11407 @item unused (@var{var} [, @var{var}]@dots{})
11408 @cindex pragma, unused
11409 This pragma declares variables to be possibly unused. GCC will not
11410 produce warnings for the listed variables. The effect is similar to
11411 that of the @code{unused} attribute, except that this pragma may appear
11412 anywhere within the variables' scopes.
11415 @node Solaris Pragmas
11416 @subsection Solaris Pragmas
11418 The Solaris target supports @code{#pragma redefine_extname}
11419 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
11420 @code{#pragma} directives for compatibility with the system compiler.
11423 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
11424 @cindex pragma, align
11426 Increase the minimum alignment of each @var{variable} to @var{alignment}.
11427 This is the same as GCC's @code{aligned} attribute @pxref{Variable
11428 Attributes}). Macro expansion occurs on the arguments to this pragma
11429 when compiling C and Objective-C@. It does not currently occur when
11430 compiling C++, but this is a bug which may be fixed in a future
11433 @item fini (@var{function} [, @var{function}]...)
11434 @cindex pragma, fini
11436 This pragma causes each listed @var{function} to be called after
11437 main, or during shared module unloading, by adding a call to the
11438 @code{.fini} section.
11440 @item init (@var{function} [, @var{function}]...)
11441 @cindex pragma, init
11443 This pragma causes each listed @var{function} to be called during
11444 initialization (before @code{main}) or during shared module loading, by
11445 adding a call to the @code{.init} section.
11449 @node Symbol-Renaming Pragmas
11450 @subsection Symbol-Renaming Pragmas
11452 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
11453 supports two @code{#pragma} directives which change the name used in
11454 assembly for a given declaration. These pragmas are only available on
11455 platforms whose system headers need them. To get this effect on all
11456 platforms supported by GCC, use the asm labels extension (@pxref{Asm
11460 @item redefine_extname @var{oldname} @var{newname}
11461 @cindex pragma, redefine_extname
11463 This pragma gives the C function @var{oldname} the assembly symbol
11464 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
11465 will be defined if this pragma is available (currently only on
11468 @item extern_prefix @var{string}
11469 @cindex pragma, extern_prefix
11471 This pragma causes all subsequent external function and variable
11472 declarations to have @var{string} prepended to their assembly symbols.
11473 This effect may be terminated with another @code{extern_prefix} pragma
11474 whose argument is an empty string. The preprocessor macro
11475 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
11476 available (currently only on Tru64 UNIX)@.
11479 These pragmas and the asm labels extension interact in a complicated
11480 manner. Here are some corner cases you may want to be aware of.
11483 @item Both pragmas silently apply only to declarations with external
11484 linkage. Asm labels do not have this restriction.
11486 @item In C++, both pragmas silently apply only to declarations with
11487 ``C'' linkage. Again, asm labels do not have this restriction.
11489 @item If any of the three ways of changing the assembly name of a
11490 declaration is applied to a declaration whose assembly name has
11491 already been determined (either by a previous use of one of these
11492 features, or because the compiler needed the assembly name in order to
11493 generate code), and the new name is different, a warning issues and
11494 the name does not change.
11496 @item The @var{oldname} used by @code{#pragma redefine_extname} is
11497 always the C-language name.
11499 @item If @code{#pragma extern_prefix} is in effect, and a declaration
11500 occurs with an asm label attached, the prefix is silently ignored for
11503 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
11504 apply to the same declaration, whichever triggered first wins, and a
11505 warning issues if they contradict each other. (We would like to have
11506 @code{#pragma redefine_extname} always win, for consistency with asm
11507 labels, but if @code{#pragma extern_prefix} triggers first we have no
11508 way of knowing that that happened.)
11511 @node Structure-Packing Pragmas
11512 @subsection Structure-Packing Pragmas
11514 For compatibility with Microsoft Windows compilers, GCC supports a
11515 set of @code{#pragma} directives which change the maximum alignment of
11516 members of structures (other than zero-width bitfields), unions, and
11517 classes subsequently defined. The @var{n} value below always is required
11518 to be a small power of two and specifies the new alignment in bytes.
11521 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
11522 @item @code{#pragma pack()} sets the alignment to the one that was in
11523 effect when compilation started (see also command line option
11524 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
11525 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
11526 setting on an internal stack and then optionally sets the new alignment.
11527 @item @code{#pragma pack(pop)} restores the alignment setting to the one
11528 saved at the top of the internal stack (and removes that stack entry).
11529 Note that @code{#pragma pack([@var{n}])} does not influence this internal
11530 stack; thus it is possible to have @code{#pragma pack(push)} followed by
11531 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
11532 @code{#pragma pack(pop)}.
11535 Some targets, e.g.@: i386 and powerpc, support the @code{ms_struct}
11536 @code{#pragma} which lays out a structure as the documented
11537 @code{__attribute__ ((ms_struct))}.
11539 @item @code{#pragma ms_struct on} turns on the layout for structures
11541 @item @code{#pragma ms_struct off} turns off the layout for structures
11543 @item @code{#pragma ms_struct reset} goes back to the default layout.
11547 @subsection Weak Pragmas
11549 For compatibility with SVR4, GCC supports a set of @code{#pragma}
11550 directives for declaring symbols to be weak, and defining weak
11554 @item #pragma weak @var{symbol}
11555 @cindex pragma, weak
11556 This pragma declares @var{symbol} to be weak, as if the declaration
11557 had the attribute of the same name. The pragma may appear before
11558 or after the declaration of @var{symbol}, but must appear before
11559 either its first use or its definition. It is not an error for
11560 @var{symbol} to never be defined at all.
11562 @item #pragma weak @var{symbol1} = @var{symbol2}
11563 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
11564 It is an error if @var{symbol2} is not defined in the current
11568 @node Diagnostic Pragmas
11569 @subsection Diagnostic Pragmas
11571 GCC allows the user to selectively enable or disable certain types of
11572 diagnostics, and change the kind of the diagnostic. For example, a
11573 project's policy might require that all sources compile with
11574 @option{-Werror} but certain files might have exceptions allowing
11575 specific types of warnings. Or, a project might selectively enable
11576 diagnostics and treat them as errors depending on which preprocessor
11577 macros are defined.
11580 @item #pragma GCC diagnostic @var{kind} @var{option}
11581 @cindex pragma, diagnostic
11583 Modifies the disposition of a diagnostic. Note that not all
11584 diagnostics are modifiable; at the moment only warnings (normally
11585 controlled by @samp{-W@dots{}}) can be controlled, and not all of them.
11586 Use @option{-fdiagnostics-show-option} to determine which diagnostics
11587 are controllable and which option controls them.
11589 @var{kind} is @samp{error} to treat this diagnostic as an error,
11590 @samp{warning} to treat it like a warning (even if @option{-Werror} is
11591 in effect), or @samp{ignored} if the diagnostic is to be ignored.
11592 @var{option} is a double quoted string which matches the command line
11596 #pragma GCC diagnostic warning "-Wformat"
11597 #pragma GCC diagnostic error "-Wformat"
11598 #pragma GCC diagnostic ignored "-Wformat"
11601 Note that these pragmas override any command line options. Also,
11602 while it is syntactically valid to put these pragmas anywhere in your
11603 sources, the only supported location for them is before any data or
11604 functions are defined. Doing otherwise may result in unpredictable
11605 results depending on how the optimizer manages your sources. If the
11606 same option is listed multiple times, the last one specified is the
11607 one that is in effect. This pragma is not intended to be a general
11608 purpose replacement for command line options, but for implementing
11609 strict control over project policies.
11613 GCC also offers a simple mechanism for printing messages during
11617 @item #pragma message @var{string}
11618 @cindex pragma, diagnostic
11620 Prints @var{string} as a compiler message on compilation. The message
11621 is informational only, and is neither a compilation warning nor an error.
11624 #pragma message "Compiling " __FILE__ "..."
11627 @var{string} may be parenthesized, and is printed with location
11628 information. For example,
11631 #define DO_PRAGMA(x) _Pragma (#x)
11632 #define TODO(x) DO_PRAGMA(message ("TODO - " #x))
11634 TODO(Remember to fix this)
11637 prints @samp{/tmp/file.c:4: note: #pragma message:
11638 TODO - Remember to fix this}.
11642 @node Visibility Pragmas
11643 @subsection Visibility Pragmas
11646 @item #pragma GCC visibility push(@var{visibility})
11647 @itemx #pragma GCC visibility pop
11648 @cindex pragma, visibility
11650 This pragma allows the user to set the visibility for multiple
11651 declarations without having to give each a visibility attribute
11652 @xref{Function Attributes}, for more information about visibility and
11653 the attribute syntax.
11655 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
11656 declarations. Class members and template specializations are not
11657 affected; if you want to override the visibility for a particular
11658 member or instantiation, you must use an attribute.
11663 @node Push/Pop Macro Pragmas
11664 @subsection Push/Pop Macro Pragmas
11666 For compatibility with Microsoft Windows compilers, GCC supports
11667 @samp{#pragma push_macro(@var{"macro_name"})}
11668 and @samp{#pragma pop_macro(@var{"macro_name"})}.
11671 @item #pragma push_macro(@var{"macro_name"})
11672 @cindex pragma, push_macro
11673 This pragma saves the value of the macro named as @var{macro_name} to
11674 the top of the stack for this macro.
11676 @item #pragma pop_macro(@var{"macro_name"})
11677 @cindex pragma, pop_macro
11678 This pragma sets the value of the macro named as @var{macro_name} to
11679 the value on top of the stack for this macro. If the stack for
11680 @var{macro_name} is empty, the value of the macro remains unchanged.
11687 #pragma push_macro("X")
11690 #pragma pop_macro("X")
11694 In this example, the definition of X as 1 is saved by @code{#pragma
11695 push_macro} and restored by @code{#pragma pop_macro}.
11697 @node Function Specific Option Pragmas
11698 @subsection Function Specific Option Pragmas
11701 @item #pragma GCC option (@var{"string"}...)
11702 @cindex pragma GCC option
11704 This pragma allows you to set target specific options for functions
11705 defined later in the source file. One or more strings can be
11706 specified. Each function that is defined after this point will be as
11707 if @code{attribute((option("STRING")))} was specified for that
11708 function. The parenthesis around the options is optional.
11709 @xref{Function Attributes}, for more information about the
11710 @code{option} attribute and the attribute syntax.
11712 The @samp{#pragma GCC option} pragma is not implemented in GCC
11713 versions earlier than 4.4, and is currently only implemented for the
11714 386 and x86_64 backend.
11718 @item #pragma GCC option (push)
11719 @itemx #pragma GCC option (pop)
11720 @cindex pragma GCC option
11722 These pragmas maintain a stack of the current options. It is
11723 intended for include files where you temporarily want to switch to
11724 using a different @samp{#pragma GCC option} and then to pop back to
11725 the previous options.
11729 @item #pragma GCC option (reset)
11730 @cindex pragma, target option
11731 @cindex pragma GCC option
11733 This pragma clears the current @code{#pragma GCC options} to use the
11734 default switches as specified on the command line.
11737 @item #pragma GCC optimize (@var{"string"}...)
11738 @cindex pragma GCC optimize
11740 This pragma allows you to set global optimization options for functions
11741 defined later in the source file. One or more strings can be
11742 specified. Each function that is defined after this point will be as
11743 if @code{attribute((optimize("STRING")))} was specified for that
11744 function. The parenthesis around the options is optional.
11745 @xref{Function Attributes}, for more information about the
11746 @code{optimize} attribute and the attribute syntax.
11748 The @samp{#pragma GCC optimize} pragma is not implemented in GCC
11749 versions earlier than 4.4.
11753 @item #pragma GCC optimize (push)
11754 @itemx #pragma GCC optimize (pop)
11755 @cindex pragma GCC optimize
11757 These pragmas maintain a stack of the current optimization options.
11758 It is intended for include files where you temporarily want to switch
11759 to using a different @code{#pragma GCC optimize} and then to pop back
11760 to the previous optimizations.
11764 @item #pragma GCC optimize reset
11765 @cindex pragma GCC optimize
11767 This pragma clears the current @code{#pragma GCC optimize} to use the
11768 default switches as specified on the command line.
11771 @node Unnamed Fields
11772 @section Unnamed struct/union fields within structs/unions
11776 For compatibility with other compilers, GCC allows you to define
11777 a structure or union that contains, as fields, structures and unions
11778 without names. For example:
11791 In this example, the user would be able to access members of the unnamed
11792 union with code like @samp{foo.b}. Note that only unnamed structs and
11793 unions are allowed, you may not have, for example, an unnamed
11796 You must never create such structures that cause ambiguous field definitions.
11797 For example, this structure:
11808 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
11809 Such constructs are not supported and must be avoided. In the future,
11810 such constructs may be detected and treated as compilation errors.
11812 @opindex fms-extensions
11813 Unless @option{-fms-extensions} is used, the unnamed field must be a
11814 structure or union definition without a tag (for example, @samp{struct
11815 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
11816 also be a definition with a tag such as @samp{struct foo @{ int a;
11817 @};}, a reference to a previously defined structure or union such as
11818 @samp{struct foo;}, or a reference to a @code{typedef} name for a
11819 previously defined structure or union type.
11822 @section Thread-Local Storage
11823 @cindex Thread-Local Storage
11824 @cindex @acronym{TLS}
11827 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
11828 are allocated such that there is one instance of the variable per extant
11829 thread. The run-time model GCC uses to implement this originates
11830 in the IA-64 processor-specific ABI, but has since been migrated
11831 to other processors as well. It requires significant support from
11832 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
11833 system libraries (@file{libc.so} and @file{libpthread.so}), so it
11834 is not available everywhere.
11836 At the user level, the extension is visible with a new storage
11837 class keyword: @code{__thread}. For example:
11841 extern __thread struct state s;
11842 static __thread char *p;
11845 The @code{__thread} specifier may be used alone, with the @code{extern}
11846 or @code{static} specifiers, but with no other storage class specifier.
11847 When used with @code{extern} or @code{static}, @code{__thread} must appear
11848 immediately after the other storage class specifier.
11850 The @code{__thread} specifier may be applied to any global, file-scoped
11851 static, function-scoped static, or static data member of a class. It may
11852 not be applied to block-scoped automatic or non-static data member.
11854 When the address-of operator is applied to a thread-local variable, it is
11855 evaluated at run-time and returns the address of the current thread's
11856 instance of that variable. An address so obtained may be used by any
11857 thread. When a thread terminates, any pointers to thread-local variables
11858 in that thread become invalid.
11860 No static initialization may refer to the address of a thread-local variable.
11862 In C++, if an initializer is present for a thread-local variable, it must
11863 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
11866 See @uref{http://people.redhat.com/drepper/tls.pdf,
11867 ELF Handling For Thread-Local Storage} for a detailed explanation of
11868 the four thread-local storage addressing models, and how the run-time
11869 is expected to function.
11872 * C99 Thread-Local Edits::
11873 * C++98 Thread-Local Edits::
11876 @node C99 Thread-Local Edits
11877 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
11879 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
11880 that document the exact semantics of the language extension.
11884 @cite{5.1.2 Execution environments}
11886 Add new text after paragraph 1
11889 Within either execution environment, a @dfn{thread} is a flow of
11890 control within a program. It is implementation defined whether
11891 or not there may be more than one thread associated with a program.
11892 It is implementation defined how threads beyond the first are
11893 created, the name and type of the function called at thread
11894 startup, and how threads may be terminated. However, objects
11895 with thread storage duration shall be initialized before thread
11900 @cite{6.2.4 Storage durations of objects}
11902 Add new text before paragraph 3
11905 An object whose identifier is declared with the storage-class
11906 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
11907 Its lifetime is the entire execution of the thread, and its
11908 stored value is initialized only once, prior to thread startup.
11912 @cite{6.4.1 Keywords}
11914 Add @code{__thread}.
11917 @cite{6.7.1 Storage-class specifiers}
11919 Add @code{__thread} to the list of storage class specifiers in
11922 Change paragraph 2 to
11925 With the exception of @code{__thread}, at most one storage-class
11926 specifier may be given [@dots{}]. The @code{__thread} specifier may
11927 be used alone, or immediately following @code{extern} or
11931 Add new text after paragraph 6
11934 The declaration of an identifier for a variable that has
11935 block scope that specifies @code{__thread} shall also
11936 specify either @code{extern} or @code{static}.
11938 The @code{__thread} specifier shall be used only with
11943 @node C++98 Thread-Local Edits
11944 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
11946 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
11947 that document the exact semantics of the language extension.
11951 @b{[intro.execution]}
11953 New text after paragraph 4
11956 A @dfn{thread} is a flow of control within the abstract machine.
11957 It is implementation defined whether or not there may be more than
11961 New text after paragraph 7
11964 It is unspecified whether additional action must be taken to
11965 ensure when and whether side effects are visible to other threads.
11971 Add @code{__thread}.
11974 @b{[basic.start.main]}
11976 Add after paragraph 5
11979 The thread that begins execution at the @code{main} function is called
11980 the @dfn{main thread}. It is implementation defined how functions
11981 beginning threads other than the main thread are designated or typed.
11982 A function so designated, as well as the @code{main} function, is called
11983 a @dfn{thread startup function}. It is implementation defined what
11984 happens if a thread startup function returns. It is implementation
11985 defined what happens to other threads when any thread calls @code{exit}.
11989 @b{[basic.start.init]}
11991 Add after paragraph 4
11994 The storage for an object of thread storage duration shall be
11995 statically initialized before the first statement of the thread startup
11996 function. An object of thread storage duration shall not require
11997 dynamic initialization.
12001 @b{[basic.start.term]}
12003 Add after paragraph 3
12006 The type of an object with thread storage duration shall not have a
12007 non-trivial destructor, nor shall it be an array type whose elements
12008 (directly or indirectly) have non-trivial destructors.
12014 Add ``thread storage duration'' to the list in paragraph 1.
12019 Thread, static, and automatic storage durations are associated with
12020 objects introduced by declarations [@dots{}].
12023 Add @code{__thread} to the list of specifiers in paragraph 3.
12026 @b{[basic.stc.thread]}
12028 New section before @b{[basic.stc.static]}
12031 The keyword @code{__thread} applied to a non-local object gives the
12032 object thread storage duration.
12034 A local variable or class data member declared both @code{static}
12035 and @code{__thread} gives the variable or member thread storage
12040 @b{[basic.stc.static]}
12045 All objects which have neither thread storage duration, dynamic
12046 storage duration nor are local [@dots{}].
12052 Add @code{__thread} to the list in paragraph 1.
12057 With the exception of @code{__thread}, at most one
12058 @var{storage-class-specifier} shall appear in a given
12059 @var{decl-specifier-seq}. The @code{__thread} specifier may
12060 be used alone, or immediately following the @code{extern} or
12061 @code{static} specifiers. [@dots{}]
12064 Add after paragraph 5
12067 The @code{__thread} specifier can be applied only to the names of objects
12068 and to anonymous unions.
12074 Add after paragraph 6
12077 Non-@code{static} members shall not be @code{__thread}.
12081 @node Binary constants
12082 @section Binary constants using the @samp{0b} prefix
12083 @cindex Binary constants using the @samp{0b} prefix
12085 Integer constants can be written as binary constants, consisting of a
12086 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
12087 @samp{0B}. This is particularly useful in environments that operate a
12088 lot on the bit-level (like microcontrollers).
12090 The following statements are identical:
12099 The type of these constants follows the same rules as for octal or
12100 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
12103 @node C++ Extensions
12104 @chapter Extensions to the C++ Language
12105 @cindex extensions, C++ language
12106 @cindex C++ language extensions
12108 The GNU compiler provides these extensions to the C++ language (and you
12109 can also use most of the C language extensions in your C++ programs). If you
12110 want to write code that checks whether these features are available, you can
12111 test for the GNU compiler the same way as for C programs: check for a
12112 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
12113 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
12114 Predefined Macros,cpp,The GNU C Preprocessor}).
12117 * Volatiles:: What constitutes an access to a volatile object.
12118 * Restricted Pointers:: C99 restricted pointers and references.
12119 * Vague Linkage:: Where G++ puts inlines, vtables and such.
12120 * C++ Interface:: You can use a single C++ header file for both
12121 declarations and definitions.
12122 * Template Instantiation:: Methods for ensuring that exactly one copy of
12123 each needed template instantiation is emitted.
12124 * Bound member functions:: You can extract a function pointer to the
12125 method denoted by a @samp{->*} or @samp{.*} expression.
12126 * C++ Attributes:: Variable, function, and type attributes for C++ only.
12127 * Namespace Association:: Strong using-directives for namespace association.
12128 * Type Traits:: Compiler support for type traits
12129 * Java Exceptions:: Tweaking exception handling to work with Java.
12130 * Deprecated Features:: Things will disappear from g++.
12131 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
12135 @section When is a Volatile Object Accessed?
12136 @cindex accessing volatiles
12137 @cindex volatile read
12138 @cindex volatile write
12139 @cindex volatile access
12141 Both the C and C++ standard have the concept of volatile objects. These
12142 are normally accessed by pointers and used for accessing hardware. The
12143 standards encourage compilers to refrain from optimizations concerning
12144 accesses to volatile objects. The C standard leaves it implementation
12145 defined as to what constitutes a volatile access. The C++ standard omits
12146 to specify this, except to say that C++ should behave in a similar manner
12147 to C with respect to volatiles, where possible. The minimum either
12148 standard specifies is that at a sequence point all previous accesses to
12149 volatile objects have stabilized and no subsequent accesses have
12150 occurred. Thus an implementation is free to reorder and combine
12151 volatile accesses which occur between sequence points, but cannot do so
12152 for accesses across a sequence point. The use of volatiles does not
12153 allow you to violate the restriction on updating objects multiple times
12154 within a sequence point.
12156 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
12158 The behavior differs slightly between C and C++ in the non-obvious cases:
12161 volatile int *src = @var{somevalue};
12165 With C, such expressions are rvalues, and GCC interprets this either as a
12166 read of the volatile object being pointed to or only as request to evaluate
12167 the side-effects. The C++ standard specifies that such expressions do not
12168 undergo lvalue to rvalue conversion, and that the type of the dereferenced
12169 object may be incomplete. The C++ standard does not specify explicitly
12170 that it is this lvalue to rvalue conversion which may be responsible for
12171 causing an access. However, there is reason to believe that it is,
12172 because otherwise certain simple expressions become undefined. However,
12173 because it would surprise most programmers, G++ treats dereferencing a
12174 pointer to volatile object of complete type when the value is unused as
12175 GCC would do for an equivalent type in C@. When the object has incomplete
12176 type, G++ issues a warning; if you wish to force an error, you must
12177 force a conversion to rvalue with, for instance, a static cast.
12179 When using a reference to volatile, G++ does not treat equivalent
12180 expressions as accesses to volatiles, but instead issues a warning that
12181 no volatile is accessed. The rationale for this is that otherwise it
12182 becomes difficult to determine where volatile access occur, and not
12183 possible to ignore the return value from functions returning volatile
12184 references. Again, if you wish to force a read, cast the reference to
12187 @node Restricted Pointers
12188 @section Restricting Pointer Aliasing
12189 @cindex restricted pointers
12190 @cindex restricted references
12191 @cindex restricted this pointer
12193 As with the C front end, G++ understands the C99 feature of restricted pointers,
12194 specified with the @code{__restrict__}, or @code{__restrict} type
12195 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
12196 language flag, @code{restrict} is not a keyword in C++.
12198 In addition to allowing restricted pointers, you can specify restricted
12199 references, which indicate that the reference is not aliased in the local
12203 void fn (int *__restrict__ rptr, int &__restrict__ rref)
12210 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
12211 @var{rref} refers to a (different) unaliased integer.
12213 You may also specify whether a member function's @var{this} pointer is
12214 unaliased by using @code{__restrict__} as a member function qualifier.
12217 void T::fn () __restrict__
12224 Within the body of @code{T::fn}, @var{this} will have the effective
12225 definition @code{T *__restrict__ const this}. Notice that the
12226 interpretation of a @code{__restrict__} member function qualifier is
12227 different to that of @code{const} or @code{volatile} qualifier, in that it
12228 is applied to the pointer rather than the object. This is consistent with
12229 other compilers which implement restricted pointers.
12231 As with all outermost parameter qualifiers, @code{__restrict__} is
12232 ignored in function definition matching. This means you only need to
12233 specify @code{__restrict__} in a function definition, rather than
12234 in a function prototype as well.
12236 @node Vague Linkage
12237 @section Vague Linkage
12238 @cindex vague linkage
12240 There are several constructs in C++ which require space in the object
12241 file but are not clearly tied to a single translation unit. We say that
12242 these constructs have ``vague linkage''. Typically such constructs are
12243 emitted wherever they are needed, though sometimes we can be more
12247 @item Inline Functions
12248 Inline functions are typically defined in a header file which can be
12249 included in many different compilations. Hopefully they can usually be
12250 inlined, but sometimes an out-of-line copy is necessary, if the address
12251 of the function is taken or if inlining fails. In general, we emit an
12252 out-of-line copy in all translation units where one is needed. As an
12253 exception, we only emit inline virtual functions with the vtable, since
12254 it will always require a copy.
12256 Local static variables and string constants used in an inline function
12257 are also considered to have vague linkage, since they must be shared
12258 between all inlined and out-of-line instances of the function.
12262 C++ virtual functions are implemented in most compilers using a lookup
12263 table, known as a vtable. The vtable contains pointers to the virtual
12264 functions provided by a class, and each object of the class contains a
12265 pointer to its vtable (or vtables, in some multiple-inheritance
12266 situations). If the class declares any non-inline, non-pure virtual
12267 functions, the first one is chosen as the ``key method'' for the class,
12268 and the vtable is only emitted in the translation unit where the key
12271 @emph{Note:} If the chosen key method is later defined as inline, the
12272 vtable will still be emitted in every translation unit which defines it.
12273 Make sure that any inline virtuals are declared inline in the class
12274 body, even if they are not defined there.
12276 @item type_info objects
12279 C++ requires information about types to be written out in order to
12280 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
12281 For polymorphic classes (classes with virtual functions), the type_info
12282 object is written out along with the vtable so that @samp{dynamic_cast}
12283 can determine the dynamic type of a class object at runtime. For all
12284 other types, we write out the type_info object when it is used: when
12285 applying @samp{typeid} to an expression, throwing an object, or
12286 referring to a type in a catch clause or exception specification.
12288 @item Template Instantiations
12289 Most everything in this section also applies to template instantiations,
12290 but there are other options as well.
12291 @xref{Template Instantiation,,Where's the Template?}.
12295 When used with GNU ld version 2.8 or later on an ELF system such as
12296 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
12297 these constructs will be discarded at link time. This is known as
12300 On targets that don't support COMDAT, but do support weak symbols, GCC
12301 will use them. This way one copy will override all the others, but
12302 the unused copies will still take up space in the executable.
12304 For targets which do not support either COMDAT or weak symbols,
12305 most entities with vague linkage will be emitted as local symbols to
12306 avoid duplicate definition errors from the linker. This will not happen
12307 for local statics in inlines, however, as having multiple copies will
12308 almost certainly break things.
12310 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
12311 another way to control placement of these constructs.
12313 @node C++ Interface
12314 @section #pragma interface and implementation
12316 @cindex interface and implementation headers, C++
12317 @cindex C++ interface and implementation headers
12318 @cindex pragmas, interface and implementation
12320 @code{#pragma interface} and @code{#pragma implementation} provide the
12321 user with a way of explicitly directing the compiler to emit entities
12322 with vague linkage (and debugging information) in a particular
12325 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
12326 most cases, because of COMDAT support and the ``key method'' heuristic
12327 mentioned in @ref{Vague Linkage}. Using them can actually cause your
12328 program to grow due to unnecessary out-of-line copies of inline
12329 functions. Currently (3.4) the only benefit of these
12330 @code{#pragma}s is reduced duplication of debugging information, and
12331 that should be addressed soon on DWARF 2 targets with the use of
12335 @item #pragma interface
12336 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
12337 @kindex #pragma interface
12338 Use this directive in @emph{header files} that define object classes, to save
12339 space in most of the object files that use those classes. Normally,
12340 local copies of certain information (backup copies of inline member
12341 functions, debugging information, and the internal tables that implement
12342 virtual functions) must be kept in each object file that includes class
12343 definitions. You can use this pragma to avoid such duplication. When a
12344 header file containing @samp{#pragma interface} is included in a
12345 compilation, this auxiliary information will not be generated (unless
12346 the main input source file itself uses @samp{#pragma implementation}).
12347 Instead, the object files will contain references to be resolved at link
12350 The second form of this directive is useful for the case where you have
12351 multiple headers with the same name in different directories. If you
12352 use this form, you must specify the same string to @samp{#pragma
12355 @item #pragma implementation
12356 @itemx #pragma implementation "@var{objects}.h"
12357 @kindex #pragma implementation
12358 Use this pragma in a @emph{main input file}, when you want full output from
12359 included header files to be generated (and made globally visible). The
12360 included header file, in turn, should use @samp{#pragma interface}.
12361 Backup copies of inline member functions, debugging information, and the
12362 internal tables used to implement virtual functions are all generated in
12363 implementation files.
12365 @cindex implied @code{#pragma implementation}
12366 @cindex @code{#pragma implementation}, implied
12367 @cindex naming convention, implementation headers
12368 If you use @samp{#pragma implementation} with no argument, it applies to
12369 an include file with the same basename@footnote{A file's @dfn{basename}
12370 was the name stripped of all leading path information and of trailing
12371 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
12372 file. For example, in @file{allclass.cc}, giving just
12373 @samp{#pragma implementation}
12374 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
12376 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
12377 an implementation file whenever you would include it from
12378 @file{allclass.cc} even if you never specified @samp{#pragma
12379 implementation}. This was deemed to be more trouble than it was worth,
12380 however, and disabled.
12382 Use the string argument if you want a single implementation file to
12383 include code from multiple header files. (You must also use
12384 @samp{#include} to include the header file; @samp{#pragma
12385 implementation} only specifies how to use the file---it doesn't actually
12388 There is no way to split up the contents of a single header file into
12389 multiple implementation files.
12392 @cindex inlining and C++ pragmas
12393 @cindex C++ pragmas, effect on inlining
12394 @cindex pragmas in C++, effect on inlining
12395 @samp{#pragma implementation} and @samp{#pragma interface} also have an
12396 effect on function inlining.
12398 If you define a class in a header file marked with @samp{#pragma
12399 interface}, the effect on an inline function defined in that class is
12400 similar to an explicit @code{extern} declaration---the compiler emits
12401 no code at all to define an independent version of the function. Its
12402 definition is used only for inlining with its callers.
12404 @opindex fno-implement-inlines
12405 Conversely, when you include the same header file in a main source file
12406 that declares it as @samp{#pragma implementation}, the compiler emits
12407 code for the function itself; this defines a version of the function
12408 that can be found via pointers (or by callers compiled without
12409 inlining). If all calls to the function can be inlined, you can avoid
12410 emitting the function by compiling with @option{-fno-implement-inlines}.
12411 If any calls were not inlined, you will get linker errors.
12413 @node Template Instantiation
12414 @section Where's the Template?
12415 @cindex template instantiation
12417 C++ templates are the first language feature to require more
12418 intelligence from the environment than one usually finds on a UNIX
12419 system. Somehow the compiler and linker have to make sure that each
12420 template instance occurs exactly once in the executable if it is needed,
12421 and not at all otherwise. There are two basic approaches to this
12422 problem, which are referred to as the Borland model and the Cfront model.
12425 @item Borland model
12426 Borland C++ solved the template instantiation problem by adding the code
12427 equivalent of common blocks to their linker; the compiler emits template
12428 instances in each translation unit that uses them, and the linker
12429 collapses them together. The advantage of this model is that the linker
12430 only has to consider the object files themselves; there is no external
12431 complexity to worry about. This disadvantage is that compilation time
12432 is increased because the template code is being compiled repeatedly.
12433 Code written for this model tends to include definitions of all
12434 templates in the header file, since they must be seen to be
12438 The AT&T C++ translator, Cfront, solved the template instantiation
12439 problem by creating the notion of a template repository, an
12440 automatically maintained place where template instances are stored. A
12441 more modern version of the repository works as follows: As individual
12442 object files are built, the compiler places any template definitions and
12443 instantiations encountered in the repository. At link time, the link
12444 wrapper adds in the objects in the repository and compiles any needed
12445 instances that were not previously emitted. The advantages of this
12446 model are more optimal compilation speed and the ability to use the
12447 system linker; to implement the Borland model a compiler vendor also
12448 needs to replace the linker. The disadvantages are vastly increased
12449 complexity, and thus potential for error; for some code this can be
12450 just as transparent, but in practice it can been very difficult to build
12451 multiple programs in one directory and one program in multiple
12452 directories. Code written for this model tends to separate definitions
12453 of non-inline member templates into a separate file, which should be
12454 compiled separately.
12457 When used with GNU ld version 2.8 or later on an ELF system such as
12458 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
12459 Borland model. On other systems, G++ implements neither automatic
12462 A future version of G++ will support a hybrid model whereby the compiler
12463 will emit any instantiations for which the template definition is
12464 included in the compile, and store template definitions and
12465 instantiation context information into the object file for the rest.
12466 The link wrapper will extract that information as necessary and invoke
12467 the compiler to produce the remaining instantiations. The linker will
12468 then combine duplicate instantiations.
12470 In the mean time, you have the following options for dealing with
12471 template instantiations:
12476 Compile your template-using code with @option{-frepo}. The compiler will
12477 generate files with the extension @samp{.rpo} listing all of the
12478 template instantiations used in the corresponding object files which
12479 could be instantiated there; the link wrapper, @samp{collect2}, will
12480 then update the @samp{.rpo} files to tell the compiler where to place
12481 those instantiations and rebuild any affected object files. The
12482 link-time overhead is negligible after the first pass, as the compiler
12483 will continue to place the instantiations in the same files.
12485 This is your best option for application code written for the Borland
12486 model, as it will just work. Code written for the Cfront model will
12487 need to be modified so that the template definitions are available at
12488 one or more points of instantiation; usually this is as simple as adding
12489 @code{#include <tmethods.cc>} to the end of each template header.
12491 For library code, if you want the library to provide all of the template
12492 instantiations it needs, just try to link all of its object files
12493 together; the link will fail, but cause the instantiations to be
12494 generated as a side effect. Be warned, however, that this may cause
12495 conflicts if multiple libraries try to provide the same instantiations.
12496 For greater control, use explicit instantiation as described in the next
12500 @opindex fno-implicit-templates
12501 Compile your code with @option{-fno-implicit-templates} to disable the
12502 implicit generation of template instances, and explicitly instantiate
12503 all the ones you use. This approach requires more knowledge of exactly
12504 which instances you need than do the others, but it's less
12505 mysterious and allows greater control. You can scatter the explicit
12506 instantiations throughout your program, perhaps putting them in the
12507 translation units where the instances are used or the translation units
12508 that define the templates themselves; you can put all of the explicit
12509 instantiations you need into one big file; or you can create small files
12516 template class Foo<int>;
12517 template ostream& operator <<
12518 (ostream&, const Foo<int>&);
12521 for each of the instances you need, and create a template instantiation
12522 library from those.
12524 If you are using Cfront-model code, you can probably get away with not
12525 using @option{-fno-implicit-templates} when compiling files that don't
12526 @samp{#include} the member template definitions.
12528 If you use one big file to do the instantiations, you may want to
12529 compile it without @option{-fno-implicit-templates} so you get all of the
12530 instances required by your explicit instantiations (but not by any
12531 other files) without having to specify them as well.
12533 G++ has extended the template instantiation syntax given in the ISO
12534 standard to allow forward declaration of explicit instantiations
12535 (with @code{extern}), instantiation of the compiler support data for a
12536 template class (i.e.@: the vtable) without instantiating any of its
12537 members (with @code{inline}), and instantiation of only the static data
12538 members of a template class, without the support data or member
12539 functions (with (@code{static}):
12542 extern template int max (int, int);
12543 inline template class Foo<int>;
12544 static template class Foo<int>;
12548 Do nothing. Pretend G++ does implement automatic instantiation
12549 management. Code written for the Borland model will work fine, but
12550 each translation unit will contain instances of each of the templates it
12551 uses. In a large program, this can lead to an unacceptable amount of code
12555 @node Bound member functions
12556 @section Extracting the function pointer from a bound pointer to member function
12558 @cindex pointer to member function
12559 @cindex bound pointer to member function
12561 In C++, pointer to member functions (PMFs) are implemented using a wide
12562 pointer of sorts to handle all the possible call mechanisms; the PMF
12563 needs to store information about how to adjust the @samp{this} pointer,
12564 and if the function pointed to is virtual, where to find the vtable, and
12565 where in the vtable to look for the member function. If you are using
12566 PMFs in an inner loop, you should really reconsider that decision. If
12567 that is not an option, you can extract the pointer to the function that
12568 would be called for a given object/PMF pair and call it directly inside
12569 the inner loop, to save a bit of time.
12571 Note that you will still be paying the penalty for the call through a
12572 function pointer; on most modern architectures, such a call defeats the
12573 branch prediction features of the CPU@. This is also true of normal
12574 virtual function calls.
12576 The syntax for this extension is
12580 extern int (A::*fp)();
12581 typedef int (*fptr)(A *);
12583 fptr p = (fptr)(a.*fp);
12586 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
12587 no object is needed to obtain the address of the function. They can be
12588 converted to function pointers directly:
12591 fptr p1 = (fptr)(&A::foo);
12594 @opindex Wno-pmf-conversions
12595 You must specify @option{-Wno-pmf-conversions} to use this extension.
12597 @node C++ Attributes
12598 @section C++-Specific Variable, Function, and Type Attributes
12600 Some attributes only make sense for C++ programs.
12603 @item init_priority (@var{priority})
12604 @cindex init_priority attribute
12607 In Standard C++, objects defined at namespace scope are guaranteed to be
12608 initialized in an order in strict accordance with that of their definitions
12609 @emph{in a given translation unit}. No guarantee is made for initializations
12610 across translation units. However, GNU C++ allows users to control the
12611 order of initialization of objects defined at namespace scope with the
12612 @code{init_priority} attribute by specifying a relative @var{priority},
12613 a constant integral expression currently bounded between 101 and 65535
12614 inclusive. Lower numbers indicate a higher priority.
12616 In the following example, @code{A} would normally be created before
12617 @code{B}, but the @code{init_priority} attribute has reversed that order:
12620 Some_Class A __attribute__ ((init_priority (2000)));
12621 Some_Class B __attribute__ ((init_priority (543)));
12625 Note that the particular values of @var{priority} do not matter; only their
12628 @item java_interface
12629 @cindex java_interface attribute
12631 This type attribute informs C++ that the class is a Java interface. It may
12632 only be applied to classes declared within an @code{extern "Java"} block.
12633 Calls to methods declared in this interface will be dispatched using GCJ's
12634 interface table mechanism, instead of regular virtual table dispatch.
12638 See also @ref{Namespace Association}.
12640 @node Namespace Association
12641 @section Namespace Association
12643 @strong{Caution:} The semantics of this extension are not fully
12644 defined. Users should refrain from using this extension as its
12645 semantics may change subtly over time. It is possible that this
12646 extension will be removed in future versions of G++.
12648 A using-directive with @code{__attribute ((strong))} is stronger
12649 than a normal using-directive in two ways:
12653 Templates from the used namespace can be specialized and explicitly
12654 instantiated as though they were members of the using namespace.
12657 The using namespace is considered an associated namespace of all
12658 templates in the used namespace for purposes of argument-dependent
12662 The used namespace must be nested within the using namespace so that
12663 normal unqualified lookup works properly.
12665 This is useful for composing a namespace transparently from
12666 implementation namespaces. For example:
12671 template <class T> struct A @{ @};
12673 using namespace debug __attribute ((__strong__));
12674 template <> struct A<int> @{ @}; // @r{ok to specialize}
12676 template <class T> void f (A<T>);
12681 f (std::A<float>()); // @r{lookup finds} std::f
12687 @section Type Traits
12689 The C++ front-end implements syntactic extensions that allow to
12690 determine at compile time various characteristics of a type (or of a
12694 @item __has_nothrow_assign (type)
12695 If @code{type} is const qualified or is a reference type then the trait is
12696 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
12697 is true, else if @code{type} is a cv class or union type with copy assignment
12698 operators that are known not to throw an exception then the trait is true,
12699 else it is false. Requires: @code{type} shall be a complete type, an array
12700 type of unknown bound, or is a @code{void} type.
12702 @item __has_nothrow_copy (type)
12703 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
12704 @code{type} is a cv class or union type with copy constructors that
12705 are known not to throw an exception then the trait is true, else it is false.
12706 Requires: @code{type} shall be a complete type, an array type of
12707 unknown bound, or is a @code{void} type.
12709 @item __has_nothrow_constructor (type)
12710 If @code{__has_trivial_constructor (type)} is true then the trait is
12711 true, else if @code{type} is a cv class or union type (or array
12712 thereof) with a default constructor that is known not to throw an
12713 exception then the trait is true, else it is false. Requires:
12714 @code{type} shall be a complete type, an array type of unknown bound,
12715 or is a @code{void} type.
12717 @item __has_trivial_assign (type)
12718 If @code{type} is const qualified or is a reference type then the trait is
12719 false. Otherwise if @code{__is_pod (type)} is true then the trait is
12720 true, else if @code{type} is a cv class or union type with a trivial
12721 copy assignment ([class.copy]) then the trait is true, else it is
12722 false. Requires: @code{type} shall be a complete type, an array type
12723 of unknown bound, or is a @code{void} type.
12725 @item __has_trivial_copy (type)
12726 If @code{__is_pod (type)} is true or @code{type} is a reference type
12727 then the trait is true, else if @code{type} is a cv class or union type
12728 with a trivial copy constructor ([class.copy]) then the trait
12729 is true, else it is false. Requires: @code{type} shall be a complete
12730 type, an array type of unknown bound, or is a @code{void} type.
12732 @item __has_trivial_constructor (type)
12733 If @code{__is_pod (type)} is true then the trait is true, else if
12734 @code{type} is a cv class or union type (or array thereof) with a
12735 trivial default constructor ([class.ctor]) then the trait is true,
12736 else it is false. Requires: @code{type} shall be a complete type, an
12737 array type of unknown bound, or is a @code{void} type.
12739 @item __has_trivial_destructor (type)
12740 If @code{__is_pod (type)} is true or @code{type} is a reference type then
12741 the trait is true, else if @code{type} is a cv class or union type (or
12742 array thereof) with a trivial destructor ([class.dtor]) then the trait
12743 is true, else it is false. Requires: @code{type} shall be a complete
12744 type, an array type of unknown bound, or is a @code{void} type.
12746 @item __has_virtual_destructor (type)
12747 If @code{type} is a class type with a virtual destructor
12748 ([class.dtor]) then the trait is true, else it is false. Requires:
12749 @code{type} shall be a complete type, an array type of unknown bound,
12750 or is a @code{void} type.
12752 @item __is_abstract (type)
12753 If @code{type} is an abstract class ([class.abstract]) then the trait
12754 is true, else it is false. Requires: @code{type} shall be a complete
12755 type, an array type of unknown bound, or is a @code{void} type.
12757 @item __is_base_of (base_type, derived_type)
12758 If @code{base_type} is a base class of @code{derived_type}
12759 ([class.derived]) then the trait is true, otherwise it is false.
12760 Top-level cv qualifications of @code{base_type} and
12761 @code{derived_type} are ignored. For the purposes of this trait, a
12762 class type is considered is own base. Requires: if @code{__is_class
12763 (base_type)} and @code{__is_class (derived_type)} are true and
12764 @code{base_type} and @code{derived_type} are not the same type
12765 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
12766 type. Diagnostic is produced if this requirement is not met.
12768 @item __is_class (type)
12769 If @code{type} is a cv class type, and not a union type
12770 ([basic.compound]) the the trait is true, else it is false.
12772 @item __is_empty (type)
12773 If @code{__is_class (type)} is false then the trait is false.
12774 Otherwise @code{type} is considered empty if and only if: @code{type}
12775 has no non-static data members, or all non-static data members, if
12776 any, are bit-fields of lenght 0, and @code{type} has no virtual
12777 members, and @code{type} has no virtual base classes, and @code{type}
12778 has no base classes @code{base_type} for which
12779 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
12780 be a complete type, an array type of unknown bound, or is a
12783 @item __is_enum (type)
12784 If @code{type} is a cv enumeration type ([basic.compound]) the the trait is
12785 true, else it is false.
12787 @item __is_pod (type)
12788 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
12789 else it is false. Requires: @code{type} shall be a complete type,
12790 an array type of unknown bound, or is a @code{void} type.
12792 @item __is_polymorphic (type)
12793 If @code{type} is a polymorphic class ([class.virtual]) then the trait
12794 is true, else it is false. Requires: @code{type} shall be a complete
12795 type, an array type of unknown bound, or is a @code{void} type.
12797 @item __is_union (type)
12798 If @code{type} is a cv union type ([basic.compound]) the the trait is
12799 true, else it is false.
12803 @node Java Exceptions
12804 @section Java Exceptions
12806 The Java language uses a slightly different exception handling model
12807 from C++. Normally, GNU C++ will automatically detect when you are
12808 writing C++ code that uses Java exceptions, and handle them
12809 appropriately. However, if C++ code only needs to execute destructors
12810 when Java exceptions are thrown through it, GCC will guess incorrectly.
12811 Sample problematic code is:
12814 struct S @{ ~S(); @};
12815 extern void bar(); // @r{is written in Java, and may throw exceptions}
12824 The usual effect of an incorrect guess is a link failure, complaining of
12825 a missing routine called @samp{__gxx_personality_v0}.
12827 You can inform the compiler that Java exceptions are to be used in a
12828 translation unit, irrespective of what it might think, by writing
12829 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
12830 @samp{#pragma} must appear before any functions that throw or catch
12831 exceptions, or run destructors when exceptions are thrown through them.
12833 You cannot mix Java and C++ exceptions in the same translation unit. It
12834 is believed to be safe to throw a C++ exception from one file through
12835 another file compiled for the Java exception model, or vice versa, but
12836 there may be bugs in this area.
12838 @node Deprecated Features
12839 @section Deprecated Features
12841 In the past, the GNU C++ compiler was extended to experiment with new
12842 features, at a time when the C++ language was still evolving. Now that
12843 the C++ standard is complete, some of those features are superseded by
12844 superior alternatives. Using the old features might cause a warning in
12845 some cases that the feature will be dropped in the future. In other
12846 cases, the feature might be gone already.
12848 While the list below is not exhaustive, it documents some of the options
12849 that are now deprecated:
12852 @item -fexternal-templates
12853 @itemx -falt-external-templates
12854 These are two of the many ways for G++ to implement template
12855 instantiation. @xref{Template Instantiation}. The C++ standard clearly
12856 defines how template definitions have to be organized across
12857 implementation units. G++ has an implicit instantiation mechanism that
12858 should work just fine for standard-conforming code.
12860 @item -fstrict-prototype
12861 @itemx -fno-strict-prototype
12862 Previously it was possible to use an empty prototype parameter list to
12863 indicate an unspecified number of parameters (like C), rather than no
12864 parameters, as C++ demands. This feature has been removed, except where
12865 it is required for backwards compatibility. @xref{Backwards Compatibility}.
12868 G++ allows a virtual function returning @samp{void *} to be overridden
12869 by one returning a different pointer type. This extension to the
12870 covariant return type rules is now deprecated and will be removed from a
12873 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
12874 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
12875 and are now removed from G++. Code using these operators should be
12876 modified to use @code{std::min} and @code{std::max} instead.
12878 The named return value extension has been deprecated, and is now
12881 The use of initializer lists with new expressions has been deprecated,
12882 and is now removed from G++.
12884 Floating and complex non-type template parameters have been deprecated,
12885 and are now removed from G++.
12887 The implicit typename extension has been deprecated and is now
12890 The use of default arguments in function pointers, function typedefs
12891 and other places where they are not permitted by the standard is
12892 deprecated and will be removed from a future version of G++.
12894 G++ allows floating-point literals to appear in integral constant expressions,
12895 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
12896 This extension is deprecated and will be removed from a future version.
12898 G++ allows static data members of const floating-point type to be declared
12899 with an initializer in a class definition. The standard only allows
12900 initializers for static members of const integral types and const
12901 enumeration types so this extension has been deprecated and will be removed
12902 from a future version.
12904 @node Backwards Compatibility
12905 @section Backwards Compatibility
12906 @cindex Backwards Compatibility
12907 @cindex ARM [Annotated C++ Reference Manual]
12909 Now that there is a definitive ISO standard C++, G++ has a specification
12910 to adhere to. The C++ language evolved over time, and features that
12911 used to be acceptable in previous drafts of the standard, such as the ARM
12912 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
12913 compilation of C++ written to such drafts, G++ contains some backwards
12914 compatibilities. @emph{All such backwards compatibility features are
12915 liable to disappear in future versions of G++.} They should be considered
12916 deprecated. @xref{Deprecated Features}.
12920 If a variable is declared at for scope, it used to remain in scope until
12921 the end of the scope which contained the for statement (rather than just
12922 within the for scope). G++ retains this, but issues a warning, if such a
12923 variable is accessed outside the for scope.
12925 @item Implicit C language
12926 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
12927 scope to set the language. On such systems, all header files are
12928 implicitly scoped inside a C language scope. Also, an empty prototype
12929 @code{()} will be treated as an unspecified number of arguments, rather
12930 than no arguments, as C++ demands.