1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006, 2007 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 @node Nested Functions
374 @section Nested Functions
375 @cindex nested functions
376 @cindex downward funargs
379 A @dfn{nested function} is a function defined inside another function.
380 (Nested functions are not supported for GNU C++.) The nested function's
381 name is local to the block where it is defined. For example, here we
382 define a nested function named @code{square}, and call it twice:
386 foo (double a, double b)
388 double square (double z) @{ return z * z; @}
390 return square (a) + square (b);
395 The nested function can access all the variables of the containing
396 function that are visible at the point of its definition. This is
397 called @dfn{lexical scoping}. For example, here we show a nested
398 function which uses an inherited variable named @code{offset}:
402 bar (int *array, int offset, int size)
404 int access (int *array, int index)
405 @{ return array[index + offset]; @}
408 for (i = 0; i < size; i++)
409 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
414 Nested function definitions are permitted within functions in the places
415 where variable definitions are allowed; that is, in any block, mixed
416 with the other declarations and statements in the block.
418 It is possible to call the nested function from outside the scope of its
419 name by storing its address or passing the address to another function:
422 hack (int *array, int size)
424 void store (int index, int value)
425 @{ array[index] = value; @}
427 intermediate (store, size);
431 Here, the function @code{intermediate} receives the address of
432 @code{store} as an argument. If @code{intermediate} calls @code{store},
433 the arguments given to @code{store} are used to store into @code{array}.
434 But this technique works only so long as the containing function
435 (@code{hack}, in this example) does not exit.
437 If you try to call the nested function through its address after the
438 containing function has exited, all hell will break loose. If you try
439 to call it after a containing scope level has exited, and if it refers
440 to some of the variables that are no longer in scope, you may be lucky,
441 but it's not wise to take the risk. If, however, the nested function
442 does not refer to anything that has gone out of scope, you should be
445 GCC implements taking the address of a nested function using a technique
446 called @dfn{trampolines}. A paper describing them is available as
449 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
451 A nested function can jump to a label inherited from a containing
452 function, provided the label was explicitly declared in the containing
453 function (@pxref{Local Labels}). Such a jump returns instantly to the
454 containing function, exiting the nested function which did the
455 @code{goto} and any intermediate functions as well. Here is an example:
459 bar (int *array, int offset, int size)
462 int access (int *array, int index)
466 return array[index + offset];
470 for (i = 0; i < size; i++)
471 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
475 /* @r{Control comes here from @code{access}
476 if it detects an error.} */
483 A nested function always has no linkage. Declaring one with
484 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
485 before its definition, use @code{auto} (which is otherwise meaningless
486 for function declarations).
489 bar (int *array, int offset, int size)
492 auto int access (int *, int);
494 int access (int *array, int index)
498 return array[index + offset];
504 @node Constructing Calls
505 @section Constructing Function Calls
506 @cindex constructing calls
507 @cindex forwarding calls
509 Using the built-in functions described below, you can record
510 the arguments a function received, and call another function
511 with the same arguments, without knowing the number or types
514 You can also record the return value of that function call,
515 and later return that value, without knowing what data type
516 the function tried to return (as long as your caller expects
519 However, these built-in functions may interact badly with some
520 sophisticated features or other extensions of the language. It
521 is, therefore, not recommended to use them outside very simple
522 functions acting as mere forwarders for their arguments.
524 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
525 This built-in function returns a pointer to data
526 describing how to perform a call with the same arguments as were passed
527 to the current function.
529 The function saves the arg pointer register, structure value address,
530 and all registers that might be used to pass arguments to a function
531 into a block of memory allocated on the stack. Then it returns the
532 address of that block.
535 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
536 This built-in function invokes @var{function}
537 with a copy of the parameters described by @var{arguments}
540 The value of @var{arguments} should be the value returned by
541 @code{__builtin_apply_args}. The argument @var{size} specifies the size
542 of the stack argument data, in bytes.
544 This function returns a pointer to data describing
545 how to return whatever value was returned by @var{function}. The data
546 is saved in a block of memory allocated on the stack.
548 It is not always simple to compute the proper value for @var{size}. The
549 value is used by @code{__builtin_apply} to compute the amount of data
550 that should be pushed on the stack and copied from the incoming argument
554 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
555 This built-in function returns the value described by @var{result} from
556 the containing function. You should specify, for @var{result}, a value
557 returned by @code{__builtin_apply}.
560 @deftypefn {Built-in Function} __builtin_va_arg_pack ()
561 This built-in function represents all anonymous arguments of an inline
562 function. It can be used only in inline functions which will be always
563 inlined, never compiled as a separate function, such as those using
564 @code{__attribute__ ((__always_inline__))} or
565 @code{__attribute__ ((__gnu_inline__))} extern inline functions.
566 It must be only passed as last argument to some other function
567 with variable arguments. This is useful for writing small wrapper
568 inlines for variable argument functions, when using preprocessor
569 macros is undesirable. For example:
571 extern int myprintf (FILE *f, const char *format, ...);
572 extern inline __attribute__ ((__gnu_inline__)) int
573 myprintf (FILE *f, const char *format, ...)
575 int r = fprintf (f, "myprintf: ");
578 int s = fprintf (f, format, __builtin_va_arg_pack ());
587 @section Referring to a Type with @code{typeof}
590 @cindex macros, types of arguments
592 Another way to refer to the type of an expression is with @code{typeof}.
593 The syntax of using of this keyword looks like @code{sizeof}, but the
594 construct acts semantically like a type name defined with @code{typedef}.
596 There are two ways of writing the argument to @code{typeof}: with an
597 expression or with a type. Here is an example with an expression:
604 This assumes that @code{x} is an array of pointers to functions;
605 the type described is that of the values of the functions.
607 Here is an example with a typename as the argument:
614 Here the type described is that of pointers to @code{int}.
616 If you are writing a header file that must work when included in ISO C
617 programs, write @code{__typeof__} instead of @code{typeof}.
618 @xref{Alternate Keywords}.
620 A @code{typeof}-construct can be used anywhere a typedef name could be
621 used. For example, you can use it in a declaration, in a cast, or inside
622 of @code{sizeof} or @code{typeof}.
624 @code{typeof} is often useful in conjunction with the
625 statements-within-expressions feature. Here is how the two together can
626 be used to define a safe ``maximum'' macro that operates on any
627 arithmetic type and evaluates each of its arguments exactly once:
631 (@{ typeof (a) _a = (a); \
632 typeof (b) _b = (b); \
633 _a > _b ? _a : _b; @})
636 @cindex underscores in variables in macros
637 @cindex @samp{_} in variables in macros
638 @cindex local variables in macros
639 @cindex variables, local, in macros
640 @cindex macros, local variables in
642 The reason for using names that start with underscores for the local
643 variables is to avoid conflicts with variable names that occur within the
644 expressions that are substituted for @code{a} and @code{b}. Eventually we
645 hope to design a new form of declaration syntax that allows you to declare
646 variables whose scopes start only after their initializers; this will be a
647 more reliable way to prevent such conflicts.
650 Some more examples of the use of @code{typeof}:
654 This declares @code{y} with the type of what @code{x} points to.
661 This declares @code{y} as an array of such values.
668 This declares @code{y} as an array of pointers to characters:
671 typeof (typeof (char *)[4]) y;
675 It is equivalent to the following traditional C declaration:
681 To see the meaning of the declaration using @code{typeof}, and why it
682 might be a useful way to write, rewrite it with these macros:
685 #define pointer(T) typeof(T *)
686 #define array(T, N) typeof(T [N])
690 Now the declaration can be rewritten this way:
693 array (pointer (char), 4) y;
697 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
698 pointers to @code{char}.
701 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
702 a more limited extension which permitted one to write
705 typedef @var{T} = @var{expr};
709 with the effect of declaring @var{T} to have the type of the expression
710 @var{expr}. This extension does not work with GCC 3 (versions between
711 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
712 relies on it should be rewritten to use @code{typeof}:
715 typedef typeof(@var{expr}) @var{T};
719 This will work with all versions of GCC@.
722 @section Conditionals with Omitted Operands
723 @cindex conditional expressions, extensions
724 @cindex omitted middle-operands
725 @cindex middle-operands, omitted
726 @cindex extensions, @code{?:}
727 @cindex @code{?:} extensions
729 The middle operand in a conditional expression may be omitted. Then
730 if the first operand is nonzero, its value is the value of the conditional
733 Therefore, the expression
740 has the value of @code{x} if that is nonzero; otherwise, the value of
743 This example is perfectly equivalent to
749 @cindex side effect in ?:
750 @cindex ?: side effect
752 In this simple case, the ability to omit the middle operand is not
753 especially useful. When it becomes useful is when the first operand does,
754 or may (if it is a macro argument), contain a side effect. Then repeating
755 the operand in the middle would perform the side effect twice. Omitting
756 the middle operand uses the value already computed without the undesirable
757 effects of recomputing it.
760 @section Double-Word Integers
761 @cindex @code{long long} data types
762 @cindex double-word arithmetic
763 @cindex multiprecision arithmetic
764 @cindex @code{LL} integer suffix
765 @cindex @code{ULL} integer suffix
767 ISO C99 supports data types for integers that are at least 64 bits wide,
768 and as an extension GCC supports them in C89 mode and in C++.
769 Simply write @code{long long int} for a signed integer, or
770 @code{unsigned long long int} for an unsigned integer. To make an
771 integer constant of type @code{long long int}, add the suffix @samp{LL}
772 to the integer. To make an integer constant of type @code{unsigned long
773 long int}, add the suffix @samp{ULL} to the integer.
775 You can use these types in arithmetic like any other integer types.
776 Addition, subtraction, and bitwise boolean operations on these types
777 are open-coded on all types of machines. Multiplication is open-coded
778 if the machine supports fullword-to-doubleword a widening multiply
779 instruction. Division and shifts are open-coded only on machines that
780 provide special support. The operations that are not open-coded use
781 special library routines that come with GCC@.
783 There may be pitfalls when you use @code{long long} types for function
784 arguments, unless you declare function prototypes. If a function
785 expects type @code{int} for its argument, and you pass a value of type
786 @code{long long int}, confusion will result because the caller and the
787 subroutine will disagree about the number of bytes for the argument.
788 Likewise, if the function expects @code{long long int} and you pass
789 @code{int}. The best way to avoid such problems is to use prototypes.
792 @section Complex Numbers
793 @cindex complex numbers
794 @cindex @code{_Complex} keyword
795 @cindex @code{__complex__} keyword
797 ISO C99 supports complex floating data types, and as an extension GCC
798 supports them in C89 mode and in C++, and supports complex integer data
799 types which are not part of ISO C99. You can declare complex types
800 using the keyword @code{_Complex}. As an extension, the older GNU
801 keyword @code{__complex__} is also supported.
803 For example, @samp{_Complex double x;} declares @code{x} as a
804 variable whose real part and imaginary part are both of type
805 @code{double}. @samp{_Complex short int y;} declares @code{y} to
806 have real and imaginary parts of type @code{short int}; this is not
807 likely to be useful, but it shows that the set of complex types is
810 To write a constant with a complex data type, use the suffix @samp{i} or
811 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
812 has type @code{_Complex float} and @code{3i} has type
813 @code{_Complex int}. Such a constant always has a pure imaginary
814 value, but you can form any complex value you like by adding one to a
815 real constant. This is a GNU extension; if you have an ISO C99
816 conforming C library (such as GNU libc), and want to construct complex
817 constants of floating type, you should include @code{<complex.h>} and
818 use the macros @code{I} or @code{_Complex_I} instead.
820 @cindex @code{__real__} keyword
821 @cindex @code{__imag__} keyword
822 To extract the real part of a complex-valued expression @var{exp}, write
823 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
824 extract the imaginary part. This is a GNU extension; for values of
825 floating type, you should use the ISO C99 functions @code{crealf},
826 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
827 @code{cimagl}, declared in @code{<complex.h>} and also provided as
828 built-in functions by GCC@.
830 @cindex complex conjugation
831 The operator @samp{~} performs complex conjugation when used on a value
832 with a complex type. This is a GNU extension; for values of
833 floating type, you should use the ISO C99 functions @code{conjf},
834 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
835 provided as built-in functions by GCC@.
837 GCC can allocate complex automatic variables in a noncontiguous
838 fashion; it's even possible for the real part to be in a register while
839 the imaginary part is on the stack (or vice-versa). Only the DWARF2
840 debug info format can represent this, so use of DWARF2 is recommended.
841 If you are using the stabs debug info format, GCC describes a noncontiguous
842 complex variable as if it were two separate variables of noncomplex type.
843 If the variable's actual name is @code{foo}, the two fictitious
844 variables are named @code{foo$real} and @code{foo$imag}. You can
845 examine and set these two fictitious variables with your debugger.
848 @section Additional Floating Types
849 @cindex additional floating types
850 @cindex @code{__float80} data type
851 @cindex @code{__float128} data type
852 @cindex @code{w} floating point suffix
853 @cindex @code{q} floating point suffix
854 @cindex @code{W} floating point suffix
855 @cindex @code{Q} floating point suffix
857 As an extension, the GNU C compiler supports additional floating
858 types, @code{__float80} and @code{__float128} to support 80bit
859 (@code{XFmode}) and 128 bit (@code{TFmode}) floating types.
860 Support for additional types includes the arithmetic operators:
861 add, subtract, multiply, divide; unary arithmetic operators;
862 relational operators; equality operators; and conversions to and from
863 integer and other floating types. Use a suffix @samp{w} or @samp{W}
864 in a literal constant of type @code{__float80} and @samp{q} or @samp{Q}
865 for @code{_float128}. You can declare complex types using the
866 corresponding internal complex type, @code{XCmode} for @code{__float80}
867 type and @code{TCmode} for @code{__float128} type:
870 typedef _Complex float __attribute__((mode(TC))) _Complex128;
871 typedef _Complex float __attribute__((mode(XC))) _Complex80;
874 Not all targets support additional floating point types. @code{__float80}
875 is supported on i386, x86_64 and ia64 targets and target @code{__float128}
876 is supported on x86_64 and ia64 targets.
879 @section Decimal Floating Types
880 @cindex decimal floating types
881 @cindex @code{_Decimal32} data type
882 @cindex @code{_Decimal64} data type
883 @cindex @code{_Decimal128} data type
884 @cindex @code{df} integer suffix
885 @cindex @code{dd} integer suffix
886 @cindex @code{dl} integer suffix
887 @cindex @code{DF} integer suffix
888 @cindex @code{DD} integer suffix
889 @cindex @code{DL} integer suffix
891 As an extension, the GNU C compiler supports decimal floating types as
892 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
893 floating types in GCC will evolve as the draft technical report changes.
894 Calling conventions for any target might also change. Not all targets
895 support decimal floating types.
897 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
898 @code{_Decimal128}. They use a radix of ten, unlike the floating types
899 @code{float}, @code{double}, and @code{long double} whose radix is not
900 specified by the C standard but is usually two.
902 Support for decimal floating types includes the arithmetic operators
903 add, subtract, multiply, divide; unary arithmetic operators;
904 relational operators; equality operators; and conversions to and from
905 integer and other floating types. Use a suffix @samp{df} or
906 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
907 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
910 GCC support of decimal float as specified by the draft technical report
915 Translation time data type (TTDT) is not supported.
918 When the value of a decimal floating type cannot be represented in the
919 integer type to which it is being converted, the result is undefined
920 rather than the result value specified by the draft technical report.
923 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
924 are supported by the DWARF2 debug information format.
930 ISO C99 supports floating-point numbers written not only in the usual
931 decimal notation, such as @code{1.55e1}, but also numbers such as
932 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
933 supports this in C89 mode (except in some cases when strictly
934 conforming) and in C++. In that format the
935 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
936 mandatory. The exponent is a decimal number that indicates the power of
937 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
944 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
945 is the same as @code{1.55e1}.
947 Unlike for floating-point numbers in the decimal notation the exponent
948 is always required in the hexadecimal notation. Otherwise the compiler
949 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
950 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
951 extension for floating-point constants of type @code{float}.
954 @section Fixed-Point Types
955 @cindex fixed-point types
956 @cindex @code{_Fract} data type
957 @cindex @code{_Accum} data type
958 @cindex @code{_Sat} data type
959 @cindex @code{hr} fixed-suffix
960 @cindex @code{r} fixed-suffix
961 @cindex @code{lr} fixed-suffix
962 @cindex @code{llr} fixed-suffix
963 @cindex @code{uhr} fixed-suffix
964 @cindex @code{ur} fixed-suffix
965 @cindex @code{ulr} fixed-suffix
966 @cindex @code{ullr} fixed-suffix
967 @cindex @code{hk} fixed-suffix
968 @cindex @code{k} fixed-suffix
969 @cindex @code{lk} fixed-suffix
970 @cindex @code{llk} fixed-suffix
971 @cindex @code{uhk} fixed-suffix
972 @cindex @code{uk} fixed-suffix
973 @cindex @code{ulk} fixed-suffix
974 @cindex @code{ullk} fixed-suffix
975 @cindex @code{HR} fixed-suffix
976 @cindex @code{R} fixed-suffix
977 @cindex @code{LR} fixed-suffix
978 @cindex @code{LLR} fixed-suffix
979 @cindex @code{UHR} fixed-suffix
980 @cindex @code{UR} fixed-suffix
981 @cindex @code{ULR} fixed-suffix
982 @cindex @code{ULLR} fixed-suffix
983 @cindex @code{HK} fixed-suffix
984 @cindex @code{K} fixed-suffix
985 @cindex @code{LK} fixed-suffix
986 @cindex @code{LLK} fixed-suffix
987 @cindex @code{UHK} fixed-suffix
988 @cindex @code{UK} fixed-suffix
989 @cindex @code{ULK} fixed-suffix
990 @cindex @code{ULLK} fixed-suffix
992 As an extension, the GNU C compiler supports fixed-point types as
993 defined in the N1169 draft of ISO/IEC DTR 18037. Support for fixed-point
994 types in GCC will evolve as the draft technical report changes.
995 Calling conventions for any target might also change. Not all targets
996 support fixed-point types.
998 The fixed-point types are
1002 @code{long long _Fract},
1003 @code{unsigned short _Fract},
1004 @code{unsigned _Fract},
1005 @code{unsigned long _Fract},
1006 @code{unsigned long long _Fract},
1007 @code{_Sat short _Fract},
1009 @code{_Sat long _Fract},
1010 @code{_Sat long long _Fract},
1011 @code{_Sat unsigned short _Fract},
1012 @code{_Sat unsigned _Fract},
1013 @code{_Sat unsigned long _Fract},
1014 @code{_Sat unsigned long long _Fract},
1015 @code{short _Accum},
1018 @code{long long _Accum},
1019 @code{unsigned short _Accum},
1020 @code{unsigned _Accum},
1021 @code{unsigned long _Accum},
1022 @code{unsigned long long _Accum},
1023 @code{_Sat short _Accum},
1025 @code{_Sat long _Accum},
1026 @code{_Sat long long _Accum},
1027 @code{_Sat unsigned short _Accum},
1028 @code{_Sat unsigned _Accum},
1029 @code{_Sat unsigned long _Accum},
1030 @code{_Sat unsigned long long _Accum}.
1031 Fixed-point data values contain fractional and optional integral parts.
1032 The format of fixed-point data varies and depends on the target machine.
1034 Support for fixed-point types includes prefix and postfix increment
1035 and decrement operators (@code{++}, @code{--}); unary arithmetic operators
1036 (@code{+}, @code{-}, @code{!}); binary arithmetic operators (@code{+},
1037 @code{-}, @code{*}, @code{/}); binary shift operators (@code{<<}, @code{>>});
1038 relational operators (@code{<}, @code{<=}, @code{>=}, @code{>});
1039 equality operators (@code{==}, @code{!=}); assignment operators
1040 (@code{+=}, @code{-=}, @code{*=}, @code{/=}, @code{<<=}, @code{>>=});
1041 and conversions to and from integer, floating-point, or fixed-point types.
1043 Use a suffix @samp{hr} or @samp{HR} in a literal constant of type
1044 @code{short _Fract} and @code{_Sat short _Fract},
1045 @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract},
1046 @samp{lr} or @samp{LR} for @code{long _Fract} and @code{_Sat long _Fract},
1047 @samp{llr} or @samp{LLR} for @code{long long _Fract} and
1048 @code{_Sat long long _Fract},
1049 @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
1050 @code{_Sat unsigned short _Fract},
1051 @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
1052 @code{_Sat unsigned _Fract},
1053 @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
1054 @code{_Sat unsigned long _Fract},
1055 @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
1056 and @code{_Sat unsigned long long _Fract},
1057 @samp{hk} or @samp{HK} for @code{short _Accum} and @code{_Sat short _Accum},
1058 @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum},
1059 @samp{lk} or @samp{LK} for @code{long _Accum} and @code{_Sat long _Accum},
1060 @samp{llk} or @samp{LLK} for @code{long long _Accum} and
1061 @code{_Sat long long _Accum},
1062 @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
1063 @code{_Sat unsigned short _Accum},
1064 @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
1065 @code{_Sat unsigned _Accum},
1066 @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
1067 @code{_Sat unsigned long _Accum},
1068 and @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
1069 and @code{_Sat unsigned long long _Accum}.
1071 GCC support of fixed-point types as specified by the draft technical report
1076 Pragmas to control overflow and rounding behaviors are not implemented.
1079 Fixed-point types are supported by the DWARF2 debug information format.
1082 @section Arrays of Length Zero
1083 @cindex arrays of length zero
1084 @cindex zero-length arrays
1085 @cindex length-zero arrays
1086 @cindex flexible array members
1088 Zero-length arrays are allowed in GNU C@. They are very useful as the
1089 last element of a structure which is really a header for a variable-length
1098 struct line *thisline = (struct line *)
1099 malloc (sizeof (struct line) + this_length);
1100 thisline->length = this_length;
1103 In ISO C90, you would have to give @code{contents} a length of 1, which
1104 means either you waste space or complicate the argument to @code{malloc}.
1106 In ISO C99, you would use a @dfn{flexible array member}, which is
1107 slightly different in syntax and semantics:
1111 Flexible array members are written as @code{contents[]} without
1115 Flexible array members have incomplete type, and so the @code{sizeof}
1116 operator may not be applied. As a quirk of the original implementation
1117 of zero-length arrays, @code{sizeof} evaluates to zero.
1120 Flexible array members may only appear as the last member of a
1121 @code{struct} that is otherwise non-empty.
1124 A structure containing a flexible array member, or a union containing
1125 such a structure (possibly recursively), may not be a member of a
1126 structure or an element of an array. (However, these uses are
1127 permitted by GCC as extensions.)
1130 GCC versions before 3.0 allowed zero-length arrays to be statically
1131 initialized, as if they were flexible arrays. In addition to those
1132 cases that were useful, it also allowed initializations in situations
1133 that would corrupt later data. Non-empty initialization of zero-length
1134 arrays is now treated like any case where there are more initializer
1135 elements than the array holds, in that a suitable warning about "excess
1136 elements in array" is given, and the excess elements (all of them, in
1137 this case) are ignored.
1139 Instead GCC allows static initialization of flexible array members.
1140 This is equivalent to defining a new structure containing the original
1141 structure followed by an array of sufficient size to contain the data.
1142 I.e.@: in the following, @code{f1} is constructed as if it were declared
1148 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
1151 struct f1 f1; int data[3];
1152 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1156 The convenience of this extension is that @code{f1} has the desired
1157 type, eliminating the need to consistently refer to @code{f2.f1}.
1159 This has symmetry with normal static arrays, in that an array of
1160 unknown size is also written with @code{[]}.
1162 Of course, this extension only makes sense if the extra data comes at
1163 the end of a top-level object, as otherwise we would be overwriting
1164 data at subsequent offsets. To avoid undue complication and confusion
1165 with initialization of deeply nested arrays, we simply disallow any
1166 non-empty initialization except when the structure is the top-level
1167 object. For example:
1170 struct foo @{ int x; int y[]; @};
1171 struct bar @{ struct foo z; @};
1173 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1174 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1175 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1176 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1179 @node Empty Structures
1180 @section Structures With No Members
1181 @cindex empty structures
1182 @cindex zero-size structures
1184 GCC permits a C structure to have no members:
1191 The structure will have size zero. In C++, empty structures are part
1192 of the language. G++ treats empty structures as if they had a single
1193 member of type @code{char}.
1195 @node Variable Length
1196 @section Arrays of Variable Length
1197 @cindex variable-length arrays
1198 @cindex arrays of variable length
1201 Variable-length automatic arrays are allowed in ISO C99, and as an
1202 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1203 implementation of variable-length arrays does not yet conform in detail
1204 to the ISO C99 standard.) These arrays are
1205 declared like any other automatic arrays, but with a length that is not
1206 a constant expression. The storage is allocated at the point of
1207 declaration and deallocated when the brace-level is exited. For
1212 concat_fopen (char *s1, char *s2, char *mode)
1214 char str[strlen (s1) + strlen (s2) + 1];
1217 return fopen (str, mode);
1221 @cindex scope of a variable length array
1222 @cindex variable-length array scope
1223 @cindex deallocating variable length arrays
1224 Jumping or breaking out of the scope of the array name deallocates the
1225 storage. Jumping into the scope is not allowed; you get an error
1228 @cindex @code{alloca} vs variable-length arrays
1229 You can use the function @code{alloca} to get an effect much like
1230 variable-length arrays. The function @code{alloca} is available in
1231 many other C implementations (but not in all). On the other hand,
1232 variable-length arrays are more elegant.
1234 There are other differences between these two methods. Space allocated
1235 with @code{alloca} exists until the containing @emph{function} returns.
1236 The space for a variable-length array is deallocated as soon as the array
1237 name's scope ends. (If you use both variable-length arrays and
1238 @code{alloca} in the same function, deallocation of a variable-length array
1239 will also deallocate anything more recently allocated with @code{alloca}.)
1241 You can also use variable-length arrays as arguments to functions:
1245 tester (int len, char data[len][len])
1251 The length of an array is computed once when the storage is allocated
1252 and is remembered for the scope of the array in case you access it with
1255 If you want to pass the array first and the length afterward, you can
1256 use a forward declaration in the parameter list---another GNU extension.
1260 tester (int len; char data[len][len], int len)
1266 @cindex parameter forward declaration
1267 The @samp{int len} before the semicolon is a @dfn{parameter forward
1268 declaration}, and it serves the purpose of making the name @code{len}
1269 known when the declaration of @code{data} is parsed.
1271 You can write any number of such parameter forward declarations in the
1272 parameter list. They can be separated by commas or semicolons, but the
1273 last one must end with a semicolon, which is followed by the ``real''
1274 parameter declarations. Each forward declaration must match a ``real''
1275 declaration in parameter name and data type. ISO C99 does not support
1276 parameter forward declarations.
1278 @node Variadic Macros
1279 @section Macros with a Variable Number of Arguments.
1280 @cindex variable number of arguments
1281 @cindex macro with variable arguments
1282 @cindex rest argument (in macro)
1283 @cindex variadic macros
1285 In the ISO C standard of 1999, a macro can be declared to accept a
1286 variable number of arguments much as a function can. The syntax for
1287 defining the macro is similar to that of a function. Here is an
1291 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1294 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1295 such a macro, it represents the zero or more tokens until the closing
1296 parenthesis that ends the invocation, including any commas. This set of
1297 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1298 wherever it appears. See the CPP manual for more information.
1300 GCC has long supported variadic macros, and used a different syntax that
1301 allowed you to give a name to the variable arguments just like any other
1302 argument. Here is an example:
1305 #define debug(format, args...) fprintf (stderr, format, args)
1308 This is in all ways equivalent to the ISO C example above, but arguably
1309 more readable and descriptive.
1311 GNU CPP has two further variadic macro extensions, and permits them to
1312 be used with either of the above forms of macro definition.
1314 In standard C, you are not allowed to leave the variable argument out
1315 entirely; but you are allowed to pass an empty argument. For example,
1316 this invocation is invalid in ISO C, because there is no comma after
1323 GNU CPP permits you to completely omit the variable arguments in this
1324 way. In the above examples, the compiler would complain, though since
1325 the expansion of the macro still has the extra comma after the format
1328 To help solve this problem, CPP behaves specially for variable arguments
1329 used with the token paste operator, @samp{##}. If instead you write
1332 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1335 and if the variable arguments are omitted or empty, the @samp{##}
1336 operator causes the preprocessor to remove the comma before it. If you
1337 do provide some variable arguments in your macro invocation, GNU CPP
1338 does not complain about the paste operation and instead places the
1339 variable arguments after the comma. Just like any other pasted macro
1340 argument, these arguments are not macro expanded.
1342 @node Escaped Newlines
1343 @section Slightly Looser Rules for Escaped Newlines
1344 @cindex escaped newlines
1345 @cindex newlines (escaped)
1347 Recently, the preprocessor has relaxed its treatment of escaped
1348 newlines. Previously, the newline had to immediately follow a
1349 backslash. The current implementation allows whitespace in the form
1350 of spaces, horizontal and vertical tabs, and form feeds between the
1351 backslash and the subsequent newline. The preprocessor issues a
1352 warning, but treats it as a valid escaped newline and combines the two
1353 lines to form a single logical line. This works within comments and
1354 tokens, as well as between tokens. Comments are @emph{not} treated as
1355 whitespace for the purposes of this relaxation, since they have not
1356 yet been replaced with spaces.
1359 @section Non-Lvalue Arrays May Have Subscripts
1360 @cindex subscripting
1361 @cindex arrays, non-lvalue
1363 @cindex subscripting and function values
1364 In ISO C99, arrays that are not lvalues still decay to pointers, and
1365 may be subscripted, although they may not be modified or used after
1366 the next sequence point and the unary @samp{&} operator may not be
1367 applied to them. As an extension, GCC allows such arrays to be
1368 subscripted in C89 mode, though otherwise they do not decay to
1369 pointers outside C99 mode. For example,
1370 this is valid in GNU C though not valid in C89:
1374 struct foo @{int a[4];@};
1380 return f().a[index];
1386 @section Arithmetic on @code{void}- and Function-Pointers
1387 @cindex void pointers, arithmetic
1388 @cindex void, size of pointer to
1389 @cindex function pointers, arithmetic
1390 @cindex function, size of pointer to
1392 In GNU C, addition and subtraction operations are supported on pointers to
1393 @code{void} and on pointers to functions. This is done by treating the
1394 size of a @code{void} or of a function as 1.
1396 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1397 and on function types, and returns 1.
1399 @opindex Wpointer-arith
1400 The option @option{-Wpointer-arith} requests a warning if these extensions
1404 @section Non-Constant Initializers
1405 @cindex initializers, non-constant
1406 @cindex non-constant initializers
1408 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1409 automatic variable are not required to be constant expressions in GNU C@.
1410 Here is an example of an initializer with run-time varying elements:
1413 foo (float f, float g)
1415 float beat_freqs[2] = @{ f-g, f+g @};
1420 @node Compound Literals
1421 @section Compound Literals
1422 @cindex constructor expressions
1423 @cindex initializations in expressions
1424 @cindex structures, constructor expression
1425 @cindex expressions, constructor
1426 @cindex compound literals
1427 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1429 ISO C99 supports compound literals. A compound literal looks like
1430 a cast containing an initializer. Its value is an object of the
1431 type specified in the cast, containing the elements specified in
1432 the initializer; it is an lvalue. As an extension, GCC supports
1433 compound literals in C89 mode and in C++.
1435 Usually, the specified type is a structure. Assume that
1436 @code{struct foo} and @code{structure} are declared as shown:
1439 struct foo @{int a; char b[2];@} structure;
1443 Here is an example of constructing a @code{struct foo} with a compound literal:
1446 structure = ((struct foo) @{x + y, 'a', 0@});
1450 This is equivalent to writing the following:
1454 struct foo temp = @{x + y, 'a', 0@};
1459 You can also construct an array. If all the elements of the compound literal
1460 are (made up of) simple constant expressions, suitable for use in
1461 initializers of objects of static storage duration, then the compound
1462 literal can be coerced to a pointer to its first element and used in
1463 such an initializer, as shown here:
1466 char **foo = (char *[]) @{ "x", "y", "z" @};
1469 Compound literals for scalar types and union types are is
1470 also allowed, but then the compound literal is equivalent
1473 As a GNU extension, GCC allows initialization of objects with static storage
1474 duration by compound literals (which is not possible in ISO C99, because
1475 the initializer is not a constant).
1476 It is handled as if the object was initialized only with the bracket
1477 enclosed list if the types of the compound literal and the object match.
1478 The initializer list of the compound literal must be constant.
1479 If the object being initialized has array type of unknown size, the size is
1480 determined by compound literal size.
1483 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1484 static int y[] = (int []) @{1, 2, 3@};
1485 static int z[] = (int [3]) @{1@};
1489 The above lines are equivalent to the following:
1491 static struct foo x = @{1, 'a', 'b'@};
1492 static int y[] = @{1, 2, 3@};
1493 static int z[] = @{1, 0, 0@};
1496 @node Designated Inits
1497 @section Designated Initializers
1498 @cindex initializers with labeled elements
1499 @cindex labeled elements in initializers
1500 @cindex case labels in initializers
1501 @cindex designated initializers
1503 Standard C89 requires the elements of an initializer to appear in a fixed
1504 order, the same as the order of the elements in the array or structure
1507 In ISO C99 you can give the elements in any order, specifying the array
1508 indices or structure field names they apply to, and GNU C allows this as
1509 an extension in C89 mode as well. This extension is not
1510 implemented in GNU C++.
1512 To specify an array index, write
1513 @samp{[@var{index}] =} before the element value. For example,
1516 int a[6] = @{ [4] = 29, [2] = 15 @};
1523 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1527 The index values must be constant expressions, even if the array being
1528 initialized is automatic.
1530 An alternative syntax for this which has been obsolete since GCC 2.5 but
1531 GCC still accepts is to write @samp{[@var{index}]} before the element
1532 value, with no @samp{=}.
1534 To initialize a range of elements to the same value, write
1535 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1536 extension. For example,
1539 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1543 If the value in it has side-effects, the side-effects will happen only once,
1544 not for each initialized field by the range initializer.
1547 Note that the length of the array is the highest value specified
1550 In a structure initializer, specify the name of a field to initialize
1551 with @samp{.@var{fieldname} =} before the element value. For example,
1552 given the following structure,
1555 struct point @{ int x, y; @};
1559 the following initialization
1562 struct point p = @{ .y = yvalue, .x = xvalue @};
1569 struct point p = @{ xvalue, yvalue @};
1572 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1573 @samp{@var{fieldname}:}, as shown here:
1576 struct point p = @{ y: yvalue, x: xvalue @};
1580 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1581 @dfn{designator}. You can also use a designator (or the obsolete colon
1582 syntax) when initializing a union, to specify which element of the union
1583 should be used. For example,
1586 union foo @{ int i; double d; @};
1588 union foo f = @{ .d = 4 @};
1592 will convert 4 to a @code{double} to store it in the union using
1593 the second element. By contrast, casting 4 to type @code{union foo}
1594 would store it into the union as the integer @code{i}, since it is
1595 an integer. (@xref{Cast to Union}.)
1597 You can combine this technique of naming elements with ordinary C
1598 initialization of successive elements. Each initializer element that
1599 does not have a designator applies to the next consecutive element of the
1600 array or structure. For example,
1603 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1610 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1613 Labeling the elements of an array initializer is especially useful
1614 when the indices are characters or belong to an @code{enum} type.
1619 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1620 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1623 @cindex designator lists
1624 You can also write a series of @samp{.@var{fieldname}} and
1625 @samp{[@var{index}]} designators before an @samp{=} to specify a
1626 nested subobject to initialize; the list is taken relative to the
1627 subobject corresponding to the closest surrounding brace pair. For
1628 example, with the @samp{struct point} declaration above:
1631 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1635 If the same field is initialized multiple times, it will have value from
1636 the last initialization. If any such overridden initialization has
1637 side-effect, it is unspecified whether the side-effect happens or not.
1638 Currently, GCC will discard them and issue a warning.
1641 @section Case Ranges
1643 @cindex ranges in case statements
1645 You can specify a range of consecutive values in a single @code{case} label,
1649 case @var{low} ... @var{high}:
1653 This has the same effect as the proper number of individual @code{case}
1654 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1656 This feature is especially useful for ranges of ASCII character codes:
1662 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1663 it may be parsed wrong when you use it with integer values. For example,
1678 @section Cast to a Union Type
1679 @cindex cast to a union
1680 @cindex union, casting to a
1682 A cast to union type is similar to other casts, except that the type
1683 specified is a union type. You can specify the type either with
1684 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1685 a constructor though, not a cast, and hence does not yield an lvalue like
1686 normal casts. (@xref{Compound Literals}.)
1688 The types that may be cast to the union type are those of the members
1689 of the union. Thus, given the following union and variables:
1692 union foo @{ int i; double d; @};
1698 both @code{x} and @code{y} can be cast to type @code{union foo}.
1700 Using the cast as the right-hand side of an assignment to a variable of
1701 union type is equivalent to storing in a member of the union:
1706 u = (union foo) x @equiv{} u.i = x
1707 u = (union foo) y @equiv{} u.d = y
1710 You can also use the union cast as a function argument:
1713 void hack (union foo);
1715 hack ((union foo) x);
1718 @node Mixed Declarations
1719 @section Mixed Declarations and Code
1720 @cindex mixed declarations and code
1721 @cindex declarations, mixed with code
1722 @cindex code, mixed with declarations
1724 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1725 within compound statements. As an extension, GCC also allows this in
1726 C89 mode. For example, you could do:
1735 Each identifier is visible from where it is declared until the end of
1736 the enclosing block.
1738 @node Function Attributes
1739 @section Declaring Attributes of Functions
1740 @cindex function attributes
1741 @cindex declaring attributes of functions
1742 @cindex functions that never return
1743 @cindex functions that return more than once
1744 @cindex functions that have no side effects
1745 @cindex functions in arbitrary sections
1746 @cindex functions that behave like malloc
1747 @cindex @code{volatile} applied to function
1748 @cindex @code{const} applied to function
1749 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1750 @cindex functions with non-null pointer arguments
1751 @cindex functions that are passed arguments in registers on the 386
1752 @cindex functions that pop the argument stack on the 386
1753 @cindex functions that do not pop the argument stack on the 386
1755 In GNU C, you declare certain things about functions called in your program
1756 which help the compiler optimize function calls and check your code more
1759 The keyword @code{__attribute__} allows you to specify special
1760 attributes when making a declaration. This keyword is followed by an
1761 attribute specification inside double parentheses. The following
1762 attributes are currently defined for functions on all targets:
1763 @code{aligned}, @code{alloc_size}, @code{noreturn},
1764 @code{returns_twice}, @code{noinline}, @code{always_inline},
1765 @code{flatten}, @code{pure}, @code{const}, @code{nothrow},
1766 @code{sentinel}, @code{format}, @code{format_arg},
1767 @code{no_instrument_function}, @code{section}, @code{constructor},
1768 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1769 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1770 @code{nonnull}, @code{gnu_inline} and @code{externally_visible},
1771 @code{hot}, @code{cold}.
1772 Several other attributes are defined for functions on particular
1773 target systems. Other attributes, including @code{section} are
1774 supported for variables declarations (@pxref{Variable Attributes}) and
1775 for types (@pxref{Type Attributes}).
1777 You may also specify attributes with @samp{__} preceding and following
1778 each keyword. This allows you to use them in header files without
1779 being concerned about a possible macro of the same name. For example,
1780 you may use @code{__noreturn__} instead of @code{noreturn}.
1782 @xref{Attribute Syntax}, for details of the exact syntax for using
1786 @c Keep this table alphabetized by attribute name. Treat _ as space.
1788 @item alias ("@var{target}")
1789 @cindex @code{alias} attribute
1790 The @code{alias} attribute causes the declaration to be emitted as an
1791 alias for another symbol, which must be specified. For instance,
1794 void __f () @{ /* @r{Do something.} */; @}
1795 void f () __attribute__ ((weak, alias ("__f")));
1798 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1799 mangled name for the target must be used. It is an error if @samp{__f}
1800 is not defined in the same translation unit.
1802 Not all target machines support this attribute.
1804 @item aligned (@var{alignment})
1805 @cindex @code{aligned} attribute
1806 This attribute specifies a minimum alignment for the function,
1809 You cannot use this attribute to decrease the alignment of a function,
1810 only to increase it. However, when you explicitly specify a function
1811 alignment this will override the effect of the
1812 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1815 Note that the effectiveness of @code{aligned} attributes may be
1816 limited by inherent limitations in your linker. On many systems, the
1817 linker is only able to arrange for functions to be aligned up to a
1818 certain maximum alignment. (For some linkers, the maximum supported
1819 alignment may be very very small.) See your linker documentation for
1820 further information.
1822 The @code{aligned} attribute can also be used for variables and fields
1823 (@pxref{Variable Attributes}.)
1826 @cindex @code{alloc_size} attribute
1827 The @code{alloc_size} attribute is used to tell the compiler that the
1828 function return value points to memory, where the size is given by
1829 one or two of the functions parameters. GCC uses this
1830 information to improve the correctness of @code{__builtin_object_size}.
1832 The function parameter(s) denoting the allocated size are specified by
1833 one or two integer arguments supplied to the attribute. The allocated size
1834 is either the value of the single function argument specified or the product
1835 of the two function arguments specified. Argument numbering starts at
1841 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1842 void my_realloc(void* size_t) __attribute__((alloc_size(2)))
1845 declares that my_calloc will return memory of the size given by
1846 the product of parameter 1 and 2 and that my_realloc will return memory
1847 of the size given by parameter 2.
1850 @cindex @code{always_inline} function attribute
1851 Generally, functions are not inlined unless optimization is specified.
1852 For functions declared inline, this attribute inlines the function even
1853 if no optimization level was specified.
1856 @cindex @code{gnu_inline} function attribute
1857 This attribute should be used with a function which is also declared
1858 with the @code{inline} keyword. It directs GCC to treat the function
1859 as if it were defined in gnu89 mode even when compiling in C99 or
1862 If the function is declared @code{extern}, then this definition of the
1863 function is used only for inlining. In no case is the function
1864 compiled as a standalone function, not even if you take its address
1865 explicitly. Such an address becomes an external reference, as if you
1866 had only declared the function, and had not defined it. This has
1867 almost the effect of a macro. The way to use this is to put a
1868 function definition in a header file with this attribute, and put
1869 another copy of the function, without @code{extern}, in a library
1870 file. The definition in the header file will cause most calls to the
1871 function to be inlined. If any uses of the function remain, they will
1872 refer to the single copy in the library. Note that the two
1873 definitions of the functions need not be precisely the same, although
1874 if they do not have the same effect your program may behave oddly.
1876 In C, if the function is neither @code{extern} nor @code{static}, then
1877 the function is compiled as a standalone function, as well as being
1878 inlined where possible.
1880 This is how GCC traditionally handled functions declared
1881 @code{inline}. Since ISO C99 specifies a different semantics for
1882 @code{inline}, this function attribute is provided as a transition
1883 measure and as a useful feature in its own right. This attribute is
1884 available in GCC 4.1.3 and later. It is available if either of the
1885 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1886 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1887 Function is As Fast As a Macro}.
1889 In C++, this attribute does not depend on @code{extern} in any way,
1890 but it still requires the @code{inline} keyword to enable its special
1893 @cindex @code{flatten} function attribute
1895 Generally, inlining into a function is limited. For a function marked with
1896 this attribute, every call inside this function will be inlined, if possible.
1897 Whether the function itself is considered for inlining depends on its size and
1898 the current inlining parameters. The @code{flatten} attribute only works
1899 reliably in unit-at-a-time mode.
1902 @cindex functions that do pop the argument stack on the 386
1904 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1905 assume that the calling function will pop off the stack space used to
1906 pass arguments. This is
1907 useful to override the effects of the @option{-mrtd} switch.
1910 @cindex @code{const} function attribute
1911 Many functions do not examine any values except their arguments, and
1912 have no effects except the return value. Basically this is just slightly
1913 more strict class than the @code{pure} attribute below, since function is not
1914 allowed to read global memory.
1916 @cindex pointer arguments
1917 Note that a function that has pointer arguments and examines the data
1918 pointed to must @emph{not} be declared @code{const}. Likewise, a
1919 function that calls a non-@code{const} function usually must not be
1920 @code{const}. It does not make sense for a @code{const} function to
1923 The attribute @code{const} is not implemented in GCC versions earlier
1924 than 2.5. An alternative way to declare that a function has no side
1925 effects, which works in the current version and in some older versions,
1929 typedef int intfn ();
1931 extern const intfn square;
1934 This approach does not work in GNU C++ from 2.6.0 on, since the language
1935 specifies that the @samp{const} must be attached to the return value.
1939 @itemx constructor (@var{priority})
1940 @itemx destructor (@var{priority})
1941 @cindex @code{constructor} function attribute
1942 @cindex @code{destructor} function attribute
1943 The @code{constructor} attribute causes the function to be called
1944 automatically before execution enters @code{main ()}. Similarly, the
1945 @code{destructor} attribute causes the function to be called
1946 automatically after @code{main ()} has completed or @code{exit ()} has
1947 been called. Functions with these attributes are useful for
1948 initializing data that will be used implicitly during the execution of
1951 You may provide an optional integer priority to control the order in
1952 which constructor and destructor functions are run. A constructor
1953 with a smaller priority number runs before a constructor with a larger
1954 priority number; the opposite relationship holds for destructors. So,
1955 if you have a constructor that allocates a resource and a destructor
1956 that deallocates the same resource, both functions typically have the
1957 same priority. The priorities for constructor and destructor
1958 functions are the same as those specified for namespace-scope C++
1959 objects (@pxref{C++ Attributes}).
1961 These attributes are not currently implemented for Objective-C@.
1964 @cindex @code{deprecated} attribute.
1965 The @code{deprecated} attribute results in a warning if the function
1966 is used anywhere in the source file. This is useful when identifying
1967 functions that are expected to be removed in a future version of a
1968 program. The warning also includes the location of the declaration
1969 of the deprecated function, to enable users to easily find further
1970 information about why the function is deprecated, or what they should
1971 do instead. Note that the warnings only occurs for uses:
1974 int old_fn () __attribute__ ((deprecated));
1976 int (*fn_ptr)() = old_fn;
1979 results in a warning on line 3 but not line 2.
1981 The @code{deprecated} attribute can also be used for variables and
1982 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1985 @cindex @code{__declspec(dllexport)}
1986 On Microsoft Windows targets and Symbian OS targets the
1987 @code{dllexport} attribute causes the compiler to provide a global
1988 pointer to a pointer in a DLL, so that it can be referenced with the
1989 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1990 name is formed by combining @code{_imp__} and the function or variable
1993 You can use @code{__declspec(dllexport)} as a synonym for
1994 @code{__attribute__ ((dllexport))} for compatibility with other
1997 On systems that support the @code{visibility} attribute, this
1998 attribute also implies ``default'' visibility. It is an error to
1999 explicitly specify any other visibility.
2001 Currently, the @code{dllexport} attribute is ignored for inlined
2002 functions, unless the @option{-fkeep-inline-functions} flag has been
2003 used. The attribute is also ignored for undefined symbols.
2005 When applied to C++ classes, the attribute marks defined non-inlined
2006 member functions and static data members as exports. Static consts
2007 initialized in-class are not marked unless they are also defined
2010 For Microsoft Windows targets there are alternative methods for
2011 including the symbol in the DLL's export table such as using a
2012 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
2013 the @option{--export-all} linker flag.
2016 @cindex @code{__declspec(dllimport)}
2017 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
2018 attribute causes the compiler to reference a function or variable via
2019 a global pointer to a pointer that is set up by the DLL exporting the
2020 symbol. The attribute implies @code{extern}. On Microsoft Windows
2021 targets, the pointer name is formed by combining @code{_imp__} and the
2022 function or variable name.
2024 You can use @code{__declspec(dllimport)} as a synonym for
2025 @code{__attribute__ ((dllimport))} for compatibility with other
2028 On systems that support the @code{visibility} attribute, this
2029 attribute also implies ``default'' visibility. It is an error to
2030 explicitly specify any other visibility.
2032 Currently, the attribute is ignored for inlined functions. If the
2033 attribute is applied to a symbol @emph{definition}, an error is reported.
2034 If a symbol previously declared @code{dllimport} is later defined, the
2035 attribute is ignored in subsequent references, and a warning is emitted.
2036 The attribute is also overridden by a subsequent declaration as
2039 When applied to C++ classes, the attribute marks non-inlined
2040 member functions and static data members as imports. However, the
2041 attribute is ignored for virtual methods to allow creation of vtables
2044 On the SH Symbian OS target the @code{dllimport} attribute also has
2045 another affect---it can cause the vtable and run-time type information
2046 for a class to be exported. This happens when the class has a
2047 dllimport'ed constructor or a non-inline, non-pure virtual function
2048 and, for either of those two conditions, the class also has a inline
2049 constructor or destructor and has a key function that is defined in
2050 the current translation unit.
2052 For Microsoft Windows based targets the use of the @code{dllimport}
2053 attribute on functions is not necessary, but provides a small
2054 performance benefit by eliminating a thunk in the DLL@. The use of the
2055 @code{dllimport} attribute on imported variables was required on older
2056 versions of the GNU linker, but can now be avoided by passing the
2057 @option{--enable-auto-import} switch to the GNU linker. As with
2058 functions, using the attribute for a variable eliminates a thunk in
2061 One drawback to using this attribute is that a pointer to a function
2062 or variable marked as @code{dllimport} cannot be used as a constant
2063 address. On Microsoft Windows targets, the attribute can be disabled
2064 for functions by setting the @option{-mnop-fun-dllimport} flag.
2067 @cindex eight bit data on the H8/300, H8/300H, and H8S
2068 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2069 variable should be placed into the eight bit data section.
2070 The compiler will generate more efficient code for certain operations
2071 on data in the eight bit data area. Note the eight bit data area is limited to
2074 You must use GAS and GLD from GNU binutils version 2.7 or later for
2075 this attribute to work correctly.
2077 @item exception_handler
2078 @cindex exception handler functions on the Blackfin processor
2079 Use this attribute on the Blackfin to indicate that the specified function
2080 is an exception handler. The compiler will generate function entry and
2081 exit sequences suitable for use in an exception handler when this
2082 attribute is present.
2085 @cindex functions which handle memory bank switching
2086 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2087 use a calling convention that takes care of switching memory banks when
2088 entering and leaving a function. This calling convention is also the
2089 default when using the @option{-mlong-calls} option.
2091 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2092 to call and return from a function.
2094 On 68HC11 the compiler will generate a sequence of instructions
2095 to invoke a board-specific routine to switch the memory bank and call the
2096 real function. The board-specific routine simulates a @code{call}.
2097 At the end of a function, it will jump to a board-specific routine
2098 instead of using @code{rts}. The board-specific return routine simulates
2102 @cindex functions that pop the argument stack on the 386
2103 On the Intel 386, the @code{fastcall} attribute causes the compiler to
2104 pass the first argument (if of integral type) in the register ECX and
2105 the second argument (if of integral type) in the register EDX@. Subsequent
2106 and other typed arguments are passed on the stack. The called function will
2107 pop the arguments off the stack. If the number of arguments is variable all
2108 arguments are pushed on the stack.
2110 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2111 @cindex @code{format} function attribute
2113 The @code{format} attribute specifies that a function takes @code{printf},
2114 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2115 should be type-checked against a format string. For example, the
2120 my_printf (void *my_object, const char *my_format, ...)
2121 __attribute__ ((format (printf, 2, 3)));
2125 causes the compiler to check the arguments in calls to @code{my_printf}
2126 for consistency with the @code{printf} style format string argument
2129 The parameter @var{archetype} determines how the format string is
2130 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2131 or @code{strfmon}. (You can also use @code{__printf__},
2132 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
2133 parameter @var{string-index} specifies which argument is the format
2134 string argument (starting from 1), while @var{first-to-check} is the
2135 number of the first argument to check against the format string. For
2136 functions where the arguments are not available to be checked (such as
2137 @code{vprintf}), specify the third parameter as zero. In this case the
2138 compiler only checks the format string for consistency. For
2139 @code{strftime} formats, the third parameter is required to be zero.
2140 Since non-static C++ methods have an implicit @code{this} argument, the
2141 arguments of such methods should be counted from two, not one, when
2142 giving values for @var{string-index} and @var{first-to-check}.
2144 In the example above, the format string (@code{my_format}) is the second
2145 argument of the function @code{my_print}, and the arguments to check
2146 start with the third argument, so the correct parameters for the format
2147 attribute are 2 and 3.
2149 @opindex ffreestanding
2150 @opindex fno-builtin
2151 The @code{format} attribute allows you to identify your own functions
2152 which take format strings as arguments, so that GCC can check the
2153 calls to these functions for errors. The compiler always (unless
2154 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
2155 for the standard library functions @code{printf}, @code{fprintf},
2156 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2157 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2158 warnings are requested (using @option{-Wformat}), so there is no need to
2159 modify the header file @file{stdio.h}. In C99 mode, the functions
2160 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2161 @code{vsscanf} are also checked. Except in strictly conforming C
2162 standard modes, the X/Open function @code{strfmon} is also checked as
2163 are @code{printf_unlocked} and @code{fprintf_unlocked}.
2164 @xref{C Dialect Options,,Options Controlling C Dialect}.
2166 The target may provide additional types of format checks.
2167 @xref{Target Format Checks,,Format Checks Specific to Particular
2170 @item format_arg (@var{string-index})
2171 @cindex @code{format_arg} function attribute
2172 @opindex Wformat-nonliteral
2173 The @code{format_arg} attribute specifies that a function takes a format
2174 string for a @code{printf}, @code{scanf}, @code{strftime} or
2175 @code{strfmon} style function and modifies it (for example, to translate
2176 it into another language), so the result can be passed to a
2177 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2178 function (with the remaining arguments to the format function the same
2179 as they would have been for the unmodified string). For example, the
2184 my_dgettext (char *my_domain, const char *my_format)
2185 __attribute__ ((format_arg (2)));
2189 causes the compiler to check the arguments in calls to a @code{printf},
2190 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2191 format string argument is a call to the @code{my_dgettext} function, for
2192 consistency with the format string argument @code{my_format}. If the
2193 @code{format_arg} attribute had not been specified, all the compiler
2194 could tell in such calls to format functions would be that the format
2195 string argument is not constant; this would generate a warning when
2196 @option{-Wformat-nonliteral} is used, but the calls could not be checked
2197 without the attribute.
2199 The parameter @var{string-index} specifies which argument is the format
2200 string argument (starting from one). Since non-static C++ methods have
2201 an implicit @code{this} argument, the arguments of such methods should
2202 be counted from two.
2204 The @code{format-arg} attribute allows you to identify your own
2205 functions which modify format strings, so that GCC can check the
2206 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2207 type function whose operands are a call to one of your own function.
2208 The compiler always treats @code{gettext}, @code{dgettext}, and
2209 @code{dcgettext} in this manner except when strict ISO C support is
2210 requested by @option{-ansi} or an appropriate @option{-std} option, or
2211 @option{-ffreestanding} or @option{-fno-builtin}
2212 is used. @xref{C Dialect Options,,Options
2213 Controlling C Dialect}.
2215 @item function_vector
2216 @cindex calling functions through the function vector on H8/300, M16C, and M32C processors
2217 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2218 function should be called through the function vector. Calling a
2219 function through the function vector will reduce code size, however;
2220 the function vector has a limited size (maximum 128 entries on the H8/300
2221 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2223 You must use GAS and GLD from GNU binutils version 2.7 or later for
2224 this attribute to work correctly.
2226 On M16C/M32C targets, the @code{function_vector} attribute declares a
2227 special page subroutine call function. Use of this attribute reduces
2228 the code size by 2 bytes for each call generated to the
2229 subroutine. The argument to the attribute is the vector number entry
2230 from the special page vector table which contains the 16 low-order
2231 bits of the subroutine's entry address. Each vector table has special
2232 page number (18 to 255) which are used in @code{jsrs} instruction.
2233 Jump addresses of the routines are generated by adding 0x0F0000 (in
2234 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2235 byte addresses set in the vector table. Therefore you need to ensure
2236 that all the special page vector routines should get mapped within the
2237 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2240 In the following example 2 bytes will be saved for each call to
2241 function @code{foo}.
2244 void foo (void) __attribute__((function_vector(0x18)));
2255 If functions are defined in one file and are called in another file,
2256 then be sure to write this declaration in both files.
2258 This attribute is ignored for R8C target.
2261 @cindex interrupt handler functions
2262 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, m68k, MS1,
2263 and Xstormy16 ports to indicate that the specified function is an
2264 interrupt handler. The compiler will generate function entry and exit
2265 sequences suitable for use in an interrupt handler when this attribute
2268 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2269 SH processors can be specified via the @code{interrupt_handler} attribute.
2271 Note, on the AVR, interrupts will be enabled inside the function.
2273 Note, for the ARM, you can specify the kind of interrupt to be handled by
2274 adding an optional parameter to the interrupt attribute like this:
2277 void f () __attribute__ ((interrupt ("IRQ")));
2280 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2282 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2283 may be called with a word aligned stack pointer.
2285 @item interrupt_handler
2286 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2287 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2288 indicate that the specified function is an interrupt handler. The compiler
2289 will generate function entry and exit sequences suitable for use in an
2290 interrupt handler when this attribute is present.
2292 @item interrupt_thread
2293 @cindex interrupt thread functions on fido
2294 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2295 that the specified function is an interrupt handler that is designed
2296 to run as a thread. The compiler omits generate prologue/epilogue
2297 sequences and replaces the return instruction with a @code{sleep}
2298 instruction. This attribute is available only on fido.
2301 @cindex User stack pointer in interrupts on the Blackfin
2302 When used together with @code{interrupt_handler}, @code{exception_handler}
2303 or @code{nmi_handler}, code will be generated to load the stack pointer
2304 from the USP register in the function prologue.
2307 @cindex @code{l1_text} function attribute
2308 This attribute specifies a function to be placed into L1 Instruction
2309 SRAM. The function will be put into a specific section named @code{.l1.text}.
2310 With @option{-mfdpic}, function calls with a such function as the callee
2311 or caller will use inlined PLT.
2313 @item long_call/short_call
2314 @cindex indirect calls on ARM
2315 This attribute specifies how a particular function is called on
2316 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2317 command line switch and @code{#pragma long_calls} settings. The
2318 @code{long_call} attribute indicates that the function might be far
2319 away from the call site and require a different (more expensive)
2320 calling sequence. The @code{short_call} attribute always places
2321 the offset to the function from the call site into the @samp{BL}
2322 instruction directly.
2324 @item longcall/shortcall
2325 @cindex functions called via pointer on the RS/6000 and PowerPC
2326 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2327 indicates that the function might be far away from the call site and
2328 require a different (more expensive) calling sequence. The
2329 @code{shortcall} attribute indicates that the function is always close
2330 enough for the shorter calling sequence to be used. These attributes
2331 override both the @option{-mlongcall} switch and, on the RS/6000 and
2332 PowerPC, the @code{#pragma longcall} setting.
2334 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2335 calls are necessary.
2337 @item long_call/near/far
2338 @cindex indirect calls on MIPS
2339 These attributes specify how a particular function is called on MIPS@.
2340 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2341 command-line switch. The @code{long_call} and @code{far} attributes are
2342 synonyms, and cause the compiler to always call
2343 the function by first loading its address into a register, and then using
2344 the contents of that register. The @code{near} attribute has the opposite
2345 effect; it specifies that non-PIC calls should be made using the more
2346 efficient @code{jal} instruction.
2349 @cindex @code{malloc} attribute
2350 The @code{malloc} attribute is used to tell the compiler that a function
2351 may be treated as if any non-@code{NULL} pointer it returns cannot
2352 alias any other pointer valid when the function returns.
2353 This will often improve optimization.
2354 Standard functions with this property include @code{malloc} and
2355 @code{calloc}. @code{realloc}-like functions have this property as
2356 long as the old pointer is never referred to (including comparing it
2357 to the new pointer) after the function returns a non-@code{NULL}
2360 @item mips16/nomips16
2361 @cindex @code{mips16} attribute
2362 @cindex @code{nomips16} attribute
2364 On MIPS targets, you can use the @code{mips16} and @code{nomips16}
2365 function attributes to locally select or turn off MIPS16 code generation.
2366 A function with the @code{mips16} attribute is emitted as MIPS16 code,
2367 while MIPS16 code generation is disabled for functions with the
2368 @code{nomips16} attribute. These attributes override the
2369 @option{-mips16} and @option{-mno-mips16} options on the command line
2370 (@pxref{MIPS Options}).
2372 When compiling files containing mixed MIPS16 and non-MIPS16 code, the
2373 preprocessor symbol @code{__mips16} reflects the setting on the command line,
2374 not that within individual functions. Mixed MIPS16 and non-MIPS16 code
2375 may interact badly with some GCC extensions such as @code{__builtin_apply}
2376 (@pxref{Constructing Calls}).
2378 @item model (@var{model-name})
2379 @cindex function addressability on the M32R/D
2380 @cindex variable addressability on the IA-64
2382 On the M32R/D, use this attribute to set the addressability of an
2383 object, and of the code generated for a function. The identifier
2384 @var{model-name} is one of @code{small}, @code{medium}, or
2385 @code{large}, representing each of the code models.
2387 Small model objects live in the lower 16MB of memory (so that their
2388 addresses can be loaded with the @code{ld24} instruction), and are
2389 callable with the @code{bl} instruction.
2391 Medium model objects may live anywhere in the 32-bit address space (the
2392 compiler will generate @code{seth/add3} instructions to load their addresses),
2393 and are callable with the @code{bl} instruction.
2395 Large model objects may live anywhere in the 32-bit address space (the
2396 compiler will generate @code{seth/add3} instructions to load their addresses),
2397 and may not be reachable with the @code{bl} instruction (the compiler will
2398 generate the much slower @code{seth/add3/jl} instruction sequence).
2400 On IA-64, use this attribute to set the addressability of an object.
2401 At present, the only supported identifier for @var{model-name} is
2402 @code{small}, indicating addressability via ``small'' (22-bit)
2403 addresses (so that their addresses can be loaded with the @code{addl}
2404 instruction). Caveat: such addressing is by definition not position
2405 independent and hence this attribute must not be used for objects
2406 defined by shared libraries.
2409 @cindex function without a prologue/epilogue code
2410 Use this attribute on the ARM, AVR, C4x, IP2K and SPU ports to indicate that
2411 the specified function does not need prologue/epilogue sequences generated by
2412 the compiler. It is up to the programmer to provide these sequences.
2415 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2416 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2417 use the normal calling convention based on @code{jsr} and @code{rts}.
2418 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2422 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2423 Use this attribute together with @code{interrupt_handler},
2424 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2425 entry code should enable nested interrupts or exceptions.
2428 @cindex NMI handler functions on the Blackfin processor
2429 Use this attribute on the Blackfin to indicate that the specified function
2430 is an NMI handler. The compiler will generate function entry and
2431 exit sequences suitable for use in an NMI handler when this
2432 attribute is present.
2434 @item no_instrument_function
2435 @cindex @code{no_instrument_function} function attribute
2436 @opindex finstrument-functions
2437 If @option{-finstrument-functions} is given, profiling function calls will
2438 be generated at entry and exit of most user-compiled functions.
2439 Functions with this attribute will not be so instrumented.
2442 @cindex @code{noinline} function attribute
2443 This function attribute prevents a function from being considered for
2446 @item nonnull (@var{arg-index}, @dots{})
2447 @cindex @code{nonnull} function attribute
2448 The @code{nonnull} attribute specifies that some function parameters should
2449 be non-null pointers. For instance, the declaration:
2453 my_memcpy (void *dest, const void *src, size_t len)
2454 __attribute__((nonnull (1, 2)));
2458 causes the compiler to check that, in calls to @code{my_memcpy},
2459 arguments @var{dest} and @var{src} are non-null. If the compiler
2460 determines that a null pointer is passed in an argument slot marked
2461 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2462 is issued. The compiler may also choose to make optimizations based
2463 on the knowledge that certain function arguments will not be null.
2465 If no argument index list is given to the @code{nonnull} attribute,
2466 all pointer arguments are marked as non-null. To illustrate, the
2467 following declaration is equivalent to the previous example:
2471 my_memcpy (void *dest, const void *src, size_t len)
2472 __attribute__((nonnull));
2476 @cindex @code{noreturn} function attribute
2477 A few standard library functions, such as @code{abort} and @code{exit},
2478 cannot return. GCC knows this automatically. Some programs define
2479 their own functions that never return. You can declare them
2480 @code{noreturn} to tell the compiler this fact. For example,
2484 void fatal () __attribute__ ((noreturn));
2487 fatal (/* @r{@dots{}} */)
2489 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2495 The @code{noreturn} keyword tells the compiler to assume that
2496 @code{fatal} cannot return. It can then optimize without regard to what
2497 would happen if @code{fatal} ever did return. This makes slightly
2498 better code. More importantly, it helps avoid spurious warnings of
2499 uninitialized variables.
2501 The @code{noreturn} keyword does not affect the exceptional path when that
2502 applies: a @code{noreturn}-marked function may still return to the caller
2503 by throwing an exception or calling @code{longjmp}.
2505 Do not assume that registers saved by the calling function are
2506 restored before calling the @code{noreturn} function.
2508 It does not make sense for a @code{noreturn} function to have a return
2509 type other than @code{void}.
2511 The attribute @code{noreturn} is not implemented in GCC versions
2512 earlier than 2.5. An alternative way to declare that a function does
2513 not return, which works in the current version and in some older
2514 versions, is as follows:
2517 typedef void voidfn ();
2519 volatile voidfn fatal;
2522 This approach does not work in GNU C++.
2525 @cindex @code{nothrow} function attribute
2526 The @code{nothrow} attribute is used to inform the compiler that a
2527 function cannot throw an exception. For example, most functions in
2528 the standard C library can be guaranteed not to throw an exception
2529 with the notable exceptions of @code{qsort} and @code{bsearch} that
2530 take function pointer arguments. The @code{nothrow} attribute is not
2531 implemented in GCC versions earlier than 3.3.
2534 @cindex @code{pure} function attribute
2535 Many functions have no effects except the return value and their
2536 return value depends only on the parameters and/or global variables.
2537 Such a function can be subject
2538 to common subexpression elimination and loop optimization just as an
2539 arithmetic operator would be. These functions should be declared
2540 with the attribute @code{pure}. For example,
2543 int square (int) __attribute__ ((pure));
2547 says that the hypothetical function @code{square} is safe to call
2548 fewer times than the program says.
2550 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2551 Interesting non-pure functions are functions with infinite loops or those
2552 depending on volatile memory or other system resource, that may change between
2553 two consecutive calls (such as @code{feof} in a multithreading environment).
2555 The attribute @code{pure} is not implemented in GCC versions earlier
2559 @cindex @code{hot} function attribute
2560 The @code{hot} attribute is used to inform the compiler that a function is a
2561 hot spot of the compiled program. The function is optimized more aggressively
2562 and on many target it is placed into special subsection of the text section so
2563 all hot functions appears close together improving locality.
2565 When profile feedback is available, via @option{-fprofile-use}, hot functions
2566 are automatically detected and this attribute is ignored.
2568 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2571 @cindex @code{cold} function attribute
2572 The @code{cold} attribute is used to inform the compiler that a function is
2573 unlikely executed. The function is optimized for size rather than speed and on
2574 many targets it is placed into special subsection of the text section so all
2575 cold functions appears close together improving code locality of non-cold parts
2576 of program. The paths leading to call of cold functions within code are marked
2577 as unlikely by the branch prediction mechanism. It is thus useful to mark
2578 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2579 improve optimization of hot functions that do call marked functions in rare
2582 When profile feedback is available, via @option{-fprofile-use}, hot functions
2583 are automatically detected and this attribute is ignored.
2585 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2587 @item regparm (@var{number})
2588 @cindex @code{regparm} attribute
2589 @cindex functions that are passed arguments in registers on the 386
2590 On the Intel 386, the @code{regparm} attribute causes the compiler to
2591 pass arguments number one to @var{number} if they are of integral type
2592 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2593 take a variable number of arguments will continue to be passed all of their
2594 arguments on the stack.
2596 Beware that on some ELF systems this attribute is unsuitable for
2597 global functions in shared libraries with lazy binding (which is the
2598 default). Lazy binding will send the first call via resolving code in
2599 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2600 per the standard calling conventions. Solaris 8 is affected by this.
2601 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2602 safe since the loaders there save all registers. (Lazy binding can be
2603 disabled with the linker or the loader if desired, to avoid the
2607 @cindex @code{sseregparm} attribute
2608 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2609 causes the compiler to pass up to 3 floating point arguments in
2610 SSE registers instead of on the stack. Functions that take a
2611 variable number of arguments will continue to pass all of their
2612 floating point arguments on the stack.
2614 @item force_align_arg_pointer
2615 @cindex @code{force_align_arg_pointer} attribute
2616 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2617 applied to individual function definitions, generating an alternate
2618 prologue and epilogue that realigns the runtime stack. This supports
2619 mixing legacy codes that run with a 4-byte aligned stack with modern
2620 codes that keep a 16-byte stack for SSE compatibility. The alternate
2621 prologue and epilogue are slower and bigger than the regular ones, and
2622 the alternate prologue requires a scratch register; this lowers the
2623 number of registers available if used in conjunction with the
2624 @code{regparm} attribute. The @code{force_align_arg_pointer}
2625 attribute is incompatible with nested functions; this is considered a
2629 @cindex @code{returns_twice} attribute
2630 The @code{returns_twice} attribute tells the compiler that a function may
2631 return more than one time. The compiler will ensure that all registers
2632 are dead before calling such a function and will emit a warning about
2633 the variables that may be clobbered after the second return from the
2634 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2635 The @code{longjmp}-like counterpart of such function, if any, might need
2636 to be marked with the @code{noreturn} attribute.
2639 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2640 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2641 all registers except the stack pointer should be saved in the prologue
2642 regardless of whether they are used or not.
2644 @item section ("@var{section-name}")
2645 @cindex @code{section} function attribute
2646 Normally, the compiler places the code it generates in the @code{text} section.
2647 Sometimes, however, you need additional sections, or you need certain
2648 particular functions to appear in special sections. The @code{section}
2649 attribute specifies that a function lives in a particular section.
2650 For example, the declaration:
2653 extern void foobar (void) __attribute__ ((section ("bar")));
2657 puts the function @code{foobar} in the @code{bar} section.
2659 Some file formats do not support arbitrary sections so the @code{section}
2660 attribute is not available on all platforms.
2661 If you need to map the entire contents of a module to a particular
2662 section, consider using the facilities of the linker instead.
2665 @cindex @code{sentinel} function attribute
2666 This function attribute ensures that a parameter in a function call is
2667 an explicit @code{NULL}. The attribute is only valid on variadic
2668 functions. By default, the sentinel is located at position zero, the
2669 last parameter of the function call. If an optional integer position
2670 argument P is supplied to the attribute, the sentinel must be located at
2671 position P counting backwards from the end of the argument list.
2674 __attribute__ ((sentinel))
2676 __attribute__ ((sentinel(0)))
2679 The attribute is automatically set with a position of 0 for the built-in
2680 functions @code{execl} and @code{execlp}. The built-in function
2681 @code{execle} has the attribute set with a position of 1.
2683 A valid @code{NULL} in this context is defined as zero with any pointer
2684 type. If your system defines the @code{NULL} macro with an integer type
2685 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2686 with a copy that redefines NULL appropriately.
2688 The warnings for missing or incorrect sentinels are enabled with
2692 See long_call/short_call.
2695 See longcall/shortcall.
2698 @cindex signal handler functions on the AVR processors
2699 Use this attribute on the AVR to indicate that the specified
2700 function is a signal handler. The compiler will generate function
2701 entry and exit sequences suitable for use in a signal handler when this
2702 attribute is present. Interrupts will be disabled inside the function.
2705 Use this attribute on the SH to indicate an @code{interrupt_handler}
2706 function should switch to an alternate stack. It expects a string
2707 argument that names a global variable holding the address of the
2712 void f () __attribute__ ((interrupt_handler,
2713 sp_switch ("alt_stack")));
2717 @cindex functions that pop the argument stack on the 386
2718 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2719 assume that the called function will pop off the stack space used to
2720 pass arguments, unless it takes a variable number of arguments.
2723 @cindex tiny data section on the H8/300H and H8S
2724 Use this attribute on the H8/300H and H8S to indicate that the specified
2725 variable should be placed into the tiny data section.
2726 The compiler will generate more efficient code for loads and stores
2727 on data in the tiny data section. Note the tiny data area is limited to
2728 slightly under 32kbytes of data.
2731 Use this attribute on the SH for an @code{interrupt_handler} to return using
2732 @code{trapa} instead of @code{rte}. This attribute expects an integer
2733 argument specifying the trap number to be used.
2736 @cindex @code{unused} attribute.
2737 This attribute, attached to a function, means that the function is meant
2738 to be possibly unused. GCC will not produce a warning for this
2742 @cindex @code{used} attribute.
2743 This attribute, attached to a function, means that code must be emitted
2744 for the function even if it appears that the function is not referenced.
2745 This is useful, for example, when the function is referenced only in
2749 @cindex @code{version_id} attribute on IA64 HP-UX
2750 This attribute, attached to a global variable or function, renames a
2751 symbol to contain a version string, thus allowing for function level
2752 versioning. HP-UX system header files may use version level functioning
2753 for some system calls.
2756 extern int foo () __attribute__((version_id ("20040821")));
2759 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
2761 @item visibility ("@var{visibility_type}")
2762 @cindex @code{visibility} attribute
2763 This attribute affects the linkage of the declaration to which it is attached.
2764 There are four supported @var{visibility_type} values: default,
2765 hidden, protected or internal visibility.
2768 void __attribute__ ((visibility ("protected")))
2769 f () @{ /* @r{Do something.} */; @}
2770 int i __attribute__ ((visibility ("hidden")));
2773 The possible values of @var{visibility_type} correspond to the
2774 visibility settings in the ELF gABI.
2777 @c keep this list of visibilities in alphabetical order.
2780 Default visibility is the normal case for the object file format.
2781 This value is available for the visibility attribute to override other
2782 options that may change the assumed visibility of entities.
2784 On ELF, default visibility means that the declaration is visible to other
2785 modules and, in shared libraries, means that the declared entity may be
2788 On Darwin, default visibility means that the declaration is visible to
2791 Default visibility corresponds to ``external linkage'' in the language.
2794 Hidden visibility indicates that the entity declared will have a new
2795 form of linkage, which we'll call ``hidden linkage''. Two
2796 declarations of an object with hidden linkage refer to the same object
2797 if they are in the same shared object.
2800 Internal visibility is like hidden visibility, but with additional
2801 processor specific semantics. Unless otherwise specified by the
2802 psABI, GCC defines internal visibility to mean that a function is
2803 @emph{never} called from another module. Compare this with hidden
2804 functions which, while they cannot be referenced directly by other
2805 modules, can be referenced indirectly via function pointers. By
2806 indicating that a function cannot be called from outside the module,
2807 GCC may for instance omit the load of a PIC register since it is known
2808 that the calling function loaded the correct value.
2811 Protected visibility is like default visibility except that it
2812 indicates that references within the defining module will bind to the
2813 definition in that module. That is, the declared entity cannot be
2814 overridden by another module.
2818 All visibilities are supported on many, but not all, ELF targets
2819 (supported when the assembler supports the @samp{.visibility}
2820 pseudo-op). Default visibility is supported everywhere. Hidden
2821 visibility is supported on Darwin targets.
2823 The visibility attribute should be applied only to declarations which
2824 would otherwise have external linkage. The attribute should be applied
2825 consistently, so that the same entity should not be declared with
2826 different settings of the attribute.
2828 In C++, the visibility attribute applies to types as well as functions
2829 and objects, because in C++ types have linkage. A class must not have
2830 greater visibility than its non-static data member types and bases,
2831 and class members default to the visibility of their class. Also, a
2832 declaration without explicit visibility is limited to the visibility
2835 In C++, you can mark member functions and static member variables of a
2836 class with the visibility attribute. This is useful if if you know a
2837 particular method or static member variable should only be used from
2838 one shared object; then you can mark it hidden while the rest of the
2839 class has default visibility. Care must be taken to avoid breaking
2840 the One Definition Rule; for example, it is usually not useful to mark
2841 an inline method as hidden without marking the whole class as hidden.
2843 A C++ namespace declaration can also have the visibility attribute.
2844 This attribute applies only to the particular namespace body, not to
2845 other definitions of the same namespace; it is equivalent to using
2846 @samp{#pragma GCC visibility} before and after the namespace
2847 definition (@pxref{Visibility Pragmas}).
2849 In C++, if a template argument has limited visibility, this
2850 restriction is implicitly propagated to the template instantiation.
2851 Otherwise, template instantiations and specializations default to the
2852 visibility of their template.
2854 If both the template and enclosing class have explicit visibility, the
2855 visibility from the template is used.
2857 @item warn_unused_result
2858 @cindex @code{warn_unused_result} attribute
2859 The @code{warn_unused_result} attribute causes a warning to be emitted
2860 if a caller of the function with this attribute does not use its
2861 return value. This is useful for functions where not checking
2862 the result is either a security problem or always a bug, such as
2866 int fn () __attribute__ ((warn_unused_result));
2869 if (fn () < 0) return -1;
2875 results in warning on line 5.
2878 @cindex @code{weak} attribute
2879 The @code{weak} attribute causes the declaration to be emitted as a weak
2880 symbol rather than a global. This is primarily useful in defining
2881 library functions which can be overridden in user code, though it can
2882 also be used with non-function declarations. Weak symbols are supported
2883 for ELF targets, and also for a.out targets when using the GNU assembler
2887 @itemx weakref ("@var{target}")
2888 @cindex @code{weakref} attribute
2889 The @code{weakref} attribute marks a declaration as a weak reference.
2890 Without arguments, it should be accompanied by an @code{alias} attribute
2891 naming the target symbol. Optionally, the @var{target} may be given as
2892 an argument to @code{weakref} itself. In either case, @code{weakref}
2893 implicitly marks the declaration as @code{weak}. Without a
2894 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2895 @code{weakref} is equivalent to @code{weak}.
2898 static int x() __attribute__ ((weakref ("y")));
2899 /* is equivalent to... */
2900 static int x() __attribute__ ((weak, weakref, alias ("y")));
2902 static int x() __attribute__ ((weakref));
2903 static int x() __attribute__ ((alias ("y")));
2906 A weak reference is an alias that does not by itself require a
2907 definition to be given for the target symbol. If the target symbol is
2908 only referenced through weak references, then the becomes a @code{weak}
2909 undefined symbol. If it is directly referenced, however, then such
2910 strong references prevail, and a definition will be required for the
2911 symbol, not necessarily in the same translation unit.
2913 The effect is equivalent to moving all references to the alias to a
2914 separate translation unit, renaming the alias to the aliased symbol,
2915 declaring it as weak, compiling the two separate translation units and
2916 performing a reloadable link on them.
2918 At present, a declaration to which @code{weakref} is attached can
2919 only be @code{static}.
2921 @item externally_visible
2922 @cindex @code{externally_visible} attribute.
2923 This attribute, attached to a global variable or function nullify
2924 effect of @option{-fwhole-program} command line option, so the object
2925 remain visible outside the current compilation unit
2929 You can specify multiple attributes in a declaration by separating them
2930 by commas within the double parentheses or by immediately following an
2931 attribute declaration with another attribute declaration.
2933 @cindex @code{#pragma}, reason for not using
2934 @cindex pragma, reason for not using
2935 Some people object to the @code{__attribute__} feature, suggesting that
2936 ISO C's @code{#pragma} should be used instead. At the time
2937 @code{__attribute__} was designed, there were two reasons for not doing
2942 It is impossible to generate @code{#pragma} commands from a macro.
2945 There is no telling what the same @code{#pragma} might mean in another
2949 These two reasons applied to almost any application that might have been
2950 proposed for @code{#pragma}. It was basically a mistake to use
2951 @code{#pragma} for @emph{anything}.
2953 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2954 to be generated from macros. In addition, a @code{#pragma GCC}
2955 namespace is now in use for GCC-specific pragmas. However, it has been
2956 found convenient to use @code{__attribute__} to achieve a natural
2957 attachment of attributes to their corresponding declarations, whereas
2958 @code{#pragma GCC} is of use for constructs that do not naturally form
2959 part of the grammar. @xref{Other Directives,,Miscellaneous
2960 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2962 @node Attribute Syntax
2963 @section Attribute Syntax
2964 @cindex attribute syntax
2966 This section describes the syntax with which @code{__attribute__} may be
2967 used, and the constructs to which attribute specifiers bind, for the C
2968 language. Some details may vary for C++ and Objective-C@. Because of
2969 infelicities in the grammar for attributes, some forms described here
2970 may not be successfully parsed in all cases.
2972 There are some problems with the semantics of attributes in C++. For
2973 example, there are no manglings for attributes, although they may affect
2974 code generation, so problems may arise when attributed types are used in
2975 conjunction with templates or overloading. Similarly, @code{typeid}
2976 does not distinguish between types with different attributes. Support
2977 for attributes in C++ may be restricted in future to attributes on
2978 declarations only, but not on nested declarators.
2980 @xref{Function Attributes}, for details of the semantics of attributes
2981 applying to functions. @xref{Variable Attributes}, for details of the
2982 semantics of attributes applying to variables. @xref{Type Attributes},
2983 for details of the semantics of attributes applying to structure, union
2984 and enumerated types.
2986 An @dfn{attribute specifier} is of the form
2987 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2988 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2989 each attribute is one of the following:
2993 Empty. Empty attributes are ignored.
2996 A word (which may be an identifier such as @code{unused}, or a reserved
2997 word such as @code{const}).
3000 A word, followed by, in parentheses, parameters for the attribute.
3001 These parameters take one of the following forms:
3005 An identifier. For example, @code{mode} attributes use this form.
3008 An identifier followed by a comma and a non-empty comma-separated list
3009 of expressions. For example, @code{format} attributes use this form.
3012 A possibly empty comma-separated list of expressions. For example,
3013 @code{format_arg} attributes use this form with the list being a single
3014 integer constant expression, and @code{alias} attributes use this form
3015 with the list being a single string constant.
3019 An @dfn{attribute specifier list} is a sequence of one or more attribute
3020 specifiers, not separated by any other tokens.
3022 In GNU C, an attribute specifier list may appear after the colon following a
3023 label, other than a @code{case} or @code{default} label. The only
3024 attribute it makes sense to use after a label is @code{unused}. This
3025 feature is intended for code generated by programs which contains labels
3026 that may be unused but which is compiled with @option{-Wall}. It would
3027 not normally be appropriate to use in it human-written code, though it
3028 could be useful in cases where the code that jumps to the label is
3029 contained within an @code{#ifdef} conditional. GNU C++ does not permit
3030 such placement of attribute lists, as it is permissible for a
3031 declaration, which could begin with an attribute list, to be labelled in
3032 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
3033 does not arise there.
3035 An attribute specifier list may appear as part of a @code{struct},
3036 @code{union} or @code{enum} specifier. It may go either immediately
3037 after the @code{struct}, @code{union} or @code{enum} keyword, or after
3038 the closing brace. The former syntax is preferred.
3039 Where attribute specifiers follow the closing brace, they are considered
3040 to relate to the structure, union or enumerated type defined, not to any
3041 enclosing declaration the type specifier appears in, and the type
3042 defined is not complete until after the attribute specifiers.
3043 @c Otherwise, there would be the following problems: a shift/reduce
3044 @c conflict between attributes binding the struct/union/enum and
3045 @c binding to the list of specifiers/qualifiers; and "aligned"
3046 @c attributes could use sizeof for the structure, but the size could be
3047 @c changed later by "packed" attributes.
3049 Otherwise, an attribute specifier appears as part of a declaration,
3050 counting declarations of unnamed parameters and type names, and relates
3051 to that declaration (which may be nested in another declaration, for
3052 example in the case of a parameter declaration), or to a particular declarator
3053 within a declaration. Where an
3054 attribute specifier is applied to a parameter declared as a function or
3055 an array, it should apply to the function or array rather than the
3056 pointer to which the parameter is implicitly converted, but this is not
3057 yet correctly implemented.
3059 Any list of specifiers and qualifiers at the start of a declaration may
3060 contain attribute specifiers, whether or not such a list may in that
3061 context contain storage class specifiers. (Some attributes, however,
3062 are essentially in the nature of storage class specifiers, and only make
3063 sense where storage class specifiers may be used; for example,
3064 @code{section}.) There is one necessary limitation to this syntax: the
3065 first old-style parameter declaration in a function definition cannot
3066 begin with an attribute specifier, because such an attribute applies to
3067 the function instead by syntax described below (which, however, is not
3068 yet implemented in this case). In some other cases, attribute
3069 specifiers are permitted by this grammar but not yet supported by the
3070 compiler. All attribute specifiers in this place relate to the
3071 declaration as a whole. In the obsolescent usage where a type of
3072 @code{int} is implied by the absence of type specifiers, such a list of
3073 specifiers and qualifiers may be an attribute specifier list with no
3074 other specifiers or qualifiers.
3076 At present, the first parameter in a function prototype must have some
3077 type specifier which is not an attribute specifier; this resolves an
3078 ambiguity in the interpretation of @code{void f(int
3079 (__attribute__((foo)) x))}, but is subject to change. At present, if
3080 the parentheses of a function declarator contain only attributes then
3081 those attributes are ignored, rather than yielding an error or warning
3082 or implying a single parameter of type int, but this is subject to
3085 An attribute specifier list may appear immediately before a declarator
3086 (other than the first) in a comma-separated list of declarators in a
3087 declaration of more than one identifier using a single list of
3088 specifiers and qualifiers. Such attribute specifiers apply
3089 only to the identifier before whose declarator they appear. For
3093 __attribute__((noreturn)) void d0 (void),
3094 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
3099 the @code{noreturn} attribute applies to all the functions
3100 declared; the @code{format} attribute only applies to @code{d1}.
3102 An attribute specifier list may appear immediately before the comma,
3103 @code{=} or semicolon terminating the declaration of an identifier other
3104 than a function definition. At present, such attribute specifiers apply
3105 to the declared object or function, but in future they may attach to the
3106 outermost adjacent declarator. In simple cases there is no difference,
3107 but, for example, in
3110 void (****f)(void) __attribute__((noreturn));
3114 at present the @code{noreturn} attribute applies to @code{f}, which
3115 causes a warning since @code{f} is not a function, but in future it may
3116 apply to the function @code{****f}. The precise semantics of what
3117 attributes in such cases will apply to are not yet specified. Where an
3118 assembler name for an object or function is specified (@pxref{Asm
3119 Labels}), at present the attribute must follow the @code{asm}
3120 specification; in future, attributes before the @code{asm} specification
3121 may apply to the adjacent declarator, and those after it to the declared
3124 An attribute specifier list may, in future, be permitted to appear after
3125 the declarator in a function definition (before any old-style parameter
3126 declarations or the function body).
3128 Attribute specifiers may be mixed with type qualifiers appearing inside
3129 the @code{[]} of a parameter array declarator, in the C99 construct by
3130 which such qualifiers are applied to the pointer to which the array is
3131 implicitly converted. Such attribute specifiers apply to the pointer,
3132 not to the array, but at present this is not implemented and they are
3135 An attribute specifier list may appear at the start of a nested
3136 declarator. At present, there are some limitations in this usage: the
3137 attributes correctly apply to the declarator, but for most individual
3138 attributes the semantics this implies are not implemented.
3139 When attribute specifiers follow the @code{*} of a pointer
3140 declarator, they may be mixed with any type qualifiers present.
3141 The following describes the formal semantics of this syntax. It will make the
3142 most sense if you are familiar with the formal specification of
3143 declarators in the ISO C standard.
3145 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
3146 D1}, where @code{T} contains declaration specifiers that specify a type
3147 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
3148 contains an identifier @var{ident}. The type specified for @var{ident}
3149 for derived declarators whose type does not include an attribute
3150 specifier is as in the ISO C standard.
3152 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
3153 and the declaration @code{T D} specifies the type
3154 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3155 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3156 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
3158 If @code{D1} has the form @code{*
3159 @var{type-qualifier-and-attribute-specifier-list} D}, and the
3160 declaration @code{T D} specifies the type
3161 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
3162 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
3163 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
3169 void (__attribute__((noreturn)) ****f) (void);
3173 specifies the type ``pointer to pointer to pointer to pointer to
3174 non-returning function returning @code{void}''. As another example,
3177 char *__attribute__((aligned(8))) *f;
3181 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
3182 Note again that this does not work with most attributes; for example,
3183 the usage of @samp{aligned} and @samp{noreturn} attributes given above
3184 is not yet supported.
3186 For compatibility with existing code written for compiler versions that
3187 did not implement attributes on nested declarators, some laxity is
3188 allowed in the placing of attributes. If an attribute that only applies
3189 to types is applied to a declaration, it will be treated as applying to
3190 the type of that declaration. If an attribute that only applies to
3191 declarations is applied to the type of a declaration, it will be treated
3192 as applying to that declaration; and, for compatibility with code
3193 placing the attributes immediately before the identifier declared, such
3194 an attribute applied to a function return type will be treated as
3195 applying to the function type, and such an attribute applied to an array
3196 element type will be treated as applying to the array type. If an
3197 attribute that only applies to function types is applied to a
3198 pointer-to-function type, it will be treated as applying to the pointer
3199 target type; if such an attribute is applied to a function return type
3200 that is not a pointer-to-function type, it will be treated as applying
3201 to the function type.
3203 @node Function Prototypes
3204 @section Prototypes and Old-Style Function Definitions
3205 @cindex function prototype declarations
3206 @cindex old-style function definitions
3207 @cindex promotion of formal parameters
3209 GNU C extends ISO C to allow a function prototype to override a later
3210 old-style non-prototype definition. Consider the following example:
3213 /* @r{Use prototypes unless the compiler is old-fashioned.} */
3220 /* @r{Prototype function declaration.} */
3221 int isroot P((uid_t));
3223 /* @r{Old-style function definition.} */
3225 isroot (x) /* @r{??? lossage here ???} */
3232 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
3233 not allow this example, because subword arguments in old-style
3234 non-prototype definitions are promoted. Therefore in this example the
3235 function definition's argument is really an @code{int}, which does not
3236 match the prototype argument type of @code{short}.
3238 This restriction of ISO C makes it hard to write code that is portable
3239 to traditional C compilers, because the programmer does not know
3240 whether the @code{uid_t} type is @code{short}, @code{int}, or
3241 @code{long}. Therefore, in cases like these GNU C allows a prototype
3242 to override a later old-style definition. More precisely, in GNU C, a
3243 function prototype argument type overrides the argument type specified
3244 by a later old-style definition if the former type is the same as the
3245 latter type before promotion. Thus in GNU C the above example is
3246 equivalent to the following:
3259 GNU C++ does not support old-style function definitions, so this
3260 extension is irrelevant.
3263 @section C++ Style Comments
3265 @cindex C++ comments
3266 @cindex comments, C++ style
3268 In GNU C, you may use C++ style comments, which start with @samp{//} and
3269 continue until the end of the line. Many other C implementations allow
3270 such comments, and they are included in the 1999 C standard. However,
3271 C++ style comments are not recognized if you specify an @option{-std}
3272 option specifying a version of ISO C before C99, or @option{-ansi}
3273 (equivalent to @option{-std=c89}).
3276 @section Dollar Signs in Identifier Names
3278 @cindex dollar signs in identifier names
3279 @cindex identifier names, dollar signs in
3281 In GNU C, you may normally use dollar signs in identifier names.
3282 This is because many traditional C implementations allow such identifiers.
3283 However, dollar signs in identifiers are not supported on a few target
3284 machines, typically because the target assembler does not allow them.
3286 @node Character Escapes
3287 @section The Character @key{ESC} in Constants
3289 You can use the sequence @samp{\e} in a string or character constant to
3290 stand for the ASCII character @key{ESC}.
3293 @section Inquiring on Alignment of Types or Variables
3295 @cindex type alignment
3296 @cindex variable alignment
3298 The keyword @code{__alignof__} allows you to inquire about how an object
3299 is aligned, or the minimum alignment usually required by a type. Its
3300 syntax is just like @code{sizeof}.
3302 For example, if the target machine requires a @code{double} value to be
3303 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3304 This is true on many RISC machines. On more traditional machine
3305 designs, @code{__alignof__ (double)} is 4 or even 2.
3307 Some machines never actually require alignment; they allow reference to any
3308 data type even at an odd address. For these machines, @code{__alignof__}
3309 reports the @emph{recommended} alignment of a type.
3311 If the operand of @code{__alignof__} is an lvalue rather than a type,
3312 its value is the required alignment for its type, taking into account
3313 any minimum alignment specified with GCC's @code{__attribute__}
3314 extension (@pxref{Variable Attributes}). For example, after this
3318 struct foo @{ int x; char y; @} foo1;
3322 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3323 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3325 It is an error to ask for the alignment of an incomplete type.
3327 @node Variable Attributes
3328 @section Specifying Attributes of Variables
3329 @cindex attribute of variables
3330 @cindex variable attributes
3332 The keyword @code{__attribute__} allows you to specify special
3333 attributes of variables or structure fields. This keyword is followed
3334 by an attribute specification inside double parentheses. Some
3335 attributes are currently defined generically for variables.
3336 Other attributes are defined for variables on particular target
3337 systems. Other attributes are available for functions
3338 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3339 Other front ends might define more attributes
3340 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3342 You may also specify attributes with @samp{__} preceding and following
3343 each keyword. This allows you to use them in header files without
3344 being concerned about a possible macro of the same name. For example,
3345 you may use @code{__aligned__} instead of @code{aligned}.
3347 @xref{Attribute Syntax}, for details of the exact syntax for using
3351 @cindex @code{aligned} attribute
3352 @item aligned (@var{alignment})
3353 This attribute specifies a minimum alignment for the variable or
3354 structure field, measured in bytes. For example, the declaration:
3357 int x __attribute__ ((aligned (16))) = 0;
3361 causes the compiler to allocate the global variable @code{x} on a
3362 16-byte boundary. On a 68040, this could be used in conjunction with
3363 an @code{asm} expression to access the @code{move16} instruction which
3364 requires 16-byte aligned operands.
3366 You can also specify the alignment of structure fields. For example, to
3367 create a double-word aligned @code{int} pair, you could write:
3370 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3374 This is an alternative to creating a union with a @code{double} member
3375 that forces the union to be double-word aligned.
3377 As in the preceding examples, you can explicitly specify the alignment
3378 (in bytes) that you wish the compiler to use for a given variable or
3379 structure field. Alternatively, you can leave out the alignment factor
3380 and just ask the compiler to align a variable or field to the maximum
3381 useful alignment for the target machine you are compiling for. For
3382 example, you could write:
3385 short array[3] __attribute__ ((aligned));
3388 Whenever you leave out the alignment factor in an @code{aligned} attribute
3389 specification, the compiler automatically sets the alignment for the declared
3390 variable or field to the largest alignment which is ever used for any data
3391 type on the target machine you are compiling for. Doing this can often make
3392 copy operations more efficient, because the compiler can use whatever
3393 instructions copy the biggest chunks of memory when performing copies to
3394 or from the variables or fields that you have aligned this way.
3396 When used on a struct, or struct member, the @code{aligned} attribute can
3397 only increase the alignment; in order to decrease it, the @code{packed}
3398 attribute must be specified as well. When used as part of a typedef, the
3399 @code{aligned} attribute can both increase and decrease alignment, and
3400 specifying the @code{packed} attribute will generate a warning.
3402 Note that the effectiveness of @code{aligned} attributes may be limited
3403 by inherent limitations in your linker. On many systems, the linker is
3404 only able to arrange for variables to be aligned up to a certain maximum
3405 alignment. (For some linkers, the maximum supported alignment may
3406 be very very small.) If your linker is only able to align variables
3407 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3408 in an @code{__attribute__} will still only provide you with 8 byte
3409 alignment. See your linker documentation for further information.
3411 The @code{aligned} attribute can also be used for functions
3412 (@pxref{Function Attributes}.)
3414 @item cleanup (@var{cleanup_function})
3415 @cindex @code{cleanup} attribute
3416 The @code{cleanup} attribute runs a function when the variable goes
3417 out of scope. This attribute can only be applied to auto function
3418 scope variables; it may not be applied to parameters or variables
3419 with static storage duration. The function must take one parameter,
3420 a pointer to a type compatible with the variable. The return value
3421 of the function (if any) is ignored.
3423 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3424 will be run during the stack unwinding that happens during the
3425 processing of the exception. Note that the @code{cleanup} attribute
3426 does not allow the exception to be caught, only to perform an action.
3427 It is undefined what happens if @var{cleanup_function} does not
3432 @cindex @code{common} attribute
3433 @cindex @code{nocommon} attribute
3436 The @code{common} attribute requests GCC to place a variable in
3437 ``common'' storage. The @code{nocommon} attribute requests the
3438 opposite---to allocate space for it directly.
3440 These attributes override the default chosen by the
3441 @option{-fno-common} and @option{-fcommon} flags respectively.
3444 @cindex @code{deprecated} attribute
3445 The @code{deprecated} attribute results in a warning if the variable
3446 is used anywhere in the source file. This is useful when identifying
3447 variables that are expected to be removed in a future version of a
3448 program. The warning also includes the location of the declaration
3449 of the deprecated variable, to enable users to easily find further
3450 information about why the variable is deprecated, or what they should
3451 do instead. Note that the warning only occurs for uses:
3454 extern int old_var __attribute__ ((deprecated));
3456 int new_fn () @{ return old_var; @}
3459 results in a warning on line 3 but not line 2.
3461 The @code{deprecated} attribute can also be used for functions and
3462 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3464 @item mode (@var{mode})
3465 @cindex @code{mode} attribute
3466 This attribute specifies the data type for the declaration---whichever
3467 type corresponds to the mode @var{mode}. This in effect lets you
3468 request an integer or floating point type according to its width.
3470 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3471 indicate the mode corresponding to a one-byte integer, @samp{word} or
3472 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3473 or @samp{__pointer__} for the mode used to represent pointers.
3476 @cindex @code{packed} attribute
3477 The @code{packed} attribute specifies that a variable or structure field
3478 should have the smallest possible alignment---one byte for a variable,
3479 and one bit for a field, unless you specify a larger value with the
3480 @code{aligned} attribute.
3482 Here is a structure in which the field @code{x} is packed, so that it
3483 immediately follows @code{a}:
3489 int x[2] __attribute__ ((packed));
3493 @item section ("@var{section-name}")
3494 @cindex @code{section} variable attribute
3495 Normally, the compiler places the objects it generates in sections like
3496 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3497 or you need certain particular variables to appear in special sections,
3498 for example to map to special hardware. The @code{section}
3499 attribute specifies that a variable (or function) lives in a particular
3500 section. For example, this small program uses several specific section names:
3503 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3504 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3505 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3506 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3510 /* @r{Initialize stack pointer} */
3511 init_sp (stack + sizeof (stack));
3513 /* @r{Initialize initialized data} */
3514 memcpy (&init_data, &data, &edata - &data);
3516 /* @r{Turn on the serial ports} */
3523 Use the @code{section} attribute with an @emph{initialized} definition
3524 of a @emph{global} variable, as shown in the example. GCC issues
3525 a warning and otherwise ignores the @code{section} attribute in
3526 uninitialized variable declarations.
3528 You may only use the @code{section} attribute with a fully initialized
3529 global definition because of the way linkers work. The linker requires
3530 each object be defined once, with the exception that uninitialized
3531 variables tentatively go in the @code{common} (or @code{bss}) section
3532 and can be multiply ``defined''. You can force a variable to be
3533 initialized with the @option{-fno-common} flag or the @code{nocommon}
3536 Some file formats do not support arbitrary sections so the @code{section}
3537 attribute is not available on all platforms.
3538 If you need to map the entire contents of a module to a particular
3539 section, consider using the facilities of the linker instead.
3542 @cindex @code{shared} variable attribute
3543 On Microsoft Windows, in addition to putting variable definitions in a named
3544 section, the section can also be shared among all running copies of an
3545 executable or DLL@. For example, this small program defines shared data
3546 by putting it in a named section @code{shared} and marking the section
3550 int foo __attribute__((section ("shared"), shared)) = 0;
3555 /* @r{Read and write foo. All running
3556 copies see the same value.} */
3562 You may only use the @code{shared} attribute along with @code{section}
3563 attribute with a fully initialized global definition because of the way
3564 linkers work. See @code{section} attribute for more information.
3566 The @code{shared} attribute is only available on Microsoft Windows@.
3568 @item tls_model ("@var{tls_model}")
3569 @cindex @code{tls_model} attribute
3570 The @code{tls_model} attribute sets thread-local storage model
3571 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3572 overriding @option{-ftls-model=} command line switch on a per-variable
3574 The @var{tls_model} argument should be one of @code{global-dynamic},
3575 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3577 Not all targets support this attribute.
3580 This attribute, attached to a variable, means that the variable is meant
3581 to be possibly unused. GCC will not produce a warning for this
3585 This attribute, attached to a variable, means that the variable must be
3586 emitted even if it appears that the variable is not referenced.
3588 @item vector_size (@var{bytes})
3589 This attribute specifies the vector size for the variable, measured in
3590 bytes. For example, the declaration:
3593 int foo __attribute__ ((vector_size (16)));
3597 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3598 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3599 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3601 This attribute is only applicable to integral and float scalars,
3602 although arrays, pointers, and function return values are allowed in
3603 conjunction with this construct.
3605 Aggregates with this attribute are invalid, even if they are of the same
3606 size as a corresponding scalar. For example, the declaration:
3609 struct S @{ int a; @};
3610 struct S __attribute__ ((vector_size (16))) foo;
3614 is invalid even if the size of the structure is the same as the size of
3618 The @code{selectany} attribute causes an initialized global variable to
3619 have link-once semantics. When multiple definitions of the variable are
3620 encountered by the linker, the first is selected and the remainder are
3621 discarded. Following usage by the Microsoft compiler, the linker is told
3622 @emph{not} to warn about size or content differences of the multiple
3625 Although the primary usage of this attribute is for POD types, the
3626 attribute can also be applied to global C++ objects that are initialized
3627 by a constructor. In this case, the static initialization and destruction
3628 code for the object is emitted in each translation defining the object,
3629 but the calls to the constructor and destructor are protected by a
3630 link-once guard variable.
3632 The @code{selectany} attribute is only available on Microsoft Windows
3633 targets. You can use @code{__declspec (selectany)} as a synonym for
3634 @code{__attribute__ ((selectany))} for compatibility with other
3638 The @code{weak} attribute is described in @xref{Function Attributes}.
3641 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3644 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3648 @subsection Blackfin Variable Attributes
3650 Three attributes are currently defined for the Blackfin.
3656 @cindex @code{l1_data} variable attribute
3657 @cindex @code{l1_data_A} variable attribute
3658 @cindex @code{l1_data_B} variable attribute
3659 Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
3660 Variables with @code{l1_data} attribute will be put into the specific section
3661 named @code{.l1.data}. Those with @code{l1_data_A} attribute will be put into
3662 the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
3663 attribute will be put into the specific section named @code{.l1.data.B}.
3666 @subsection M32R/D Variable Attributes
3668 One attribute is currently defined for the M32R/D@.
3671 @item model (@var{model-name})
3672 @cindex variable addressability on the M32R/D
3673 Use this attribute on the M32R/D to set the addressability of an object.
3674 The identifier @var{model-name} is one of @code{small}, @code{medium},
3675 or @code{large}, representing each of the code models.
3677 Small model objects live in the lower 16MB of memory (so that their
3678 addresses can be loaded with the @code{ld24} instruction).
3680 Medium and large model objects may live anywhere in the 32-bit address space
3681 (the compiler will generate @code{seth/add3} instructions to load their
3685 @anchor{i386 Variable Attributes}
3686 @subsection i386 Variable Attributes
3688 Two attributes are currently defined for i386 configurations:
3689 @code{ms_struct} and @code{gcc_struct}
3694 @cindex @code{ms_struct} attribute
3695 @cindex @code{gcc_struct} attribute
3697 If @code{packed} is used on a structure, or if bit-fields are used
3698 it may be that the Microsoft ABI packs them differently
3699 than GCC would normally pack them. Particularly when moving packed
3700 data between functions compiled with GCC and the native Microsoft compiler
3701 (either via function call or as data in a file), it may be necessary to access
3704 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3705 compilers to match the native Microsoft compiler.
3707 The Microsoft structure layout algorithm is fairly simple with the exception
3708 of the bitfield packing:
3710 The padding and alignment of members of structures and whether a bit field
3711 can straddle a storage-unit boundary
3714 @item Structure members are stored sequentially in the order in which they are
3715 declared: the first member has the lowest memory address and the last member
3718 @item Every data object has an alignment-requirement. The alignment-requirement
3719 for all data except structures, unions, and arrays is either the size of the
3720 object or the current packing size (specified with either the aligned attribute
3721 or the pack pragma), whichever is less. For structures, unions, and arrays,
3722 the alignment-requirement is the largest alignment-requirement of its members.
3723 Every object is allocated an offset so that:
3725 offset % alignment-requirement == 0
3727 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3728 unit if the integral types are the same size and if the next bit field fits
3729 into the current allocation unit without crossing the boundary imposed by the
3730 common alignment requirements of the bit fields.
3733 Handling of zero-length bitfields:
3735 MSVC interprets zero-length bitfields in the following ways:
3738 @item If a zero-length bitfield is inserted between two bitfields that would
3739 normally be coalesced, the bitfields will not be coalesced.
3746 unsigned long bf_1 : 12;
3748 unsigned long bf_2 : 12;
3752 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3753 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3755 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3756 alignment of the zero-length bitfield is greater than the member that follows it,
3757 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3777 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3778 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3779 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3782 Taking this into account, it is important to note the following:
3785 @item If a zero-length bitfield follows a normal bitfield, the type of the
3786 zero-length bitfield may affect the alignment of the structure as whole. For
3787 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3788 normal bitfield, and is of type short.
3790 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3791 still affect the alignment of the structure:
3801 Here, @code{t4} will take up 4 bytes.
3804 @item Zero-length bitfields following non-bitfield members are ignored:
3815 Here, @code{t5} will take up 2 bytes.
3819 @subsection PowerPC Variable Attributes
3821 Three attributes currently are defined for PowerPC configurations:
3822 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3824 For full documentation of the struct attributes please see the
3825 documentation in the @xref{i386 Variable Attributes}, section.
3827 For documentation of @code{altivec} attribute please see the
3828 documentation in the @xref{PowerPC Type Attributes}, section.
3830 @subsection SPU Variable Attributes
3832 The SPU supports the @code{spu_vector} attribute for variables. For
3833 documentation of this attribute please see the documentation in the
3834 @xref{SPU Type Attributes}, section.
3836 @subsection Xstormy16 Variable Attributes
3838 One attribute is currently defined for xstormy16 configurations:
3843 @cindex @code{below100} attribute
3845 If a variable has the @code{below100} attribute (@code{BELOW100} is
3846 allowed also), GCC will place the variable in the first 0x100 bytes of
3847 memory and use special opcodes to access it. Such variables will be
3848 placed in either the @code{.bss_below100} section or the
3849 @code{.data_below100} section.
3853 @subsection AVR Variable Attributes
3857 @cindex @code{progmem} variable attribute
3858 The @code{progmem} attribute is used on the AVR to place data in the Program
3859 Memory address space. The AVR is a Harvard Architecture processor and data
3860 normally resides in the Data Memory address space.
3863 @node Type Attributes
3864 @section Specifying Attributes of Types
3865 @cindex attribute of types
3866 @cindex type attributes
3868 The keyword @code{__attribute__} allows you to specify special
3869 attributes of @code{struct} and @code{union} types when you define
3870 such types. This keyword is followed by an attribute specification
3871 inside double parentheses. Seven attributes are currently defined for
3872 types: @code{aligned}, @code{packed}, @code{transparent_union},
3873 @code{unused}, @code{deprecated}, @code{visibility}, and
3874 @code{may_alias}. Other attributes are defined for functions
3875 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3878 You may also specify any one of these attributes with @samp{__}
3879 preceding and following its keyword. This allows you to use these
3880 attributes in header files without being concerned about a possible
3881 macro of the same name. For example, you may use @code{__aligned__}
3882 instead of @code{aligned}.
3884 You may specify type attributes either in a @code{typedef} declaration
3885 or in an enum, struct or union type declaration or definition.
3887 For an enum, struct or union type, you may specify attributes either
3888 between the enum, struct or union tag and the name of the type, or
3889 just past the closing curly brace of the @emph{definition}. The
3890 former syntax is preferred.
3892 @xref{Attribute Syntax}, for details of the exact syntax for using
3896 @cindex @code{aligned} attribute
3897 @item aligned (@var{alignment})
3898 This attribute specifies a minimum alignment (in bytes) for variables
3899 of the specified type. For example, the declarations:
3902 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3903 typedef int more_aligned_int __attribute__ ((aligned (8)));
3907 force the compiler to insure (as far as it can) that each variable whose
3908 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3909 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3910 variables of type @code{struct S} aligned to 8-byte boundaries allows
3911 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3912 store) instructions when copying one variable of type @code{struct S} to
3913 another, thus improving run-time efficiency.
3915 Note that the alignment of any given @code{struct} or @code{union} type
3916 is required by the ISO C standard to be at least a perfect multiple of
3917 the lowest common multiple of the alignments of all of the members of
3918 the @code{struct} or @code{union} in question. This means that you @emph{can}
3919 effectively adjust the alignment of a @code{struct} or @code{union}
3920 type by attaching an @code{aligned} attribute to any one of the members
3921 of such a type, but the notation illustrated in the example above is a
3922 more obvious, intuitive, and readable way to request the compiler to
3923 adjust the alignment of an entire @code{struct} or @code{union} type.
3925 As in the preceding example, you can explicitly specify the alignment
3926 (in bytes) that you wish the compiler to use for a given @code{struct}
3927 or @code{union} type. Alternatively, you can leave out the alignment factor
3928 and just ask the compiler to align a type to the maximum
3929 useful alignment for the target machine you are compiling for. For
3930 example, you could write:
3933 struct S @{ short f[3]; @} __attribute__ ((aligned));
3936 Whenever you leave out the alignment factor in an @code{aligned}
3937 attribute specification, the compiler automatically sets the alignment
3938 for the type to the largest alignment which is ever used for any data
3939 type on the target machine you are compiling for. Doing this can often
3940 make copy operations more efficient, because the compiler can use
3941 whatever instructions copy the biggest chunks of memory when performing
3942 copies to or from the variables which have types that you have aligned
3945 In the example above, if the size of each @code{short} is 2 bytes, then
3946 the size of the entire @code{struct S} type is 6 bytes. The smallest
3947 power of two which is greater than or equal to that is 8, so the
3948 compiler sets the alignment for the entire @code{struct S} type to 8
3951 Note that although you can ask the compiler to select a time-efficient
3952 alignment for a given type and then declare only individual stand-alone
3953 objects of that type, the compiler's ability to select a time-efficient
3954 alignment is primarily useful only when you plan to create arrays of
3955 variables having the relevant (efficiently aligned) type. If you
3956 declare or use arrays of variables of an efficiently-aligned type, then
3957 it is likely that your program will also be doing pointer arithmetic (or
3958 subscripting, which amounts to the same thing) on pointers to the
3959 relevant type, and the code that the compiler generates for these
3960 pointer arithmetic operations will often be more efficient for
3961 efficiently-aligned types than for other types.
3963 The @code{aligned} attribute can only increase the alignment; but you
3964 can decrease it by specifying @code{packed} as well. See below.
3966 Note that the effectiveness of @code{aligned} attributes may be limited
3967 by inherent limitations in your linker. On many systems, the linker is
3968 only able to arrange for variables to be aligned up to a certain maximum
3969 alignment. (For some linkers, the maximum supported alignment may
3970 be very very small.) If your linker is only able to align variables
3971 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3972 in an @code{__attribute__} will still only provide you with 8 byte
3973 alignment. See your linker documentation for further information.
3976 This attribute, attached to @code{struct} or @code{union} type
3977 definition, specifies that each member (other than zero-width bitfields)
3978 of the structure or union is placed to minimize the memory required. When
3979 attached to an @code{enum} definition, it indicates that the smallest
3980 integral type should be used.
3982 @opindex fshort-enums
3983 Specifying this attribute for @code{struct} and @code{union} types is
3984 equivalent to specifying the @code{packed} attribute on each of the
3985 structure or union members. Specifying the @option{-fshort-enums}
3986 flag on the line is equivalent to specifying the @code{packed}
3987 attribute on all @code{enum} definitions.
3989 In the following example @code{struct my_packed_struct}'s members are
3990 packed closely together, but the internal layout of its @code{s} member
3991 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3995 struct my_unpacked_struct
4001 struct __attribute__ ((__packed__)) my_packed_struct
4005 struct my_unpacked_struct s;
4009 You may only specify this attribute on the definition of a @code{enum},
4010 @code{struct} or @code{union}, not on a @code{typedef} which does not
4011 also define the enumerated type, structure or union.
4013 @item transparent_union
4014 This attribute, attached to a @code{union} type definition, indicates
4015 that any function parameter having that union type causes calls to that
4016 function to be treated in a special way.
4018 First, the argument corresponding to a transparent union type can be of
4019 any type in the union; no cast is required. Also, if the union contains
4020 a pointer type, the corresponding argument can be a null pointer
4021 constant or a void pointer expression; and if the union contains a void
4022 pointer type, the corresponding argument can be any pointer expression.
4023 If the union member type is a pointer, qualifiers like @code{const} on
4024 the referenced type must be respected, just as with normal pointer
4027 Second, the argument is passed to the function using the calling
4028 conventions of the first member of the transparent union, not the calling
4029 conventions of the union itself. All members of the union must have the
4030 same machine representation; this is necessary for this argument passing
4033 Transparent unions are designed for library functions that have multiple
4034 interfaces for compatibility reasons. For example, suppose the
4035 @code{wait} function must accept either a value of type @code{int *} to
4036 comply with Posix, or a value of type @code{union wait *} to comply with
4037 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
4038 @code{wait} would accept both kinds of arguments, but it would also
4039 accept any other pointer type and this would make argument type checking
4040 less useful. Instead, @code{<sys/wait.h>} might define the interface
4048 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
4050 pid_t wait (wait_status_ptr_t);
4053 This interface allows either @code{int *} or @code{union wait *}
4054 arguments to be passed, using the @code{int *} calling convention.
4055 The program can call @code{wait} with arguments of either type:
4058 int w1 () @{ int w; return wait (&w); @}
4059 int w2 () @{ union wait w; return wait (&w); @}
4062 With this interface, @code{wait}'s implementation might look like this:
4065 pid_t wait (wait_status_ptr_t p)
4067 return waitpid (-1, p.__ip, 0);
4072 When attached to a type (including a @code{union} or a @code{struct}),
4073 this attribute means that variables of that type are meant to appear
4074 possibly unused. GCC will not produce a warning for any variables of
4075 that type, even if the variable appears to do nothing. This is often
4076 the case with lock or thread classes, which are usually defined and then
4077 not referenced, but contain constructors and destructors that have
4078 nontrivial bookkeeping functions.
4081 The @code{deprecated} attribute results in a warning if the type
4082 is used anywhere in the source file. This is useful when identifying
4083 types that are expected to be removed in a future version of a program.
4084 If possible, the warning also includes the location of the declaration
4085 of the deprecated type, to enable users to easily find further
4086 information about why the type is deprecated, or what they should do
4087 instead. Note that the warnings only occur for uses and then only
4088 if the type is being applied to an identifier that itself is not being
4089 declared as deprecated.
4092 typedef int T1 __attribute__ ((deprecated));
4096 typedef T1 T3 __attribute__ ((deprecated));
4097 T3 z __attribute__ ((deprecated));
4100 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
4101 warning is issued for line 4 because T2 is not explicitly
4102 deprecated. Line 5 has no warning because T3 is explicitly
4103 deprecated. Similarly for line 6.
4105 The @code{deprecated} attribute can also be used for functions and
4106 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
4109 Accesses to objects with types with this attribute are not subjected to
4110 type-based alias analysis, but are instead assumed to be able to alias
4111 any other type of objects, just like the @code{char} type. See
4112 @option{-fstrict-aliasing} for more information on aliasing issues.
4117 typedef short __attribute__((__may_alias__)) short_a;
4123 short_a *b = (short_a *) &a;
4127 if (a == 0x12345678)
4134 If you replaced @code{short_a} with @code{short} in the variable
4135 declaration, the above program would abort when compiled with
4136 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
4137 above in recent GCC versions.
4140 In C++, attribute visibility (@pxref{Function Attributes}) can also be
4141 applied to class, struct, union and enum types. Unlike other type
4142 attributes, the attribute must appear between the initial keyword and
4143 the name of the type; it cannot appear after the body of the type.
4145 Note that the type visibility is applied to vague linkage entities
4146 associated with the class (vtable, typeinfo node, etc.). In
4147 particular, if a class is thrown as an exception in one shared object
4148 and caught in another, the class must have default visibility.
4149 Otherwise the two shared objects will be unable to use the same
4150 typeinfo node and exception handling will break.
4152 @subsection ARM Type Attributes
4154 On those ARM targets that support @code{dllimport} (such as Symbian
4155 OS), you can use the @code{notshared} attribute to indicate that the
4156 virtual table and other similar data for a class should not be
4157 exported from a DLL@. For example:
4160 class __declspec(notshared) C @{
4162 __declspec(dllimport) C();
4166 __declspec(dllexport)
4170 In this code, @code{C::C} is exported from the current DLL, but the
4171 virtual table for @code{C} is not exported. (You can use
4172 @code{__attribute__} instead of @code{__declspec} if you prefer, but
4173 most Symbian OS code uses @code{__declspec}.)
4175 @anchor{i386 Type Attributes}
4176 @subsection i386 Type Attributes
4178 Two attributes are currently defined for i386 configurations:
4179 @code{ms_struct} and @code{gcc_struct}
4183 @cindex @code{ms_struct}
4184 @cindex @code{gcc_struct}
4186 If @code{packed} is used on a structure, or if bit-fields are used
4187 it may be that the Microsoft ABI packs them differently
4188 than GCC would normally pack them. Particularly when moving packed
4189 data between functions compiled with GCC and the native Microsoft compiler
4190 (either via function call or as data in a file), it may be necessary to access
4193 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
4194 compilers to match the native Microsoft compiler.
4197 To specify multiple attributes, separate them by commas within the
4198 double parentheses: for example, @samp{__attribute__ ((aligned (16),
4201 @anchor{PowerPC Type Attributes}
4202 @subsection PowerPC Type Attributes
4204 Three attributes currently are defined for PowerPC configurations:
4205 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
4207 For full documentation of the struct attributes please see the
4208 documentation in the @xref{i386 Type Attributes}, section.
4210 The @code{altivec} attribute allows one to declare AltiVec vector data
4211 types supported by the AltiVec Programming Interface Manual. The
4212 attribute requires an argument to specify one of three vector types:
4213 @code{vector__}, @code{pixel__} (always followed by unsigned short),
4214 and @code{bool__} (always followed by unsigned).
4217 __attribute__((altivec(vector__)))
4218 __attribute__((altivec(pixel__))) unsigned short
4219 __attribute__((altivec(bool__))) unsigned
4222 These attributes mainly are intended to support the @code{__vector},
4223 @code{__pixel}, and @code{__bool} AltiVec keywords.
4225 @anchor{SPU Type Attributes}
4226 @subsection SPU Type Attributes
4228 The SPU supports the @code{spu_vector} attribute for types. This attribute
4229 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
4230 Language Extensions Specification. It is intended to support the
4231 @code{__vector} keyword.
4235 @section An Inline Function is As Fast As a Macro
4236 @cindex inline functions
4237 @cindex integrating function code
4239 @cindex macros, inline alternative
4241 By declaring a function inline, you can direct GCC to make
4242 calls to that function faster. One way GCC can achieve this is to
4243 integrate that function's code into the code for its callers. This
4244 makes execution faster by eliminating the function-call overhead; in
4245 addition, if any of the actual argument values are constant, their
4246 known values may permit simplifications at compile time so that not
4247 all of the inline function's code needs to be included. The effect on
4248 code size is less predictable; object code may be larger or smaller
4249 with function inlining, depending on the particular case. You can
4250 also direct GCC to try to integrate all ``simple enough'' functions
4251 into their callers with the option @option{-finline-functions}.
4253 GCC implements three different semantics of declaring a function
4254 inline. One is available with @option{-std=gnu89} or
4255 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
4256 on all inline declarations, another when @option{-std=c99} or
4257 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
4258 is used when compiling C++.
4260 To declare a function inline, use the @code{inline} keyword in its
4261 declaration, like this:
4271 If you are writing a header file to be included in ISO C89 programs, write
4272 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4274 The three types of inlining behave similarly in two important cases:
4275 when the @code{inline} keyword is used on a @code{static} function,
4276 like the example above, and when a function is first declared without
4277 using the @code{inline} keyword and then is defined with
4278 @code{inline}, like this:
4281 extern int inc (int *a);
4289 In both of these common cases, the program behaves the same as if you
4290 had not used the @code{inline} keyword, except for its speed.
4292 @cindex inline functions, omission of
4293 @opindex fkeep-inline-functions
4294 When a function is both inline and @code{static}, if all calls to the
4295 function are integrated into the caller, and the function's address is
4296 never used, then the function's own assembler code is never referenced.
4297 In this case, GCC does not actually output assembler code for the
4298 function, unless you specify the option @option{-fkeep-inline-functions}.
4299 Some calls cannot be integrated for various reasons (in particular,
4300 calls that precede the function's definition cannot be integrated, and
4301 neither can recursive calls within the definition). If there is a
4302 nonintegrated call, then the function is compiled to assembler code as
4303 usual. The function must also be compiled as usual if the program
4304 refers to its address, because that can't be inlined.
4307 Note that certain usages in a function definition can make it unsuitable
4308 for inline substitution. Among these usages are: use of varargs, use of
4309 alloca, use of variable sized data types (@pxref{Variable Length}),
4310 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4311 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4312 will warn when a function marked @code{inline} could not be substituted,
4313 and will give the reason for the failure.
4315 @cindex automatic @code{inline} for C++ member fns
4316 @cindex @code{inline} automatic for C++ member fns
4317 @cindex member fns, automatically @code{inline}
4318 @cindex C++ member fns, automatically @code{inline}
4319 @opindex fno-default-inline
4320 As required by ISO C++, GCC considers member functions defined within
4321 the body of a class to be marked inline even if they are
4322 not explicitly declared with the @code{inline} keyword. You can
4323 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4324 Options,,Options Controlling C++ Dialect}.
4326 GCC does not inline any functions when not optimizing unless you specify
4327 the @samp{always_inline} attribute for the function, like this:
4330 /* @r{Prototype.} */
4331 inline void foo (const char) __attribute__((always_inline));
4334 The remainder of this section is specific to GNU C89 inlining.
4336 @cindex non-static inline function
4337 When an inline function is not @code{static}, then the compiler must assume
4338 that there may be calls from other source files; since a global symbol can
4339 be defined only once in any program, the function must not be defined in
4340 the other source files, so the calls therein cannot be integrated.
4341 Therefore, a non-@code{static} inline function is always compiled on its
4342 own in the usual fashion.
4344 If you specify both @code{inline} and @code{extern} in the function
4345 definition, then the definition is used only for inlining. In no case
4346 is the function compiled on its own, not even if you refer to its
4347 address explicitly. Such an address becomes an external reference, as
4348 if you had only declared the function, and had not defined it.
4350 This combination of @code{inline} and @code{extern} has almost the
4351 effect of a macro. The way to use it is to put a function definition in
4352 a header file with these keywords, and put another copy of the
4353 definition (lacking @code{inline} and @code{extern}) in a library file.
4354 The definition in the header file will cause most calls to the function
4355 to be inlined. If any uses of the function remain, they will refer to
4356 the single copy in the library.
4359 @section Assembler Instructions with C Expression Operands
4360 @cindex extended @code{asm}
4361 @cindex @code{asm} expressions
4362 @cindex assembler instructions
4365 In an assembler instruction using @code{asm}, you can specify the
4366 operands of the instruction using C expressions. This means you need not
4367 guess which registers or memory locations will contain the data you want
4370 You must specify an assembler instruction template much like what
4371 appears in a machine description, plus an operand constraint string for
4374 For example, here is how to use the 68881's @code{fsinx} instruction:
4377 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4381 Here @code{angle} is the C expression for the input operand while
4382 @code{result} is that of the output operand. Each has @samp{"f"} as its
4383 operand constraint, saying that a floating point register is required.
4384 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4385 output operands' constraints must use @samp{=}. The constraints use the
4386 same language used in the machine description (@pxref{Constraints}).
4388 Each operand is described by an operand-constraint string followed by
4389 the C expression in parentheses. A colon separates the assembler
4390 template from the first output operand and another separates the last
4391 output operand from the first input, if any. Commas separate the
4392 operands within each group. The total number of operands is currently
4393 limited to 30; this limitation may be lifted in some future version of
4396 If there are no output operands but there are input operands, you must
4397 place two consecutive colons surrounding the place where the output
4400 As of GCC version 3.1, it is also possible to specify input and output
4401 operands using symbolic names which can be referenced within the
4402 assembler code. These names are specified inside square brackets
4403 preceding the constraint string, and can be referenced inside the
4404 assembler code using @code{%[@var{name}]} instead of a percentage sign
4405 followed by the operand number. Using named operands the above example
4409 asm ("fsinx %[angle],%[output]"
4410 : [output] "=f" (result)
4411 : [angle] "f" (angle));
4415 Note that the symbolic operand names have no relation whatsoever to
4416 other C identifiers. You may use any name you like, even those of
4417 existing C symbols, but you must ensure that no two operands within the same
4418 assembler construct use the same symbolic name.
4420 Output operand expressions must be lvalues; the compiler can check this.
4421 The input operands need not be lvalues. The compiler cannot check
4422 whether the operands have data types that are reasonable for the
4423 instruction being executed. It does not parse the assembler instruction
4424 template and does not know what it means or even whether it is valid
4425 assembler input. The extended @code{asm} feature is most often used for
4426 machine instructions the compiler itself does not know exist. If
4427 the output expression cannot be directly addressed (for example, it is a
4428 bit-field), your constraint must allow a register. In that case, GCC
4429 will use the register as the output of the @code{asm}, and then store
4430 that register into the output.
4432 The ordinary output operands must be write-only; GCC will assume that
4433 the values in these operands before the instruction are dead and need
4434 not be generated. Extended asm supports input-output or read-write
4435 operands. Use the constraint character @samp{+} to indicate such an
4436 operand and list it with the output operands. You should only use
4437 read-write operands when the constraints for the operand (or the
4438 operand in which only some of the bits are to be changed) allow a
4441 You may, as an alternative, logically split its function into two
4442 separate operands, one input operand and one write-only output
4443 operand. The connection between them is expressed by constraints
4444 which say they need to be in the same location when the instruction
4445 executes. You can use the same C expression for both operands, or
4446 different expressions. For example, here we write the (fictitious)
4447 @samp{combine} instruction with @code{bar} as its read-only source
4448 operand and @code{foo} as its read-write destination:
4451 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4455 The constraint @samp{"0"} for operand 1 says that it must occupy the
4456 same location as operand 0. A number in constraint is allowed only in
4457 an input operand and it must refer to an output operand.
4459 Only a number in the constraint can guarantee that one operand will be in
4460 the same place as another. The mere fact that @code{foo} is the value
4461 of both operands is not enough to guarantee that they will be in the
4462 same place in the generated assembler code. The following would not
4466 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4469 Various optimizations or reloading could cause operands 0 and 1 to be in
4470 different registers; GCC knows no reason not to do so. For example, the
4471 compiler might find a copy of the value of @code{foo} in one register and
4472 use it for operand 1, but generate the output operand 0 in a different
4473 register (copying it afterward to @code{foo}'s own address). Of course,
4474 since the register for operand 1 is not even mentioned in the assembler
4475 code, the result will not work, but GCC can't tell that.
4477 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4478 the operand number for a matching constraint. For example:
4481 asm ("cmoveq %1,%2,%[result]"
4482 : [result] "=r"(result)
4483 : "r" (test), "r"(new), "[result]"(old));
4486 Sometimes you need to make an @code{asm} operand be a specific register,
4487 but there's no matching constraint letter for that register @emph{by
4488 itself}. To force the operand into that register, use a local variable
4489 for the operand and specify the register in the variable declaration.
4490 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4491 register constraint letter that matches the register:
4494 register int *p1 asm ("r0") = @dots{};
4495 register int *p2 asm ("r1") = @dots{};
4496 register int *result asm ("r0");
4497 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4500 @anchor{Example of asm with clobbered asm reg}
4501 In the above example, beware that a register that is call-clobbered by
4502 the target ABI will be overwritten by any function call in the
4503 assignment, including library calls for arithmetic operators.
4504 Assuming it is a call-clobbered register, this may happen to @code{r0}
4505 above by the assignment to @code{p2}. If you have to use such a
4506 register, use temporary variables for expressions between the register
4511 register int *p1 asm ("r0") = @dots{};
4512 register int *p2 asm ("r1") = t1;
4513 register int *result asm ("r0");
4514 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4517 Some instructions clobber specific hard registers. To describe this,
4518 write a third colon after the input operands, followed by the names of
4519 the clobbered hard registers (given as strings). Here is a realistic
4520 example for the VAX:
4523 asm volatile ("movc3 %0,%1,%2"
4524 : /* @r{no outputs} */
4525 : "g" (from), "g" (to), "g" (count)
4526 : "r0", "r1", "r2", "r3", "r4", "r5");
4529 You may not write a clobber description in a way that overlaps with an
4530 input or output operand. For example, you may not have an operand
4531 describing a register class with one member if you mention that register
4532 in the clobber list. Variables declared to live in specific registers
4533 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4534 have no part mentioned in the clobber description.
4535 There is no way for you to specify that an input
4536 operand is modified without also specifying it as an output
4537 operand. Note that if all the output operands you specify are for this
4538 purpose (and hence unused), you will then also need to specify
4539 @code{volatile} for the @code{asm} construct, as described below, to
4540 prevent GCC from deleting the @code{asm} statement as unused.
4542 If you refer to a particular hardware register from the assembler code,
4543 you will probably have to list the register after the third colon to
4544 tell the compiler the register's value is modified. In some assemblers,
4545 the register names begin with @samp{%}; to produce one @samp{%} in the
4546 assembler code, you must write @samp{%%} in the input.
4548 If your assembler instruction can alter the condition code register, add
4549 @samp{cc} to the list of clobbered registers. GCC on some machines
4550 represents the condition codes as a specific hardware register;
4551 @samp{cc} serves to name this register. On other machines, the
4552 condition code is handled differently, and specifying @samp{cc} has no
4553 effect. But it is valid no matter what the machine.
4555 If your assembler instructions access memory in an unpredictable
4556 fashion, add @samp{memory} to the list of clobbered registers. This
4557 will cause GCC to not keep memory values cached in registers across the
4558 assembler instruction and not optimize stores or loads to that memory.
4559 You will also want to add the @code{volatile} keyword if the memory
4560 affected is not listed in the inputs or outputs of the @code{asm}, as
4561 the @samp{memory} clobber does not count as a side-effect of the
4562 @code{asm}. If you know how large the accessed memory is, you can add
4563 it as input or output but if this is not known, you should add
4564 @samp{memory}. As an example, if you access ten bytes of a string, you
4565 can use a memory input like:
4568 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4571 Note that in the following example the memory input is necessary,
4572 otherwise GCC might optimize the store to @code{x} away:
4579 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4580 "=&d" (r) : "a" (y), "m" (*y));
4585 You can put multiple assembler instructions together in a single
4586 @code{asm} template, separated by the characters normally used in assembly
4587 code for the system. A combination that works in most places is a newline
4588 to break the line, plus a tab character to move to the instruction field
4589 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4590 assembler allows semicolons as a line-breaking character. Note that some
4591 assembler dialects use semicolons to start a comment.
4592 The input operands are guaranteed not to use any of the clobbered
4593 registers, and neither will the output operands' addresses, so you can
4594 read and write the clobbered registers as many times as you like. Here
4595 is an example of multiple instructions in a template; it assumes the
4596 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4599 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4601 : "g" (from), "g" (to)
4605 Unless an output operand has the @samp{&} constraint modifier, GCC
4606 may allocate it in the same register as an unrelated input operand, on
4607 the assumption the inputs are consumed before the outputs are produced.
4608 This assumption may be false if the assembler code actually consists of
4609 more than one instruction. In such a case, use @samp{&} for each output
4610 operand that may not overlap an input. @xref{Modifiers}.
4612 If you want to test the condition code produced by an assembler
4613 instruction, you must include a branch and a label in the @code{asm}
4614 construct, as follows:
4617 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4623 This assumes your assembler supports local labels, as the GNU assembler
4624 and most Unix assemblers do.
4626 Speaking of labels, jumps from one @code{asm} to another are not
4627 supported. The compiler's optimizers do not know about these jumps, and
4628 therefore they cannot take account of them when deciding how to
4631 @cindex macros containing @code{asm}
4632 Usually the most convenient way to use these @code{asm} instructions is to
4633 encapsulate them in macros that look like functions. For example,
4637 (@{ double __value, __arg = (x); \
4638 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4643 Here the variable @code{__arg} is used to make sure that the instruction
4644 operates on a proper @code{double} value, and to accept only those
4645 arguments @code{x} which can convert automatically to a @code{double}.
4647 Another way to make sure the instruction operates on the correct data
4648 type is to use a cast in the @code{asm}. This is different from using a
4649 variable @code{__arg} in that it converts more different types. For
4650 example, if the desired type were @code{int}, casting the argument to
4651 @code{int} would accept a pointer with no complaint, while assigning the
4652 argument to an @code{int} variable named @code{__arg} would warn about
4653 using a pointer unless the caller explicitly casts it.
4655 If an @code{asm} has output operands, GCC assumes for optimization
4656 purposes the instruction has no side effects except to change the output
4657 operands. This does not mean instructions with a side effect cannot be
4658 used, but you must be careful, because the compiler may eliminate them
4659 if the output operands aren't used, or move them out of loops, or
4660 replace two with one if they constitute a common subexpression. Also,
4661 if your instruction does have a side effect on a variable that otherwise
4662 appears not to change, the old value of the variable may be reused later
4663 if it happens to be found in a register.
4665 You can prevent an @code{asm} instruction from being deleted
4666 by writing the keyword @code{volatile} after
4667 the @code{asm}. For example:
4670 #define get_and_set_priority(new) \
4672 asm volatile ("get_and_set_priority %0, %1" \
4673 : "=g" (__old) : "g" (new)); \
4678 The @code{volatile} keyword indicates that the instruction has
4679 important side-effects. GCC will not delete a volatile @code{asm} if
4680 it is reachable. (The instruction can still be deleted if GCC can
4681 prove that control-flow will never reach the location of the
4682 instruction.) Note that even a volatile @code{asm} instruction
4683 can be moved relative to other code, including across jump
4684 instructions. For example, on many targets there is a system
4685 register which can be set to control the rounding mode of
4686 floating point operations. You might try
4687 setting it with a volatile @code{asm}, like this PowerPC example:
4690 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4695 This will not work reliably, as the compiler may move the addition back
4696 before the volatile @code{asm}. To make it work you need to add an
4697 artificial dependency to the @code{asm} referencing a variable in the code
4698 you don't want moved, for example:
4701 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4705 Similarly, you can't expect a
4706 sequence of volatile @code{asm} instructions to remain perfectly
4707 consecutive. If you want consecutive output, use a single @code{asm}.
4708 Also, GCC will perform some optimizations across a volatile @code{asm}
4709 instruction; GCC does not ``forget everything'' when it encounters
4710 a volatile @code{asm} instruction the way some other compilers do.
4712 An @code{asm} instruction without any output operands will be treated
4713 identically to a volatile @code{asm} instruction.
4715 It is a natural idea to look for a way to give access to the condition
4716 code left by the assembler instruction. However, when we attempted to
4717 implement this, we found no way to make it work reliably. The problem
4718 is that output operands might need reloading, which would result in
4719 additional following ``store'' instructions. On most machines, these
4720 instructions would alter the condition code before there was time to
4721 test it. This problem doesn't arise for ordinary ``test'' and
4722 ``compare'' instructions because they don't have any output operands.
4724 For reasons similar to those described above, it is not possible to give
4725 an assembler instruction access to the condition code left by previous
4728 If you are writing a header file that should be includable in ISO C
4729 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4732 @subsection Size of an @code{asm}
4734 Some targets require that GCC track the size of each instruction used in
4735 order to generate correct code. Because the final length of an
4736 @code{asm} is only known by the assembler, GCC must make an estimate as
4737 to how big it will be. The estimate is formed by counting the number of
4738 statements in the pattern of the @code{asm} and multiplying that by the
4739 length of the longest instruction on that processor. Statements in the
4740 @code{asm} are identified by newline characters and whatever statement
4741 separator characters are supported by the assembler; on most processors
4742 this is the `@code{;}' character.
4744 Normally, GCC's estimate is perfectly adequate to ensure that correct
4745 code is generated, but it is possible to confuse the compiler if you use
4746 pseudo instructions or assembler macros that expand into multiple real
4747 instructions or if you use assembler directives that expand to more
4748 space in the object file than would be needed for a single instruction.
4749 If this happens then the assembler will produce a diagnostic saying that
4750 a label is unreachable.
4752 @subsection i386 floating point asm operands
4754 There are several rules on the usage of stack-like regs in
4755 asm_operands insns. These rules apply only to the operands that are
4760 Given a set of input regs that die in an asm_operands, it is
4761 necessary to know which are implicitly popped by the asm, and
4762 which must be explicitly popped by gcc.
4764 An input reg that is implicitly popped by the asm must be
4765 explicitly clobbered, unless it is constrained to match an
4769 For any input reg that is implicitly popped by an asm, it is
4770 necessary to know how to adjust the stack to compensate for the pop.
4771 If any non-popped input is closer to the top of the reg-stack than
4772 the implicitly popped reg, it would not be possible to know what the
4773 stack looked like---it's not clear how the rest of the stack ``slides
4776 All implicitly popped input regs must be closer to the top of
4777 the reg-stack than any input that is not implicitly popped.
4779 It is possible that if an input dies in an insn, reload might
4780 use the input reg for an output reload. Consider this example:
4783 asm ("foo" : "=t" (a) : "f" (b));
4786 This asm says that input B is not popped by the asm, and that
4787 the asm pushes a result onto the reg-stack, i.e., the stack is one
4788 deeper after the asm than it was before. But, it is possible that
4789 reload will think that it can use the same reg for both the input and
4790 the output, if input B dies in this insn.
4792 If any input operand uses the @code{f} constraint, all output reg
4793 constraints must use the @code{&} earlyclobber.
4795 The asm above would be written as
4798 asm ("foo" : "=&t" (a) : "f" (b));
4802 Some operands need to be in particular places on the stack. All
4803 output operands fall in this category---there is no other way to
4804 know which regs the outputs appear in unless the user indicates
4805 this in the constraints.
4807 Output operands must specifically indicate which reg an output
4808 appears in after an asm. @code{=f} is not allowed: the operand
4809 constraints must select a class with a single reg.
4812 Output operands may not be ``inserted'' between existing stack regs.
4813 Since no 387 opcode uses a read/write operand, all output operands
4814 are dead before the asm_operands, and are pushed by the asm_operands.
4815 It makes no sense to push anywhere but the top of the reg-stack.
4817 Output operands must start at the top of the reg-stack: output
4818 operands may not ``skip'' a reg.
4821 Some asm statements may need extra stack space for internal
4822 calculations. This can be guaranteed by clobbering stack registers
4823 unrelated to the inputs and outputs.
4827 Here are a couple of reasonable asms to want to write. This asm
4828 takes one input, which is internally popped, and produces two outputs.
4831 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4834 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4835 and replaces them with one output. The user must code the @code{st(1)}
4836 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4839 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4845 @section Controlling Names Used in Assembler Code
4846 @cindex assembler names for identifiers
4847 @cindex names used in assembler code
4848 @cindex identifiers, names in assembler code
4850 You can specify the name to be used in the assembler code for a C
4851 function or variable by writing the @code{asm} (or @code{__asm__})
4852 keyword after the declarator as follows:
4855 int foo asm ("myfoo") = 2;
4859 This specifies that the name to be used for the variable @code{foo} in
4860 the assembler code should be @samp{myfoo} rather than the usual
4863 On systems where an underscore is normally prepended to the name of a C
4864 function or variable, this feature allows you to define names for the
4865 linker that do not start with an underscore.
4867 It does not make sense to use this feature with a non-static local
4868 variable since such variables do not have assembler names. If you are
4869 trying to put the variable in a particular register, see @ref{Explicit
4870 Reg Vars}. GCC presently accepts such code with a warning, but will
4871 probably be changed to issue an error, rather than a warning, in the
4874 You cannot use @code{asm} in this way in a function @emph{definition}; but
4875 you can get the same effect by writing a declaration for the function
4876 before its definition and putting @code{asm} there, like this:
4879 extern func () asm ("FUNC");
4886 It is up to you to make sure that the assembler names you choose do not
4887 conflict with any other assembler symbols. Also, you must not use a
4888 register name; that would produce completely invalid assembler code. GCC
4889 does not as yet have the ability to store static variables in registers.
4890 Perhaps that will be added.
4892 @node Explicit Reg Vars
4893 @section Variables in Specified Registers
4894 @cindex explicit register variables
4895 @cindex variables in specified registers
4896 @cindex specified registers
4897 @cindex registers, global allocation
4899 GNU C allows you to put a few global variables into specified hardware
4900 registers. You can also specify the register in which an ordinary
4901 register variable should be allocated.
4905 Global register variables reserve registers throughout the program.
4906 This may be useful in programs such as programming language
4907 interpreters which have a couple of global variables that are accessed
4911 Local register variables in specific registers do not reserve the
4912 registers, except at the point where they are used as input or output
4913 operands in an @code{asm} statement and the @code{asm} statement itself is
4914 not deleted. The compiler's data flow analysis is capable of determining
4915 where the specified registers contain live values, and where they are
4916 available for other uses. Stores into local register variables may be deleted
4917 when they appear to be dead according to dataflow analysis. References
4918 to local register variables may be deleted or moved or simplified.
4920 These local variables are sometimes convenient for use with the extended
4921 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4922 output of the assembler instruction directly into a particular register.
4923 (This will work provided the register you specify fits the constraints
4924 specified for that operand in the @code{asm}.)
4932 @node Global Reg Vars
4933 @subsection Defining Global Register Variables
4934 @cindex global register variables
4935 @cindex registers, global variables in
4937 You can define a global register variable in GNU C like this:
4940 register int *foo asm ("a5");
4944 Here @code{a5} is the name of the register which should be used. Choose a
4945 register which is normally saved and restored by function calls on your
4946 machine, so that library routines will not clobber it.
4948 Naturally the register name is cpu-dependent, so you would need to
4949 conditionalize your program according to cpu type. The register
4950 @code{a5} would be a good choice on a 68000 for a variable of pointer
4951 type. On machines with register windows, be sure to choose a ``global''
4952 register that is not affected magically by the function call mechanism.
4954 In addition, operating systems on one type of cpu may differ in how they
4955 name the registers; then you would need additional conditionals. For
4956 example, some 68000 operating systems call this register @code{%a5}.
4958 Eventually there may be a way of asking the compiler to choose a register
4959 automatically, but first we need to figure out how it should choose and
4960 how to enable you to guide the choice. No solution is evident.
4962 Defining a global register variable in a certain register reserves that
4963 register entirely for this use, at least within the current compilation.
4964 The register will not be allocated for any other purpose in the functions
4965 in the current compilation. The register will not be saved and restored by
4966 these functions. Stores into this register are never deleted even if they
4967 would appear to be dead, but references may be deleted or moved or
4970 It is not safe to access the global register variables from signal
4971 handlers, or from more than one thread of control, because the system
4972 library routines may temporarily use the register for other things (unless
4973 you recompile them specially for the task at hand).
4975 @cindex @code{qsort}, and global register variables
4976 It is not safe for one function that uses a global register variable to
4977 call another such function @code{foo} by way of a third function
4978 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4979 different source file in which the variable wasn't declared). This is
4980 because @code{lose} might save the register and put some other value there.
4981 For example, you can't expect a global register variable to be available in
4982 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4983 might have put something else in that register. (If you are prepared to
4984 recompile @code{qsort} with the same global register variable, you can
4985 solve this problem.)
4987 If you want to recompile @code{qsort} or other source files which do not
4988 actually use your global register variable, so that they will not use that
4989 register for any other purpose, then it suffices to specify the compiler
4990 option @option{-ffixed-@var{reg}}. You need not actually add a global
4991 register declaration to their source code.
4993 A function which can alter the value of a global register variable cannot
4994 safely be called from a function compiled without this variable, because it
4995 could clobber the value the caller expects to find there on return.
4996 Therefore, the function which is the entry point into the part of the
4997 program that uses the global register variable must explicitly save and
4998 restore the value which belongs to its caller.
5000 @cindex register variable after @code{longjmp}
5001 @cindex global register after @code{longjmp}
5002 @cindex value after @code{longjmp}
5005 On most machines, @code{longjmp} will restore to each global register
5006 variable the value it had at the time of the @code{setjmp}. On some
5007 machines, however, @code{longjmp} will not change the value of global
5008 register variables. To be portable, the function that called @code{setjmp}
5009 should make other arrangements to save the values of the global register
5010 variables, and to restore them in a @code{longjmp}. This way, the same
5011 thing will happen regardless of what @code{longjmp} does.
5013 All global register variable declarations must precede all function
5014 definitions. If such a declaration could appear after function
5015 definitions, the declaration would be too late to prevent the register from
5016 being used for other purposes in the preceding functions.
5018 Global register variables may not have initial values, because an
5019 executable file has no means to supply initial contents for a register.
5021 On the SPARC, there are reports that g3 @dots{} g7 are suitable
5022 registers, but certain library functions, such as @code{getwd}, as well
5023 as the subroutines for division and remainder, modify g3 and g4. g1 and
5024 g2 are local temporaries.
5026 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
5027 Of course, it will not do to use more than a few of those.
5029 @node Local Reg Vars
5030 @subsection Specifying Registers for Local Variables
5031 @cindex local variables, specifying registers
5032 @cindex specifying registers for local variables
5033 @cindex registers for local variables
5035 You can define a local register variable with a specified register
5039 register int *foo asm ("a5");
5043 Here @code{a5} is the name of the register which should be used. Note
5044 that this is the same syntax used for defining global register
5045 variables, but for a local variable it would appear within a function.
5047 Naturally the register name is cpu-dependent, but this is not a
5048 problem, since specific registers are most often useful with explicit
5049 assembler instructions (@pxref{Extended Asm}). Both of these things
5050 generally require that you conditionalize your program according to
5053 In addition, operating systems on one type of cpu may differ in how they
5054 name the registers; then you would need additional conditionals. For
5055 example, some 68000 operating systems call this register @code{%a5}.
5057 Defining such a register variable does not reserve the register; it
5058 remains available for other uses in places where flow control determines
5059 the variable's value is not live.
5061 This option does not guarantee that GCC will generate code that has
5062 this variable in the register you specify at all times. You may not
5063 code an explicit reference to this register in the @emph{assembler
5064 instruction template} part of an @code{asm} statement and assume it will
5065 always refer to this variable. However, using the variable as an
5066 @code{asm} @emph{operand} guarantees that the specified register is used
5069 Stores into local register variables may be deleted when they appear to be dead
5070 according to dataflow analysis. References to local register variables may
5071 be deleted or moved or simplified.
5073 As for global register variables, it's recommended that you choose a
5074 register which is normally saved and restored by function calls on
5075 your machine, so that library routines will not clobber it. A common
5076 pitfall is to initialize multiple call-clobbered registers with
5077 arbitrary expressions, where a function call or library call for an
5078 arithmetic operator will overwrite a register value from a previous
5079 assignment, for example @code{r0} below:
5081 register int *p1 asm ("r0") = @dots{};
5082 register int *p2 asm ("r1") = @dots{};
5084 In those cases, a solution is to use a temporary variable for
5085 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
5087 @node Alternate Keywords
5088 @section Alternate Keywords
5089 @cindex alternate keywords
5090 @cindex keywords, alternate
5092 @option{-ansi} and the various @option{-std} options disable certain
5093 keywords. This causes trouble when you want to use GNU C extensions, or
5094 a general-purpose header file that should be usable by all programs,
5095 including ISO C programs. The keywords @code{asm}, @code{typeof} and
5096 @code{inline} are not available in programs compiled with
5097 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
5098 program compiled with @option{-std=c99}). The ISO C99 keyword
5099 @code{restrict} is only available when @option{-std=gnu99} (which will
5100 eventually be the default) or @option{-std=c99} (or the equivalent
5101 @option{-std=iso9899:1999}) is used.
5103 The way to solve these problems is to put @samp{__} at the beginning and
5104 end of each problematical keyword. For example, use @code{__asm__}
5105 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
5107 Other C compilers won't accept these alternative keywords; if you want to
5108 compile with another compiler, you can define the alternate keywords as
5109 macros to replace them with the customary keywords. It looks like this:
5117 @findex __extension__
5119 @option{-pedantic} and other options cause warnings for many GNU C extensions.
5121 prevent such warnings within one expression by writing
5122 @code{__extension__} before the expression. @code{__extension__} has no
5123 effect aside from this.
5125 @node Incomplete Enums
5126 @section Incomplete @code{enum} Types
5128 You can define an @code{enum} tag without specifying its possible values.
5129 This results in an incomplete type, much like what you get if you write
5130 @code{struct foo} without describing the elements. A later declaration
5131 which does specify the possible values completes the type.
5133 You can't allocate variables or storage using the type while it is
5134 incomplete. However, you can work with pointers to that type.
5136 This extension may not be very useful, but it makes the handling of
5137 @code{enum} more consistent with the way @code{struct} and @code{union}
5140 This extension is not supported by GNU C++.
5142 @node Function Names
5143 @section Function Names as Strings
5144 @cindex @code{__func__} identifier
5145 @cindex @code{__FUNCTION__} identifier
5146 @cindex @code{__PRETTY_FUNCTION__} identifier
5148 GCC provides three magic variables which hold the name of the current
5149 function, as a string. The first of these is @code{__func__}, which
5150 is part of the C99 standard:
5153 The identifier @code{__func__} is implicitly declared by the translator
5154 as if, immediately following the opening brace of each function
5155 definition, the declaration
5158 static const char __func__[] = "function-name";
5161 appeared, where function-name is the name of the lexically-enclosing
5162 function. This name is the unadorned name of the function.
5165 @code{__FUNCTION__} is another name for @code{__func__}. Older
5166 versions of GCC recognize only this name. However, it is not
5167 standardized. For maximum portability, we recommend you use
5168 @code{__func__}, but provide a fallback definition with the
5172 #if __STDC_VERSION__ < 199901L
5174 # define __func__ __FUNCTION__
5176 # define __func__ "<unknown>"
5181 In C, @code{__PRETTY_FUNCTION__} is yet another name for
5182 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
5183 the type signature of the function as well as its bare name. For
5184 example, this program:
5188 extern int printf (char *, ...);
5195 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
5196 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
5214 __PRETTY_FUNCTION__ = void a::sub(int)
5217 These identifiers are not preprocessor macros. In GCC 3.3 and
5218 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
5219 were treated as string literals; they could be used to initialize
5220 @code{char} arrays, and they could be concatenated with other string
5221 literals. GCC 3.4 and later treat them as variables, like
5222 @code{__func__}. In C++, @code{__FUNCTION__} and
5223 @code{__PRETTY_FUNCTION__} have always been variables.
5225 @node Return Address
5226 @section Getting the Return or Frame Address of a Function
5228 These functions may be used to get information about the callers of a
5231 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
5232 This function returns the return address of the current function, or of
5233 one of its callers. The @var{level} argument is number of frames to
5234 scan up the call stack. A value of @code{0} yields the return address
5235 of the current function, a value of @code{1} yields the return address
5236 of the caller of the current function, and so forth. When inlining
5237 the expected behavior is that the function will return the address of
5238 the function that will be returned to. To work around this behavior use
5239 the @code{noinline} function attribute.
5241 The @var{level} argument must be a constant integer.
5243 On some machines it may be impossible to determine the return address of
5244 any function other than the current one; in such cases, or when the top
5245 of the stack has been reached, this function will return @code{0} or a
5246 random value. In addition, @code{__builtin_frame_address} may be used
5247 to determine if the top of the stack has been reached.
5249 This function should only be used with a nonzero argument for debugging
5253 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
5254 This function is similar to @code{__builtin_return_address}, but it
5255 returns the address of the function frame rather than the return address
5256 of the function. Calling @code{__builtin_frame_address} with a value of
5257 @code{0} yields the frame address of the current function, a value of
5258 @code{1} yields the frame address of the caller of the current function,
5261 The frame is the area on the stack which holds local variables and saved
5262 registers. The frame address is normally the address of the first word
5263 pushed on to the stack by the function. However, the exact definition
5264 depends upon the processor and the calling convention. If the processor
5265 has a dedicated frame pointer register, and the function has a frame,
5266 then @code{__builtin_frame_address} will return the value of the frame
5269 On some machines it may be impossible to determine the frame address of
5270 any function other than the current one; in such cases, or when the top
5271 of the stack has been reached, this function will return @code{0} if
5272 the first frame pointer is properly initialized by the startup code.
5274 This function should only be used with a nonzero argument for debugging
5278 @node Vector Extensions
5279 @section Using vector instructions through built-in functions
5281 On some targets, the instruction set contains SIMD vector instructions that
5282 operate on multiple values contained in one large register at the same time.
5283 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5286 The first step in using these extensions is to provide the necessary data
5287 types. This should be done using an appropriate @code{typedef}:
5290 typedef int v4si __attribute__ ((vector_size (16)));
5293 The @code{int} type specifies the base type, while the attribute specifies
5294 the vector size for the variable, measured in bytes. For example, the
5295 declaration above causes the compiler to set the mode for the @code{v4si}
5296 type to be 16 bytes wide and divided into @code{int} sized units. For
5297 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5298 corresponding mode of @code{foo} will be @acronym{V4SI}.
5300 The @code{vector_size} attribute is only applicable to integral and
5301 float scalars, although arrays, pointers, and function return values
5302 are allowed in conjunction with this construct.
5304 All the basic integer types can be used as base types, both as signed
5305 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5306 @code{long long}. In addition, @code{float} and @code{double} can be
5307 used to build floating-point vector types.
5309 Specifying a combination that is not valid for the current architecture
5310 will cause GCC to synthesize the instructions using a narrower mode.
5311 For example, if you specify a variable of type @code{V4SI} and your
5312 architecture does not allow for this specific SIMD type, GCC will
5313 produce code that uses 4 @code{SIs}.
5315 The types defined in this manner can be used with a subset of normal C
5316 operations. Currently, GCC will allow using the following operators
5317 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5319 The operations behave like C++ @code{valarrays}. Addition is defined as
5320 the addition of the corresponding elements of the operands. For
5321 example, in the code below, each of the 4 elements in @var{a} will be
5322 added to the corresponding 4 elements in @var{b} and the resulting
5323 vector will be stored in @var{c}.
5326 typedef int v4si __attribute__ ((vector_size (16)));
5333 Subtraction, multiplication, division, and the logical operations
5334 operate in a similar manner. Likewise, the result of using the unary
5335 minus or complement operators on a vector type is a vector whose
5336 elements are the negative or complemented values of the corresponding
5337 elements in the operand.
5339 You can declare variables and use them in function calls and returns, as
5340 well as in assignments and some casts. You can specify a vector type as
5341 a return type for a function. Vector types can also be used as function
5342 arguments. It is possible to cast from one vector type to another,
5343 provided they are of the same size (in fact, you can also cast vectors
5344 to and from other datatypes of the same size).
5346 You cannot operate between vectors of different lengths or different
5347 signedness without a cast.
5349 A port that supports hardware vector operations, usually provides a set
5350 of built-in functions that can be used to operate on vectors. For
5351 example, a function to add two vectors and multiply the result by a
5352 third could look like this:
5355 v4si f (v4si a, v4si b, v4si c)
5357 v4si tmp = __builtin_addv4si (a, b);
5358 return __builtin_mulv4si (tmp, c);
5365 @findex __builtin_offsetof
5367 GCC implements for both C and C++ a syntactic extension to implement
5368 the @code{offsetof} macro.
5372 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5374 offsetof_member_designator:
5376 | offsetof_member_designator "." @code{identifier}
5377 | offsetof_member_designator "[" @code{expr} "]"
5380 This extension is sufficient such that
5383 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5386 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5387 may be dependent. In either case, @var{member} may consist of a single
5388 identifier, or a sequence of member accesses and array references.
5390 @node Atomic Builtins
5391 @section Built-in functions for atomic memory access
5393 The following builtins are intended to be compatible with those described
5394 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5395 section 7.4. As such, they depart from the normal GCC practice of using
5396 the ``__builtin_'' prefix, and further that they are overloaded such that
5397 they work on multiple types.
5399 The definition given in the Intel documentation allows only for the use of
5400 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5401 counterparts. GCC will allow any integral scalar or pointer type that is
5402 1, 2, 4 or 8 bytes in length.
5404 Not all operations are supported by all target processors. If a particular
5405 operation cannot be implemented on the target processor, a warning will be
5406 generated and a call an external function will be generated. The external
5407 function will carry the same name as the builtin, with an additional suffix
5408 @samp{_@var{n}} where @var{n} is the size of the data type.
5410 @c ??? Should we have a mechanism to suppress this warning? This is almost
5411 @c useful for implementing the operation under the control of an external
5414 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5415 no memory operand will be moved across the operation, either forward or
5416 backward. Further, instructions will be issued as necessary to prevent the
5417 processor from speculating loads across the operation and from queuing stores
5418 after the operation.
5420 All of the routines are are described in the Intel documentation to take
5421 ``an optional list of variables protected by the memory barrier''. It's
5422 not clear what is meant by that; it could mean that @emph{only} the
5423 following variables are protected, or it could mean that these variables
5424 should in addition be protected. At present GCC ignores this list and
5425 protects all variables which are globally accessible. If in the future
5426 we make some use of this list, an empty list will continue to mean all
5427 globally accessible variables.
5430 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5431 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5432 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5433 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5434 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5435 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5436 @findex __sync_fetch_and_add
5437 @findex __sync_fetch_and_sub
5438 @findex __sync_fetch_and_or
5439 @findex __sync_fetch_and_and
5440 @findex __sync_fetch_and_xor
5441 @findex __sync_fetch_and_nand
5442 These builtins perform the operation suggested by the name, and
5443 returns the value that had previously been in memory. That is,
5446 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5447 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5450 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5451 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5452 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5453 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5454 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5455 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5456 @findex __sync_add_and_fetch
5457 @findex __sync_sub_and_fetch
5458 @findex __sync_or_and_fetch
5459 @findex __sync_and_and_fetch
5460 @findex __sync_xor_and_fetch
5461 @findex __sync_nand_and_fetch
5462 These builtins perform the operation suggested by the name, and
5463 return the new value. That is,
5466 @{ *ptr @var{op}= value; return *ptr; @}
5467 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5470 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5471 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5472 @findex __sync_bool_compare_and_swap
5473 @findex __sync_val_compare_and_swap
5474 These builtins perform an atomic compare and swap. That is, if the current
5475 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5478 The ``bool'' version returns true if the comparison is successful and
5479 @var{newval} was written. The ``val'' version returns the contents
5480 of @code{*@var{ptr}} before the operation.
5482 @item __sync_synchronize (...)
5483 @findex __sync_synchronize
5484 This builtin issues a full memory barrier.
5486 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5487 @findex __sync_lock_test_and_set
5488 This builtin, as described by Intel, is not a traditional test-and-set
5489 operation, but rather an atomic exchange operation. It writes @var{value}
5490 into @code{*@var{ptr}}, and returns the previous contents of
5493 Many targets have only minimal support for such locks, and do not support
5494 a full exchange operation. In this case, a target may support reduced
5495 functionality here by which the @emph{only} valid value to store is the
5496 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5497 is implementation defined.
5499 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5500 This means that references after the builtin cannot move to (or be
5501 speculated to) before the builtin, but previous memory stores may not
5502 be globally visible yet, and previous memory loads may not yet be
5505 @item void __sync_lock_release (@var{type} *ptr, ...)
5506 @findex __sync_lock_release
5507 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5508 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5510 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5511 This means that all previous memory stores are globally visible, and all
5512 previous memory loads have been satisfied, but following memory reads
5513 are not prevented from being speculated to before the barrier.
5516 @node Object Size Checking
5517 @section Object Size Checking Builtins
5518 @findex __builtin_object_size
5519 @findex __builtin___memcpy_chk
5520 @findex __builtin___mempcpy_chk
5521 @findex __builtin___memmove_chk
5522 @findex __builtin___memset_chk
5523 @findex __builtin___strcpy_chk
5524 @findex __builtin___stpcpy_chk
5525 @findex __builtin___strncpy_chk
5526 @findex __builtin___strcat_chk
5527 @findex __builtin___strncat_chk
5528 @findex __builtin___sprintf_chk
5529 @findex __builtin___snprintf_chk
5530 @findex __builtin___vsprintf_chk
5531 @findex __builtin___vsnprintf_chk
5532 @findex __builtin___printf_chk
5533 @findex __builtin___vprintf_chk
5534 @findex __builtin___fprintf_chk
5535 @findex __builtin___vfprintf_chk
5537 GCC implements a limited buffer overflow protection mechanism
5538 that can prevent some buffer overflow attacks.
5540 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5541 is a built-in construct that returns a constant number of bytes from
5542 @var{ptr} to the end of the object @var{ptr} pointer points to
5543 (if known at compile time). @code{__builtin_object_size} never evaluates
5544 its arguments for side-effects. If there are any side-effects in them, it
5545 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5546 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5547 point to and all of them are known at compile time, the returned number
5548 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5549 0 and minimum if nonzero. If it is not possible to determine which objects
5550 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5551 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5552 for @var{type} 2 or 3.
5554 @var{type} is an integer constant from 0 to 3. If the least significant
5555 bit is clear, objects are whole variables, if it is set, a closest
5556 surrounding subobject is considered the object a pointer points to.
5557 The second bit determines if maximum or minimum of remaining bytes
5561 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5562 char *p = &var.buf1[1], *q = &var.b;
5564 /* Here the object p points to is var. */
5565 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5566 /* The subobject p points to is var.buf1. */
5567 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5568 /* The object q points to is var. */
5569 assert (__builtin_object_size (q, 0)
5570 == (char *) (&var + 1) - (char *) &var.b);
5571 /* The subobject q points to is var.b. */
5572 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5576 There are built-in functions added for many common string operation
5577 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5578 built-in is provided. This built-in has an additional last argument,
5579 which is the number of bytes remaining in object the @var{dest}
5580 argument points to or @code{(size_t) -1} if the size is not known.
5582 The built-in functions are optimized into the normal string functions
5583 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5584 it is known at compile time that the destination object will not
5585 be overflown. If the compiler can determine at compile time the
5586 object will be always overflown, it issues a warning.
5588 The intended use can be e.g.
5592 #define bos0(dest) __builtin_object_size (dest, 0)
5593 #define memcpy(dest, src, n) \
5594 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5598 /* It is unknown what object p points to, so this is optimized
5599 into plain memcpy - no checking is possible. */
5600 memcpy (p, "abcde", n);
5601 /* Destination is known and length too. It is known at compile
5602 time there will be no overflow. */
5603 memcpy (&buf[5], "abcde", 5);
5604 /* Destination is known, but the length is not known at compile time.
5605 This will result in __memcpy_chk call that can check for overflow
5607 memcpy (&buf[5], "abcde", n);
5608 /* Destination is known and it is known at compile time there will
5609 be overflow. There will be a warning and __memcpy_chk call that
5610 will abort the program at runtime. */
5611 memcpy (&buf[6], "abcde", 5);
5614 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5615 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5616 @code{strcat} and @code{strncat}.
5618 There are also checking built-in functions for formatted output functions.
5620 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5621 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5622 const char *fmt, ...);
5623 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5625 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5626 const char *fmt, va_list ap);
5629 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5630 etc. functions and can contain implementation specific flags on what
5631 additional security measures the checking function might take, such as
5632 handling @code{%n} differently.
5634 The @var{os} argument is the object size @var{s} points to, like in the
5635 other built-in functions. There is a small difference in the behavior
5636 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5637 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5638 the checking function is called with @var{os} argument set to
5641 In addition to this, there are checking built-in functions
5642 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5643 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5644 These have just one additional argument, @var{flag}, right before
5645 format string @var{fmt}. If the compiler is able to optimize them to
5646 @code{fputc} etc. functions, it will, otherwise the checking function
5647 should be called and the @var{flag} argument passed to it.
5649 @node Other Builtins
5650 @section Other built-in functions provided by GCC
5651 @cindex built-in functions
5652 @findex __builtin_isfinite
5653 @findex __builtin_isnormal
5654 @findex __builtin_isgreater
5655 @findex __builtin_isgreaterequal
5656 @findex __builtin_isless
5657 @findex __builtin_islessequal
5658 @findex __builtin_islessgreater
5659 @findex __builtin_isunordered
5660 @findex __builtin_powi
5661 @findex __builtin_powif
5662 @findex __builtin_powil
5820 @findex fprintf_unlocked
5822 @findex fputs_unlocked
5939 @findex printf_unlocked
5971 @findex significandf
5972 @findex significandl
6043 GCC provides a large number of built-in functions other than the ones
6044 mentioned above. Some of these are for internal use in the processing
6045 of exceptions or variable-length argument lists and will not be
6046 documented here because they may change from time to time; we do not
6047 recommend general use of these functions.
6049 The remaining functions are provided for optimization purposes.
6051 @opindex fno-builtin
6052 GCC includes built-in versions of many of the functions in the standard
6053 C library. The versions prefixed with @code{__builtin_} will always be
6054 treated as having the same meaning as the C library function even if you
6055 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
6056 Many of these functions are only optimized in certain cases; if they are
6057 not optimized in a particular case, a call to the library function will
6062 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
6063 @option{-std=c99}), the functions
6064 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
6065 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
6066 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
6067 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
6068 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
6069 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
6070 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
6071 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
6072 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
6073 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
6074 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
6075 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
6076 @code{signbitd64}, @code{signbitd128}, @code{significandf},
6077 @code{significandl}, @code{significand}, @code{sincosf},
6078 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
6079 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
6080 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
6081 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
6083 may be handled as built-in functions.
6084 All these functions have corresponding versions
6085 prefixed with @code{__builtin_}, which may be used even in strict C89
6088 The ISO C99 functions
6089 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
6090 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
6091 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
6092 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
6093 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
6094 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
6095 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
6096 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
6097 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
6098 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
6099 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
6100 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
6101 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
6102 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
6103 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
6104 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
6105 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
6106 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
6107 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
6108 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
6109 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
6110 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
6111 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
6112 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
6113 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
6114 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
6115 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
6116 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
6117 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
6118 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
6119 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
6120 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
6121 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
6122 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
6123 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
6124 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
6125 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
6126 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
6127 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
6128 are handled as built-in functions
6129 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6131 There are also built-in versions of the ISO C99 functions
6132 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
6133 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
6134 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
6135 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
6136 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
6137 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
6138 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
6139 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
6140 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
6141 that are recognized in any mode since ISO C90 reserves these names for
6142 the purpose to which ISO C99 puts them. All these functions have
6143 corresponding versions prefixed with @code{__builtin_}.
6145 The ISO C94 functions
6146 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
6147 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
6148 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
6150 are handled as built-in functions
6151 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
6153 The ISO C90 functions
6154 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
6155 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
6156 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
6157 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
6158 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
6159 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
6160 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
6161 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
6162 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
6163 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
6164 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
6165 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
6166 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
6167 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
6168 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
6169 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
6170 are all recognized as built-in functions unless
6171 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
6172 is specified for an individual function). All of these functions have
6173 corresponding versions prefixed with @code{__builtin_}.
6175 GCC provides built-in versions of the ISO C99 floating point comparison
6176 macros that avoid raising exceptions for unordered operands. They have
6177 the same names as the standard macros ( @code{isgreater},
6178 @code{isgreaterequal}, @code{isless}, @code{islessequal},
6179 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
6180 prefixed. We intend for a library implementor to be able to simply
6181 @code{#define} each standard macro to its built-in equivalent.
6182 In the same fashion, GCC provides @code{isfinite} and @code{isnormal}
6183 built-ins used with @code{__builtin_} prefixed.
6185 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
6187 You can use the built-in function @code{__builtin_types_compatible_p} to
6188 determine whether two types are the same.
6190 This built-in function returns 1 if the unqualified versions of the
6191 types @var{type1} and @var{type2} (which are types, not expressions) are
6192 compatible, 0 otherwise. The result of this built-in function can be
6193 used in integer constant expressions.
6195 This built-in function ignores top level qualifiers (e.g., @code{const},
6196 @code{volatile}). For example, @code{int} is equivalent to @code{const
6199 The type @code{int[]} and @code{int[5]} are compatible. On the other
6200 hand, @code{int} and @code{char *} are not compatible, even if the size
6201 of their types, on the particular architecture are the same. Also, the
6202 amount of pointer indirection is taken into account when determining
6203 similarity. Consequently, @code{short *} is not similar to
6204 @code{short **}. Furthermore, two types that are typedefed are
6205 considered compatible if their underlying types are compatible.
6207 An @code{enum} type is not considered to be compatible with another
6208 @code{enum} type even if both are compatible with the same integer
6209 type; this is what the C standard specifies.
6210 For example, @code{enum @{foo, bar@}} is not similar to
6211 @code{enum @{hot, dog@}}.
6213 You would typically use this function in code whose execution varies
6214 depending on the arguments' types. For example:
6219 typeof (x) tmp = (x); \
6220 if (__builtin_types_compatible_p (typeof (x), long double)) \
6221 tmp = foo_long_double (tmp); \
6222 else if (__builtin_types_compatible_p (typeof (x), double)) \
6223 tmp = foo_double (tmp); \
6224 else if (__builtin_types_compatible_p (typeof (x), float)) \
6225 tmp = foo_float (tmp); \
6232 @emph{Note:} This construct is only available for C@.
6236 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
6238 You can use the built-in function @code{__builtin_choose_expr} to
6239 evaluate code depending on the value of a constant expression. This
6240 built-in function returns @var{exp1} if @var{const_exp}, which is a
6241 constant expression that must be able to be determined at compile time,
6242 is nonzero. Otherwise it returns 0.
6244 This built-in function is analogous to the @samp{? :} operator in C,
6245 except that the expression returned has its type unaltered by promotion
6246 rules. Also, the built-in function does not evaluate the expression
6247 that was not chosen. For example, if @var{const_exp} evaluates to true,
6248 @var{exp2} is not evaluated even if it has side-effects.
6250 This built-in function can return an lvalue if the chosen argument is an
6253 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
6254 type. Similarly, if @var{exp2} is returned, its return type is the same
6261 __builtin_choose_expr ( \
6262 __builtin_types_compatible_p (typeof (x), double), \
6264 __builtin_choose_expr ( \
6265 __builtin_types_compatible_p (typeof (x), float), \
6267 /* @r{The void expression results in a compile-time error} \
6268 @r{when assigning the result to something.} */ \
6272 @emph{Note:} This construct is only available for C@. Furthermore, the
6273 unused expression (@var{exp1} or @var{exp2} depending on the value of
6274 @var{const_exp}) may still generate syntax errors. This may change in
6279 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6280 You can use the built-in function @code{__builtin_constant_p} to
6281 determine if a value is known to be constant at compile-time and hence
6282 that GCC can perform constant-folding on expressions involving that
6283 value. The argument of the function is the value to test. The function
6284 returns the integer 1 if the argument is known to be a compile-time
6285 constant and 0 if it is not known to be a compile-time constant. A
6286 return of 0 does not indicate that the value is @emph{not} a constant,
6287 but merely that GCC cannot prove it is a constant with the specified
6288 value of the @option{-O} option.
6290 You would typically use this function in an embedded application where
6291 memory was a critical resource. If you have some complex calculation,
6292 you may want it to be folded if it involves constants, but need to call
6293 a function if it does not. For example:
6296 #define Scale_Value(X) \
6297 (__builtin_constant_p (X) \
6298 ? ((X) * SCALE + OFFSET) : Scale (X))
6301 You may use this built-in function in either a macro or an inline
6302 function. However, if you use it in an inlined function and pass an
6303 argument of the function as the argument to the built-in, GCC will
6304 never return 1 when you call the inline function with a string constant
6305 or compound literal (@pxref{Compound Literals}) and will not return 1
6306 when you pass a constant numeric value to the inline function unless you
6307 specify the @option{-O} option.
6309 You may also use @code{__builtin_constant_p} in initializers for static
6310 data. For instance, you can write
6313 static const int table[] = @{
6314 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6320 This is an acceptable initializer even if @var{EXPRESSION} is not a
6321 constant expression. GCC must be more conservative about evaluating the
6322 built-in in this case, because it has no opportunity to perform
6325 Previous versions of GCC did not accept this built-in in data
6326 initializers. The earliest version where it is completely safe is
6330 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6331 @opindex fprofile-arcs
6332 You may use @code{__builtin_expect} to provide the compiler with
6333 branch prediction information. In general, you should prefer to
6334 use actual profile feedback for this (@option{-fprofile-arcs}), as
6335 programmers are notoriously bad at predicting how their programs
6336 actually perform. However, there are applications in which this
6337 data is hard to collect.
6339 The return value is the value of @var{exp}, which should be an integral
6340 expression. The semantics of the built-in are that it is expected that
6341 @var{exp} == @var{c}. For example:
6344 if (__builtin_expect (x, 0))
6349 would indicate that we do not expect to call @code{foo}, since
6350 we expect @code{x} to be zero. Since you are limited to integral
6351 expressions for @var{exp}, you should use constructions such as
6354 if (__builtin_expect (ptr != NULL, 1))
6359 when testing pointer or floating-point values.
6362 @deftypefn {Built-in Function} void __builtin___clear_cache (char *@var{begin}, char *@var{end})
6363 This function is used to flush the processor's instruction cache for
6364 the region of memory between @var{begin} inclusive and @var{end}
6365 exclusive. Some targets require that the instruction cache be
6366 flushed, after modifying memory containing code, in order to obtain
6367 deterministic behavior.
6369 If the target does not require instruction cache flushes,
6370 @code{__builtin___clear_cache} has no effect. Otherwise either
6371 instructions are emitted in-line to clear the instruction cache or a
6372 call to the @code{__clear_cache} function in libgcc is made.
6375 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6376 This function is used to minimize cache-miss latency by moving data into
6377 a cache before it is accessed.
6378 You can insert calls to @code{__builtin_prefetch} into code for which
6379 you know addresses of data in memory that is likely to be accessed soon.
6380 If the target supports them, data prefetch instructions will be generated.
6381 If the prefetch is done early enough before the access then the data will
6382 be in the cache by the time it is accessed.
6384 The value of @var{addr} is the address of the memory to prefetch.
6385 There are two optional arguments, @var{rw} and @var{locality}.
6386 The value of @var{rw} is a compile-time constant one or zero; one
6387 means that the prefetch is preparing for a write to the memory address
6388 and zero, the default, means that the prefetch is preparing for a read.
6389 The value @var{locality} must be a compile-time constant integer between
6390 zero and three. A value of zero means that the data has no temporal
6391 locality, so it need not be left in the cache after the access. A value
6392 of three means that the data has a high degree of temporal locality and
6393 should be left in all levels of cache possible. Values of one and two
6394 mean, respectively, a low or moderate degree of temporal locality. The
6398 for (i = 0; i < n; i++)
6401 __builtin_prefetch (&a[i+j], 1, 1);
6402 __builtin_prefetch (&b[i+j], 0, 1);
6407 Data prefetch does not generate faults if @var{addr} is invalid, but
6408 the address expression itself must be valid. For example, a prefetch
6409 of @code{p->next} will not fault if @code{p->next} is not a valid
6410 address, but evaluation will fault if @code{p} is not a valid address.
6412 If the target does not support data prefetch, the address expression
6413 is evaluated if it includes side effects but no other code is generated
6414 and GCC does not issue a warning.
6417 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6418 Returns a positive infinity, if supported by the floating-point format,
6419 else @code{DBL_MAX}. This function is suitable for implementing the
6420 ISO C macro @code{HUGE_VAL}.
6423 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6424 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6427 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6428 Similar to @code{__builtin_huge_val}, except the return
6429 type is @code{long double}.
6432 @deftypefn {Built-in Function} double __builtin_inf (void)
6433 Similar to @code{__builtin_huge_val}, except a warning is generated
6434 if the target floating-point format does not support infinities.
6437 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6438 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6441 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6442 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6445 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6446 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6449 @deftypefn {Built-in Function} float __builtin_inff (void)
6450 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6451 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6454 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6455 Similar to @code{__builtin_inf}, except the return
6456 type is @code{long double}.
6459 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6460 This is an implementation of the ISO C99 function @code{nan}.
6462 Since ISO C99 defines this function in terms of @code{strtod}, which we
6463 do not implement, a description of the parsing is in order. The string
6464 is parsed as by @code{strtol}; that is, the base is recognized by
6465 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6466 in the significand such that the least significant bit of the number
6467 is at the least significant bit of the significand. The number is
6468 truncated to fit the significand field provided. The significand is
6469 forced to be a quiet NaN@.
6471 This function, if given a string literal all of which would have been
6472 consumed by strtol, is evaluated early enough that it is considered a
6473 compile-time constant.
6476 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6477 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6480 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6481 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6484 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6485 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6488 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6489 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6492 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6493 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6496 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6497 Similar to @code{__builtin_nan}, except the significand is forced
6498 to be a signaling NaN@. The @code{nans} function is proposed by
6499 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6502 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6503 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6506 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6507 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6510 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6511 Returns one plus the index of the least significant 1-bit of @var{x}, or
6512 if @var{x} is zero, returns zero.
6515 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6516 Returns the number of leading 0-bits in @var{x}, starting at the most
6517 significant bit position. If @var{x} is 0, the result is undefined.
6520 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6521 Returns the number of trailing 0-bits in @var{x}, starting at the least
6522 significant bit position. If @var{x} is 0, the result is undefined.
6525 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6526 Returns the number of 1-bits in @var{x}.
6529 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6530 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6534 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6535 Similar to @code{__builtin_ffs}, except the argument type is
6536 @code{unsigned long}.
6539 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6540 Similar to @code{__builtin_clz}, except the argument type is
6541 @code{unsigned long}.
6544 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6545 Similar to @code{__builtin_ctz}, except the argument type is
6546 @code{unsigned long}.
6549 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6550 Similar to @code{__builtin_popcount}, except the argument type is
6551 @code{unsigned long}.
6554 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6555 Similar to @code{__builtin_parity}, except the argument type is
6556 @code{unsigned long}.
6559 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6560 Similar to @code{__builtin_ffs}, except the argument type is
6561 @code{unsigned long long}.
6564 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6565 Similar to @code{__builtin_clz}, except the argument type is
6566 @code{unsigned long long}.
6569 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6570 Similar to @code{__builtin_ctz}, except the argument type is
6571 @code{unsigned long long}.
6574 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6575 Similar to @code{__builtin_popcount}, except the argument type is
6576 @code{unsigned long long}.
6579 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6580 Similar to @code{__builtin_parity}, except the argument type is
6581 @code{unsigned long long}.
6584 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6585 Returns the first argument raised to the power of the second. Unlike the
6586 @code{pow} function no guarantees about precision and rounding are made.
6589 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6590 Similar to @code{__builtin_powi}, except the argument and return types
6594 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6595 Similar to @code{__builtin_powi}, except the argument and return types
6596 are @code{long double}.
6599 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6600 Returns @var{x} with the order of the bytes reversed; for example,
6601 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6605 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6606 Similar to @code{__builtin_bswap32}, except the argument and return types
6610 @node Target Builtins
6611 @section Built-in Functions Specific to Particular Target Machines
6613 On some target machines, GCC supports many built-in functions specific
6614 to those machines. Generally these generate calls to specific machine
6615 instructions, but allow the compiler to schedule those calls.
6618 * Alpha Built-in Functions::
6619 * ARM iWMMXt Built-in Functions::
6620 * ARM NEON Intrinsics::
6621 * Blackfin Built-in Functions::
6622 * FR-V Built-in Functions::
6623 * X86 Built-in Functions::
6624 * MIPS DSP Built-in Functions::
6625 * MIPS Paired-Single Support::
6626 * PowerPC AltiVec Built-in Functions::
6627 * SPARC VIS Built-in Functions::
6628 * SPU Built-in Functions::
6631 @node Alpha Built-in Functions
6632 @subsection Alpha Built-in Functions
6634 These built-in functions are available for the Alpha family of
6635 processors, depending on the command-line switches used.
6637 The following built-in functions are always available. They
6638 all generate the machine instruction that is part of the name.
6641 long __builtin_alpha_implver (void)
6642 long __builtin_alpha_rpcc (void)
6643 long __builtin_alpha_amask (long)
6644 long __builtin_alpha_cmpbge (long, long)
6645 long __builtin_alpha_extbl (long, long)
6646 long __builtin_alpha_extwl (long, long)
6647 long __builtin_alpha_extll (long, long)
6648 long __builtin_alpha_extql (long, long)
6649 long __builtin_alpha_extwh (long, long)
6650 long __builtin_alpha_extlh (long, long)
6651 long __builtin_alpha_extqh (long, long)
6652 long __builtin_alpha_insbl (long, long)
6653 long __builtin_alpha_inswl (long, long)
6654 long __builtin_alpha_insll (long, long)
6655 long __builtin_alpha_insql (long, long)
6656 long __builtin_alpha_inswh (long, long)
6657 long __builtin_alpha_inslh (long, long)
6658 long __builtin_alpha_insqh (long, long)
6659 long __builtin_alpha_mskbl (long, long)
6660 long __builtin_alpha_mskwl (long, long)
6661 long __builtin_alpha_mskll (long, long)
6662 long __builtin_alpha_mskql (long, long)
6663 long __builtin_alpha_mskwh (long, long)
6664 long __builtin_alpha_msklh (long, long)
6665 long __builtin_alpha_mskqh (long, long)
6666 long __builtin_alpha_umulh (long, long)
6667 long __builtin_alpha_zap (long, long)
6668 long __builtin_alpha_zapnot (long, long)
6671 The following built-in functions are always with @option{-mmax}
6672 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6673 later. They all generate the machine instruction that is part
6677 long __builtin_alpha_pklb (long)
6678 long __builtin_alpha_pkwb (long)
6679 long __builtin_alpha_unpkbl (long)
6680 long __builtin_alpha_unpkbw (long)
6681 long __builtin_alpha_minub8 (long, long)
6682 long __builtin_alpha_minsb8 (long, long)
6683 long __builtin_alpha_minuw4 (long, long)
6684 long __builtin_alpha_minsw4 (long, long)
6685 long __builtin_alpha_maxub8 (long, long)
6686 long __builtin_alpha_maxsb8 (long, long)
6687 long __builtin_alpha_maxuw4 (long, long)
6688 long __builtin_alpha_maxsw4 (long, long)
6689 long __builtin_alpha_perr (long, long)
6692 The following built-in functions are always with @option{-mcix}
6693 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6694 later. They all generate the machine instruction that is part
6698 long __builtin_alpha_cttz (long)
6699 long __builtin_alpha_ctlz (long)
6700 long __builtin_alpha_ctpop (long)
6703 The following builtins are available on systems that use the OSF/1
6704 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6705 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6706 @code{rdval} and @code{wrval}.
6709 void *__builtin_thread_pointer (void)
6710 void __builtin_set_thread_pointer (void *)
6713 @node ARM iWMMXt Built-in Functions
6714 @subsection ARM iWMMXt Built-in Functions
6716 These built-in functions are available for the ARM family of
6717 processors when the @option{-mcpu=iwmmxt} switch is used:
6720 typedef int v2si __attribute__ ((vector_size (8)));
6721 typedef short v4hi __attribute__ ((vector_size (8)));
6722 typedef char v8qi __attribute__ ((vector_size (8)));
6724 int __builtin_arm_getwcx (int)
6725 void __builtin_arm_setwcx (int, int)
6726 int __builtin_arm_textrmsb (v8qi, int)
6727 int __builtin_arm_textrmsh (v4hi, int)
6728 int __builtin_arm_textrmsw (v2si, int)
6729 int __builtin_arm_textrmub (v8qi, int)
6730 int __builtin_arm_textrmuh (v4hi, int)
6731 int __builtin_arm_textrmuw (v2si, int)
6732 v8qi __builtin_arm_tinsrb (v8qi, int)
6733 v4hi __builtin_arm_tinsrh (v4hi, int)
6734 v2si __builtin_arm_tinsrw (v2si, int)
6735 long long __builtin_arm_tmia (long long, int, int)
6736 long long __builtin_arm_tmiabb (long long, int, int)
6737 long long __builtin_arm_tmiabt (long long, int, int)
6738 long long __builtin_arm_tmiaph (long long, int, int)
6739 long long __builtin_arm_tmiatb (long long, int, int)
6740 long long __builtin_arm_tmiatt (long long, int, int)
6741 int __builtin_arm_tmovmskb (v8qi)
6742 int __builtin_arm_tmovmskh (v4hi)
6743 int __builtin_arm_tmovmskw (v2si)
6744 long long __builtin_arm_waccb (v8qi)
6745 long long __builtin_arm_wacch (v4hi)
6746 long long __builtin_arm_waccw (v2si)
6747 v8qi __builtin_arm_waddb (v8qi, v8qi)
6748 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6749 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6750 v4hi __builtin_arm_waddh (v4hi, v4hi)
6751 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6752 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6753 v2si __builtin_arm_waddw (v2si, v2si)
6754 v2si __builtin_arm_waddwss (v2si, v2si)
6755 v2si __builtin_arm_waddwus (v2si, v2si)
6756 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6757 long long __builtin_arm_wand(long long, long long)
6758 long long __builtin_arm_wandn (long long, long long)
6759 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6760 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6761 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6762 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6763 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6764 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6765 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6766 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6767 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6768 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6769 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6770 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6771 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6772 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6773 long long __builtin_arm_wmacsz (v4hi, v4hi)
6774 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6775 long long __builtin_arm_wmacuz (v4hi, v4hi)
6776 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6777 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6778 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6779 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6780 v2si __builtin_arm_wmaxsw (v2si, v2si)
6781 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6782 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6783 v2si __builtin_arm_wmaxuw (v2si, v2si)
6784 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6785 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6786 v2si __builtin_arm_wminsw (v2si, v2si)
6787 v8qi __builtin_arm_wminub (v8qi, v8qi)
6788 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6789 v2si __builtin_arm_wminuw (v2si, v2si)
6790 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6791 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6792 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6793 long long __builtin_arm_wor (long long, long long)
6794 v2si __builtin_arm_wpackdss (long long, long long)
6795 v2si __builtin_arm_wpackdus (long long, long long)
6796 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6797 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6798 v4hi __builtin_arm_wpackwss (v2si, v2si)
6799 v4hi __builtin_arm_wpackwus (v2si, v2si)
6800 long long __builtin_arm_wrord (long long, long long)
6801 long long __builtin_arm_wrordi (long long, int)
6802 v4hi __builtin_arm_wrorh (v4hi, long long)
6803 v4hi __builtin_arm_wrorhi (v4hi, int)
6804 v2si __builtin_arm_wrorw (v2si, long long)
6805 v2si __builtin_arm_wrorwi (v2si, int)
6806 v2si __builtin_arm_wsadb (v8qi, v8qi)
6807 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6808 v2si __builtin_arm_wsadh (v4hi, v4hi)
6809 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6810 v4hi __builtin_arm_wshufh (v4hi, int)
6811 long long __builtin_arm_wslld (long long, long long)
6812 long long __builtin_arm_wslldi (long long, int)
6813 v4hi __builtin_arm_wsllh (v4hi, long long)
6814 v4hi __builtin_arm_wsllhi (v4hi, int)
6815 v2si __builtin_arm_wsllw (v2si, long long)
6816 v2si __builtin_arm_wsllwi (v2si, int)
6817 long long __builtin_arm_wsrad (long long, long long)
6818 long long __builtin_arm_wsradi (long long, int)
6819 v4hi __builtin_arm_wsrah (v4hi, long long)
6820 v4hi __builtin_arm_wsrahi (v4hi, int)
6821 v2si __builtin_arm_wsraw (v2si, long long)
6822 v2si __builtin_arm_wsrawi (v2si, int)
6823 long long __builtin_arm_wsrld (long long, long long)
6824 long long __builtin_arm_wsrldi (long long, int)
6825 v4hi __builtin_arm_wsrlh (v4hi, long long)
6826 v4hi __builtin_arm_wsrlhi (v4hi, int)
6827 v2si __builtin_arm_wsrlw (v2si, long long)
6828 v2si __builtin_arm_wsrlwi (v2si, int)
6829 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6830 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6831 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6832 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6833 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6834 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6835 v2si __builtin_arm_wsubw (v2si, v2si)
6836 v2si __builtin_arm_wsubwss (v2si, v2si)
6837 v2si __builtin_arm_wsubwus (v2si, v2si)
6838 v4hi __builtin_arm_wunpckehsb (v8qi)
6839 v2si __builtin_arm_wunpckehsh (v4hi)
6840 long long __builtin_arm_wunpckehsw (v2si)
6841 v4hi __builtin_arm_wunpckehub (v8qi)
6842 v2si __builtin_arm_wunpckehuh (v4hi)
6843 long long __builtin_arm_wunpckehuw (v2si)
6844 v4hi __builtin_arm_wunpckelsb (v8qi)
6845 v2si __builtin_arm_wunpckelsh (v4hi)
6846 long long __builtin_arm_wunpckelsw (v2si)
6847 v4hi __builtin_arm_wunpckelub (v8qi)
6848 v2si __builtin_arm_wunpckeluh (v4hi)
6849 long long __builtin_arm_wunpckeluw (v2si)
6850 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6851 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6852 v2si __builtin_arm_wunpckihw (v2si, v2si)
6853 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6854 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6855 v2si __builtin_arm_wunpckilw (v2si, v2si)
6856 long long __builtin_arm_wxor (long long, long long)
6857 long long __builtin_arm_wzero ()
6860 @node ARM NEON Intrinsics
6861 @subsection ARM NEON Intrinsics
6863 These built-in intrinsics for the ARM Advanced SIMD extension are available
6864 when the @option{-mfpu=neon} switch is used:
6866 @include arm-neon-intrinsics.texi
6868 @node Blackfin Built-in Functions
6869 @subsection Blackfin Built-in Functions
6871 Currently, there are two Blackfin-specific built-in functions. These are
6872 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6873 using inline assembly; by using these built-in functions the compiler can
6874 automatically add workarounds for hardware errata involving these
6875 instructions. These functions are named as follows:
6878 void __builtin_bfin_csync (void)
6879 void __builtin_bfin_ssync (void)
6882 @node FR-V Built-in Functions
6883 @subsection FR-V Built-in Functions
6885 GCC provides many FR-V-specific built-in functions. In general,
6886 these functions are intended to be compatible with those described
6887 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6888 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6889 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6890 pointer rather than by value.
6892 Most of the functions are named after specific FR-V instructions.
6893 Such functions are said to be ``directly mapped'' and are summarized
6894 here in tabular form.
6898 * Directly-mapped Integer Functions::
6899 * Directly-mapped Media Functions::
6900 * Raw read/write Functions::
6901 * Other Built-in Functions::
6904 @node Argument Types
6905 @subsubsection Argument Types
6907 The arguments to the built-in functions can be divided into three groups:
6908 register numbers, compile-time constants and run-time values. In order
6909 to make this classification clear at a glance, the arguments and return
6910 values are given the following pseudo types:
6912 @multitable @columnfractions .20 .30 .15 .35
6913 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6914 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6915 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6916 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6917 @item @code{uw2} @tab @code{unsigned long long} @tab No
6918 @tab an unsigned doubleword
6919 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6920 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6921 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6922 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6925 These pseudo types are not defined by GCC, they are simply a notational
6926 convenience used in this manual.
6928 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6929 and @code{sw2} are evaluated at run time. They correspond to
6930 register operands in the underlying FR-V instructions.
6932 @code{const} arguments represent immediate operands in the underlying
6933 FR-V instructions. They must be compile-time constants.
6935 @code{acc} arguments are evaluated at compile time and specify the number
6936 of an accumulator register. For example, an @code{acc} argument of 2
6937 will select the ACC2 register.
6939 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6940 number of an IACC register. See @pxref{Other Built-in Functions}
6943 @node Directly-mapped Integer Functions
6944 @subsubsection Directly-mapped Integer Functions
6946 The functions listed below map directly to FR-V I-type instructions.
6948 @multitable @columnfractions .45 .32 .23
6949 @item Function prototype @tab Example usage @tab Assembly output
6950 @item @code{sw1 __ADDSS (sw1, sw1)}
6951 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6952 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6953 @item @code{sw1 __SCAN (sw1, sw1)}
6954 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6955 @tab @code{SCAN @var{a},@var{b},@var{c}}
6956 @item @code{sw1 __SCUTSS (sw1)}
6957 @tab @code{@var{b} = __SCUTSS (@var{a})}
6958 @tab @code{SCUTSS @var{a},@var{b}}
6959 @item @code{sw1 __SLASS (sw1, sw1)}
6960 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6961 @tab @code{SLASS @var{a},@var{b},@var{c}}
6962 @item @code{void __SMASS (sw1, sw1)}
6963 @tab @code{__SMASS (@var{a}, @var{b})}
6964 @tab @code{SMASS @var{a},@var{b}}
6965 @item @code{void __SMSSS (sw1, sw1)}
6966 @tab @code{__SMSSS (@var{a}, @var{b})}
6967 @tab @code{SMSSS @var{a},@var{b}}
6968 @item @code{void __SMU (sw1, sw1)}
6969 @tab @code{__SMU (@var{a}, @var{b})}
6970 @tab @code{SMU @var{a},@var{b}}
6971 @item @code{sw2 __SMUL (sw1, sw1)}
6972 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6973 @tab @code{SMUL @var{a},@var{b},@var{c}}
6974 @item @code{sw1 __SUBSS (sw1, sw1)}
6975 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6976 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6977 @item @code{uw2 __UMUL (uw1, uw1)}
6978 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6979 @tab @code{UMUL @var{a},@var{b},@var{c}}
6982 @node Directly-mapped Media Functions
6983 @subsubsection Directly-mapped Media Functions
6985 The functions listed below map directly to FR-V M-type instructions.
6987 @multitable @columnfractions .45 .32 .23
6988 @item Function prototype @tab Example usage @tab Assembly output
6989 @item @code{uw1 __MABSHS (sw1)}
6990 @tab @code{@var{b} = __MABSHS (@var{a})}
6991 @tab @code{MABSHS @var{a},@var{b}}
6992 @item @code{void __MADDACCS (acc, acc)}
6993 @tab @code{__MADDACCS (@var{b}, @var{a})}
6994 @tab @code{MADDACCS @var{a},@var{b}}
6995 @item @code{sw1 __MADDHSS (sw1, sw1)}
6996 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6997 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6998 @item @code{uw1 __MADDHUS (uw1, uw1)}
6999 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
7000 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
7001 @item @code{uw1 __MAND (uw1, uw1)}
7002 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
7003 @tab @code{MAND @var{a},@var{b},@var{c}}
7004 @item @code{void __MASACCS (acc, acc)}
7005 @tab @code{__MASACCS (@var{b}, @var{a})}
7006 @tab @code{MASACCS @var{a},@var{b}}
7007 @item @code{uw1 __MAVEH (uw1, uw1)}
7008 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
7009 @tab @code{MAVEH @var{a},@var{b},@var{c}}
7010 @item @code{uw2 __MBTOH (uw1)}
7011 @tab @code{@var{b} = __MBTOH (@var{a})}
7012 @tab @code{MBTOH @var{a},@var{b}}
7013 @item @code{void __MBTOHE (uw1 *, uw1)}
7014 @tab @code{__MBTOHE (&@var{b}, @var{a})}
7015 @tab @code{MBTOHE @var{a},@var{b}}
7016 @item @code{void __MCLRACC (acc)}
7017 @tab @code{__MCLRACC (@var{a})}
7018 @tab @code{MCLRACC @var{a}}
7019 @item @code{void __MCLRACCA (void)}
7020 @tab @code{__MCLRACCA ()}
7021 @tab @code{MCLRACCA}
7022 @item @code{uw1 __Mcop1 (uw1, uw1)}
7023 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
7024 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
7025 @item @code{uw1 __Mcop2 (uw1, uw1)}
7026 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
7027 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
7028 @item @code{uw1 __MCPLHI (uw2, const)}
7029 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
7030 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
7031 @item @code{uw1 __MCPLI (uw2, const)}
7032 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
7033 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
7034 @item @code{void __MCPXIS (acc, sw1, sw1)}
7035 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
7036 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
7037 @item @code{void __MCPXIU (acc, uw1, uw1)}
7038 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
7039 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
7040 @item @code{void __MCPXRS (acc, sw1, sw1)}
7041 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
7042 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
7043 @item @code{void __MCPXRU (acc, uw1, uw1)}
7044 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
7045 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
7046 @item @code{uw1 __MCUT (acc, uw1)}
7047 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
7048 @tab @code{MCUT @var{a},@var{b},@var{c}}
7049 @item @code{uw1 __MCUTSS (acc, sw1)}
7050 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
7051 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
7052 @item @code{void __MDADDACCS (acc, acc)}
7053 @tab @code{__MDADDACCS (@var{b}, @var{a})}
7054 @tab @code{MDADDACCS @var{a},@var{b}}
7055 @item @code{void __MDASACCS (acc, acc)}
7056 @tab @code{__MDASACCS (@var{b}, @var{a})}
7057 @tab @code{MDASACCS @var{a},@var{b}}
7058 @item @code{uw2 __MDCUTSSI (acc, const)}
7059 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
7060 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
7061 @item @code{uw2 __MDPACKH (uw2, uw2)}
7062 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
7063 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
7064 @item @code{uw2 __MDROTLI (uw2, const)}
7065 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
7066 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
7067 @item @code{void __MDSUBACCS (acc, acc)}
7068 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
7069 @tab @code{MDSUBACCS @var{a},@var{b}}
7070 @item @code{void __MDUNPACKH (uw1 *, uw2)}
7071 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
7072 @tab @code{MDUNPACKH @var{a},@var{b}}
7073 @item @code{uw2 __MEXPDHD (uw1, const)}
7074 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
7075 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
7076 @item @code{uw1 __MEXPDHW (uw1, const)}
7077 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
7078 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
7079 @item @code{uw1 __MHDSETH (uw1, const)}
7080 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
7081 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
7082 @item @code{sw1 __MHDSETS (const)}
7083 @tab @code{@var{b} = __MHDSETS (@var{a})}
7084 @tab @code{MHDSETS #@var{a},@var{b}}
7085 @item @code{uw1 __MHSETHIH (uw1, const)}
7086 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
7087 @tab @code{MHSETHIH #@var{a},@var{b}}
7088 @item @code{sw1 __MHSETHIS (sw1, const)}
7089 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
7090 @tab @code{MHSETHIS #@var{a},@var{b}}
7091 @item @code{uw1 __MHSETLOH (uw1, const)}
7092 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
7093 @tab @code{MHSETLOH #@var{a},@var{b}}
7094 @item @code{sw1 __MHSETLOS (sw1, const)}
7095 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
7096 @tab @code{MHSETLOS #@var{a},@var{b}}
7097 @item @code{uw1 __MHTOB (uw2)}
7098 @tab @code{@var{b} = __MHTOB (@var{a})}
7099 @tab @code{MHTOB @var{a},@var{b}}
7100 @item @code{void __MMACHS (acc, sw1, sw1)}
7101 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
7102 @tab @code{MMACHS @var{a},@var{b},@var{c}}
7103 @item @code{void __MMACHU (acc, uw1, uw1)}
7104 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
7105 @tab @code{MMACHU @var{a},@var{b},@var{c}}
7106 @item @code{void __MMRDHS (acc, sw1, sw1)}
7107 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
7108 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
7109 @item @code{void __MMRDHU (acc, uw1, uw1)}
7110 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
7111 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
7112 @item @code{void __MMULHS (acc, sw1, sw1)}
7113 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
7114 @tab @code{MMULHS @var{a},@var{b},@var{c}}
7115 @item @code{void __MMULHU (acc, uw1, uw1)}
7116 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
7117 @tab @code{MMULHU @var{a},@var{b},@var{c}}
7118 @item @code{void __MMULXHS (acc, sw1, sw1)}
7119 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
7120 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
7121 @item @code{void __MMULXHU (acc, uw1, uw1)}
7122 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
7123 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
7124 @item @code{uw1 __MNOT (uw1)}
7125 @tab @code{@var{b} = __MNOT (@var{a})}
7126 @tab @code{MNOT @var{a},@var{b}}
7127 @item @code{uw1 __MOR (uw1, uw1)}
7128 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
7129 @tab @code{MOR @var{a},@var{b},@var{c}}
7130 @item @code{uw1 __MPACKH (uh, uh)}
7131 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
7132 @tab @code{MPACKH @var{a},@var{b},@var{c}}
7133 @item @code{sw2 __MQADDHSS (sw2, sw2)}
7134 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
7135 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
7136 @item @code{uw2 __MQADDHUS (uw2, uw2)}
7137 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
7138 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
7139 @item @code{void __MQCPXIS (acc, sw2, sw2)}
7140 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
7141 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
7142 @item @code{void __MQCPXIU (acc, uw2, uw2)}
7143 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
7144 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
7145 @item @code{void __MQCPXRS (acc, sw2, sw2)}
7146 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
7147 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
7148 @item @code{void __MQCPXRU (acc, uw2, uw2)}
7149 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
7150 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
7151 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
7152 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
7153 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
7154 @item @code{sw2 __MQLMTHS (sw2, sw2)}
7155 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
7156 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
7157 @item @code{void __MQMACHS (acc, sw2, sw2)}
7158 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
7159 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
7160 @item @code{void __MQMACHU (acc, uw2, uw2)}
7161 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
7162 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
7163 @item @code{void __MQMACXHS (acc, sw2, sw2)}
7164 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
7165 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
7166 @item @code{void __MQMULHS (acc, sw2, sw2)}
7167 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
7168 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
7169 @item @code{void __MQMULHU (acc, uw2, uw2)}
7170 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
7171 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
7172 @item @code{void __MQMULXHS (acc, sw2, sw2)}
7173 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
7174 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
7175 @item @code{void __MQMULXHU (acc, uw2, uw2)}
7176 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
7177 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
7178 @item @code{sw2 __MQSATHS (sw2, sw2)}
7179 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
7180 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
7181 @item @code{uw2 __MQSLLHI (uw2, int)}
7182 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
7183 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
7184 @item @code{sw2 __MQSRAHI (sw2, int)}
7185 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
7186 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
7187 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
7188 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
7189 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
7190 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
7191 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
7192 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
7193 @item @code{void __MQXMACHS (acc, sw2, sw2)}
7194 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
7195 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
7196 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
7197 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
7198 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
7199 @item @code{uw1 __MRDACC (acc)}
7200 @tab @code{@var{b} = __MRDACC (@var{a})}
7201 @tab @code{MRDACC @var{a},@var{b}}
7202 @item @code{uw1 __MRDACCG (acc)}
7203 @tab @code{@var{b} = __MRDACCG (@var{a})}
7204 @tab @code{MRDACCG @var{a},@var{b}}
7205 @item @code{uw1 __MROTLI (uw1, const)}
7206 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
7207 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
7208 @item @code{uw1 __MROTRI (uw1, const)}
7209 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
7210 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
7211 @item @code{sw1 __MSATHS (sw1, sw1)}
7212 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
7213 @tab @code{MSATHS @var{a},@var{b},@var{c}}
7214 @item @code{uw1 __MSATHU (uw1, uw1)}
7215 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
7216 @tab @code{MSATHU @var{a},@var{b},@var{c}}
7217 @item @code{uw1 __MSLLHI (uw1, const)}
7218 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
7219 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
7220 @item @code{sw1 __MSRAHI (sw1, const)}
7221 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
7222 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
7223 @item @code{uw1 __MSRLHI (uw1, const)}
7224 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
7225 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
7226 @item @code{void __MSUBACCS (acc, acc)}
7227 @tab @code{__MSUBACCS (@var{b}, @var{a})}
7228 @tab @code{MSUBACCS @var{a},@var{b}}
7229 @item @code{sw1 __MSUBHSS (sw1, sw1)}
7230 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
7231 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
7232 @item @code{uw1 __MSUBHUS (uw1, uw1)}
7233 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
7234 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
7235 @item @code{void __MTRAP (void)}
7236 @tab @code{__MTRAP ()}
7238 @item @code{uw2 __MUNPACKH (uw1)}
7239 @tab @code{@var{b} = __MUNPACKH (@var{a})}
7240 @tab @code{MUNPACKH @var{a},@var{b}}
7241 @item @code{uw1 __MWCUT (uw2, uw1)}
7242 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
7243 @tab @code{MWCUT @var{a},@var{b},@var{c}}
7244 @item @code{void __MWTACC (acc, uw1)}
7245 @tab @code{__MWTACC (@var{b}, @var{a})}
7246 @tab @code{MWTACC @var{a},@var{b}}
7247 @item @code{void __MWTACCG (acc, uw1)}
7248 @tab @code{__MWTACCG (@var{b}, @var{a})}
7249 @tab @code{MWTACCG @var{a},@var{b}}
7250 @item @code{uw1 __MXOR (uw1, uw1)}
7251 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
7252 @tab @code{MXOR @var{a},@var{b},@var{c}}
7255 @node Raw read/write Functions
7256 @subsubsection Raw read/write Functions
7258 This sections describes built-in functions related to read and write
7259 instructions to access memory. These functions generate
7260 @code{membar} instructions to flush the I/O load and stores where
7261 appropriate, as described in Fujitsu's manual described above.
7265 @item unsigned char __builtin_read8 (void *@var{data})
7266 @item unsigned short __builtin_read16 (void *@var{data})
7267 @item unsigned long __builtin_read32 (void *@var{data})
7268 @item unsigned long long __builtin_read64 (void *@var{data})
7270 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
7271 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
7272 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
7273 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
7276 @node Other Built-in Functions
7277 @subsubsection Other Built-in Functions
7279 This section describes built-in functions that are not named after
7280 a specific FR-V instruction.
7283 @item sw2 __IACCreadll (iacc @var{reg})
7284 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
7285 for future expansion and must be 0.
7287 @item sw1 __IACCreadl (iacc @var{reg})
7288 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
7289 Other values of @var{reg} are rejected as invalid.
7291 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
7292 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
7293 is reserved for future expansion and must be 0.
7295 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7296 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7297 is 1. Other values of @var{reg} are rejected as invalid.
7299 @item void __data_prefetch0 (const void *@var{x})
7300 Use the @code{dcpl} instruction to load the contents of address @var{x}
7301 into the data cache.
7303 @item void __data_prefetch (const void *@var{x})
7304 Use the @code{nldub} instruction to load the contents of address @var{x}
7305 into the data cache. The instruction will be issued in slot I1@.
7308 @node X86 Built-in Functions
7309 @subsection X86 Built-in Functions
7311 These built-in functions are available for the i386 and x86-64 family
7312 of computers, depending on the command-line switches used.
7314 Note that, if you specify command-line switches such as @option{-msse},
7315 the compiler could use the extended instruction sets even if the built-ins
7316 are not used explicitly in the program. For this reason, applications
7317 which perform runtime CPU detection must compile separate files for each
7318 supported architecture, using the appropriate flags. In particular,
7319 the file containing the CPU detection code should be compiled without
7322 The following machine modes are available for use with MMX built-in functions
7323 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7324 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7325 vector of eight 8-bit integers. Some of the built-in functions operate on
7326 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
7328 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7329 of two 32-bit floating point values.
7331 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7332 floating point values. Some instructions use a vector of four 32-bit
7333 integers, these use @code{V4SI}. Finally, some instructions operate on an
7334 entire vector register, interpreting it as a 128-bit integer, these use mode
7337 In the 64-bit mode, x86-64 family of processors uses additional built-in
7338 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7339 floating point and @code{TC} 128-bit complex floating point values.
7341 The following floating point built-in functions are made available in the
7342 64-bit mode. All of them implement the function that is part of the name.
7345 __float128 __builtin_fabsq (__float128)
7346 __float128 __builtin_copysignq (__float128, __float128)
7349 The following floating point built-in functions are made available in the
7353 @item __float128 __builtin_infq (void)
7354 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7357 The following built-in functions are made available by @option{-mmmx}.
7358 All of them generate the machine instruction that is part of the name.
7361 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7362 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7363 v2si __builtin_ia32_paddd (v2si, v2si)
7364 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7365 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7366 v2si __builtin_ia32_psubd (v2si, v2si)
7367 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7368 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7369 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7370 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7371 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7372 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7373 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7374 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7375 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7376 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7377 di __builtin_ia32_pand (di, di)
7378 di __builtin_ia32_pandn (di,di)
7379 di __builtin_ia32_por (di, di)
7380 di __builtin_ia32_pxor (di, di)
7381 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7382 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7383 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7384 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7385 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7386 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7387 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7388 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7389 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7390 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7391 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7392 v2si __builtin_ia32_punpckldq (v2si, v2si)
7393 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7394 v4hi __builtin_ia32_packssdw (v2si, v2si)
7395 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7398 The following built-in functions are made available either with
7399 @option{-msse}, or with a combination of @option{-m3dnow} and
7400 @option{-march=athlon}. All of them generate the machine
7401 instruction that is part of the name.
7404 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7405 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7406 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7407 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7408 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7409 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7410 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7411 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7412 int __builtin_ia32_pextrw (v4hi, int)
7413 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7414 int __builtin_ia32_pmovmskb (v8qi)
7415 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7416 void __builtin_ia32_movntq (di *, di)
7417 void __builtin_ia32_sfence (void)
7420 The following built-in functions are available when @option{-msse} is used.
7421 All of them generate the machine instruction that is part of the name.
7424 int __builtin_ia32_comieq (v4sf, v4sf)
7425 int __builtin_ia32_comineq (v4sf, v4sf)
7426 int __builtin_ia32_comilt (v4sf, v4sf)
7427 int __builtin_ia32_comile (v4sf, v4sf)
7428 int __builtin_ia32_comigt (v4sf, v4sf)
7429 int __builtin_ia32_comige (v4sf, v4sf)
7430 int __builtin_ia32_ucomieq (v4sf, v4sf)
7431 int __builtin_ia32_ucomineq (v4sf, v4sf)
7432 int __builtin_ia32_ucomilt (v4sf, v4sf)
7433 int __builtin_ia32_ucomile (v4sf, v4sf)
7434 int __builtin_ia32_ucomigt (v4sf, v4sf)
7435 int __builtin_ia32_ucomige (v4sf, v4sf)
7436 v4sf __builtin_ia32_addps (v4sf, v4sf)
7437 v4sf __builtin_ia32_subps (v4sf, v4sf)
7438 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7439 v4sf __builtin_ia32_divps (v4sf, v4sf)
7440 v4sf __builtin_ia32_addss (v4sf, v4sf)
7441 v4sf __builtin_ia32_subss (v4sf, v4sf)
7442 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7443 v4sf __builtin_ia32_divss (v4sf, v4sf)
7444 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7445 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7446 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7447 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7448 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7449 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7450 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7451 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7452 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7453 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7454 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7455 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7456 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7457 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7458 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7459 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7460 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7461 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7462 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7463 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7464 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7465 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7466 v4sf __builtin_ia32_minps (v4sf, v4sf)
7467 v4sf __builtin_ia32_minss (v4sf, v4sf)
7468 v4sf __builtin_ia32_andps (v4sf, v4sf)
7469 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7470 v4sf __builtin_ia32_orps (v4sf, v4sf)
7471 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7472 v4sf __builtin_ia32_movss (v4sf, v4sf)
7473 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7474 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7475 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7476 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7477 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7478 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7479 v2si __builtin_ia32_cvtps2pi (v4sf)
7480 int __builtin_ia32_cvtss2si (v4sf)
7481 v2si __builtin_ia32_cvttps2pi (v4sf)
7482 int __builtin_ia32_cvttss2si (v4sf)
7483 v4sf __builtin_ia32_rcpps (v4sf)
7484 v4sf __builtin_ia32_rsqrtps (v4sf)
7485 v4sf __builtin_ia32_sqrtps (v4sf)
7486 v4sf __builtin_ia32_rcpss (v4sf)
7487 v4sf __builtin_ia32_rsqrtss (v4sf)
7488 v4sf __builtin_ia32_sqrtss (v4sf)
7489 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7490 void __builtin_ia32_movntps (float *, v4sf)
7491 int __builtin_ia32_movmskps (v4sf)
7494 The following built-in functions are available when @option{-msse} is used.
7497 @item v4sf __builtin_ia32_loadaps (float *)
7498 Generates the @code{movaps} machine instruction as a load from memory.
7499 @item void __builtin_ia32_storeaps (float *, v4sf)
7500 Generates the @code{movaps} machine instruction as a store to memory.
7501 @item v4sf __builtin_ia32_loadups (float *)
7502 Generates the @code{movups} machine instruction as a load from memory.
7503 @item void __builtin_ia32_storeups (float *, v4sf)
7504 Generates the @code{movups} machine instruction as a store to memory.
7505 @item v4sf __builtin_ia32_loadsss (float *)
7506 Generates the @code{movss} machine instruction as a load from memory.
7507 @item void __builtin_ia32_storess (float *, v4sf)
7508 Generates the @code{movss} machine instruction as a store to memory.
7509 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7510 Generates the @code{movhps} machine instruction as a load from memory.
7511 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7512 Generates the @code{movlps} machine instruction as a load from memory
7513 @item void __builtin_ia32_storehps (v4sf, v2si *)
7514 Generates the @code{movhps} machine instruction as a store to memory.
7515 @item void __builtin_ia32_storelps (v4sf, v2si *)
7516 Generates the @code{movlps} machine instruction as a store to memory.
7519 The following built-in functions are available when @option{-msse2} is used.
7520 All of them generate the machine instruction that is part of the name.
7523 int __builtin_ia32_comisdeq (v2df, v2df)
7524 int __builtin_ia32_comisdlt (v2df, v2df)
7525 int __builtin_ia32_comisdle (v2df, v2df)
7526 int __builtin_ia32_comisdgt (v2df, v2df)
7527 int __builtin_ia32_comisdge (v2df, v2df)
7528 int __builtin_ia32_comisdneq (v2df, v2df)
7529 int __builtin_ia32_ucomisdeq (v2df, v2df)
7530 int __builtin_ia32_ucomisdlt (v2df, v2df)
7531 int __builtin_ia32_ucomisdle (v2df, v2df)
7532 int __builtin_ia32_ucomisdgt (v2df, v2df)
7533 int __builtin_ia32_ucomisdge (v2df, v2df)
7534 int __builtin_ia32_ucomisdneq (v2df, v2df)
7535 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7536 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7537 v2df __builtin_ia32_cmplepd (v2df, v2df)
7538 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7539 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7540 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7541 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7542 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7543 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7544 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7545 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7546 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7547 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7548 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7549 v2df __builtin_ia32_cmplesd (v2df, v2df)
7550 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7551 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7552 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7553 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7554 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7555 v2di __builtin_ia32_paddq (v2di, v2di)
7556 v2di __builtin_ia32_psubq (v2di, v2di)
7557 v2df __builtin_ia32_addpd (v2df, v2df)
7558 v2df __builtin_ia32_subpd (v2df, v2df)
7559 v2df __builtin_ia32_mulpd (v2df, v2df)
7560 v2df __builtin_ia32_divpd (v2df, v2df)
7561 v2df __builtin_ia32_addsd (v2df, v2df)
7562 v2df __builtin_ia32_subsd (v2df, v2df)
7563 v2df __builtin_ia32_mulsd (v2df, v2df)
7564 v2df __builtin_ia32_divsd (v2df, v2df)
7565 v2df __builtin_ia32_minpd (v2df, v2df)
7566 v2df __builtin_ia32_maxpd (v2df, v2df)
7567 v2df __builtin_ia32_minsd (v2df, v2df)
7568 v2df __builtin_ia32_maxsd (v2df, v2df)
7569 v2df __builtin_ia32_andpd (v2df, v2df)
7570 v2df __builtin_ia32_andnpd (v2df, v2df)
7571 v2df __builtin_ia32_orpd (v2df, v2df)
7572 v2df __builtin_ia32_xorpd (v2df, v2df)
7573 v2df __builtin_ia32_movsd (v2df, v2df)
7574 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7575 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7576 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7577 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7578 v4si __builtin_ia32_paddd128 (v4si, v4si)
7579 v2di __builtin_ia32_paddq128 (v2di, v2di)
7580 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7581 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7582 v4si __builtin_ia32_psubd128 (v4si, v4si)
7583 v2di __builtin_ia32_psubq128 (v2di, v2di)
7584 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7585 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7586 v2di __builtin_ia32_pand128 (v2di, v2di)
7587 v2di __builtin_ia32_pandn128 (v2di, v2di)
7588 v2di __builtin_ia32_por128 (v2di, v2di)
7589 v2di __builtin_ia32_pxor128 (v2di, v2di)
7590 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7591 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7592 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7593 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7594 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7595 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7596 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7597 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7598 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7599 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7600 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7601 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7602 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7603 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7604 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7605 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7606 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7607 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7608 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7609 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7610 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7611 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7612 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7613 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7614 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7615 v2df __builtin_ia32_loadupd (double *)
7616 void __builtin_ia32_storeupd (double *, v2df)
7617 v2df __builtin_ia32_loadhpd (v2df, double *)
7618 v2df __builtin_ia32_loadlpd (v2df, double *)
7619 int __builtin_ia32_movmskpd (v2df)
7620 int __builtin_ia32_pmovmskb128 (v16qi)
7621 void __builtin_ia32_movnti (int *, int)
7622 void __builtin_ia32_movntpd (double *, v2df)
7623 void __builtin_ia32_movntdq (v2df *, v2df)
7624 v4si __builtin_ia32_pshufd (v4si, int)
7625 v8hi __builtin_ia32_pshuflw (v8hi, int)
7626 v8hi __builtin_ia32_pshufhw (v8hi, int)
7627 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7628 v2df __builtin_ia32_sqrtpd (v2df)
7629 v2df __builtin_ia32_sqrtsd (v2df)
7630 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7631 v2df __builtin_ia32_cvtdq2pd (v4si)
7632 v4sf __builtin_ia32_cvtdq2ps (v4si)
7633 v4si __builtin_ia32_cvtpd2dq (v2df)
7634 v2si __builtin_ia32_cvtpd2pi (v2df)
7635 v4sf __builtin_ia32_cvtpd2ps (v2df)
7636 v4si __builtin_ia32_cvttpd2dq (v2df)
7637 v2si __builtin_ia32_cvttpd2pi (v2df)
7638 v2df __builtin_ia32_cvtpi2pd (v2si)
7639 int __builtin_ia32_cvtsd2si (v2df)
7640 int __builtin_ia32_cvttsd2si (v2df)
7641 long long __builtin_ia32_cvtsd2si64 (v2df)
7642 long long __builtin_ia32_cvttsd2si64 (v2df)
7643 v4si __builtin_ia32_cvtps2dq (v4sf)
7644 v2df __builtin_ia32_cvtps2pd (v4sf)
7645 v4si __builtin_ia32_cvttps2dq (v4sf)
7646 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7647 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7648 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7649 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7650 void __builtin_ia32_clflush (const void *)
7651 void __builtin_ia32_lfence (void)
7652 void __builtin_ia32_mfence (void)
7653 v16qi __builtin_ia32_loaddqu (const char *)
7654 void __builtin_ia32_storedqu (char *, v16qi)
7655 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7656 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7657 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7658 v4si __builtin_ia32_pslld128 (v4si, v2di)
7659 v2di __builtin_ia32_psllq128 (v4si, v2di)
7660 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7661 v4si __builtin_ia32_psrld128 (v4si, v2di)
7662 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7663 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7664 v4si __builtin_ia32_psrad128 (v4si, v2di)
7665 v2di __builtin_ia32_pslldqi128 (v2di, int)
7666 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7667 v4si __builtin_ia32_pslldi128 (v4si, int)
7668 v2di __builtin_ia32_psllqi128 (v2di, int)
7669 v2di __builtin_ia32_psrldqi128 (v2di, int)
7670 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7671 v4si __builtin_ia32_psrldi128 (v4si, int)
7672 v2di __builtin_ia32_psrlqi128 (v2di, int)
7673 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7674 v4si __builtin_ia32_psradi128 (v4si, int)
7675 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7678 The following built-in functions are available when @option{-msse3} is used.
7679 All of them generate the machine instruction that is part of the name.
7682 v2df __builtin_ia32_addsubpd (v2df, v2df)
7683 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7684 v2df __builtin_ia32_haddpd (v2df, v2df)
7685 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7686 v2df __builtin_ia32_hsubpd (v2df, v2df)
7687 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7688 v16qi __builtin_ia32_lddqu (char const *)
7689 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7690 v2df __builtin_ia32_movddup (v2df)
7691 v4sf __builtin_ia32_movshdup (v4sf)
7692 v4sf __builtin_ia32_movsldup (v4sf)
7693 void __builtin_ia32_mwait (unsigned int, unsigned int)
7696 The following built-in functions are available when @option{-msse3} is used.
7699 @item v2df __builtin_ia32_loadddup (double const *)
7700 Generates the @code{movddup} machine instruction as a load from memory.
7703 The following built-in functions are available when @option{-mssse3} is used.
7704 All of them generate the machine instruction that is part of the name
7708 v2si __builtin_ia32_phaddd (v2si, v2si)
7709 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7710 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7711 v2si __builtin_ia32_phsubd (v2si, v2si)
7712 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7713 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7714 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7715 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7716 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7717 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7718 v2si __builtin_ia32_psignd (v2si, v2si)
7719 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7720 long long __builtin_ia32_palignr (long long, long long, int)
7721 v8qi __builtin_ia32_pabsb (v8qi)
7722 v2si __builtin_ia32_pabsd (v2si)
7723 v4hi __builtin_ia32_pabsw (v4hi)
7726 The following built-in functions are available when @option{-mssse3} is used.
7727 All of them generate the machine instruction that is part of the name
7731 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7732 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7733 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7734 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7735 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7736 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7737 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7738 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7739 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7740 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7741 v4si __builtin_ia32_psignd128 (v4si, v4si)
7742 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7743 v2di __builtin_ia32_palignr (v2di, v2di, int)
7744 v16qi __builtin_ia32_pabsb128 (v16qi)
7745 v4si __builtin_ia32_pabsd128 (v4si)
7746 v8hi __builtin_ia32_pabsw128 (v8hi)
7749 The following built-in functions are available when @option{-msse4.1} is
7750 used. All of them generate the machine instruction that is part of the
7754 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
7755 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
7756 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
7757 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
7758 v2df __builtin_ia32_dppd (v2df, v2df, const int)
7759 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
7760 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
7761 v2di __builtin_ia32_movntdqa (v2di *);
7762 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
7763 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
7764 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
7765 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
7766 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
7767 v8hi __builtin_ia32_phminposuw128 (v8hi)
7768 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
7769 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
7770 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
7771 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
7772 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
7773 v4si __builtin_ia32_pminsd128 (v4si, v4si)
7774 v4si __builtin_ia32_pminud128 (v4si, v4si)
7775 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
7776 v4si __builtin_ia32_pmovsxbd128 (v16qi)
7777 v2di __builtin_ia32_pmovsxbq128 (v16qi)
7778 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
7779 v2di __builtin_ia32_pmovsxdq128 (v4si)
7780 v4si __builtin_ia32_pmovsxwd128 (v8hi)
7781 v2di __builtin_ia32_pmovsxwq128 (v8hi)
7782 v4si __builtin_ia32_pmovzxbd128 (v16qi)
7783 v2di __builtin_ia32_pmovzxbq128 (v16qi)
7784 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
7785 v2di __builtin_ia32_pmovzxdq128 (v4si)
7786 v4si __builtin_ia32_pmovzxwd128 (v8hi)
7787 v2di __builtin_ia32_pmovzxwq128 (v8hi)
7788 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
7789 v4si __builtin_ia32_pmulld128 (v4si, v4si)
7790 int __builtin_ia32_ptestc128 (v2di, v2di)
7791 int __builtin_ia32_ptestnzc128 (v2di, v2di)
7792 int __builtin_ia32_ptestz128 (v2di, v2di)
7793 v2df __builtin_ia32_roundpd (v2df, const int)
7794 v4sf __builtin_ia32_roundps (v4sf, const int)
7795 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
7796 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
7799 The following built-in functions are available when @option{-msse4.1} is
7803 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
7804 Generates the @code{insertps} machine instruction.
7805 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
7806 Generates the @code{pextrb} machine instruction.
7807 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
7808 Generates the @code{pinsrb} machine instruction.
7809 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
7810 Generates the @code{pinsrd} machine instruction.
7811 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
7812 Generates the @code{pinsrq} machine instruction in 64bit mode.
7815 The following built-in functions are changed to generate new SSE4.1
7816 instructions when @option{-msse4.1} is used.
7819 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
7820 Generates the @code{extractps} machine instruction.
7821 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
7822 Generates the @code{pextrd} machine instruction.
7823 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
7824 Generates the @code{pextrq} machine instruction in 64bit mode.
7827 The following built-in functions are available when @option{-msse4.2} is
7828 used. All of them generate the machine instruction that is part of the
7832 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
7833 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
7834 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
7835 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
7836 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
7837 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
7838 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
7839 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
7840 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
7841 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
7842 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
7843 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
7844 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
7845 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
7846 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
7849 The following built-in functions are available when @option{-msse4.2} is
7853 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
7854 Generates the @code{crc32b} machine instruction.
7855 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
7856 Generates the @code{crc32w} machine instruction.
7857 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
7858 Generates the @code{crc32l} machine instruction.
7859 @item unsigned long long __builtin_ia32_crc32di (unsigned int, unsigned long long)
7862 The following built-in functions are changed to generate new SSE4.2
7863 instructions when @option{-msse4.2} is used.
7866 @item int __builtin_popcount (unsigned int)
7867 Generates the @code{popcntl} machine instruction.
7868 @item int __builtin_popcountl (unsigned long)
7869 Generates the @code{popcntl} or @code{popcntq} machine instruction,
7870 depending on the size of @code{unsigned long}.
7871 @item int __builtin_popcountll (unsigned long long)
7872 Generates the @code{popcntq} machine instruction.
7875 The following built-in functions are available when @option{-msse4a} is used.
7876 All of them generate the machine instruction that is part of the name.
7879 void __builtin_ia32_movntsd (double *, v2df)
7880 void __builtin_ia32_movntss (float *, v4sf)
7881 v2di __builtin_ia32_extrq (v2di, v16qi)
7882 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
7883 v2di __builtin_ia32_insertq (v2di, v2di)
7884 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
7887 The following built-in functions are available when @option{-m3dnow} is used.
7888 All of them generate the machine instruction that is part of the name.
7891 void __builtin_ia32_femms (void)
7892 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7893 v2si __builtin_ia32_pf2id (v2sf)
7894 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7895 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7896 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7897 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7898 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7899 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7900 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7901 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7902 v2sf __builtin_ia32_pfrcp (v2sf)
7903 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7904 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7905 v2sf __builtin_ia32_pfrsqrt (v2sf)
7906 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7907 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7908 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7909 v2sf __builtin_ia32_pi2fd (v2si)
7910 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7913 The following built-in functions are available when both @option{-m3dnow}
7914 and @option{-march=athlon} are used. All of them generate the machine
7915 instruction that is part of the name.
7918 v2si __builtin_ia32_pf2iw (v2sf)
7919 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7920 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7921 v2sf __builtin_ia32_pi2fw (v2si)
7922 v2sf __builtin_ia32_pswapdsf (v2sf)
7923 v2si __builtin_ia32_pswapdsi (v2si)
7926 @node MIPS DSP Built-in Functions
7927 @subsection MIPS DSP Built-in Functions
7929 The MIPS DSP Application-Specific Extension (ASE) includes new
7930 instructions that are designed to improve the performance of DSP and
7931 media applications. It provides instructions that operate on packed
7932 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
7934 GCC supports MIPS DSP operations using both the generic
7935 vector extensions (@pxref{Vector Extensions}) and a collection of
7936 MIPS-specific built-in functions. Both kinds of support are
7937 enabled by the @option{-mdsp} command-line option.
7939 Revision 2 of the ASE was introduced in the second half of 2006.
7940 This revision adds extra instructions to the original ASE, but is
7941 otherwise backwards-compatible with it. You can select revision 2
7942 using the command-line option @option{-mdspr2}; this option implies
7945 At present, GCC only provides support for operations on 32-bit
7946 vectors. The vector type associated with 8-bit integer data is
7947 usually called @code{v4i8}, the vector type associated with Q7
7948 is usually called @code{v4q7}, the vector type associated with 16-bit
7949 integer data is usually called @code{v2i16}, and the vector type
7950 associated with Q15 is usually called @code{v2q15}. They can be
7951 defined in C as follows:
7954 typedef signed char v4i8 __attribute__ ((vector_size(4)));
7955 typedef signed char v4q7 __attribute__ ((vector_size(4)));
7956 typedef short v2i16 __attribute__ ((vector_size(4)));
7957 typedef short v2q15 __attribute__ ((vector_size(4)));
7960 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
7961 initialized in the same way as aggregates. For example:
7964 v4i8 a = @{1, 2, 3, 4@};
7966 b = (v4i8) @{5, 6, 7, 8@};
7968 v2q15 c = @{0x0fcb, 0x3a75@};
7970 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7973 @emph{Note:} The CPU's endianness determines the order in which values
7974 are packed. On little-endian targets, the first value is the least
7975 significant and the last value is the most significant. The opposite
7976 order applies to big-endian targets. For example, the code above will
7977 set the lowest byte of @code{a} to @code{1} on little-endian targets
7978 and @code{4} on big-endian targets.
7980 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
7981 representation. As shown in this example, the integer representation
7982 of a Q7 value can be obtained by multiplying the fractional value by
7983 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
7984 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7987 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7988 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7989 and @code{c} and @code{d} are @code{v2q15} values.
7991 @multitable @columnfractions .50 .50
7992 @item C code @tab MIPS instruction
7993 @item @code{a + b} @tab @code{addu.qb}
7994 @item @code{c + d} @tab @code{addq.ph}
7995 @item @code{a - b} @tab @code{subu.qb}
7996 @item @code{c - d} @tab @code{subq.ph}
7999 The table below lists the @code{v2i16} operation for which
8000 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
8001 @code{v2i16} values.
8003 @multitable @columnfractions .50 .50
8004 @item C code @tab MIPS instruction
8005 @item @code{e * f} @tab @code{mul.ph}
8008 It is easier to describe the DSP built-in functions if we first define
8009 the following types:
8014 typedef unsigned int ui32;
8015 typedef long long a64;
8018 @code{q31} and @code{i32} are actually the same as @code{int}, but we
8019 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
8020 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
8021 @code{long long}, but we use @code{a64} to indicate values that will
8022 be placed in one of the four DSP accumulators (@code{$ac0},
8023 @code{$ac1}, @code{$ac2} or @code{$ac3}).
8025 Also, some built-in functions prefer or require immediate numbers as
8026 parameters, because the corresponding DSP instructions accept both immediate
8027 numbers and register operands, or accept immediate numbers only. The
8028 immediate parameters are listed as follows.
8037 imm_n32_31: -32 to 31.
8038 imm_n512_511: -512 to 511.
8041 The following built-in functions map directly to a particular MIPS DSP
8042 instruction. Please refer to the architecture specification
8043 for details on what each instruction does.
8046 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
8047 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
8048 q31 __builtin_mips_addq_s_w (q31, q31)
8049 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
8050 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
8051 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
8052 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
8053 q31 __builtin_mips_subq_s_w (q31, q31)
8054 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
8055 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
8056 i32 __builtin_mips_addsc (i32, i32)
8057 i32 __builtin_mips_addwc (i32, i32)
8058 i32 __builtin_mips_modsub (i32, i32)
8059 i32 __builtin_mips_raddu_w_qb (v4i8)
8060 v2q15 __builtin_mips_absq_s_ph (v2q15)
8061 q31 __builtin_mips_absq_s_w (q31)
8062 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
8063 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
8064 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
8065 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
8066 q31 __builtin_mips_preceq_w_phl (v2q15)
8067 q31 __builtin_mips_preceq_w_phr (v2q15)
8068 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
8069 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
8070 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
8071 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
8072 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
8073 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
8074 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
8075 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
8076 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
8077 v4i8 __builtin_mips_shll_qb (v4i8, i32)
8078 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
8079 v2q15 __builtin_mips_shll_ph (v2q15, i32)
8080 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
8081 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
8082 q31 __builtin_mips_shll_s_w (q31, imm0_31)
8083 q31 __builtin_mips_shll_s_w (q31, i32)
8084 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
8085 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
8086 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
8087 v2q15 __builtin_mips_shra_ph (v2q15, i32)
8088 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
8089 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
8090 q31 __builtin_mips_shra_r_w (q31, imm0_31)
8091 q31 __builtin_mips_shra_r_w (q31, i32)
8092 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
8093 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
8094 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
8095 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
8096 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
8097 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
8098 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
8099 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
8100 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
8101 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
8102 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
8103 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
8104 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
8105 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
8106 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
8107 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
8108 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
8109 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
8110 i32 __builtin_mips_bitrev (i32)
8111 i32 __builtin_mips_insv (i32, i32)
8112 v4i8 __builtin_mips_repl_qb (imm0_255)
8113 v4i8 __builtin_mips_repl_qb (i32)
8114 v2q15 __builtin_mips_repl_ph (imm_n512_511)
8115 v2q15 __builtin_mips_repl_ph (i32)
8116 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
8117 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
8118 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
8119 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
8120 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
8121 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
8122 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
8123 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
8124 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
8125 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
8126 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
8127 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
8128 i32 __builtin_mips_extr_w (a64, imm0_31)
8129 i32 __builtin_mips_extr_w (a64, i32)
8130 i32 __builtin_mips_extr_r_w (a64, imm0_31)
8131 i32 __builtin_mips_extr_s_h (a64, i32)
8132 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
8133 i32 __builtin_mips_extr_rs_w (a64, i32)
8134 i32 __builtin_mips_extr_s_h (a64, imm0_31)
8135 i32 __builtin_mips_extr_r_w (a64, i32)
8136 i32 __builtin_mips_extp (a64, imm0_31)
8137 i32 __builtin_mips_extp (a64, i32)
8138 i32 __builtin_mips_extpdp (a64, imm0_31)
8139 i32 __builtin_mips_extpdp (a64, i32)
8140 a64 __builtin_mips_shilo (a64, imm_n32_31)
8141 a64 __builtin_mips_shilo (a64, i32)
8142 a64 __builtin_mips_mthlip (a64, i32)
8143 void __builtin_mips_wrdsp (i32, imm0_63)
8144 i32 __builtin_mips_rddsp (imm0_63)
8145 i32 __builtin_mips_lbux (void *, i32)
8146 i32 __builtin_mips_lhx (void *, i32)
8147 i32 __builtin_mips_lwx (void *, i32)
8148 i32 __builtin_mips_bposge32 (void)
8151 The following built-in functions map directly to a particular MIPS DSP REV 2
8152 instruction. Please refer to the architecture specification
8153 for details on what each instruction does.
8156 v4q7 __builtin_mips_absq_s_qb (v4q7);
8157 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
8158 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
8159 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
8160 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
8161 i32 __builtin_mips_append (i32, i32, imm0_31);
8162 i32 __builtin_mips_balign (i32, i32, imm0_3);
8163 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
8164 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
8165 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
8166 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
8167 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
8168 a64 __builtin_mips_madd (a64, i32, i32);
8169 a64 __builtin_mips_maddu (a64, ui32, ui32);
8170 a64 __builtin_mips_msub (a64, i32, i32);
8171 a64 __builtin_mips_msubu (a64, ui32, ui32);
8172 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
8173 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
8174 q31 __builtin_mips_mulq_rs_w (q31, q31);
8175 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
8176 q31 __builtin_mips_mulq_s_w (q31, q31);
8177 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
8178 a64 __builtin_mips_mult (i32, i32);
8179 a64 __builtin_mips_multu (ui32, ui32);
8180 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
8181 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
8182 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
8183 i32 __builtin_mips_prepend (i32, i32, imm0_31);
8184 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
8185 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
8186 v4i8 __builtin_mips_shra_qb (v4i8, i32);
8187 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
8188 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
8189 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
8190 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
8191 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
8192 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
8193 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
8194 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
8195 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
8196 q31 __builtin_mips_addqh_w (q31, q31);
8197 q31 __builtin_mips_addqh_r_w (q31, q31);
8198 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
8199 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
8200 q31 __builtin_mips_subqh_w (q31, q31);
8201 q31 __builtin_mips_subqh_r_w (q31, q31);
8202 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
8203 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
8204 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
8205 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
8206 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
8207 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
8211 @node MIPS Paired-Single Support
8212 @subsection MIPS Paired-Single Support
8214 The MIPS64 architecture includes a number of instructions that
8215 operate on pairs of single-precision floating-point values.
8216 Each pair is packed into a 64-bit floating-point register,
8217 with one element being designated the ``upper half'' and
8218 the other being designated the ``lower half''.
8220 GCC supports paired-single operations using both the generic
8221 vector extensions (@pxref{Vector Extensions}) and a collection of
8222 MIPS-specific built-in functions. Both kinds of support are
8223 enabled by the @option{-mpaired-single} command-line option.
8225 The vector type associated with paired-single values is usually
8226 called @code{v2sf}. It can be defined in C as follows:
8229 typedef float v2sf __attribute__ ((vector_size (8)));
8232 @code{v2sf} values are initialized in the same way as aggregates.
8236 v2sf a = @{1.5, 9.1@};
8239 b = (v2sf) @{e, f@};
8242 @emph{Note:} The CPU's endianness determines which value is stored in
8243 the upper half of a register and which value is stored in the lower half.
8244 On little-endian targets, the first value is the lower one and the second
8245 value is the upper one. The opposite order applies to big-endian targets.
8246 For example, the code above will set the lower half of @code{a} to
8247 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
8250 * Paired-Single Arithmetic::
8251 * Paired-Single Built-in Functions::
8252 * MIPS-3D Built-in Functions::
8255 @node Paired-Single Arithmetic
8256 @subsubsection Paired-Single Arithmetic
8258 The table below lists the @code{v2sf} operations for which hardware
8259 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
8260 values and @code{x} is an integral value.
8262 @multitable @columnfractions .50 .50
8263 @item C code @tab MIPS instruction
8264 @item @code{a + b} @tab @code{add.ps}
8265 @item @code{a - b} @tab @code{sub.ps}
8266 @item @code{-a} @tab @code{neg.ps}
8267 @item @code{a * b} @tab @code{mul.ps}
8268 @item @code{a * b + c} @tab @code{madd.ps}
8269 @item @code{a * b - c} @tab @code{msub.ps}
8270 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
8271 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
8272 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
8275 Note that the multiply-accumulate instructions can be disabled
8276 using the command-line option @code{-mno-fused-madd}.
8278 @node Paired-Single Built-in Functions
8279 @subsubsection Paired-Single Built-in Functions
8281 The following paired-single functions map directly to a particular
8282 MIPS instruction. Please refer to the architecture specification
8283 for details on what each instruction does.
8286 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
8287 Pair lower lower (@code{pll.ps}).
8289 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
8290 Pair upper lower (@code{pul.ps}).
8292 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
8293 Pair lower upper (@code{plu.ps}).
8295 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
8296 Pair upper upper (@code{puu.ps}).
8298 @item v2sf __builtin_mips_cvt_ps_s (float, float)
8299 Convert pair to paired single (@code{cvt.ps.s}).
8301 @item float __builtin_mips_cvt_s_pl (v2sf)
8302 Convert pair lower to single (@code{cvt.s.pl}).
8304 @item float __builtin_mips_cvt_s_pu (v2sf)
8305 Convert pair upper to single (@code{cvt.s.pu}).
8307 @item v2sf __builtin_mips_abs_ps (v2sf)
8308 Absolute value (@code{abs.ps}).
8310 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
8311 Align variable (@code{alnv.ps}).
8313 @emph{Note:} The value of the third parameter must be 0 or 4
8314 modulo 8, otherwise the result will be unpredictable. Please read the
8315 instruction description for details.
8318 The following multi-instruction functions are also available.
8319 In each case, @var{cond} can be any of the 16 floating-point conditions:
8320 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
8321 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
8322 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
8325 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8326 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8327 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
8328 @code{movt.ps}/@code{movf.ps}).
8330 The @code{movt} functions return the value @var{x} computed by:
8333 c.@var{cond}.ps @var{cc},@var{a},@var{b}
8334 mov.ps @var{x},@var{c}
8335 movt.ps @var{x},@var{d},@var{cc}
8338 The @code{movf} functions are similar but use @code{movf.ps} instead
8341 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8342 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8343 Comparison of two paired-single values (@code{c.@var{cond}.ps},
8344 @code{bc1t}/@code{bc1f}).
8346 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
8347 and return either the upper or lower half of the result. For example:
8351 if (__builtin_mips_upper_c_eq_ps (a, b))
8352 upper_halves_are_equal ();
8354 upper_halves_are_unequal ();
8356 if (__builtin_mips_lower_c_eq_ps (a, b))
8357 lower_halves_are_equal ();
8359 lower_halves_are_unequal ();
8363 @node MIPS-3D Built-in Functions
8364 @subsubsection MIPS-3D Built-in Functions
8366 The MIPS-3D Application-Specific Extension (ASE) includes additional
8367 paired-single instructions that are designed to improve the performance
8368 of 3D graphics operations. Support for these instructions is controlled
8369 by the @option{-mips3d} command-line option.
8371 The functions listed below map directly to a particular MIPS-3D
8372 instruction. Please refer to the architecture specification for
8373 more details on what each instruction does.
8376 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
8377 Reduction add (@code{addr.ps}).
8379 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
8380 Reduction multiply (@code{mulr.ps}).
8382 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
8383 Convert paired single to paired word (@code{cvt.pw.ps}).
8385 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
8386 Convert paired word to paired single (@code{cvt.ps.pw}).
8388 @item float __builtin_mips_recip1_s (float)
8389 @itemx double __builtin_mips_recip1_d (double)
8390 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
8391 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
8393 @item float __builtin_mips_recip2_s (float, float)
8394 @itemx double __builtin_mips_recip2_d (double, double)
8395 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
8396 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
8398 @item float __builtin_mips_rsqrt1_s (float)
8399 @itemx double __builtin_mips_rsqrt1_d (double)
8400 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
8401 Reduced precision reciprocal square root (sequence step 1)
8402 (@code{rsqrt1.@var{fmt}}).
8404 @item float __builtin_mips_rsqrt2_s (float, float)
8405 @itemx double __builtin_mips_rsqrt2_d (double, double)
8406 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
8407 Reduced precision reciprocal square root (sequence step 2)
8408 (@code{rsqrt2.@var{fmt}}).
8411 The following multi-instruction functions are also available.
8412 In each case, @var{cond} can be any of the 16 floating-point conditions:
8413 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
8414 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
8415 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
8418 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
8419 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
8420 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
8421 @code{bc1t}/@code{bc1f}).
8423 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
8424 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
8429 if (__builtin_mips_cabs_eq_s (a, b))
8435 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8436 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8437 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
8438 @code{bc1t}/@code{bc1f}).
8440 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
8441 and return either the upper or lower half of the result. For example:
8445 if (__builtin_mips_upper_cabs_eq_ps (a, b))
8446 upper_halves_are_equal ();
8448 upper_halves_are_unequal ();
8450 if (__builtin_mips_lower_cabs_eq_ps (a, b))
8451 lower_halves_are_equal ();
8453 lower_halves_are_unequal ();
8456 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8457 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8458 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
8459 @code{movt.ps}/@code{movf.ps}).
8461 The @code{movt} functions return the value @var{x} computed by:
8464 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
8465 mov.ps @var{x},@var{c}
8466 movt.ps @var{x},@var{d},@var{cc}
8469 The @code{movf} functions are similar but use @code{movf.ps} instead
8472 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8473 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8474 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8475 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8476 Comparison of two paired-single values
8477 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
8478 @code{bc1any2t}/@code{bc1any2f}).
8480 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
8481 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
8482 result is true and the @code{all} forms return true if both results are true.
8487 if (__builtin_mips_any_c_eq_ps (a, b))
8492 if (__builtin_mips_all_c_eq_ps (a, b))
8498 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8499 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8500 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8501 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8502 Comparison of four paired-single values
8503 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
8504 @code{bc1any4t}/@code{bc1any4f}).
8506 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
8507 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
8508 The @code{any} forms return true if any of the four results are true
8509 and the @code{all} forms return true if all four results are true.
8514 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
8519 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
8526 @node PowerPC AltiVec Built-in Functions
8527 @subsection PowerPC AltiVec Built-in Functions
8529 GCC provides an interface for the PowerPC family of processors to access
8530 the AltiVec operations described in Motorola's AltiVec Programming
8531 Interface Manual. The interface is made available by including
8532 @code{<altivec.h>} and using @option{-maltivec} and
8533 @option{-mabi=altivec}. The interface supports the following vector
8537 vector unsigned char
8541 vector unsigned short
8552 GCC's implementation of the high-level language interface available from
8553 C and C++ code differs from Motorola's documentation in several ways.
8558 A vector constant is a list of constant expressions within curly braces.
8561 A vector initializer requires no cast if the vector constant is of the
8562 same type as the variable it is initializing.
8565 If @code{signed} or @code{unsigned} is omitted, the signedness of the
8566 vector type is the default signedness of the base type. The default
8567 varies depending on the operating system, so a portable program should
8568 always specify the signedness.
8571 Compiling with @option{-maltivec} adds keywords @code{__vector},
8572 @code{__pixel}, and @code{__bool}. Macros @option{vector},
8573 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
8577 GCC allows using a @code{typedef} name as the type specifier for a
8581 For C, overloaded functions are implemented with macros so the following
8585 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
8588 Since @code{vec_add} is a macro, the vector constant in the example
8589 is treated as four separate arguments. Wrap the entire argument in
8590 parentheses for this to work.
8593 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
8594 Internally, GCC uses built-in functions to achieve the functionality in
8595 the aforementioned header file, but they are not supported and are
8596 subject to change without notice.
8598 The following interfaces are supported for the generic and specific
8599 AltiVec operations and the AltiVec predicates. In cases where there
8600 is a direct mapping between generic and specific operations, only the
8601 generic names are shown here, although the specific operations can also
8604 Arguments that are documented as @code{const int} require literal
8605 integral values within the range required for that operation.
8608 vector signed char vec_abs (vector signed char);
8609 vector signed short vec_abs (vector signed short);
8610 vector signed int vec_abs (vector signed int);
8611 vector float vec_abs (vector float);
8613 vector signed char vec_abss (vector signed char);
8614 vector signed short vec_abss (vector signed short);
8615 vector signed int vec_abss (vector signed int);
8617 vector signed char vec_add (vector bool char, vector signed char);
8618 vector signed char vec_add (vector signed char, vector bool char);
8619 vector signed char vec_add (vector signed char, vector signed char);
8620 vector unsigned char vec_add (vector bool char, vector unsigned char);
8621 vector unsigned char vec_add (vector unsigned char, vector bool char);
8622 vector unsigned char vec_add (vector unsigned char,
8623 vector unsigned char);
8624 vector signed short vec_add (vector bool short, vector signed short);
8625 vector signed short vec_add (vector signed short, vector bool short);
8626 vector signed short vec_add (vector signed short, vector signed short);
8627 vector unsigned short vec_add (vector bool short,
8628 vector unsigned short);
8629 vector unsigned short vec_add (vector unsigned short,
8631 vector unsigned short vec_add (vector unsigned short,
8632 vector unsigned short);
8633 vector signed int vec_add (vector bool int, vector signed int);
8634 vector signed int vec_add (vector signed int, vector bool int);
8635 vector signed int vec_add (vector signed int, vector signed int);
8636 vector unsigned int vec_add (vector bool int, vector unsigned int);
8637 vector unsigned int vec_add (vector unsigned int, vector bool int);
8638 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8639 vector float vec_add (vector float, vector float);
8641 vector float vec_vaddfp (vector float, vector float);
8643 vector signed int vec_vadduwm (vector bool int, vector signed int);
8644 vector signed int vec_vadduwm (vector signed int, vector bool int);
8645 vector signed int vec_vadduwm (vector signed int, vector signed int);
8646 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8647 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8648 vector unsigned int vec_vadduwm (vector unsigned int,
8649 vector unsigned int);
8651 vector signed short vec_vadduhm (vector bool short,
8652 vector signed short);
8653 vector signed short vec_vadduhm (vector signed short,
8655 vector signed short vec_vadduhm (vector signed short,
8656 vector signed short);
8657 vector unsigned short vec_vadduhm (vector bool short,
8658 vector unsigned short);
8659 vector unsigned short vec_vadduhm (vector unsigned short,
8661 vector unsigned short vec_vadduhm (vector unsigned short,
8662 vector unsigned short);
8664 vector signed char vec_vaddubm (vector bool char, vector signed char);
8665 vector signed char vec_vaddubm (vector signed char, vector bool char);
8666 vector signed char vec_vaddubm (vector signed char, vector signed char);
8667 vector unsigned char vec_vaddubm (vector bool char,
8668 vector unsigned char);
8669 vector unsigned char vec_vaddubm (vector unsigned char,
8671 vector unsigned char vec_vaddubm (vector unsigned char,
8672 vector unsigned char);
8674 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8676 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8677 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8678 vector unsigned char vec_adds (vector unsigned char,
8679 vector unsigned char);
8680 vector signed char vec_adds (vector bool char, vector signed char);
8681 vector signed char vec_adds (vector signed char, vector bool char);
8682 vector signed char vec_adds (vector signed char, vector signed char);
8683 vector unsigned short vec_adds (vector bool short,
8684 vector unsigned short);
8685 vector unsigned short vec_adds (vector unsigned short,
8687 vector unsigned short vec_adds (vector unsigned short,
8688 vector unsigned short);
8689 vector signed short vec_adds (vector bool short, vector signed short);
8690 vector signed short vec_adds (vector signed short, vector bool short);
8691 vector signed short vec_adds (vector signed short, vector signed short);
8692 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8693 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8694 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8695 vector signed int vec_adds (vector bool int, vector signed int);
8696 vector signed int vec_adds (vector signed int, vector bool int);
8697 vector signed int vec_adds (vector signed int, vector signed int);
8699 vector signed int vec_vaddsws (vector bool int, vector signed int);
8700 vector signed int vec_vaddsws (vector signed int, vector bool int);
8701 vector signed int vec_vaddsws (vector signed int, vector signed int);
8703 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8704 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8705 vector unsigned int vec_vadduws (vector unsigned int,
8706 vector unsigned int);
8708 vector signed short vec_vaddshs (vector bool short,
8709 vector signed short);
8710 vector signed short vec_vaddshs (vector signed short,
8712 vector signed short vec_vaddshs (vector signed short,
8713 vector signed short);
8715 vector unsigned short vec_vadduhs (vector bool short,
8716 vector unsigned short);
8717 vector unsigned short vec_vadduhs (vector unsigned short,
8719 vector unsigned short vec_vadduhs (vector unsigned short,
8720 vector unsigned short);
8722 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8723 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8724 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8726 vector unsigned char vec_vaddubs (vector bool char,
8727 vector unsigned char);
8728 vector unsigned char vec_vaddubs (vector unsigned char,
8730 vector unsigned char vec_vaddubs (vector unsigned char,
8731 vector unsigned char);
8733 vector float vec_and (vector float, vector float);
8734 vector float vec_and (vector float, vector bool int);
8735 vector float vec_and (vector bool int, vector float);
8736 vector bool int vec_and (vector bool int, vector bool int);
8737 vector signed int vec_and (vector bool int, vector signed int);
8738 vector signed int vec_and (vector signed int, vector bool int);
8739 vector signed int vec_and (vector signed int, vector signed int);
8740 vector unsigned int vec_and (vector bool int, vector unsigned int);
8741 vector unsigned int vec_and (vector unsigned int, vector bool int);
8742 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8743 vector bool short vec_and (vector bool short, vector bool short);
8744 vector signed short vec_and (vector bool short, vector signed short);
8745 vector signed short vec_and (vector signed short, vector bool short);
8746 vector signed short vec_and (vector signed short, vector signed short);
8747 vector unsigned short vec_and (vector bool short,
8748 vector unsigned short);
8749 vector unsigned short vec_and (vector unsigned short,
8751 vector unsigned short vec_and (vector unsigned short,
8752 vector unsigned short);
8753 vector signed char vec_and (vector bool char, vector signed char);
8754 vector bool char vec_and (vector bool char, vector bool char);
8755 vector signed char vec_and (vector signed char, vector bool char);
8756 vector signed char vec_and (vector signed char, vector signed char);
8757 vector unsigned char vec_and (vector bool char, vector unsigned char);
8758 vector unsigned char vec_and (vector unsigned char, vector bool char);
8759 vector unsigned char vec_and (vector unsigned char,
8760 vector unsigned char);
8762 vector float vec_andc (vector float, vector float);
8763 vector float vec_andc (vector float, vector bool int);
8764 vector float vec_andc (vector bool int, vector float);
8765 vector bool int vec_andc (vector bool int, vector bool int);
8766 vector signed int vec_andc (vector bool int, vector signed int);
8767 vector signed int vec_andc (vector signed int, vector bool int);
8768 vector signed int vec_andc (vector signed int, vector signed int);
8769 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8770 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8771 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8772 vector bool short vec_andc (vector bool short, vector bool short);
8773 vector signed short vec_andc (vector bool short, vector signed short);
8774 vector signed short vec_andc (vector signed short, vector bool short);
8775 vector signed short vec_andc (vector signed short, vector signed short);
8776 vector unsigned short vec_andc (vector bool short,
8777 vector unsigned short);
8778 vector unsigned short vec_andc (vector unsigned short,
8780 vector unsigned short vec_andc (vector unsigned short,
8781 vector unsigned short);
8782 vector signed char vec_andc (vector bool char, vector signed char);
8783 vector bool char vec_andc (vector bool char, vector bool char);
8784 vector signed char vec_andc (vector signed char, vector bool char);
8785 vector signed char vec_andc (vector signed char, vector signed char);
8786 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8787 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8788 vector unsigned char vec_andc (vector unsigned char,
8789 vector unsigned char);
8791 vector unsigned char vec_avg (vector unsigned char,
8792 vector unsigned char);
8793 vector signed char vec_avg (vector signed char, vector signed char);
8794 vector unsigned short vec_avg (vector unsigned short,
8795 vector unsigned short);
8796 vector signed short vec_avg (vector signed short, vector signed short);
8797 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8798 vector signed int vec_avg (vector signed int, vector signed int);
8800 vector signed int vec_vavgsw (vector signed int, vector signed int);
8802 vector unsigned int vec_vavguw (vector unsigned int,
8803 vector unsigned int);
8805 vector signed short vec_vavgsh (vector signed short,
8806 vector signed short);
8808 vector unsigned short vec_vavguh (vector unsigned short,
8809 vector unsigned short);
8811 vector signed char vec_vavgsb (vector signed char, vector signed char);
8813 vector unsigned char vec_vavgub (vector unsigned char,
8814 vector unsigned char);
8816 vector float vec_ceil (vector float);
8818 vector signed int vec_cmpb (vector float, vector float);
8820 vector bool char vec_cmpeq (vector signed char, vector signed char);
8821 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8822 vector bool short vec_cmpeq (vector signed short, vector signed short);
8823 vector bool short vec_cmpeq (vector unsigned short,
8824 vector unsigned short);
8825 vector bool int vec_cmpeq (vector signed int, vector signed int);
8826 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8827 vector bool int vec_cmpeq (vector float, vector float);
8829 vector bool int vec_vcmpeqfp (vector float, vector float);
8831 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8832 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8834 vector bool short vec_vcmpequh (vector signed short,
8835 vector signed short);
8836 vector bool short vec_vcmpequh (vector unsigned short,
8837 vector unsigned short);
8839 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8840 vector bool char vec_vcmpequb (vector unsigned char,
8841 vector unsigned char);
8843 vector bool int vec_cmpge (vector float, vector float);
8845 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8846 vector bool char vec_cmpgt (vector signed char, vector signed char);
8847 vector bool short vec_cmpgt (vector unsigned short,
8848 vector unsigned short);
8849 vector bool short vec_cmpgt (vector signed short, vector signed short);
8850 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8851 vector bool int vec_cmpgt (vector signed int, vector signed int);
8852 vector bool int vec_cmpgt (vector float, vector float);
8854 vector bool int vec_vcmpgtfp (vector float, vector float);
8856 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8858 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8860 vector bool short vec_vcmpgtsh (vector signed short,
8861 vector signed short);
8863 vector bool short vec_vcmpgtuh (vector unsigned short,
8864 vector unsigned short);
8866 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8868 vector bool char vec_vcmpgtub (vector unsigned char,
8869 vector unsigned char);
8871 vector bool int vec_cmple (vector float, vector float);
8873 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8874 vector bool char vec_cmplt (vector signed char, vector signed char);
8875 vector bool short vec_cmplt (vector unsigned short,
8876 vector unsigned short);
8877 vector bool short vec_cmplt (vector signed short, vector signed short);
8878 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8879 vector bool int vec_cmplt (vector signed int, vector signed int);
8880 vector bool int vec_cmplt (vector float, vector float);
8882 vector float vec_ctf (vector unsigned int, const int);
8883 vector float vec_ctf (vector signed int, const int);
8885 vector float vec_vcfsx (vector signed int, const int);
8887 vector float vec_vcfux (vector unsigned int, const int);
8889 vector signed int vec_cts (vector float, const int);
8891 vector unsigned int vec_ctu (vector float, const int);
8893 void vec_dss (const int);
8895 void vec_dssall (void);
8897 void vec_dst (const vector unsigned char *, int, const int);
8898 void vec_dst (const vector signed char *, int, const int);
8899 void vec_dst (const vector bool char *, int, const int);
8900 void vec_dst (const vector unsigned short *, int, const int);
8901 void vec_dst (const vector signed short *, int, const int);
8902 void vec_dst (const vector bool short *, int, const int);
8903 void vec_dst (const vector pixel *, int, const int);
8904 void vec_dst (const vector unsigned int *, int, const int);
8905 void vec_dst (const vector signed int *, int, const int);
8906 void vec_dst (const vector bool int *, int, const int);
8907 void vec_dst (const vector float *, int, const int);
8908 void vec_dst (const unsigned char *, int, const int);
8909 void vec_dst (const signed char *, int, const int);
8910 void vec_dst (const unsigned short *, int, const int);
8911 void vec_dst (const short *, int, const int);
8912 void vec_dst (const unsigned int *, int, const int);
8913 void vec_dst (const int *, int, const int);
8914 void vec_dst (const unsigned long *, int, const int);
8915 void vec_dst (const long *, int, const int);
8916 void vec_dst (const float *, int, const int);
8918 void vec_dstst (const vector unsigned char *, int, const int);
8919 void vec_dstst (const vector signed char *, int, const int);
8920 void vec_dstst (const vector bool char *, int, const int);
8921 void vec_dstst (const vector unsigned short *, int, const int);
8922 void vec_dstst (const vector signed short *, int, const int);
8923 void vec_dstst (const vector bool short *, int, const int);
8924 void vec_dstst (const vector pixel *, int, const int);
8925 void vec_dstst (const vector unsigned int *, int, const int);
8926 void vec_dstst (const vector signed int *, int, const int);
8927 void vec_dstst (const vector bool int *, int, const int);
8928 void vec_dstst (const vector float *, int, const int);
8929 void vec_dstst (const unsigned char *, int, const int);
8930 void vec_dstst (const signed char *, int, const int);
8931 void vec_dstst (const unsigned short *, int, const int);
8932 void vec_dstst (const short *, int, const int);
8933 void vec_dstst (const unsigned int *, int, const int);
8934 void vec_dstst (const int *, int, const int);
8935 void vec_dstst (const unsigned long *, int, const int);
8936 void vec_dstst (const long *, int, const int);
8937 void vec_dstst (const float *, int, const int);
8939 void vec_dststt (const vector unsigned char *, int, const int);
8940 void vec_dststt (const vector signed char *, int, const int);
8941 void vec_dststt (const vector bool char *, int, const int);
8942 void vec_dststt (const vector unsigned short *, int, const int);
8943 void vec_dststt (const vector signed short *, int, const int);
8944 void vec_dststt (const vector bool short *, int, const int);
8945 void vec_dststt (const vector pixel *, int, const int);
8946 void vec_dststt (const vector unsigned int *, int, const int);
8947 void vec_dststt (const vector signed int *, int, const int);
8948 void vec_dststt (const vector bool int *, int, const int);
8949 void vec_dststt (const vector float *, int, const int);
8950 void vec_dststt (const unsigned char *, int, const int);
8951 void vec_dststt (const signed char *, int, const int);
8952 void vec_dststt (const unsigned short *, int, const int);
8953 void vec_dststt (const short *, int, const int);
8954 void vec_dststt (const unsigned int *, int, const int);
8955 void vec_dststt (const int *, int, const int);
8956 void vec_dststt (const unsigned long *, int, const int);
8957 void vec_dststt (const long *, int, const int);
8958 void vec_dststt (const float *, int, const int);
8960 void vec_dstt (const vector unsigned char *, int, const int);
8961 void vec_dstt (const vector signed char *, int, const int);
8962 void vec_dstt (const vector bool char *, int, const int);
8963 void vec_dstt (const vector unsigned short *, int, const int);
8964 void vec_dstt (const vector signed short *, int, const int);
8965 void vec_dstt (const vector bool short *, int, const int);
8966 void vec_dstt (const vector pixel *, int, const int);
8967 void vec_dstt (const vector unsigned int *, int, const int);
8968 void vec_dstt (const vector signed int *, int, const int);
8969 void vec_dstt (const vector bool int *, int, const int);
8970 void vec_dstt (const vector float *, int, const int);
8971 void vec_dstt (const unsigned char *, int, const int);
8972 void vec_dstt (const signed char *, int, const int);
8973 void vec_dstt (const unsigned short *, int, const int);
8974 void vec_dstt (const short *, int, const int);
8975 void vec_dstt (const unsigned int *, int, const int);
8976 void vec_dstt (const int *, int, const int);
8977 void vec_dstt (const unsigned long *, int, const int);
8978 void vec_dstt (const long *, int, const int);
8979 void vec_dstt (const float *, int, const int);
8981 vector float vec_expte (vector float);
8983 vector float vec_floor (vector float);
8985 vector float vec_ld (int, const vector float *);
8986 vector float vec_ld (int, const float *);
8987 vector bool int vec_ld (int, const vector bool int *);
8988 vector signed int vec_ld (int, const vector signed int *);
8989 vector signed int vec_ld (int, const int *);
8990 vector signed int vec_ld (int, const long *);
8991 vector unsigned int vec_ld (int, const vector unsigned int *);
8992 vector unsigned int vec_ld (int, const unsigned int *);
8993 vector unsigned int vec_ld (int, const unsigned long *);
8994 vector bool short vec_ld (int, const vector bool short *);
8995 vector pixel vec_ld (int, const vector pixel *);
8996 vector signed short vec_ld (int, const vector signed short *);
8997 vector signed short vec_ld (int, const short *);
8998 vector unsigned short vec_ld (int, const vector unsigned short *);
8999 vector unsigned short vec_ld (int, const unsigned short *);
9000 vector bool char vec_ld (int, const vector bool char *);
9001 vector signed char vec_ld (int, const vector signed char *);
9002 vector signed char vec_ld (int, const signed char *);
9003 vector unsigned char vec_ld (int, const vector unsigned char *);
9004 vector unsigned char vec_ld (int, const unsigned char *);
9006 vector signed char vec_lde (int, const signed char *);
9007 vector unsigned char vec_lde (int, const unsigned char *);
9008 vector signed short vec_lde (int, const short *);
9009 vector unsigned short vec_lde (int, const unsigned short *);
9010 vector float vec_lde (int, const float *);
9011 vector signed int vec_lde (int, const int *);
9012 vector unsigned int vec_lde (int, const unsigned int *);
9013 vector signed int vec_lde (int, const long *);
9014 vector unsigned int vec_lde (int, const unsigned long *);
9016 vector float vec_lvewx (int, float *);
9017 vector signed int vec_lvewx (int, int *);
9018 vector unsigned int vec_lvewx (int, unsigned int *);
9019 vector signed int vec_lvewx (int, long *);
9020 vector unsigned int vec_lvewx (int, unsigned long *);
9022 vector signed short vec_lvehx (int, short *);
9023 vector unsigned short vec_lvehx (int, unsigned short *);
9025 vector signed char vec_lvebx (int, char *);
9026 vector unsigned char vec_lvebx (int, unsigned char *);
9028 vector float vec_ldl (int, const vector float *);
9029 vector float vec_ldl (int, const float *);
9030 vector bool int vec_ldl (int, const vector bool int *);
9031 vector signed int vec_ldl (int, const vector signed int *);
9032 vector signed int vec_ldl (int, const int *);
9033 vector signed int vec_ldl (int, const long *);
9034 vector unsigned int vec_ldl (int, const vector unsigned int *);
9035 vector unsigned int vec_ldl (int, const unsigned int *);
9036 vector unsigned int vec_ldl (int, const unsigned long *);
9037 vector bool short vec_ldl (int, const vector bool short *);
9038 vector pixel vec_ldl (int, const vector pixel *);
9039 vector signed short vec_ldl (int, const vector signed short *);
9040 vector signed short vec_ldl (int, const short *);
9041 vector unsigned short vec_ldl (int, const vector unsigned short *);
9042 vector unsigned short vec_ldl (int, const unsigned short *);
9043 vector bool char vec_ldl (int, const vector bool char *);
9044 vector signed char vec_ldl (int, const vector signed char *);
9045 vector signed char vec_ldl (int, const signed char *);
9046 vector unsigned char vec_ldl (int, const vector unsigned char *);
9047 vector unsigned char vec_ldl (int, const unsigned char *);
9049 vector float vec_loge (vector float);
9051 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
9052 vector unsigned char vec_lvsl (int, const volatile signed char *);
9053 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
9054 vector unsigned char vec_lvsl (int, const volatile short *);
9055 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
9056 vector unsigned char vec_lvsl (int, const volatile int *);
9057 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
9058 vector unsigned char vec_lvsl (int, const volatile long *);
9059 vector unsigned char vec_lvsl (int, const volatile float *);
9061 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
9062 vector unsigned char vec_lvsr (int, const volatile signed char *);
9063 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
9064 vector unsigned char vec_lvsr (int, const volatile short *);
9065 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
9066 vector unsigned char vec_lvsr (int, const volatile int *);
9067 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
9068 vector unsigned char vec_lvsr (int, const volatile long *);
9069 vector unsigned char vec_lvsr (int, const volatile float *);
9071 vector float vec_madd (vector float, vector float, vector float);
9073 vector signed short vec_madds (vector signed short,
9074 vector signed short,
9075 vector signed short);
9077 vector unsigned char vec_max (vector bool char, vector unsigned char);
9078 vector unsigned char vec_max (vector unsigned char, vector bool char);
9079 vector unsigned char vec_max (vector unsigned char,
9080 vector unsigned char);
9081 vector signed char vec_max (vector bool char, vector signed char);
9082 vector signed char vec_max (vector signed char, vector bool char);
9083 vector signed char vec_max (vector signed char, vector signed char);
9084 vector unsigned short vec_max (vector bool short,
9085 vector unsigned short);
9086 vector unsigned short vec_max (vector unsigned short,
9088 vector unsigned short vec_max (vector unsigned short,
9089 vector unsigned short);
9090 vector signed short vec_max (vector bool short, vector signed short);
9091 vector signed short vec_max (vector signed short, vector bool short);
9092 vector signed short vec_max (vector signed short, vector signed short);
9093 vector unsigned int vec_max (vector bool int, vector unsigned int);
9094 vector unsigned int vec_max (vector unsigned int, vector bool int);
9095 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
9096 vector signed int vec_max (vector bool int, vector signed int);
9097 vector signed int vec_max (vector signed int, vector bool int);
9098 vector signed int vec_max (vector signed int, vector signed int);
9099 vector float vec_max (vector float, vector float);
9101 vector float vec_vmaxfp (vector float, vector float);
9103 vector signed int vec_vmaxsw (vector bool int, vector signed int);
9104 vector signed int vec_vmaxsw (vector signed int, vector bool int);
9105 vector signed int vec_vmaxsw (vector signed int, vector signed int);
9107 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
9108 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
9109 vector unsigned int vec_vmaxuw (vector unsigned int,
9110 vector unsigned int);
9112 vector signed short vec_vmaxsh (vector bool short, vector signed short);
9113 vector signed short vec_vmaxsh (vector signed short, vector bool short);
9114 vector signed short vec_vmaxsh (vector signed short,
9115 vector signed short);
9117 vector unsigned short vec_vmaxuh (vector bool short,
9118 vector unsigned short);
9119 vector unsigned short vec_vmaxuh (vector unsigned short,
9121 vector unsigned short vec_vmaxuh (vector unsigned short,
9122 vector unsigned short);
9124 vector signed char vec_vmaxsb (vector bool char, vector signed char);
9125 vector signed char vec_vmaxsb (vector signed char, vector bool char);
9126 vector signed char vec_vmaxsb (vector signed char, vector signed char);
9128 vector unsigned char vec_vmaxub (vector bool char,
9129 vector unsigned char);
9130 vector unsigned char vec_vmaxub (vector unsigned char,
9132 vector unsigned char vec_vmaxub (vector unsigned char,
9133 vector unsigned char);
9135 vector bool char vec_mergeh (vector bool char, vector bool char);
9136 vector signed char vec_mergeh (vector signed char, vector signed char);
9137 vector unsigned char vec_mergeh (vector unsigned char,
9138 vector unsigned char);
9139 vector bool short vec_mergeh (vector bool short, vector bool short);
9140 vector pixel vec_mergeh (vector pixel, vector pixel);
9141 vector signed short vec_mergeh (vector signed short,
9142 vector signed short);
9143 vector unsigned short vec_mergeh (vector unsigned short,
9144 vector unsigned short);
9145 vector float vec_mergeh (vector float, vector float);
9146 vector bool int vec_mergeh (vector bool int, vector bool int);
9147 vector signed int vec_mergeh (vector signed int, vector signed int);
9148 vector unsigned int vec_mergeh (vector unsigned int,
9149 vector unsigned int);
9151 vector float vec_vmrghw (vector float, vector float);
9152 vector bool int vec_vmrghw (vector bool int, vector bool int);
9153 vector signed int vec_vmrghw (vector signed int, vector signed int);
9154 vector unsigned int vec_vmrghw (vector unsigned int,
9155 vector unsigned int);
9157 vector bool short vec_vmrghh (vector bool short, vector bool short);
9158 vector signed short vec_vmrghh (vector signed short,
9159 vector signed short);
9160 vector unsigned short vec_vmrghh (vector unsigned short,
9161 vector unsigned short);
9162 vector pixel vec_vmrghh (vector pixel, vector pixel);
9164 vector bool char vec_vmrghb (vector bool char, vector bool char);
9165 vector signed char vec_vmrghb (vector signed char, vector signed char);
9166 vector unsigned char vec_vmrghb (vector unsigned char,
9167 vector unsigned char);
9169 vector bool char vec_mergel (vector bool char, vector bool char);
9170 vector signed char vec_mergel (vector signed char, vector signed char);
9171 vector unsigned char vec_mergel (vector unsigned char,
9172 vector unsigned char);
9173 vector bool short vec_mergel (vector bool short, vector bool short);
9174 vector pixel vec_mergel (vector pixel, vector pixel);
9175 vector signed short vec_mergel (vector signed short,
9176 vector signed short);
9177 vector unsigned short vec_mergel (vector unsigned short,
9178 vector unsigned short);
9179 vector float vec_mergel (vector float, vector float);
9180 vector bool int vec_mergel (vector bool int, vector bool int);
9181 vector signed int vec_mergel (vector signed int, vector signed int);
9182 vector unsigned int vec_mergel (vector unsigned int,
9183 vector unsigned int);
9185 vector float vec_vmrglw (vector float, vector float);
9186 vector signed int vec_vmrglw (vector signed int, vector signed int);
9187 vector unsigned int vec_vmrglw (vector unsigned int,
9188 vector unsigned int);
9189 vector bool int vec_vmrglw (vector bool int, vector bool int);
9191 vector bool short vec_vmrglh (vector bool short, vector bool short);
9192 vector signed short vec_vmrglh (vector signed short,
9193 vector signed short);
9194 vector unsigned short vec_vmrglh (vector unsigned short,
9195 vector unsigned short);
9196 vector pixel vec_vmrglh (vector pixel, vector pixel);
9198 vector bool char vec_vmrglb (vector bool char, vector bool char);
9199 vector signed char vec_vmrglb (vector signed char, vector signed char);
9200 vector unsigned char vec_vmrglb (vector unsigned char,
9201 vector unsigned char);
9203 vector unsigned short vec_mfvscr (void);
9205 vector unsigned char vec_min (vector bool char, vector unsigned char);
9206 vector unsigned char vec_min (vector unsigned char, vector bool char);
9207 vector unsigned char vec_min (vector unsigned char,
9208 vector unsigned char);
9209 vector signed char vec_min (vector bool char, vector signed char);
9210 vector signed char vec_min (vector signed char, vector bool char);
9211 vector signed char vec_min (vector signed char, vector signed char);
9212 vector unsigned short vec_min (vector bool short,
9213 vector unsigned short);
9214 vector unsigned short vec_min (vector unsigned short,
9216 vector unsigned short vec_min (vector unsigned short,
9217 vector unsigned short);
9218 vector signed short vec_min (vector bool short, vector signed short);
9219 vector signed short vec_min (vector signed short, vector bool short);
9220 vector signed short vec_min (vector signed short, vector signed short);
9221 vector unsigned int vec_min (vector bool int, vector unsigned int);
9222 vector unsigned int vec_min (vector unsigned int, vector bool int);
9223 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
9224 vector signed int vec_min (vector bool int, vector signed int);
9225 vector signed int vec_min (vector signed int, vector bool int);
9226 vector signed int vec_min (vector signed int, vector signed int);
9227 vector float vec_min (vector float, vector float);
9229 vector float vec_vminfp (vector float, vector float);
9231 vector signed int vec_vminsw (vector bool int, vector signed int);
9232 vector signed int vec_vminsw (vector signed int, vector bool int);
9233 vector signed int vec_vminsw (vector signed int, vector signed int);
9235 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
9236 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
9237 vector unsigned int vec_vminuw (vector unsigned int,
9238 vector unsigned int);
9240 vector signed short vec_vminsh (vector bool short, vector signed short);
9241 vector signed short vec_vminsh (vector signed short, vector bool short);
9242 vector signed short vec_vminsh (vector signed short,
9243 vector signed short);
9245 vector unsigned short vec_vminuh (vector bool short,
9246 vector unsigned short);
9247 vector unsigned short vec_vminuh (vector unsigned short,
9249 vector unsigned short vec_vminuh (vector unsigned short,
9250 vector unsigned short);
9252 vector signed char vec_vminsb (vector bool char, vector signed char);
9253 vector signed char vec_vminsb (vector signed char, vector bool char);
9254 vector signed char vec_vminsb (vector signed char, vector signed char);
9256 vector unsigned char vec_vminub (vector bool char,
9257 vector unsigned char);
9258 vector unsigned char vec_vminub (vector unsigned char,
9260 vector unsigned char vec_vminub (vector unsigned char,
9261 vector unsigned char);
9263 vector signed short vec_mladd (vector signed short,
9264 vector signed short,
9265 vector signed short);
9266 vector signed short vec_mladd (vector signed short,
9267 vector unsigned short,
9268 vector unsigned short);
9269 vector signed short vec_mladd (vector unsigned short,
9270 vector signed short,
9271 vector signed short);
9272 vector unsigned short vec_mladd (vector unsigned short,
9273 vector unsigned short,
9274 vector unsigned short);
9276 vector signed short vec_mradds (vector signed short,
9277 vector signed short,
9278 vector signed short);
9280 vector unsigned int vec_msum (vector unsigned char,
9281 vector unsigned char,
9282 vector unsigned int);
9283 vector signed int vec_msum (vector signed char,
9284 vector unsigned char,
9286 vector unsigned int vec_msum (vector unsigned short,
9287 vector unsigned short,
9288 vector unsigned int);
9289 vector signed int vec_msum (vector signed short,
9290 vector signed short,
9293 vector signed int vec_vmsumshm (vector signed short,
9294 vector signed short,
9297 vector unsigned int vec_vmsumuhm (vector unsigned short,
9298 vector unsigned short,
9299 vector unsigned int);
9301 vector signed int vec_vmsummbm (vector signed char,
9302 vector unsigned char,
9305 vector unsigned int vec_vmsumubm (vector unsigned char,
9306 vector unsigned char,
9307 vector unsigned int);
9309 vector unsigned int vec_msums (vector unsigned short,
9310 vector unsigned short,
9311 vector unsigned int);
9312 vector signed int vec_msums (vector signed short,
9313 vector signed short,
9316 vector signed int vec_vmsumshs (vector signed short,
9317 vector signed short,
9320 vector unsigned int vec_vmsumuhs (vector unsigned short,
9321 vector unsigned short,
9322 vector unsigned int);
9324 void vec_mtvscr (vector signed int);
9325 void vec_mtvscr (vector unsigned int);
9326 void vec_mtvscr (vector bool int);
9327 void vec_mtvscr (vector signed short);
9328 void vec_mtvscr (vector unsigned short);
9329 void vec_mtvscr (vector bool short);
9330 void vec_mtvscr (vector pixel);
9331 void vec_mtvscr (vector signed char);
9332 void vec_mtvscr (vector unsigned char);
9333 void vec_mtvscr (vector bool char);
9335 vector unsigned short vec_mule (vector unsigned char,
9336 vector unsigned char);
9337 vector signed short vec_mule (vector signed char,
9338 vector signed char);
9339 vector unsigned int vec_mule (vector unsigned short,
9340 vector unsigned short);
9341 vector signed int vec_mule (vector signed short, vector signed short);
9343 vector signed int vec_vmulesh (vector signed short,
9344 vector signed short);
9346 vector unsigned int vec_vmuleuh (vector unsigned short,
9347 vector unsigned short);
9349 vector signed short vec_vmulesb (vector signed char,
9350 vector signed char);
9352 vector unsigned short vec_vmuleub (vector unsigned char,
9353 vector unsigned char);
9355 vector unsigned short vec_mulo (vector unsigned char,
9356 vector unsigned char);
9357 vector signed short vec_mulo (vector signed char, vector signed char);
9358 vector unsigned int vec_mulo (vector unsigned short,
9359 vector unsigned short);
9360 vector signed int vec_mulo (vector signed short, vector signed short);
9362 vector signed int vec_vmulosh (vector signed short,
9363 vector signed short);
9365 vector unsigned int vec_vmulouh (vector unsigned short,
9366 vector unsigned short);
9368 vector signed short vec_vmulosb (vector signed char,
9369 vector signed char);
9371 vector unsigned short vec_vmuloub (vector unsigned char,
9372 vector unsigned char);
9374 vector float vec_nmsub (vector float, vector float, vector float);
9376 vector float vec_nor (vector float, vector float);
9377 vector signed int vec_nor (vector signed int, vector signed int);
9378 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
9379 vector bool int vec_nor (vector bool int, vector bool int);
9380 vector signed short vec_nor (vector signed short, vector signed short);
9381 vector unsigned short vec_nor (vector unsigned short,
9382 vector unsigned short);
9383 vector bool short vec_nor (vector bool short, vector bool short);
9384 vector signed char vec_nor (vector signed char, vector signed char);
9385 vector unsigned char vec_nor (vector unsigned char,
9386 vector unsigned char);
9387 vector bool char vec_nor (vector bool char, vector bool char);
9389 vector float vec_or (vector float, vector float);
9390 vector float vec_or (vector float, vector bool int);
9391 vector float vec_or (vector bool int, vector float);
9392 vector bool int vec_or (vector bool int, vector bool int);
9393 vector signed int vec_or (vector bool int, vector signed int);
9394 vector signed int vec_or (vector signed int, vector bool int);
9395 vector signed int vec_or (vector signed int, vector signed int);
9396 vector unsigned int vec_or (vector bool int, vector unsigned int);
9397 vector unsigned int vec_or (vector unsigned int, vector bool int);
9398 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
9399 vector bool short vec_or (vector bool short, vector bool short);
9400 vector signed short vec_or (vector bool short, vector signed short);
9401 vector signed short vec_or (vector signed short, vector bool short);
9402 vector signed short vec_or (vector signed short, vector signed short);
9403 vector unsigned short vec_or (vector bool short, vector unsigned short);
9404 vector unsigned short vec_or (vector unsigned short, vector bool short);
9405 vector unsigned short vec_or (vector unsigned short,
9406 vector unsigned short);
9407 vector signed char vec_or (vector bool char, vector signed char);
9408 vector bool char vec_or (vector bool char, vector bool char);
9409 vector signed char vec_or (vector signed char, vector bool char);
9410 vector signed char vec_or (vector signed char, vector signed char);
9411 vector unsigned char vec_or (vector bool char, vector unsigned char);
9412 vector unsigned char vec_or (vector unsigned char, vector bool char);
9413 vector unsigned char vec_or (vector unsigned char,
9414 vector unsigned char);
9416 vector signed char vec_pack (vector signed short, vector signed short);
9417 vector unsigned char vec_pack (vector unsigned short,
9418 vector unsigned short);
9419 vector bool char vec_pack (vector bool short, vector bool short);
9420 vector signed short vec_pack (vector signed int, vector signed int);
9421 vector unsigned short vec_pack (vector unsigned int,
9422 vector unsigned int);
9423 vector bool short vec_pack (vector bool int, vector bool int);
9425 vector bool short vec_vpkuwum (vector bool int, vector bool int);
9426 vector signed short vec_vpkuwum (vector signed int, vector signed int);
9427 vector unsigned short vec_vpkuwum (vector unsigned int,
9428 vector unsigned int);
9430 vector bool char vec_vpkuhum (vector bool short, vector bool short);
9431 vector signed char vec_vpkuhum (vector signed short,
9432 vector signed short);
9433 vector unsigned char vec_vpkuhum (vector unsigned short,
9434 vector unsigned short);
9436 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
9438 vector unsigned char vec_packs (vector unsigned short,
9439 vector unsigned short);
9440 vector signed char vec_packs (vector signed short, vector signed short);
9441 vector unsigned short vec_packs (vector unsigned int,
9442 vector unsigned int);
9443 vector signed short vec_packs (vector signed int, vector signed int);
9445 vector signed short vec_vpkswss (vector signed int, vector signed int);
9447 vector unsigned short vec_vpkuwus (vector unsigned int,
9448 vector unsigned int);
9450 vector signed char vec_vpkshss (vector signed short,
9451 vector signed short);
9453 vector unsigned char vec_vpkuhus (vector unsigned short,
9454 vector unsigned short);
9456 vector unsigned char vec_packsu (vector unsigned short,
9457 vector unsigned short);
9458 vector unsigned char vec_packsu (vector signed short,
9459 vector signed short);
9460 vector unsigned short vec_packsu (vector unsigned int,
9461 vector unsigned int);
9462 vector unsigned short vec_packsu (vector signed int, vector signed int);
9464 vector unsigned short vec_vpkswus (vector signed int,
9467 vector unsigned char vec_vpkshus (vector signed short,
9468 vector signed short);
9470 vector float vec_perm (vector float,
9472 vector unsigned char);
9473 vector signed int vec_perm (vector signed int,
9475 vector unsigned char);
9476 vector unsigned int vec_perm (vector unsigned int,
9477 vector unsigned int,
9478 vector unsigned char);
9479 vector bool int vec_perm (vector bool int,
9481 vector unsigned char);
9482 vector signed short vec_perm (vector signed short,
9483 vector signed short,
9484 vector unsigned char);
9485 vector unsigned short vec_perm (vector unsigned short,
9486 vector unsigned short,
9487 vector unsigned char);
9488 vector bool short vec_perm (vector bool short,
9490 vector unsigned char);
9491 vector pixel vec_perm (vector pixel,
9493 vector unsigned char);
9494 vector signed char vec_perm (vector signed char,
9496 vector unsigned char);
9497 vector unsigned char vec_perm (vector unsigned char,
9498 vector unsigned char,
9499 vector unsigned char);
9500 vector bool char vec_perm (vector bool char,
9502 vector unsigned char);
9504 vector float vec_re (vector float);
9506 vector signed char vec_rl (vector signed char,
9507 vector unsigned char);
9508 vector unsigned char vec_rl (vector unsigned char,
9509 vector unsigned char);
9510 vector signed short vec_rl (vector signed short, vector unsigned short);
9511 vector unsigned short vec_rl (vector unsigned short,
9512 vector unsigned short);
9513 vector signed int vec_rl (vector signed int, vector unsigned int);
9514 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
9516 vector signed int vec_vrlw (vector signed int, vector unsigned int);
9517 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
9519 vector signed short vec_vrlh (vector signed short,
9520 vector unsigned short);
9521 vector unsigned short vec_vrlh (vector unsigned short,
9522 vector unsigned short);
9524 vector signed char vec_vrlb (vector signed char, vector unsigned char);
9525 vector unsigned char vec_vrlb (vector unsigned char,
9526 vector unsigned char);
9528 vector float vec_round (vector float);
9530 vector float vec_rsqrte (vector float);
9532 vector float vec_sel (vector float, vector float, vector bool int);
9533 vector float vec_sel (vector float, vector float, vector unsigned int);
9534 vector signed int vec_sel (vector signed int,
9537 vector signed int vec_sel (vector signed int,
9539 vector unsigned int);
9540 vector unsigned int vec_sel (vector unsigned int,
9541 vector unsigned int,
9543 vector unsigned int vec_sel (vector unsigned int,
9544 vector unsigned int,
9545 vector unsigned int);
9546 vector bool int vec_sel (vector bool int,
9549 vector bool int vec_sel (vector bool int,
9551 vector unsigned int);
9552 vector signed short vec_sel (vector signed short,
9553 vector signed short,
9555 vector signed short vec_sel (vector signed short,
9556 vector signed short,
9557 vector unsigned short);
9558 vector unsigned short vec_sel (vector unsigned short,
9559 vector unsigned short,
9561 vector unsigned short vec_sel (vector unsigned short,
9562 vector unsigned short,
9563 vector unsigned short);
9564 vector bool short vec_sel (vector bool short,
9567 vector bool short vec_sel (vector bool short,
9569 vector unsigned short);
9570 vector signed char vec_sel (vector signed char,
9573 vector signed char vec_sel (vector signed char,
9575 vector unsigned char);
9576 vector unsigned char vec_sel (vector unsigned char,
9577 vector unsigned char,
9579 vector unsigned char vec_sel (vector unsigned char,
9580 vector unsigned char,
9581 vector unsigned char);
9582 vector bool char vec_sel (vector bool char,
9585 vector bool char vec_sel (vector bool char,
9587 vector unsigned char);
9589 vector signed char vec_sl (vector signed char,
9590 vector unsigned char);
9591 vector unsigned char vec_sl (vector unsigned char,
9592 vector unsigned char);
9593 vector signed short vec_sl (vector signed short, vector unsigned short);
9594 vector unsigned short vec_sl (vector unsigned short,
9595 vector unsigned short);
9596 vector signed int vec_sl (vector signed int, vector unsigned int);
9597 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
9599 vector signed int vec_vslw (vector signed int, vector unsigned int);
9600 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
9602 vector signed short vec_vslh (vector signed short,
9603 vector unsigned short);
9604 vector unsigned short vec_vslh (vector unsigned short,
9605 vector unsigned short);
9607 vector signed char vec_vslb (vector signed char, vector unsigned char);
9608 vector unsigned char vec_vslb (vector unsigned char,
9609 vector unsigned char);
9611 vector float vec_sld (vector float, vector float, const int);
9612 vector signed int vec_sld (vector signed int,
9615 vector unsigned int vec_sld (vector unsigned int,
9616 vector unsigned int,
9618 vector bool int vec_sld (vector bool int,
9621 vector signed short vec_sld (vector signed short,
9622 vector signed short,
9624 vector unsigned short vec_sld (vector unsigned short,
9625 vector unsigned short,
9627 vector bool short vec_sld (vector bool short,
9630 vector pixel vec_sld (vector pixel,
9633 vector signed char vec_sld (vector signed char,
9636 vector unsigned char vec_sld (vector unsigned char,
9637 vector unsigned char,
9639 vector bool char vec_sld (vector bool char,
9643 vector signed int vec_sll (vector signed int,
9644 vector unsigned int);
9645 vector signed int vec_sll (vector signed int,
9646 vector unsigned short);
9647 vector signed int vec_sll (vector signed int,
9648 vector unsigned char);
9649 vector unsigned int vec_sll (vector unsigned int,
9650 vector unsigned int);
9651 vector unsigned int vec_sll (vector unsigned int,
9652 vector unsigned short);
9653 vector unsigned int vec_sll (vector unsigned int,
9654 vector unsigned char);
9655 vector bool int vec_sll (vector bool int,
9656 vector unsigned int);
9657 vector bool int vec_sll (vector bool int,
9658 vector unsigned short);
9659 vector bool int vec_sll (vector bool int,
9660 vector unsigned char);
9661 vector signed short vec_sll (vector signed short,
9662 vector unsigned int);
9663 vector signed short vec_sll (vector signed short,
9664 vector unsigned short);
9665 vector signed short vec_sll (vector signed short,
9666 vector unsigned char);
9667 vector unsigned short vec_sll (vector unsigned short,
9668 vector unsigned int);
9669 vector unsigned short vec_sll (vector unsigned short,
9670 vector unsigned short);
9671 vector unsigned short vec_sll (vector unsigned short,
9672 vector unsigned char);
9673 vector bool short vec_sll (vector bool short, vector unsigned int);
9674 vector bool short vec_sll (vector bool short, vector unsigned short);
9675 vector bool short vec_sll (vector bool short, vector unsigned char);
9676 vector pixel vec_sll (vector pixel, vector unsigned int);
9677 vector pixel vec_sll (vector pixel, vector unsigned short);
9678 vector pixel vec_sll (vector pixel, vector unsigned char);
9679 vector signed char vec_sll (vector signed char, vector unsigned int);
9680 vector signed char vec_sll (vector signed char, vector unsigned short);
9681 vector signed char vec_sll (vector signed char, vector unsigned char);
9682 vector unsigned char vec_sll (vector unsigned char,
9683 vector unsigned int);
9684 vector unsigned char vec_sll (vector unsigned char,
9685 vector unsigned short);
9686 vector unsigned char vec_sll (vector unsigned char,
9687 vector unsigned char);
9688 vector bool char vec_sll (vector bool char, vector unsigned int);
9689 vector bool char vec_sll (vector bool char, vector unsigned short);
9690 vector bool char vec_sll (vector bool char, vector unsigned char);
9692 vector float vec_slo (vector float, vector signed char);
9693 vector float vec_slo (vector float, vector unsigned char);
9694 vector signed int vec_slo (vector signed int, vector signed char);
9695 vector signed int vec_slo (vector signed int, vector unsigned char);
9696 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9697 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9698 vector signed short vec_slo (vector signed short, vector signed char);
9699 vector signed short vec_slo (vector signed short, vector unsigned char);
9700 vector unsigned short vec_slo (vector unsigned short,
9701 vector signed char);
9702 vector unsigned short vec_slo (vector unsigned short,
9703 vector unsigned char);
9704 vector pixel vec_slo (vector pixel, vector signed char);
9705 vector pixel vec_slo (vector pixel, vector unsigned char);
9706 vector signed char vec_slo (vector signed char, vector signed char);
9707 vector signed char vec_slo (vector signed char, vector unsigned char);
9708 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9709 vector unsigned char vec_slo (vector unsigned char,
9710 vector unsigned char);
9712 vector signed char vec_splat (vector signed char, const int);
9713 vector unsigned char vec_splat (vector unsigned char, const int);
9714 vector bool char vec_splat (vector bool char, const int);
9715 vector signed short vec_splat (vector signed short, const int);
9716 vector unsigned short vec_splat (vector unsigned short, const int);
9717 vector bool short vec_splat (vector bool short, const int);
9718 vector pixel vec_splat (vector pixel, const int);
9719 vector float vec_splat (vector float, const int);
9720 vector signed int vec_splat (vector signed int, const int);
9721 vector unsigned int vec_splat (vector unsigned int, const int);
9722 vector bool int vec_splat (vector bool int, const int);
9724 vector float vec_vspltw (vector float, const int);
9725 vector signed int vec_vspltw (vector signed int, const int);
9726 vector unsigned int vec_vspltw (vector unsigned int, const int);
9727 vector bool int vec_vspltw (vector bool int, const int);
9729 vector bool short vec_vsplth (vector bool short, const int);
9730 vector signed short vec_vsplth (vector signed short, const int);
9731 vector unsigned short vec_vsplth (vector unsigned short, const int);
9732 vector pixel vec_vsplth (vector pixel, const int);
9734 vector signed char vec_vspltb (vector signed char, const int);
9735 vector unsigned char vec_vspltb (vector unsigned char, const int);
9736 vector bool char vec_vspltb (vector bool char, const int);
9738 vector signed char vec_splat_s8 (const int);
9740 vector signed short vec_splat_s16 (const int);
9742 vector signed int vec_splat_s32 (const int);
9744 vector unsigned char vec_splat_u8 (const int);
9746 vector unsigned short vec_splat_u16 (const int);
9748 vector unsigned int vec_splat_u32 (const int);
9750 vector signed char vec_sr (vector signed char, vector unsigned char);
9751 vector unsigned char vec_sr (vector unsigned char,
9752 vector unsigned char);
9753 vector signed short vec_sr (vector signed short,
9754 vector unsigned short);
9755 vector unsigned short vec_sr (vector unsigned short,
9756 vector unsigned short);
9757 vector signed int vec_sr (vector signed int, vector unsigned int);
9758 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9760 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9761 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9763 vector signed short vec_vsrh (vector signed short,
9764 vector unsigned short);
9765 vector unsigned short vec_vsrh (vector unsigned short,
9766 vector unsigned short);
9768 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9769 vector unsigned char vec_vsrb (vector unsigned char,
9770 vector unsigned char);
9772 vector signed char vec_sra (vector signed char, vector unsigned char);
9773 vector unsigned char vec_sra (vector unsigned char,
9774 vector unsigned char);
9775 vector signed short vec_sra (vector signed short,
9776 vector unsigned short);
9777 vector unsigned short vec_sra (vector unsigned short,
9778 vector unsigned short);
9779 vector signed int vec_sra (vector signed int, vector unsigned int);
9780 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9782 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9783 vector unsigned int vec_vsraw (vector unsigned int,
9784 vector unsigned int);
9786 vector signed short vec_vsrah (vector signed short,
9787 vector unsigned short);
9788 vector unsigned short vec_vsrah (vector unsigned short,
9789 vector unsigned short);
9791 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9792 vector unsigned char vec_vsrab (vector unsigned char,
9793 vector unsigned char);
9795 vector signed int vec_srl (vector signed int, vector unsigned int);
9796 vector signed int vec_srl (vector signed int, vector unsigned short);
9797 vector signed int vec_srl (vector signed int, vector unsigned char);
9798 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9799 vector unsigned int vec_srl (vector unsigned int,
9800 vector unsigned short);
9801 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9802 vector bool int vec_srl (vector bool int, vector unsigned int);
9803 vector bool int vec_srl (vector bool int, vector unsigned short);
9804 vector bool int vec_srl (vector bool int, vector unsigned char);
9805 vector signed short vec_srl (vector signed short, vector unsigned int);
9806 vector signed short vec_srl (vector signed short,
9807 vector unsigned short);
9808 vector signed short vec_srl (vector signed short, vector unsigned char);
9809 vector unsigned short vec_srl (vector unsigned short,
9810 vector unsigned int);
9811 vector unsigned short vec_srl (vector unsigned short,
9812 vector unsigned short);
9813 vector unsigned short vec_srl (vector unsigned short,
9814 vector unsigned char);
9815 vector bool short vec_srl (vector bool short, vector unsigned int);
9816 vector bool short vec_srl (vector bool short, vector unsigned short);
9817 vector bool short vec_srl (vector bool short, vector unsigned char);
9818 vector pixel vec_srl (vector pixel, vector unsigned int);
9819 vector pixel vec_srl (vector pixel, vector unsigned short);
9820 vector pixel vec_srl (vector pixel, vector unsigned char);
9821 vector signed char vec_srl (vector signed char, vector unsigned int);
9822 vector signed char vec_srl (vector signed char, vector unsigned short);
9823 vector signed char vec_srl (vector signed char, vector unsigned char);
9824 vector unsigned char vec_srl (vector unsigned char,
9825 vector unsigned int);
9826 vector unsigned char vec_srl (vector unsigned char,
9827 vector unsigned short);
9828 vector unsigned char vec_srl (vector unsigned char,
9829 vector unsigned char);
9830 vector bool char vec_srl (vector bool char, vector unsigned int);
9831 vector bool char vec_srl (vector bool char, vector unsigned short);
9832 vector bool char vec_srl (vector bool char, vector unsigned char);
9834 vector float vec_sro (vector float, vector signed char);
9835 vector float vec_sro (vector float, vector unsigned char);
9836 vector signed int vec_sro (vector signed int, vector signed char);
9837 vector signed int vec_sro (vector signed int, vector unsigned char);
9838 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9839 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9840 vector signed short vec_sro (vector signed short, vector signed char);
9841 vector signed short vec_sro (vector signed short, vector unsigned char);
9842 vector unsigned short vec_sro (vector unsigned short,
9843 vector signed char);
9844 vector unsigned short vec_sro (vector unsigned short,
9845 vector unsigned char);
9846 vector pixel vec_sro (vector pixel, vector signed char);
9847 vector pixel vec_sro (vector pixel, vector unsigned char);
9848 vector signed char vec_sro (vector signed char, vector signed char);
9849 vector signed char vec_sro (vector signed char, vector unsigned char);
9850 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9851 vector unsigned char vec_sro (vector unsigned char,
9852 vector unsigned char);
9854 void vec_st (vector float, int, vector float *);
9855 void vec_st (vector float, int, float *);
9856 void vec_st (vector signed int, int, vector signed int *);
9857 void vec_st (vector signed int, int, int *);
9858 void vec_st (vector unsigned int, int, vector unsigned int *);
9859 void vec_st (vector unsigned int, int, unsigned int *);
9860 void vec_st (vector bool int, int, vector bool int *);
9861 void vec_st (vector bool int, int, unsigned int *);
9862 void vec_st (vector bool int, int, int *);
9863 void vec_st (vector signed short, int, vector signed short *);
9864 void vec_st (vector signed short, int, short *);
9865 void vec_st (vector unsigned short, int, vector unsigned short *);
9866 void vec_st (vector unsigned short, int, unsigned short *);
9867 void vec_st (vector bool short, int, vector bool short *);
9868 void vec_st (vector bool short, int, unsigned short *);
9869 void vec_st (vector pixel, int, vector pixel *);
9870 void vec_st (vector pixel, int, unsigned short *);
9871 void vec_st (vector pixel, int, short *);
9872 void vec_st (vector bool short, int, short *);
9873 void vec_st (vector signed char, int, vector signed char *);
9874 void vec_st (vector signed char, int, signed char *);
9875 void vec_st (vector unsigned char, int, vector unsigned char *);
9876 void vec_st (vector unsigned char, int, unsigned char *);
9877 void vec_st (vector bool char, int, vector bool char *);
9878 void vec_st (vector bool char, int, unsigned char *);
9879 void vec_st (vector bool char, int, signed char *);
9881 void vec_ste (vector signed char, int, signed char *);
9882 void vec_ste (vector unsigned char, int, unsigned char *);
9883 void vec_ste (vector bool char, int, signed char *);
9884 void vec_ste (vector bool char, int, unsigned char *);
9885 void vec_ste (vector signed short, int, short *);
9886 void vec_ste (vector unsigned short, int, unsigned short *);
9887 void vec_ste (vector bool short, int, short *);
9888 void vec_ste (vector bool short, int, unsigned short *);
9889 void vec_ste (vector pixel, int, short *);
9890 void vec_ste (vector pixel, int, unsigned short *);
9891 void vec_ste (vector float, int, float *);
9892 void vec_ste (vector signed int, int, int *);
9893 void vec_ste (vector unsigned int, int, unsigned int *);
9894 void vec_ste (vector bool int, int, int *);
9895 void vec_ste (vector bool int, int, unsigned int *);
9897 void vec_stvewx (vector float, int, float *);
9898 void vec_stvewx (vector signed int, int, int *);
9899 void vec_stvewx (vector unsigned int, int, unsigned int *);
9900 void vec_stvewx (vector bool int, int, int *);
9901 void vec_stvewx (vector bool int, int, unsigned int *);
9903 void vec_stvehx (vector signed short, int, short *);
9904 void vec_stvehx (vector unsigned short, int, unsigned short *);
9905 void vec_stvehx (vector bool short, int, short *);
9906 void vec_stvehx (vector bool short, int, unsigned short *);
9907 void vec_stvehx (vector pixel, int, short *);
9908 void vec_stvehx (vector pixel, int, unsigned short *);
9910 void vec_stvebx (vector signed char, int, signed char *);
9911 void vec_stvebx (vector unsigned char, int, unsigned char *);
9912 void vec_stvebx (vector bool char, int, signed char *);
9913 void vec_stvebx (vector bool char, int, unsigned char *);
9915 void vec_stl (vector float, int, vector float *);
9916 void vec_stl (vector float, int, float *);
9917 void vec_stl (vector signed int, int, vector signed int *);
9918 void vec_stl (vector signed int, int, int *);
9919 void vec_stl (vector unsigned int, int, vector unsigned int *);
9920 void vec_stl (vector unsigned int, int, unsigned int *);
9921 void vec_stl (vector bool int, int, vector bool int *);
9922 void vec_stl (vector bool int, int, unsigned int *);
9923 void vec_stl (vector bool int, int, int *);
9924 void vec_stl (vector signed short, int, vector signed short *);
9925 void vec_stl (vector signed short, int, short *);
9926 void vec_stl (vector unsigned short, int, vector unsigned short *);
9927 void vec_stl (vector unsigned short, int, unsigned short *);
9928 void vec_stl (vector bool short, int, vector bool short *);
9929 void vec_stl (vector bool short, int, unsigned short *);
9930 void vec_stl (vector bool short, int, short *);
9931 void vec_stl (vector pixel, int, vector pixel *);
9932 void vec_stl (vector pixel, int, unsigned short *);
9933 void vec_stl (vector pixel, int, short *);
9934 void vec_stl (vector signed char, int, vector signed char *);
9935 void vec_stl (vector signed char, int, signed char *);
9936 void vec_stl (vector unsigned char, int, vector unsigned char *);
9937 void vec_stl (vector unsigned char, int, unsigned char *);
9938 void vec_stl (vector bool char, int, vector bool char *);
9939 void vec_stl (vector bool char, int, unsigned char *);
9940 void vec_stl (vector bool char, int, signed char *);
9942 vector signed char vec_sub (vector bool char, vector signed char);
9943 vector signed char vec_sub (vector signed char, vector bool char);
9944 vector signed char vec_sub (vector signed char, vector signed char);
9945 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9946 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9947 vector unsigned char vec_sub (vector unsigned char,
9948 vector unsigned char);
9949 vector signed short vec_sub (vector bool short, vector signed short);
9950 vector signed short vec_sub (vector signed short, vector bool short);
9951 vector signed short vec_sub (vector signed short, vector signed short);
9952 vector unsigned short vec_sub (vector bool short,
9953 vector unsigned short);
9954 vector unsigned short vec_sub (vector unsigned short,
9956 vector unsigned short vec_sub (vector unsigned short,
9957 vector unsigned short);
9958 vector signed int vec_sub (vector bool int, vector signed int);
9959 vector signed int vec_sub (vector signed int, vector bool int);
9960 vector signed int vec_sub (vector signed int, vector signed int);
9961 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9962 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9963 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9964 vector float vec_sub (vector float, vector float);
9966 vector float vec_vsubfp (vector float, vector float);
9968 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9969 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9970 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9971 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9972 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9973 vector unsigned int vec_vsubuwm (vector unsigned int,
9974 vector unsigned int);
9976 vector signed short vec_vsubuhm (vector bool short,
9977 vector signed short);
9978 vector signed short vec_vsubuhm (vector signed short,
9980 vector signed short vec_vsubuhm (vector signed short,
9981 vector signed short);
9982 vector unsigned short vec_vsubuhm (vector bool short,
9983 vector unsigned short);
9984 vector unsigned short vec_vsubuhm (vector unsigned short,
9986 vector unsigned short vec_vsubuhm (vector unsigned short,
9987 vector unsigned short);
9989 vector signed char vec_vsububm (vector bool char, vector signed char);
9990 vector signed char vec_vsububm (vector signed char, vector bool char);
9991 vector signed char vec_vsububm (vector signed char, vector signed char);
9992 vector unsigned char vec_vsububm (vector bool char,
9993 vector unsigned char);
9994 vector unsigned char vec_vsububm (vector unsigned char,
9996 vector unsigned char vec_vsububm (vector unsigned char,
9997 vector unsigned char);
9999 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
10001 vector unsigned char vec_subs (vector bool char, vector unsigned char);
10002 vector unsigned char vec_subs (vector unsigned char, vector bool char);
10003 vector unsigned char vec_subs (vector unsigned char,
10004 vector unsigned char);
10005 vector signed char vec_subs (vector bool char, vector signed char);
10006 vector signed char vec_subs (vector signed char, vector bool char);
10007 vector signed char vec_subs (vector signed char, vector signed char);
10008 vector unsigned short vec_subs (vector bool short,
10009 vector unsigned short);
10010 vector unsigned short vec_subs (vector unsigned short,
10011 vector bool short);
10012 vector unsigned short vec_subs (vector unsigned short,
10013 vector unsigned short);
10014 vector signed short vec_subs (vector bool short, vector signed short);
10015 vector signed short vec_subs (vector signed short, vector bool short);
10016 vector signed short vec_subs (vector signed short, vector signed short);
10017 vector unsigned int vec_subs (vector bool int, vector unsigned int);
10018 vector unsigned int vec_subs (vector unsigned int, vector bool int);
10019 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
10020 vector signed int vec_subs (vector bool int, vector signed int);
10021 vector signed int vec_subs (vector signed int, vector bool int);
10022 vector signed int vec_subs (vector signed int, vector signed int);
10024 vector signed int vec_vsubsws (vector bool int, vector signed int);
10025 vector signed int vec_vsubsws (vector signed int, vector bool int);
10026 vector signed int vec_vsubsws (vector signed int, vector signed int);
10028 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
10029 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
10030 vector unsigned int vec_vsubuws (vector unsigned int,
10031 vector unsigned int);
10033 vector signed short vec_vsubshs (vector bool short,
10034 vector signed short);
10035 vector signed short vec_vsubshs (vector signed short,
10036 vector bool short);
10037 vector signed short vec_vsubshs (vector signed short,
10038 vector signed short);
10040 vector unsigned short vec_vsubuhs (vector bool short,
10041 vector unsigned short);
10042 vector unsigned short vec_vsubuhs (vector unsigned short,
10043 vector bool short);
10044 vector unsigned short vec_vsubuhs (vector unsigned short,
10045 vector unsigned short);
10047 vector signed char vec_vsubsbs (vector bool char, vector signed char);
10048 vector signed char vec_vsubsbs (vector signed char, vector bool char);
10049 vector signed char vec_vsubsbs (vector signed char, vector signed char);
10051 vector unsigned char vec_vsububs (vector bool char,
10052 vector unsigned char);
10053 vector unsigned char vec_vsububs (vector unsigned char,
10055 vector unsigned char vec_vsububs (vector unsigned char,
10056 vector unsigned char);
10058 vector unsigned int vec_sum4s (vector unsigned char,
10059 vector unsigned int);
10060 vector signed int vec_sum4s (vector signed char, vector signed int);
10061 vector signed int vec_sum4s (vector signed short, vector signed int);
10063 vector signed int vec_vsum4shs (vector signed short, vector signed int);
10065 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
10067 vector unsigned int vec_vsum4ubs (vector unsigned char,
10068 vector unsigned int);
10070 vector signed int vec_sum2s (vector signed int, vector signed int);
10072 vector signed int vec_sums (vector signed int, vector signed int);
10074 vector float vec_trunc (vector float);
10076 vector signed short vec_unpackh (vector signed char);
10077 vector bool short vec_unpackh (vector bool char);
10078 vector signed int vec_unpackh (vector signed short);
10079 vector bool int vec_unpackh (vector bool short);
10080 vector unsigned int vec_unpackh (vector pixel);
10082 vector bool int vec_vupkhsh (vector bool short);
10083 vector signed int vec_vupkhsh (vector signed short);
10085 vector unsigned int vec_vupkhpx (vector pixel);
10087 vector bool short vec_vupkhsb (vector bool char);
10088 vector signed short vec_vupkhsb (vector signed char);
10090 vector signed short vec_unpackl (vector signed char);
10091 vector bool short vec_unpackl (vector bool char);
10092 vector unsigned int vec_unpackl (vector pixel);
10093 vector signed int vec_unpackl (vector signed short);
10094 vector bool int vec_unpackl (vector bool short);
10096 vector unsigned int vec_vupklpx (vector pixel);
10098 vector bool int vec_vupklsh (vector bool short);
10099 vector signed int vec_vupklsh (vector signed short);
10101 vector bool short vec_vupklsb (vector bool char);
10102 vector signed short vec_vupklsb (vector signed char);
10104 vector float vec_xor (vector float, vector float);
10105 vector float vec_xor (vector float, vector bool int);
10106 vector float vec_xor (vector bool int, vector float);
10107 vector bool int vec_xor (vector bool int, vector bool int);
10108 vector signed int vec_xor (vector bool int, vector signed int);
10109 vector signed int vec_xor (vector signed int, vector bool int);
10110 vector signed int vec_xor (vector signed int, vector signed int);
10111 vector unsigned int vec_xor (vector bool int, vector unsigned int);
10112 vector unsigned int vec_xor (vector unsigned int, vector bool int);
10113 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
10114 vector bool short vec_xor (vector bool short, vector bool short);
10115 vector signed short vec_xor (vector bool short, vector signed short);
10116 vector signed short vec_xor (vector signed short, vector bool short);
10117 vector signed short vec_xor (vector signed short, vector signed short);
10118 vector unsigned short vec_xor (vector bool short,
10119 vector unsigned short);
10120 vector unsigned short vec_xor (vector unsigned short,
10121 vector bool short);
10122 vector unsigned short vec_xor (vector unsigned short,
10123 vector unsigned short);
10124 vector signed char vec_xor (vector bool char, vector signed char);
10125 vector bool char vec_xor (vector bool char, vector bool char);
10126 vector signed char vec_xor (vector signed char, vector bool char);
10127 vector signed char vec_xor (vector signed char, vector signed char);
10128 vector unsigned char vec_xor (vector bool char, vector unsigned char);
10129 vector unsigned char vec_xor (vector unsigned char, vector bool char);
10130 vector unsigned char vec_xor (vector unsigned char,
10131 vector unsigned char);
10133 int vec_all_eq (vector signed char, vector bool char);
10134 int vec_all_eq (vector signed char, vector signed char);
10135 int vec_all_eq (vector unsigned char, vector bool char);
10136 int vec_all_eq (vector unsigned char, vector unsigned char);
10137 int vec_all_eq (vector bool char, vector bool char);
10138 int vec_all_eq (vector bool char, vector unsigned char);
10139 int vec_all_eq (vector bool char, vector signed char);
10140 int vec_all_eq (vector signed short, vector bool short);
10141 int vec_all_eq (vector signed short, vector signed short);
10142 int vec_all_eq (vector unsigned short, vector bool short);
10143 int vec_all_eq (vector unsigned short, vector unsigned short);
10144 int vec_all_eq (vector bool short, vector bool short);
10145 int vec_all_eq (vector bool short, vector unsigned short);
10146 int vec_all_eq (vector bool short, vector signed short);
10147 int vec_all_eq (vector pixel, vector pixel);
10148 int vec_all_eq (vector signed int, vector bool int);
10149 int vec_all_eq (vector signed int, vector signed int);
10150 int vec_all_eq (vector unsigned int, vector bool int);
10151 int vec_all_eq (vector unsigned int, vector unsigned int);
10152 int vec_all_eq (vector bool int, vector bool int);
10153 int vec_all_eq (vector bool int, vector unsigned int);
10154 int vec_all_eq (vector bool int, vector signed int);
10155 int vec_all_eq (vector float, vector float);
10157 int vec_all_ge (vector bool char, vector unsigned char);
10158 int vec_all_ge (vector unsigned char, vector bool char);
10159 int vec_all_ge (vector unsigned char, vector unsigned char);
10160 int vec_all_ge (vector bool char, vector signed char);
10161 int vec_all_ge (vector signed char, vector bool char);
10162 int vec_all_ge (vector signed char, vector signed char);
10163 int vec_all_ge (vector bool short, vector unsigned short);
10164 int vec_all_ge (vector unsigned short, vector bool short);
10165 int vec_all_ge (vector unsigned short, vector unsigned short);
10166 int vec_all_ge (vector signed short, vector signed short);
10167 int vec_all_ge (vector bool short, vector signed short);
10168 int vec_all_ge (vector signed short, vector bool short);
10169 int vec_all_ge (vector bool int, vector unsigned int);
10170 int vec_all_ge (vector unsigned int, vector bool int);
10171 int vec_all_ge (vector unsigned int, vector unsigned int);
10172 int vec_all_ge (vector bool int, vector signed int);
10173 int vec_all_ge (vector signed int, vector bool int);
10174 int vec_all_ge (vector signed int, vector signed int);
10175 int vec_all_ge (vector float, vector float);
10177 int vec_all_gt (vector bool char, vector unsigned char);
10178 int vec_all_gt (vector unsigned char, vector bool char);
10179 int vec_all_gt (vector unsigned char, vector unsigned char);
10180 int vec_all_gt (vector bool char, vector signed char);
10181 int vec_all_gt (vector signed char, vector bool char);
10182 int vec_all_gt (vector signed char, vector signed char);
10183 int vec_all_gt (vector bool short, vector unsigned short);
10184 int vec_all_gt (vector unsigned short, vector bool short);
10185 int vec_all_gt (vector unsigned short, vector unsigned short);
10186 int vec_all_gt (vector bool short, vector signed short);
10187 int vec_all_gt (vector signed short, vector bool short);
10188 int vec_all_gt (vector signed short, vector signed short);
10189 int vec_all_gt (vector bool int, vector unsigned int);
10190 int vec_all_gt (vector unsigned int, vector bool int);
10191 int vec_all_gt (vector unsigned int, vector unsigned int);
10192 int vec_all_gt (vector bool int, vector signed int);
10193 int vec_all_gt (vector signed int, vector bool int);
10194 int vec_all_gt (vector signed int, vector signed int);
10195 int vec_all_gt (vector float, vector float);
10197 int vec_all_in (vector float, vector float);
10199 int vec_all_le (vector bool char, vector unsigned char);
10200 int vec_all_le (vector unsigned char, vector bool char);
10201 int vec_all_le (vector unsigned char, vector unsigned char);
10202 int vec_all_le (vector bool char, vector signed char);
10203 int vec_all_le (vector signed char, vector bool char);
10204 int vec_all_le (vector signed char, vector signed char);
10205 int vec_all_le (vector bool short, vector unsigned short);
10206 int vec_all_le (vector unsigned short, vector bool short);
10207 int vec_all_le (vector unsigned short, vector unsigned short);
10208 int vec_all_le (vector bool short, vector signed short);
10209 int vec_all_le (vector signed short, vector bool short);
10210 int vec_all_le (vector signed short, vector signed short);
10211 int vec_all_le (vector bool int, vector unsigned int);
10212 int vec_all_le (vector unsigned int, vector bool int);
10213 int vec_all_le (vector unsigned int, vector unsigned int);
10214 int vec_all_le (vector bool int, vector signed int);
10215 int vec_all_le (vector signed int, vector bool int);
10216 int vec_all_le (vector signed int, vector signed int);
10217 int vec_all_le (vector float, vector float);
10219 int vec_all_lt (vector bool char, vector unsigned char);
10220 int vec_all_lt (vector unsigned char, vector bool char);
10221 int vec_all_lt (vector unsigned char, vector unsigned char);
10222 int vec_all_lt (vector bool char, vector signed char);
10223 int vec_all_lt (vector signed char, vector bool char);
10224 int vec_all_lt (vector signed char, vector signed char);
10225 int vec_all_lt (vector bool short, vector unsigned short);
10226 int vec_all_lt (vector unsigned short, vector bool short);
10227 int vec_all_lt (vector unsigned short, vector unsigned short);
10228 int vec_all_lt (vector bool short, vector signed short);
10229 int vec_all_lt (vector signed short, vector bool short);
10230 int vec_all_lt (vector signed short, vector signed short);
10231 int vec_all_lt (vector bool int, vector unsigned int);
10232 int vec_all_lt (vector unsigned int, vector bool int);
10233 int vec_all_lt (vector unsigned int, vector unsigned int);
10234 int vec_all_lt (vector bool int, vector signed int);
10235 int vec_all_lt (vector signed int, vector bool int);
10236 int vec_all_lt (vector signed int, vector signed int);
10237 int vec_all_lt (vector float, vector float);
10239 int vec_all_nan (vector float);
10241 int vec_all_ne (vector signed char, vector bool char);
10242 int vec_all_ne (vector signed char, vector signed char);
10243 int vec_all_ne (vector unsigned char, vector bool char);
10244 int vec_all_ne (vector unsigned char, vector unsigned char);
10245 int vec_all_ne (vector bool char, vector bool char);
10246 int vec_all_ne (vector bool char, vector unsigned char);
10247 int vec_all_ne (vector bool char, vector signed char);
10248 int vec_all_ne (vector signed short, vector bool short);
10249 int vec_all_ne (vector signed short, vector signed short);
10250 int vec_all_ne (vector unsigned short, vector bool short);
10251 int vec_all_ne (vector unsigned short, vector unsigned short);
10252 int vec_all_ne (vector bool short, vector bool short);
10253 int vec_all_ne (vector bool short, vector unsigned short);
10254 int vec_all_ne (vector bool short, vector signed short);
10255 int vec_all_ne (vector pixel, vector pixel);
10256 int vec_all_ne (vector signed int, vector bool int);
10257 int vec_all_ne (vector signed int, vector signed int);
10258 int vec_all_ne (vector unsigned int, vector bool int);
10259 int vec_all_ne (vector unsigned int, vector unsigned int);
10260 int vec_all_ne (vector bool int, vector bool int);
10261 int vec_all_ne (vector bool int, vector unsigned int);
10262 int vec_all_ne (vector bool int, vector signed int);
10263 int vec_all_ne (vector float, vector float);
10265 int vec_all_nge (vector float, vector float);
10267 int vec_all_ngt (vector float, vector float);
10269 int vec_all_nle (vector float, vector float);
10271 int vec_all_nlt (vector float, vector float);
10273 int vec_all_numeric (vector float);
10275 int vec_any_eq (vector signed char, vector bool char);
10276 int vec_any_eq (vector signed char, vector signed char);
10277 int vec_any_eq (vector unsigned char, vector bool char);
10278 int vec_any_eq (vector unsigned char, vector unsigned char);
10279 int vec_any_eq (vector bool char, vector bool char);
10280 int vec_any_eq (vector bool char, vector unsigned char);
10281 int vec_any_eq (vector bool char, vector signed char);
10282 int vec_any_eq (vector signed short, vector bool short);
10283 int vec_any_eq (vector signed short, vector signed short);
10284 int vec_any_eq (vector unsigned short, vector bool short);
10285 int vec_any_eq (vector unsigned short, vector unsigned short);
10286 int vec_any_eq (vector bool short, vector bool short);
10287 int vec_any_eq (vector bool short, vector unsigned short);
10288 int vec_any_eq (vector bool short, vector signed short);
10289 int vec_any_eq (vector pixel, vector pixel);
10290 int vec_any_eq (vector signed int, vector bool int);
10291 int vec_any_eq (vector signed int, vector signed int);
10292 int vec_any_eq (vector unsigned int, vector bool int);
10293 int vec_any_eq (vector unsigned int, vector unsigned int);
10294 int vec_any_eq (vector bool int, vector bool int);
10295 int vec_any_eq (vector bool int, vector unsigned int);
10296 int vec_any_eq (vector bool int, vector signed int);
10297 int vec_any_eq (vector float, vector float);
10299 int vec_any_ge (vector signed char, vector bool char);
10300 int vec_any_ge (vector unsigned char, vector bool char);
10301 int vec_any_ge (vector unsigned char, vector unsigned char);
10302 int vec_any_ge (vector signed char, vector signed char);
10303 int vec_any_ge (vector bool char, vector unsigned char);
10304 int vec_any_ge (vector bool char, vector signed char);
10305 int vec_any_ge (vector unsigned short, vector bool short);
10306 int vec_any_ge (vector unsigned short, vector unsigned short);
10307 int vec_any_ge (vector signed short, vector signed short);
10308 int vec_any_ge (vector signed short, vector bool short);
10309 int vec_any_ge (vector bool short, vector unsigned short);
10310 int vec_any_ge (vector bool short, vector signed short);
10311 int vec_any_ge (vector signed int, vector bool int);
10312 int vec_any_ge (vector unsigned int, vector bool int);
10313 int vec_any_ge (vector unsigned int, vector unsigned int);
10314 int vec_any_ge (vector signed int, vector signed int);
10315 int vec_any_ge (vector bool int, vector unsigned int);
10316 int vec_any_ge (vector bool int, vector signed int);
10317 int vec_any_ge (vector float, vector float);
10319 int vec_any_gt (vector bool char, vector unsigned char);
10320 int vec_any_gt (vector unsigned char, vector bool char);
10321 int vec_any_gt (vector unsigned char, vector unsigned char);
10322 int vec_any_gt (vector bool char, vector signed char);
10323 int vec_any_gt (vector signed char, vector bool char);
10324 int vec_any_gt (vector signed char, vector signed char);
10325 int vec_any_gt (vector bool short, vector unsigned short);
10326 int vec_any_gt (vector unsigned short, vector bool short);
10327 int vec_any_gt (vector unsigned short, vector unsigned short);
10328 int vec_any_gt (vector bool short, vector signed short);
10329 int vec_any_gt (vector signed short, vector bool short);
10330 int vec_any_gt (vector signed short, vector signed short);
10331 int vec_any_gt (vector bool int, vector unsigned int);
10332 int vec_any_gt (vector unsigned int, vector bool int);
10333 int vec_any_gt (vector unsigned int, vector unsigned int);
10334 int vec_any_gt (vector bool int, vector signed int);
10335 int vec_any_gt (vector signed int, vector bool int);
10336 int vec_any_gt (vector signed int, vector signed int);
10337 int vec_any_gt (vector float, vector float);
10339 int vec_any_le (vector bool char, vector unsigned char);
10340 int vec_any_le (vector unsigned char, vector bool char);
10341 int vec_any_le (vector unsigned char, vector unsigned char);
10342 int vec_any_le (vector bool char, vector signed char);
10343 int vec_any_le (vector signed char, vector bool char);
10344 int vec_any_le (vector signed char, vector signed char);
10345 int vec_any_le (vector bool short, vector unsigned short);
10346 int vec_any_le (vector unsigned short, vector bool short);
10347 int vec_any_le (vector unsigned short, vector unsigned short);
10348 int vec_any_le (vector bool short, vector signed short);
10349 int vec_any_le (vector signed short, vector bool short);
10350 int vec_any_le (vector signed short, vector signed short);
10351 int vec_any_le (vector bool int, vector unsigned int);
10352 int vec_any_le (vector unsigned int, vector bool int);
10353 int vec_any_le (vector unsigned int, vector unsigned int);
10354 int vec_any_le (vector bool int, vector signed int);
10355 int vec_any_le (vector signed int, vector bool int);
10356 int vec_any_le (vector signed int, vector signed int);
10357 int vec_any_le (vector float, vector float);
10359 int vec_any_lt (vector bool char, vector unsigned char);
10360 int vec_any_lt (vector unsigned char, vector bool char);
10361 int vec_any_lt (vector unsigned char, vector unsigned char);
10362 int vec_any_lt (vector bool char, vector signed char);
10363 int vec_any_lt (vector signed char, vector bool char);
10364 int vec_any_lt (vector signed char, vector signed char);
10365 int vec_any_lt (vector bool short, vector unsigned short);
10366 int vec_any_lt (vector unsigned short, vector bool short);
10367 int vec_any_lt (vector unsigned short, vector unsigned short);
10368 int vec_any_lt (vector bool short, vector signed short);
10369 int vec_any_lt (vector signed short, vector bool short);
10370 int vec_any_lt (vector signed short, vector signed short);
10371 int vec_any_lt (vector bool int, vector unsigned int);
10372 int vec_any_lt (vector unsigned int, vector bool int);
10373 int vec_any_lt (vector unsigned int, vector unsigned int);
10374 int vec_any_lt (vector bool int, vector signed int);
10375 int vec_any_lt (vector signed int, vector bool int);
10376 int vec_any_lt (vector signed int, vector signed int);
10377 int vec_any_lt (vector float, vector float);
10379 int vec_any_nan (vector float);
10381 int vec_any_ne (vector signed char, vector bool char);
10382 int vec_any_ne (vector signed char, vector signed char);
10383 int vec_any_ne (vector unsigned char, vector bool char);
10384 int vec_any_ne (vector unsigned char, vector unsigned char);
10385 int vec_any_ne (vector bool char, vector bool char);
10386 int vec_any_ne (vector bool char, vector unsigned char);
10387 int vec_any_ne (vector bool char, vector signed char);
10388 int vec_any_ne (vector signed short, vector bool short);
10389 int vec_any_ne (vector signed short, vector signed short);
10390 int vec_any_ne (vector unsigned short, vector bool short);
10391 int vec_any_ne (vector unsigned short, vector unsigned short);
10392 int vec_any_ne (vector bool short, vector bool short);
10393 int vec_any_ne (vector bool short, vector unsigned short);
10394 int vec_any_ne (vector bool short, vector signed short);
10395 int vec_any_ne (vector pixel, vector pixel);
10396 int vec_any_ne (vector signed int, vector bool int);
10397 int vec_any_ne (vector signed int, vector signed int);
10398 int vec_any_ne (vector unsigned int, vector bool int);
10399 int vec_any_ne (vector unsigned int, vector unsigned int);
10400 int vec_any_ne (vector bool int, vector bool int);
10401 int vec_any_ne (vector bool int, vector unsigned int);
10402 int vec_any_ne (vector bool int, vector signed int);
10403 int vec_any_ne (vector float, vector float);
10405 int vec_any_nge (vector float, vector float);
10407 int vec_any_ngt (vector float, vector float);
10409 int vec_any_nle (vector float, vector float);
10411 int vec_any_nlt (vector float, vector float);
10413 int vec_any_numeric (vector float);
10415 int vec_any_out (vector float, vector float);
10418 @node SPARC VIS Built-in Functions
10419 @subsection SPARC VIS Built-in Functions
10421 GCC supports SIMD operations on the SPARC using both the generic vector
10422 extensions (@pxref{Vector Extensions}) as well as built-in functions for
10423 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
10424 switch, the VIS extension is exposed as the following built-in functions:
10427 typedef int v2si __attribute__ ((vector_size (8)));
10428 typedef short v4hi __attribute__ ((vector_size (8)));
10429 typedef short v2hi __attribute__ ((vector_size (4)));
10430 typedef char v8qi __attribute__ ((vector_size (8)));
10431 typedef char v4qi __attribute__ ((vector_size (4)));
10433 void * __builtin_vis_alignaddr (void *, long);
10434 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
10435 v2si __builtin_vis_faligndatav2si (v2si, v2si);
10436 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
10437 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
10439 v4hi __builtin_vis_fexpand (v4qi);
10441 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
10442 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
10443 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
10444 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
10445 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
10446 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
10447 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
10449 v4qi __builtin_vis_fpack16 (v4hi);
10450 v8qi __builtin_vis_fpack32 (v2si, v2si);
10451 v2hi __builtin_vis_fpackfix (v2si);
10452 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
10454 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
10457 @node SPU Built-in Functions
10458 @subsection SPU Built-in Functions
10460 GCC provides extensions for the SPU processor as described in the
10461 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
10462 found at @uref{http://cell.scei.co.jp/} or
10463 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
10464 implementation differs in several ways.
10469 The optional extension of specifying vector constants in parentheses is
10473 A vector initializer requires no cast if the vector constant is of the
10474 same type as the variable it is initializing.
10477 If @code{signed} or @code{unsigned} is omitted, the signedness of the
10478 vector type is the default signedness of the base type. The default
10479 varies depending on the operating system, so a portable program should
10480 always specify the signedness.
10483 By default, the keyword @code{__vector} is added. The macro
10484 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
10488 GCC allows using a @code{typedef} name as the type specifier for a
10492 For C, overloaded functions are implemented with macros so the following
10496 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
10499 Since @code{spu_add} is a macro, the vector constant in the example
10500 is treated as four separate arguments. Wrap the entire argument in
10501 parentheses for this to work.
10504 The extended version of @code{__builtin_expect} is not supported.
10508 @emph{Note:} Only the interface described in the aforementioned
10509 specification is supported. Internally, GCC uses built-in functions to
10510 implement the required functionality, but these are not supported and
10511 are subject to change without notice.
10513 @node Target Format Checks
10514 @section Format Checks Specific to Particular Target Machines
10516 For some target machines, GCC supports additional options to the
10518 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
10521 * Solaris Format Checks::
10524 @node Solaris Format Checks
10525 @subsection Solaris Format Checks
10527 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
10528 check. @code{cmn_err} accepts a subset of the standard @code{printf}
10529 conversions, and the two-argument @code{%b} conversion for displaying
10530 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
10533 @section Pragmas Accepted by GCC
10537 GCC supports several types of pragmas, primarily in order to compile
10538 code originally written for other compilers. Note that in general
10539 we do not recommend the use of pragmas; @xref{Function Attributes},
10540 for further explanation.
10545 * RS/6000 and PowerPC Pragmas::
10547 * Solaris Pragmas::
10548 * Symbol-Renaming Pragmas::
10549 * Structure-Packing Pragmas::
10551 * Diagnostic Pragmas::
10552 * Visibility Pragmas::
10556 @subsection ARM Pragmas
10558 The ARM target defines pragmas for controlling the default addition of
10559 @code{long_call} and @code{short_call} attributes to functions.
10560 @xref{Function Attributes}, for information about the effects of these
10565 @cindex pragma, long_calls
10566 Set all subsequent functions to have the @code{long_call} attribute.
10568 @item no_long_calls
10569 @cindex pragma, no_long_calls
10570 Set all subsequent functions to have the @code{short_call} attribute.
10572 @item long_calls_off
10573 @cindex pragma, long_calls_off
10574 Do not affect the @code{long_call} or @code{short_call} attributes of
10575 subsequent functions.
10579 @subsection M32C Pragmas
10582 @item memregs @var{number}
10583 @cindex pragma, memregs
10584 Overrides the command line option @code{-memregs=} for the current
10585 file. Use with care! This pragma must be before any function in the
10586 file, and mixing different memregs values in different objects may
10587 make them incompatible. This pragma is useful when a
10588 performance-critical function uses a memreg for temporary values,
10589 as it may allow you to reduce the number of memregs used.
10593 @node RS/6000 and PowerPC Pragmas
10594 @subsection RS/6000 and PowerPC Pragmas
10596 The RS/6000 and PowerPC targets define one pragma for controlling
10597 whether or not the @code{longcall} attribute is added to function
10598 declarations by default. This pragma overrides the @option{-mlongcall}
10599 option, but not the @code{longcall} and @code{shortcall} attributes.
10600 @xref{RS/6000 and PowerPC Options}, for more information about when long
10601 calls are and are not necessary.
10605 @cindex pragma, longcall
10606 Apply the @code{longcall} attribute to all subsequent function
10610 Do not apply the @code{longcall} attribute to subsequent function
10614 @c Describe c4x pragmas here.
10615 @c Describe h8300 pragmas here.
10616 @c Describe sh pragmas here.
10617 @c Describe v850 pragmas here.
10619 @node Darwin Pragmas
10620 @subsection Darwin Pragmas
10622 The following pragmas are available for all architectures running the
10623 Darwin operating system. These are useful for compatibility with other
10627 @item mark @var{tokens}@dots{}
10628 @cindex pragma, mark
10629 This pragma is accepted, but has no effect.
10631 @item options align=@var{alignment}
10632 @cindex pragma, options align
10633 This pragma sets the alignment of fields in structures. The values of
10634 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
10635 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
10636 properly; to restore the previous setting, use @code{reset} for the
10639 @item segment @var{tokens}@dots{}
10640 @cindex pragma, segment
10641 This pragma is accepted, but has no effect.
10643 @item unused (@var{var} [, @var{var}]@dots{})
10644 @cindex pragma, unused
10645 This pragma declares variables to be possibly unused. GCC will not
10646 produce warnings for the listed variables. The effect is similar to
10647 that of the @code{unused} attribute, except that this pragma may appear
10648 anywhere within the variables' scopes.
10651 @node Solaris Pragmas
10652 @subsection Solaris Pragmas
10654 The Solaris target supports @code{#pragma redefine_extname}
10655 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10656 @code{#pragma} directives for compatibility with the system compiler.
10659 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10660 @cindex pragma, align
10662 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10663 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10664 Attributes}). Macro expansion occurs on the arguments to this pragma
10665 when compiling C and Objective-C. It does not currently occur when
10666 compiling C++, but this is a bug which may be fixed in a future
10669 @item fini (@var{function} [, @var{function}]...)
10670 @cindex pragma, fini
10672 This pragma causes each listed @var{function} to be called after
10673 main, or during shared module unloading, by adding a call to the
10674 @code{.fini} section.
10676 @item init (@var{function} [, @var{function}]...)
10677 @cindex pragma, init
10679 This pragma causes each listed @var{function} to be called during
10680 initialization (before @code{main}) or during shared module loading, by
10681 adding a call to the @code{.init} section.
10685 @node Symbol-Renaming Pragmas
10686 @subsection Symbol-Renaming Pragmas
10688 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10689 supports two @code{#pragma} directives which change the name used in
10690 assembly for a given declaration. These pragmas are only available on
10691 platforms whose system headers need them. To get this effect on all
10692 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10696 @item redefine_extname @var{oldname} @var{newname}
10697 @cindex pragma, redefine_extname
10699 This pragma gives the C function @var{oldname} the assembly symbol
10700 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10701 will be defined if this pragma is available (currently only on
10704 @item extern_prefix @var{string}
10705 @cindex pragma, extern_prefix
10707 This pragma causes all subsequent external function and variable
10708 declarations to have @var{string} prepended to their assembly symbols.
10709 This effect may be terminated with another @code{extern_prefix} pragma
10710 whose argument is an empty string. The preprocessor macro
10711 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10712 available (currently only on Tru64 UNIX)@.
10715 These pragmas and the asm labels extension interact in a complicated
10716 manner. Here are some corner cases you may want to be aware of.
10719 @item Both pragmas silently apply only to declarations with external
10720 linkage. Asm labels do not have this restriction.
10722 @item In C++, both pragmas silently apply only to declarations with
10723 ``C'' linkage. Again, asm labels do not have this restriction.
10725 @item If any of the three ways of changing the assembly name of a
10726 declaration is applied to a declaration whose assembly name has
10727 already been determined (either by a previous use of one of these
10728 features, or because the compiler needed the assembly name in order to
10729 generate code), and the new name is different, a warning issues and
10730 the name does not change.
10732 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10733 always the C-language name.
10735 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10736 occurs with an asm label attached, the prefix is silently ignored for
10739 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10740 apply to the same declaration, whichever triggered first wins, and a
10741 warning issues if they contradict each other. (We would like to have
10742 @code{#pragma redefine_extname} always win, for consistency with asm
10743 labels, but if @code{#pragma extern_prefix} triggers first we have no
10744 way of knowing that that happened.)
10747 @node Structure-Packing Pragmas
10748 @subsection Structure-Packing Pragmas
10750 For compatibility with Win32, GCC supports a set of @code{#pragma}
10751 directives which change the maximum alignment of members of structures
10752 (other than zero-width bitfields), unions, and classes subsequently
10753 defined. The @var{n} value below always is required to be a small power
10754 of two and specifies the new alignment in bytes.
10757 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10758 @item @code{#pragma pack()} sets the alignment to the one that was in
10759 effect when compilation started (see also command line option
10760 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10761 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10762 setting on an internal stack and then optionally sets the new alignment.
10763 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10764 saved at the top of the internal stack (and removes that stack entry).
10765 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10766 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10767 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10768 @code{#pragma pack(pop)}.
10771 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10772 @code{#pragma} which lays out a structure as the documented
10773 @code{__attribute__ ((ms_struct))}.
10775 @item @code{#pragma ms_struct on} turns on the layout for structures
10777 @item @code{#pragma ms_struct off} turns off the layout for structures
10779 @item @code{#pragma ms_struct reset} goes back to the default layout.
10783 @subsection Weak Pragmas
10785 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10786 directives for declaring symbols to be weak, and defining weak
10790 @item #pragma weak @var{symbol}
10791 @cindex pragma, weak
10792 This pragma declares @var{symbol} to be weak, as if the declaration
10793 had the attribute of the same name. The pragma may appear before
10794 or after the declaration of @var{symbol}, but must appear before
10795 either its first use or its definition. It is not an error for
10796 @var{symbol} to never be defined at all.
10798 @item #pragma weak @var{symbol1} = @var{symbol2}
10799 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10800 It is an error if @var{symbol2} is not defined in the current
10804 @node Diagnostic Pragmas
10805 @subsection Diagnostic Pragmas
10807 GCC allows the user to selectively enable or disable certain types of
10808 diagnostics, and change the kind of the diagnostic. For example, a
10809 project's policy might require that all sources compile with
10810 @option{-Werror} but certain files might have exceptions allowing
10811 specific types of warnings. Or, a project might selectively enable
10812 diagnostics and treat them as errors depending on which preprocessor
10813 macros are defined.
10816 @item #pragma GCC diagnostic @var{kind} @var{option}
10817 @cindex pragma, diagnostic
10819 Modifies the disposition of a diagnostic. Note that not all
10820 diagnostics are modifiable; at the moment only warnings (normally
10821 controlled by @samp{-W...}) can be controlled, and not all of them.
10822 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10823 are controllable and which option controls them.
10825 @var{kind} is @samp{error} to treat this diagnostic as an error,
10826 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10827 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10828 @var{option} is a double quoted string which matches the command line
10832 #pragma GCC diagnostic warning "-Wformat"
10833 #pragma GCC diagnostic error "-Wformat"
10834 #pragma GCC diagnostic ignored "-Wformat"
10837 Note that these pragmas override any command line options. Also,
10838 while it is syntactically valid to put these pragmas anywhere in your
10839 sources, the only supported location for them is before any data or
10840 functions are defined. Doing otherwise may result in unpredictable
10841 results depending on how the optimizer manages your sources. If the
10842 same option is listed multiple times, the last one specified is the
10843 one that is in effect. This pragma is not intended to be a general
10844 purpose replacement for command line options, but for implementing
10845 strict control over project policies.
10849 @node Visibility Pragmas
10850 @subsection Visibility Pragmas
10853 @item #pragma GCC visibility push(@var{visibility})
10854 @itemx #pragma GCC visibility pop
10855 @cindex pragma, visibility
10857 This pragma allows the user to set the visibility for multiple
10858 declarations without having to give each a visibility attribute
10859 @xref{Function Attributes}, for more information about visibility and
10860 the attribute syntax.
10862 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10863 declarations. Class members and template specializations are not
10864 affected; if you want to override the visibility for a particular
10865 member or instantiation, you must use an attribute.
10869 @node Unnamed Fields
10870 @section Unnamed struct/union fields within structs/unions
10874 For compatibility with other compilers, GCC allows you to define
10875 a structure or union that contains, as fields, structures and unions
10876 without names. For example:
10889 In this example, the user would be able to access members of the unnamed
10890 union with code like @samp{foo.b}. Note that only unnamed structs and
10891 unions are allowed, you may not have, for example, an unnamed
10894 You must never create such structures that cause ambiguous field definitions.
10895 For example, this structure:
10906 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10907 Such constructs are not supported and must be avoided. In the future,
10908 such constructs may be detected and treated as compilation errors.
10910 @opindex fms-extensions
10911 Unless @option{-fms-extensions} is used, the unnamed field must be a
10912 structure or union definition without a tag (for example, @samp{struct
10913 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10914 also be a definition with a tag such as @samp{struct foo @{ int a;
10915 @};}, a reference to a previously defined structure or union such as
10916 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10917 previously defined structure or union type.
10920 @section Thread-Local Storage
10921 @cindex Thread-Local Storage
10922 @cindex @acronym{TLS}
10925 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10926 are allocated such that there is one instance of the variable per extant
10927 thread. The run-time model GCC uses to implement this originates
10928 in the IA-64 processor-specific ABI, but has since been migrated
10929 to other processors as well. It requires significant support from
10930 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10931 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10932 is not available everywhere.
10934 At the user level, the extension is visible with a new storage
10935 class keyword: @code{__thread}. For example:
10939 extern __thread struct state s;
10940 static __thread char *p;
10943 The @code{__thread} specifier may be used alone, with the @code{extern}
10944 or @code{static} specifiers, but with no other storage class specifier.
10945 When used with @code{extern} or @code{static}, @code{__thread} must appear
10946 immediately after the other storage class specifier.
10948 The @code{__thread} specifier may be applied to any global, file-scoped
10949 static, function-scoped static, or static data member of a class. It may
10950 not be applied to block-scoped automatic or non-static data member.
10952 When the address-of operator is applied to a thread-local variable, it is
10953 evaluated at run-time and returns the address of the current thread's
10954 instance of that variable. An address so obtained may be used by any
10955 thread. When a thread terminates, any pointers to thread-local variables
10956 in that thread become invalid.
10958 No static initialization may refer to the address of a thread-local variable.
10960 In C++, if an initializer is present for a thread-local variable, it must
10961 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10964 See @uref{http://people.redhat.com/drepper/tls.pdf,
10965 ELF Handling For Thread-Local Storage} for a detailed explanation of
10966 the four thread-local storage addressing models, and how the run-time
10967 is expected to function.
10970 * C99 Thread-Local Edits::
10971 * C++98 Thread-Local Edits::
10974 @node C99 Thread-Local Edits
10975 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10977 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10978 that document the exact semantics of the language extension.
10982 @cite{5.1.2 Execution environments}
10984 Add new text after paragraph 1
10987 Within either execution environment, a @dfn{thread} is a flow of
10988 control within a program. It is implementation defined whether
10989 or not there may be more than one thread associated with a program.
10990 It is implementation defined how threads beyond the first are
10991 created, the name and type of the function called at thread
10992 startup, and how threads may be terminated. However, objects
10993 with thread storage duration shall be initialized before thread
10998 @cite{6.2.4 Storage durations of objects}
11000 Add new text before paragraph 3
11003 An object whose identifier is declared with the storage-class
11004 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
11005 Its lifetime is the entire execution of the thread, and its
11006 stored value is initialized only once, prior to thread startup.
11010 @cite{6.4.1 Keywords}
11012 Add @code{__thread}.
11015 @cite{6.7.1 Storage-class specifiers}
11017 Add @code{__thread} to the list of storage class specifiers in
11020 Change paragraph 2 to
11023 With the exception of @code{__thread}, at most one storage-class
11024 specifier may be given [@dots{}]. The @code{__thread} specifier may
11025 be used alone, or immediately following @code{extern} or
11029 Add new text after paragraph 6
11032 The declaration of an identifier for a variable that has
11033 block scope that specifies @code{__thread} shall also
11034 specify either @code{extern} or @code{static}.
11036 The @code{__thread} specifier shall be used only with
11041 @node C++98 Thread-Local Edits
11042 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
11044 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
11045 that document the exact semantics of the language extension.
11049 @b{[intro.execution]}
11051 New text after paragraph 4
11054 A @dfn{thread} is a flow of control within the abstract machine.
11055 It is implementation defined whether or not there may be more than
11059 New text after paragraph 7
11062 It is unspecified whether additional action must be taken to
11063 ensure when and whether side effects are visible to other threads.
11069 Add @code{__thread}.
11072 @b{[basic.start.main]}
11074 Add after paragraph 5
11077 The thread that begins execution at the @code{main} function is called
11078 the @dfn{main thread}. It is implementation defined how functions
11079 beginning threads other than the main thread are designated or typed.
11080 A function so designated, as well as the @code{main} function, is called
11081 a @dfn{thread startup function}. It is implementation defined what
11082 happens if a thread startup function returns. It is implementation
11083 defined what happens to other threads when any thread calls @code{exit}.
11087 @b{[basic.start.init]}
11089 Add after paragraph 4
11092 The storage for an object of thread storage duration shall be
11093 statically initialized before the first statement of the thread startup
11094 function. An object of thread storage duration shall not require
11095 dynamic initialization.
11099 @b{[basic.start.term]}
11101 Add after paragraph 3
11104 The type of an object with thread storage duration shall not have a
11105 non-trivial destructor, nor shall it be an array type whose elements
11106 (directly or indirectly) have non-trivial destructors.
11112 Add ``thread storage duration'' to the list in paragraph 1.
11117 Thread, static, and automatic storage durations are associated with
11118 objects introduced by declarations [@dots{}].
11121 Add @code{__thread} to the list of specifiers in paragraph 3.
11124 @b{[basic.stc.thread]}
11126 New section before @b{[basic.stc.static]}
11129 The keyword @code{__thread} applied to a non-local object gives the
11130 object thread storage duration.
11132 A local variable or class data member declared both @code{static}
11133 and @code{__thread} gives the variable or member thread storage
11138 @b{[basic.stc.static]}
11143 All objects which have neither thread storage duration, dynamic
11144 storage duration nor are local [@dots{}].
11150 Add @code{__thread} to the list in paragraph 1.
11155 With the exception of @code{__thread}, at most one
11156 @var{storage-class-specifier} shall appear in a given
11157 @var{decl-specifier-seq}. The @code{__thread} specifier may
11158 be used alone, or immediately following the @code{extern} or
11159 @code{static} specifiers. [@dots{}]
11162 Add after paragraph 5
11165 The @code{__thread} specifier can be applied only to the names of objects
11166 and to anonymous unions.
11172 Add after paragraph 6
11175 Non-@code{static} members shall not be @code{__thread}.
11179 @node Binary constants
11180 @section Binary constants using the @samp{0b} prefix
11181 @cindex Binary constants using the @samp{0b} prefix
11183 Integer constants can be written as binary constants, consisting of a
11184 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
11185 @samp{0B}. This is particularly useful in environments that operate a
11186 lot on the bit-level (like microcontrollers).
11188 The following statements are identical:
11197 The type of these constants follows the same rules as for octal or
11198 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
11201 @node C++ Extensions
11202 @chapter Extensions to the C++ Language
11203 @cindex extensions, C++ language
11204 @cindex C++ language extensions
11206 The GNU compiler provides these extensions to the C++ language (and you
11207 can also use most of the C language extensions in your C++ programs). If you
11208 want to write code that checks whether these features are available, you can
11209 test for the GNU compiler the same way as for C programs: check for a
11210 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
11211 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
11212 Predefined Macros,cpp,The GNU C Preprocessor}).
11215 * Volatiles:: What constitutes an access to a volatile object.
11216 * Restricted Pointers:: C99 restricted pointers and references.
11217 * Vague Linkage:: Where G++ puts inlines, vtables and such.
11218 * C++ Interface:: You can use a single C++ header file for both
11219 declarations and definitions.
11220 * Template Instantiation:: Methods for ensuring that exactly one copy of
11221 each needed template instantiation is emitted.
11222 * Bound member functions:: You can extract a function pointer to the
11223 method denoted by a @samp{->*} or @samp{.*} expression.
11224 * C++ Attributes:: Variable, function, and type attributes for C++ only.
11225 * Namespace Association:: Strong using-directives for namespace association.
11226 * Type Traits:: Compiler support for type traits
11227 * Java Exceptions:: Tweaking exception handling to work with Java.
11228 * Deprecated Features:: Things will disappear from g++.
11229 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
11233 @section When is a Volatile Object Accessed?
11234 @cindex accessing volatiles
11235 @cindex volatile read
11236 @cindex volatile write
11237 @cindex volatile access
11239 Both the C and C++ standard have the concept of volatile objects. These
11240 are normally accessed by pointers and used for accessing hardware. The
11241 standards encourage compilers to refrain from optimizations concerning
11242 accesses to volatile objects. The C standard leaves it implementation
11243 defined as to what constitutes a volatile access. The C++ standard omits
11244 to specify this, except to say that C++ should behave in a similar manner
11245 to C with respect to volatiles, where possible. The minimum either
11246 standard specifies is that at a sequence point all previous accesses to
11247 volatile objects have stabilized and no subsequent accesses have
11248 occurred. Thus an implementation is free to reorder and combine
11249 volatile accesses which occur between sequence points, but cannot do so
11250 for accesses across a sequence point. The use of volatiles does not
11251 allow you to violate the restriction on updating objects multiple times
11252 within a sequence point.
11254 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
11256 The behavior differs slightly between C and C++ in the non-obvious cases:
11259 volatile int *src = @var{somevalue};
11263 With C, such expressions are rvalues, and GCC interprets this either as a
11264 read of the volatile object being pointed to or only as request to evaluate
11265 the side-effects. The C++ standard specifies that such expressions do not
11266 undergo lvalue to rvalue conversion, and that the type of the dereferenced
11267 object may be incomplete. The C++ standard does not specify explicitly
11268 that it is this lvalue to rvalue conversion which may be responsible for
11269 causing an access. However, there is reason to believe that it is,
11270 because otherwise certain simple expressions become undefined. However,
11271 because it would surprise most programmers, G++ treats dereferencing a
11272 pointer to volatile object of complete type when the value is unused as
11273 GCC would do for an equivalent type in C. When the object has incomplete
11274 type, G++ issues a warning; if you wish to force an error, you must
11275 force a conversion to rvalue with, for instance, a static cast.
11277 When using a reference to volatile, G++ does not treat equivalent
11278 expressions as accesses to volatiles, but instead issues a warning that
11279 no volatile is accessed. The rationale for this is that otherwise it
11280 becomes difficult to determine where volatile access occur, and not
11281 possible to ignore the return value from functions returning volatile
11282 references. Again, if you wish to force a read, cast the reference to
11285 @node Restricted Pointers
11286 @section Restricting Pointer Aliasing
11287 @cindex restricted pointers
11288 @cindex restricted references
11289 @cindex restricted this pointer
11291 As with the C front end, G++ understands the C99 feature of restricted pointers,
11292 specified with the @code{__restrict__}, or @code{__restrict} type
11293 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
11294 language flag, @code{restrict} is not a keyword in C++.
11296 In addition to allowing restricted pointers, you can specify restricted
11297 references, which indicate that the reference is not aliased in the local
11301 void fn (int *__restrict__ rptr, int &__restrict__ rref)
11308 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
11309 @var{rref} refers to a (different) unaliased integer.
11311 You may also specify whether a member function's @var{this} pointer is
11312 unaliased by using @code{__restrict__} as a member function qualifier.
11315 void T::fn () __restrict__
11322 Within the body of @code{T::fn}, @var{this} will have the effective
11323 definition @code{T *__restrict__ const this}. Notice that the
11324 interpretation of a @code{__restrict__} member function qualifier is
11325 different to that of @code{const} or @code{volatile} qualifier, in that it
11326 is applied to the pointer rather than the object. This is consistent with
11327 other compilers which implement restricted pointers.
11329 As with all outermost parameter qualifiers, @code{__restrict__} is
11330 ignored in function definition matching. This means you only need to
11331 specify @code{__restrict__} in a function definition, rather than
11332 in a function prototype as well.
11334 @node Vague Linkage
11335 @section Vague Linkage
11336 @cindex vague linkage
11338 There are several constructs in C++ which require space in the object
11339 file but are not clearly tied to a single translation unit. We say that
11340 these constructs have ``vague linkage''. Typically such constructs are
11341 emitted wherever they are needed, though sometimes we can be more
11345 @item Inline Functions
11346 Inline functions are typically defined in a header file which can be
11347 included in many different compilations. Hopefully they can usually be
11348 inlined, but sometimes an out-of-line copy is necessary, if the address
11349 of the function is taken or if inlining fails. In general, we emit an
11350 out-of-line copy in all translation units where one is needed. As an
11351 exception, we only emit inline virtual functions with the vtable, since
11352 it will always require a copy.
11354 Local static variables and string constants used in an inline function
11355 are also considered to have vague linkage, since they must be shared
11356 between all inlined and out-of-line instances of the function.
11360 C++ virtual functions are implemented in most compilers using a lookup
11361 table, known as a vtable. The vtable contains pointers to the virtual
11362 functions provided by a class, and each object of the class contains a
11363 pointer to its vtable (or vtables, in some multiple-inheritance
11364 situations). If the class declares any non-inline, non-pure virtual
11365 functions, the first one is chosen as the ``key method'' for the class,
11366 and the vtable is only emitted in the translation unit where the key
11369 @emph{Note:} If the chosen key method is later defined as inline, the
11370 vtable will still be emitted in every translation unit which defines it.
11371 Make sure that any inline virtuals are declared inline in the class
11372 body, even if they are not defined there.
11374 @item type_info objects
11377 C++ requires information about types to be written out in order to
11378 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
11379 For polymorphic classes (classes with virtual functions), the type_info
11380 object is written out along with the vtable so that @samp{dynamic_cast}
11381 can determine the dynamic type of a class object at runtime. For all
11382 other types, we write out the type_info object when it is used: when
11383 applying @samp{typeid} to an expression, throwing an object, or
11384 referring to a type in a catch clause or exception specification.
11386 @item Template Instantiations
11387 Most everything in this section also applies to template instantiations,
11388 but there are other options as well.
11389 @xref{Template Instantiation,,Where's the Template?}.
11393 When used with GNU ld version 2.8 or later on an ELF system such as
11394 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
11395 these constructs will be discarded at link time. This is known as
11398 On targets that don't support COMDAT, but do support weak symbols, GCC
11399 will use them. This way one copy will override all the others, but
11400 the unused copies will still take up space in the executable.
11402 For targets which do not support either COMDAT or weak symbols,
11403 most entities with vague linkage will be emitted as local symbols to
11404 avoid duplicate definition errors from the linker. This will not happen
11405 for local statics in inlines, however, as having multiple copies will
11406 almost certainly break things.
11408 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
11409 another way to control placement of these constructs.
11411 @node C++ Interface
11412 @section #pragma interface and implementation
11414 @cindex interface and implementation headers, C++
11415 @cindex C++ interface and implementation headers
11416 @cindex pragmas, interface and implementation
11418 @code{#pragma interface} and @code{#pragma implementation} provide the
11419 user with a way of explicitly directing the compiler to emit entities
11420 with vague linkage (and debugging information) in a particular
11423 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
11424 most cases, because of COMDAT support and the ``key method'' heuristic
11425 mentioned in @ref{Vague Linkage}. Using them can actually cause your
11426 program to grow due to unnecessary out-of-line copies of inline
11427 functions. Currently (3.4) the only benefit of these
11428 @code{#pragma}s is reduced duplication of debugging information, and
11429 that should be addressed soon on DWARF 2 targets with the use of
11433 @item #pragma interface
11434 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
11435 @kindex #pragma interface
11436 Use this directive in @emph{header files} that define object classes, to save
11437 space in most of the object files that use those classes. Normally,
11438 local copies of certain information (backup copies of inline member
11439 functions, debugging information, and the internal tables that implement
11440 virtual functions) must be kept in each object file that includes class
11441 definitions. You can use this pragma to avoid such duplication. When a
11442 header file containing @samp{#pragma interface} is included in a
11443 compilation, this auxiliary information will not be generated (unless
11444 the main input source file itself uses @samp{#pragma implementation}).
11445 Instead, the object files will contain references to be resolved at link
11448 The second form of this directive is useful for the case where you have
11449 multiple headers with the same name in different directories. If you
11450 use this form, you must specify the same string to @samp{#pragma
11453 @item #pragma implementation
11454 @itemx #pragma implementation "@var{objects}.h"
11455 @kindex #pragma implementation
11456 Use this pragma in a @emph{main input file}, when you want full output from
11457 included header files to be generated (and made globally visible). The
11458 included header file, in turn, should use @samp{#pragma interface}.
11459 Backup copies of inline member functions, debugging information, and the
11460 internal tables used to implement virtual functions are all generated in
11461 implementation files.
11463 @cindex implied @code{#pragma implementation}
11464 @cindex @code{#pragma implementation}, implied
11465 @cindex naming convention, implementation headers
11466 If you use @samp{#pragma implementation} with no argument, it applies to
11467 an include file with the same basename@footnote{A file's @dfn{basename}
11468 was the name stripped of all leading path information and of trailing
11469 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
11470 file. For example, in @file{allclass.cc}, giving just
11471 @samp{#pragma implementation}
11472 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
11474 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
11475 an implementation file whenever you would include it from
11476 @file{allclass.cc} even if you never specified @samp{#pragma
11477 implementation}. This was deemed to be more trouble than it was worth,
11478 however, and disabled.
11480 Use the string argument if you want a single implementation file to
11481 include code from multiple header files. (You must also use
11482 @samp{#include} to include the header file; @samp{#pragma
11483 implementation} only specifies how to use the file---it doesn't actually
11486 There is no way to split up the contents of a single header file into
11487 multiple implementation files.
11490 @cindex inlining and C++ pragmas
11491 @cindex C++ pragmas, effect on inlining
11492 @cindex pragmas in C++, effect on inlining
11493 @samp{#pragma implementation} and @samp{#pragma interface} also have an
11494 effect on function inlining.
11496 If you define a class in a header file marked with @samp{#pragma
11497 interface}, the effect on an inline function defined in that class is
11498 similar to an explicit @code{extern} declaration---the compiler emits
11499 no code at all to define an independent version of the function. Its
11500 definition is used only for inlining with its callers.
11502 @opindex fno-implement-inlines
11503 Conversely, when you include the same header file in a main source file
11504 that declares it as @samp{#pragma implementation}, the compiler emits
11505 code for the function itself; this defines a version of the function
11506 that can be found via pointers (or by callers compiled without
11507 inlining). If all calls to the function can be inlined, you can avoid
11508 emitting the function by compiling with @option{-fno-implement-inlines}.
11509 If any calls were not inlined, you will get linker errors.
11511 @node Template Instantiation
11512 @section Where's the Template?
11513 @cindex template instantiation
11515 C++ templates are the first language feature to require more
11516 intelligence from the environment than one usually finds on a UNIX
11517 system. Somehow the compiler and linker have to make sure that each
11518 template instance occurs exactly once in the executable if it is needed,
11519 and not at all otherwise. There are two basic approaches to this
11520 problem, which are referred to as the Borland model and the Cfront model.
11523 @item Borland model
11524 Borland C++ solved the template instantiation problem by adding the code
11525 equivalent of common blocks to their linker; the compiler emits template
11526 instances in each translation unit that uses them, and the linker
11527 collapses them together. The advantage of this model is that the linker
11528 only has to consider the object files themselves; there is no external
11529 complexity to worry about. This disadvantage is that compilation time
11530 is increased because the template code is being compiled repeatedly.
11531 Code written for this model tends to include definitions of all
11532 templates in the header file, since they must be seen to be
11536 The AT&T C++ translator, Cfront, solved the template instantiation
11537 problem by creating the notion of a template repository, an
11538 automatically maintained place where template instances are stored. A
11539 more modern version of the repository works as follows: As individual
11540 object files are built, the compiler places any template definitions and
11541 instantiations encountered in the repository. At link time, the link
11542 wrapper adds in the objects in the repository and compiles any needed
11543 instances that were not previously emitted. The advantages of this
11544 model are more optimal compilation speed and the ability to use the
11545 system linker; to implement the Borland model a compiler vendor also
11546 needs to replace the linker. The disadvantages are vastly increased
11547 complexity, and thus potential for error; for some code this can be
11548 just as transparent, but in practice it can been very difficult to build
11549 multiple programs in one directory and one program in multiple
11550 directories. Code written for this model tends to separate definitions
11551 of non-inline member templates into a separate file, which should be
11552 compiled separately.
11555 When used with GNU ld version 2.8 or later on an ELF system such as
11556 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
11557 Borland model. On other systems, G++ implements neither automatic
11560 A future version of G++ will support a hybrid model whereby the compiler
11561 will emit any instantiations for which the template definition is
11562 included in the compile, and store template definitions and
11563 instantiation context information into the object file for the rest.
11564 The link wrapper will extract that information as necessary and invoke
11565 the compiler to produce the remaining instantiations. The linker will
11566 then combine duplicate instantiations.
11568 In the mean time, you have the following options for dealing with
11569 template instantiations:
11574 Compile your template-using code with @option{-frepo}. The compiler will
11575 generate files with the extension @samp{.rpo} listing all of the
11576 template instantiations used in the corresponding object files which
11577 could be instantiated there; the link wrapper, @samp{collect2}, will
11578 then update the @samp{.rpo} files to tell the compiler where to place
11579 those instantiations and rebuild any affected object files. The
11580 link-time overhead is negligible after the first pass, as the compiler
11581 will continue to place the instantiations in the same files.
11583 This is your best option for application code written for the Borland
11584 model, as it will just work. Code written for the Cfront model will
11585 need to be modified so that the template definitions are available at
11586 one or more points of instantiation; usually this is as simple as adding
11587 @code{#include <tmethods.cc>} to the end of each template header.
11589 For library code, if you want the library to provide all of the template
11590 instantiations it needs, just try to link all of its object files
11591 together; the link will fail, but cause the instantiations to be
11592 generated as a side effect. Be warned, however, that this may cause
11593 conflicts if multiple libraries try to provide the same instantiations.
11594 For greater control, use explicit instantiation as described in the next
11598 @opindex fno-implicit-templates
11599 Compile your code with @option{-fno-implicit-templates} to disable the
11600 implicit generation of template instances, and explicitly instantiate
11601 all the ones you use. This approach requires more knowledge of exactly
11602 which instances you need than do the others, but it's less
11603 mysterious and allows greater control. You can scatter the explicit
11604 instantiations throughout your program, perhaps putting them in the
11605 translation units where the instances are used or the translation units
11606 that define the templates themselves; you can put all of the explicit
11607 instantiations you need into one big file; or you can create small files
11614 template class Foo<int>;
11615 template ostream& operator <<
11616 (ostream&, const Foo<int>&);
11619 for each of the instances you need, and create a template instantiation
11620 library from those.
11622 If you are using Cfront-model code, you can probably get away with not
11623 using @option{-fno-implicit-templates} when compiling files that don't
11624 @samp{#include} the member template definitions.
11626 If you use one big file to do the instantiations, you may want to
11627 compile it without @option{-fno-implicit-templates} so you get all of the
11628 instances required by your explicit instantiations (but not by any
11629 other files) without having to specify them as well.
11631 G++ has extended the template instantiation syntax given in the ISO
11632 standard to allow forward declaration of explicit instantiations
11633 (with @code{extern}), instantiation of the compiler support data for a
11634 template class (i.e.@: the vtable) without instantiating any of its
11635 members (with @code{inline}), and instantiation of only the static data
11636 members of a template class, without the support data or member
11637 functions (with (@code{static}):
11640 extern template int max (int, int);
11641 inline template class Foo<int>;
11642 static template class Foo<int>;
11646 Do nothing. Pretend G++ does implement automatic instantiation
11647 management. Code written for the Borland model will work fine, but
11648 each translation unit will contain instances of each of the templates it
11649 uses. In a large program, this can lead to an unacceptable amount of code
11653 @node Bound member functions
11654 @section Extracting the function pointer from a bound pointer to member function
11656 @cindex pointer to member function
11657 @cindex bound pointer to member function
11659 In C++, pointer to member functions (PMFs) are implemented using a wide
11660 pointer of sorts to handle all the possible call mechanisms; the PMF
11661 needs to store information about how to adjust the @samp{this} pointer,
11662 and if the function pointed to is virtual, where to find the vtable, and
11663 where in the vtable to look for the member function. If you are using
11664 PMFs in an inner loop, you should really reconsider that decision. If
11665 that is not an option, you can extract the pointer to the function that
11666 would be called for a given object/PMF pair and call it directly inside
11667 the inner loop, to save a bit of time.
11669 Note that you will still be paying the penalty for the call through a
11670 function pointer; on most modern architectures, such a call defeats the
11671 branch prediction features of the CPU@. This is also true of normal
11672 virtual function calls.
11674 The syntax for this extension is
11678 extern int (A::*fp)();
11679 typedef int (*fptr)(A *);
11681 fptr p = (fptr)(a.*fp);
11684 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11685 no object is needed to obtain the address of the function. They can be
11686 converted to function pointers directly:
11689 fptr p1 = (fptr)(&A::foo);
11692 @opindex Wno-pmf-conversions
11693 You must specify @option{-Wno-pmf-conversions} to use this extension.
11695 @node C++ Attributes
11696 @section C++-Specific Variable, Function, and Type Attributes
11698 Some attributes only make sense for C++ programs.
11701 @item init_priority (@var{priority})
11702 @cindex init_priority attribute
11705 In Standard C++, objects defined at namespace scope are guaranteed to be
11706 initialized in an order in strict accordance with that of their definitions
11707 @emph{in a given translation unit}. No guarantee is made for initializations
11708 across translation units. However, GNU C++ allows users to control the
11709 order of initialization of objects defined at namespace scope with the
11710 @code{init_priority} attribute by specifying a relative @var{priority},
11711 a constant integral expression currently bounded between 101 and 65535
11712 inclusive. Lower numbers indicate a higher priority.
11714 In the following example, @code{A} would normally be created before
11715 @code{B}, but the @code{init_priority} attribute has reversed that order:
11718 Some_Class A __attribute__ ((init_priority (2000)));
11719 Some_Class B __attribute__ ((init_priority (543)));
11723 Note that the particular values of @var{priority} do not matter; only their
11726 @item java_interface
11727 @cindex java_interface attribute
11729 This type attribute informs C++ that the class is a Java interface. It may
11730 only be applied to classes declared within an @code{extern "Java"} block.
11731 Calls to methods declared in this interface will be dispatched using GCJ's
11732 interface table mechanism, instead of regular virtual table dispatch.
11736 See also @xref{Namespace Association}.
11738 @node Namespace Association
11739 @section Namespace Association
11741 @strong{Caution:} The semantics of this extension are not fully
11742 defined. Users should refrain from using this extension as its
11743 semantics may change subtly over time. It is possible that this
11744 extension will be removed in future versions of G++.
11746 A using-directive with @code{__attribute ((strong))} is stronger
11747 than a normal using-directive in two ways:
11751 Templates from the used namespace can be specialized and explicitly
11752 instantiated as though they were members of the using namespace.
11755 The using namespace is considered an associated namespace of all
11756 templates in the used namespace for purposes of argument-dependent
11760 The used namespace must be nested within the using namespace so that
11761 normal unqualified lookup works properly.
11763 This is useful for composing a namespace transparently from
11764 implementation namespaces. For example:
11769 template <class T> struct A @{ @};
11771 using namespace debug __attribute ((__strong__));
11772 template <> struct A<int> @{ @}; // @r{ok to specialize}
11774 template <class T> void f (A<T>);
11779 f (std::A<float>()); // @r{lookup finds} std::f
11785 @section Type Traits
11787 The C++ front-end implements syntactic extensions that allow to
11788 determine at compile time various characteristics of a type (or of a
11792 @item __has_nothrow_assign (type)
11793 If @code{type} is const qualified or is a reference type then the trait is
11794 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
11795 is true, else if @code{type} is a cv class or union type with copy assignment
11796 operators that are known not to throw an exception then the trait is true,
11797 else it is false. Requires: @code{type} shall be a complete type, an array
11798 type of unknown bound, or is a @code{void} type.
11800 @item __has_nothrow_copy (type)
11801 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
11802 @code{type} is a cv class or union type with copy constructors that
11803 are known not to throw an exception then the trait is true, else it is false.
11804 Requires: @code{type} shall be a complete type, an array type of
11805 unknown bound, or is a @code{void} type.
11807 @item __has_nothrow_constructor (type)
11808 If @code{__has_trivial_constructor (type)} is true then the trait is
11809 true, else if @code{type} is a cv class or union type (or array
11810 thereof) with a default constructor that is known not to throw an
11811 exception then the trait is true, else it is false. Requires:
11812 @code{type} shall be a complete type, an array type of unknown bound,
11813 or is a @code{void} type.
11815 @item __has_trivial_assign (type)
11816 If @code{type} is const qualified or is a reference type then the trait is
11817 false. Otherwise if @code{__is_pod (type)} is true then the trait is
11818 true, else if @code{type} is a cv class or union type with a trivial
11819 copy assignment ([class.copy]) then the trait is true, else it is
11820 false. Requires: @code{type} shall be a complete type, an array type
11821 of unknown bound, or is a @code{void} type.
11823 @item __has_trivial_copy (type)
11824 If @code{__is_pod (type)} is true or @code{type} is a reference type
11825 then the trait is true, else if @code{type} is a cv class or union type
11826 with a trivial copy constructor ([class.copy]) then the trait
11827 is true, else it is false. Requires: @code{type} shall be a complete
11828 type, an array type of unknown bound, or is a @code{void} type.
11830 @item __has_trivial_constructor (type)
11831 If @code{__is_pod (type)} is true then the trait is true, else if
11832 @code{type} is a cv class or union type (or array thereof) with a
11833 trivial default constructor ([class.ctor]) then the trait is true,
11834 else it is false. Requires: @code{type} shall be a complete type, an
11835 array type of unknown bound, or is a @code{void} type.
11837 @item __has_trivial_destructor (type)
11838 If @code{__is_pod (type)} is true or @code{type} is a reference type then
11839 the trait is true, else if @code{type} is a cv class or union type (or
11840 array thereof) with a trivial destructor ([class.dtor]) then the trait
11841 is true, else it is false. Requires: @code{type} shall be a complete
11842 type, an array type of unknown bound, or is a @code{void} type.
11844 @item __has_virtual_destructor (type)
11845 If @code{type} is a class type with a virtual destructor
11846 ([class.dtor]) then the trait is true, else it is false. Requires:
11847 @code{type} shall be a complete type, an array type of unknown bound,
11848 or is a @code{void} type.
11850 @item __is_abstract (type)
11851 If @code{type} is an abstract class ([class.abstract]) then the trait
11852 is true, else it is false. Requires: @code{type} shall be a complete
11853 type, an array type of unknown bound, or is a @code{void} type.
11855 @item __is_base_of (base_type, derived_type)
11856 If @code{base_type} is a base class of @code{derived_type}
11857 ([class.derived]) then the trait is true, otherwise it is false.
11858 Top-level cv qualifications of @code{base_type} and
11859 @code{derived_type} are ignored. For the purposes of this trait, a
11860 class type is considered is own base. Requires: if @code{__is_class
11861 (base_type)} and @code{__is_class (derived_type)} are true and
11862 @code{base_type} and @code{derived_type} are not the same type
11863 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
11864 type. Diagnostic is produced if this requirement is not met.
11866 @item __is_class (type)
11867 If @code{type} is a cv class type, and not a union type
11868 ([basic.compound]) the the trait is true, else it is false.
11870 @item __is_empty (type)
11871 If @code{__is_class (type)} is false then the trait is false.
11872 Otherwise @code{type} is considered empty if and only if: @code{type}
11873 has no non-static data members, or all non-static data members, if
11874 any, are bit-fields of lenght 0, and @code{type} has no virtual
11875 members, and @code{type} has no virtual base classes, and @code{type}
11876 has no base classes @code{base_type} for which
11877 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
11878 be a complete type, an array type of unknown bound, or is a
11881 @item __is_enum (type)
11882 If @code{type} is a cv enumeration type ([basic.compound]) the the trait is
11883 true, else it is false.
11885 @item __is_pod (type)
11886 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
11887 else it is false. Requires: @code{type} shall be a complete type,
11888 an array type of unknown bound, or is a @code{void} type.
11890 @item __is_polymorphic (type)
11891 If @code{type} is a polymorphic class ([class.virtual]) then the trait
11892 is true, else it is false. Requires: @code{type} shall be a complete
11893 type, an array type of unknown bound, or is a @code{void} type.
11895 @item __is_union (type)
11896 If @code{type} is a cv union type ([basic.compound]) the the trait is
11897 true, else it is false.
11901 @node Java Exceptions
11902 @section Java Exceptions
11904 The Java language uses a slightly different exception handling model
11905 from C++. Normally, GNU C++ will automatically detect when you are
11906 writing C++ code that uses Java exceptions, and handle them
11907 appropriately. However, if C++ code only needs to execute destructors
11908 when Java exceptions are thrown through it, GCC will guess incorrectly.
11909 Sample problematic code is:
11912 struct S @{ ~S(); @};
11913 extern void bar(); // @r{is written in Java, and may throw exceptions}
11922 The usual effect of an incorrect guess is a link failure, complaining of
11923 a missing routine called @samp{__gxx_personality_v0}.
11925 You can inform the compiler that Java exceptions are to be used in a
11926 translation unit, irrespective of what it might think, by writing
11927 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11928 @samp{#pragma} must appear before any functions that throw or catch
11929 exceptions, or run destructors when exceptions are thrown through them.
11931 You cannot mix Java and C++ exceptions in the same translation unit. It
11932 is believed to be safe to throw a C++ exception from one file through
11933 another file compiled for the Java exception model, or vice versa, but
11934 there may be bugs in this area.
11936 @node Deprecated Features
11937 @section Deprecated Features
11939 In the past, the GNU C++ compiler was extended to experiment with new
11940 features, at a time when the C++ language was still evolving. Now that
11941 the C++ standard is complete, some of those features are superseded by
11942 superior alternatives. Using the old features might cause a warning in
11943 some cases that the feature will be dropped in the future. In other
11944 cases, the feature might be gone already.
11946 While the list below is not exhaustive, it documents some of the options
11947 that are now deprecated:
11950 @item -fexternal-templates
11951 @itemx -falt-external-templates
11952 These are two of the many ways for G++ to implement template
11953 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11954 defines how template definitions have to be organized across
11955 implementation units. G++ has an implicit instantiation mechanism that
11956 should work just fine for standard-conforming code.
11958 @item -fstrict-prototype
11959 @itemx -fno-strict-prototype
11960 Previously it was possible to use an empty prototype parameter list to
11961 indicate an unspecified number of parameters (like C), rather than no
11962 parameters, as C++ demands. This feature has been removed, except where
11963 it is required for backwards compatibility @xref{Backwards Compatibility}.
11966 G++ allows a virtual function returning @samp{void *} to be overridden
11967 by one returning a different pointer type. This extension to the
11968 covariant return type rules is now deprecated and will be removed from a
11971 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11972 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11973 and will be removed in a future version. Code using these operators
11974 should be modified to use @code{std::min} and @code{std::max} instead.
11976 The named return value extension has been deprecated, and is now
11979 The use of initializer lists with new expressions has been deprecated,
11980 and is now removed from G++.
11982 Floating and complex non-type template parameters have been deprecated,
11983 and are now removed from G++.
11985 The implicit typename extension has been deprecated and is now
11988 The use of default arguments in function pointers, function typedefs
11989 and other places where they are not permitted by the standard is
11990 deprecated and will be removed from a future version of G++.
11992 G++ allows floating-point literals to appear in integral constant expressions,
11993 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11994 This extension is deprecated and will be removed from a future version.
11996 G++ allows static data members of const floating-point type to be declared
11997 with an initializer in a class definition. The standard only allows
11998 initializers for static members of const integral types and const
11999 enumeration types so this extension has been deprecated and will be removed
12000 from a future version.
12002 @node Backwards Compatibility
12003 @section Backwards Compatibility
12004 @cindex Backwards Compatibility
12005 @cindex ARM [Annotated C++ Reference Manual]
12007 Now that there is a definitive ISO standard C++, G++ has a specification
12008 to adhere to. The C++ language evolved over time, and features that
12009 used to be acceptable in previous drafts of the standard, such as the ARM
12010 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
12011 compilation of C++ written to such drafts, G++ contains some backwards
12012 compatibilities. @emph{All such backwards compatibility features are
12013 liable to disappear in future versions of G++.} They should be considered
12014 deprecated @xref{Deprecated Features}.
12018 If a variable is declared at for scope, it used to remain in scope until
12019 the end of the scope which contained the for statement (rather than just
12020 within the for scope). G++ retains this, but issues a warning, if such a
12021 variable is accessed outside the for scope.
12023 @item Implicit C language
12024 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
12025 scope to set the language. On such systems, all header files are
12026 implicitly scoped inside a C language scope. Also, an empty prototype
12027 @code{()} will be treated as an unspecified number of arguments, rather
12028 than no arguments, as C++ demands.